Temperature-Assisted Flash Memory Programming for Neuromorphic Computing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current flash memory technologies for neuromorphic computing face challenges in reducing operating voltage and speeding up programming while maintaining long retention, especially in standby conditions, due to limitations in carrier transport mechanisms.

Innovation Solution

The implementation of temperature-assisted programming using a structurally defective high-k tunneling dielectric with a high carbon concentration, which enhances carrier transport through Frenkel-Poole conduction, allowing for reduced operation voltage and accelerated programming by varying temperatures during training and inference.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Use of energy by moving object

If conventional flash memory programming is used, then programming can be performed, but operating voltage is high and programming speed is slow

Engineering Contradiction:

Improveoperating voltageVSAvoidprogramming speed

Core Design Contradiction:

Use of energy by moving objectVSProductivity

Solution Approach 1:

The patent changes the temperature parameter during programming operations. By heating the flash memory device to elevated temperatures (e.g., 85°C to 125°C) during programming, carrier transport through the tunnel dielectric is enhanced, enabling programming at lower voltages and faster speeds. After programming, the device is cooled to ambient temperature for normal operation and data retention.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent implements periodic temperature cycling where the device is heated during programming phases and cooled during retention phases. This periodic thermal action allows the system to achieve fast programming when needed while maintaining data retention by cooling down, effectively separating the programming and retention operational requirements.

Inventive Principle:

Principle #19Periodic action

2Loss of time

If programming speed is increased, then training time is reduced, but retention performance deteriorates

Engineering Contradiction:

Improvetraining timeVSAvoiddata retention

Core Design Contradiction:

Loss of timeVSReliability

Solution Approach 1:

The patent applies temperature as a controllable parameter that is changed based on the operational phase. During training/programming, high temperature is applied to accelerate carrier transport and reduce programming time. During retention/inference, the device is cooled to ambient temperature to ensure long-term data retention, thus resolving the contradiction between speed and reliability.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent makes the thermal state of the device dynamic rather than static. The temperature is actively adjusted based on the operational requirements: elevated during programming to enhance speed, and reduced during retention to ensure reliability. This dynamic thermal management allows the system to optimize performance for each operational phase.

Inventive Principle:

Principle #15Dynamics

3Productivity

If high temperature is used during programming, then carrier transport is enhanced, but device reliability during operation may be compromised

Engineering Contradiction:

Improveprogramming efficiencyVSAvoiddevice stability

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent employs periodic thermal cycles where high temperature is applied only during programming operations and the device is cooled to ambient temperature during normal operation and data retention. This periodic heating and cooling ensures that the benefits of enhanced carrier transport are achieved when needed while the device maintains stability and reliability during operational phases.

Inventive Principle:

Principle #19Periodic action

Solution Approach 2:

The patent applies heat in advance during the programming phase to enhance carrier transport before the actual data is written and needs to be retained. By pre-heating the device before programming operations, the system prepares the optimal conditions for fast programming, then cools down afterward to ensure stability during retention, preventing reliability issues.

Inventive Principle:

Principle #10Preliminary action

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach effectively reduces operating voltage and training time during programming while maintaining long-term retention, enabling efficient operation of flash memory in neuromorphic computing systems by leveraging thermally-enhanced carrier transport mechanisms.

Implementation Method 1

Increasing the temperature enhances carrier transport through the structurally defective high-k tunneling dielectric thus facilitating the programming of the flash memory

Methodology Applied

Scientific EffectFrenkel-Poole conduction: Pool-Frenkel Effect

Data Source

PatentUS11631462B2Temperature assisted programming of flash memory for neuromorphic computing

Publication Date: 2023.04.18 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

A method is presented for temperature assisted programming of flash memory for neuromorphic computing. The method includes training a chip in an environment having a first temperature, adjusting the first temperature to a second temperature in the environment, and employing the chip for inference in the second temperature environment. The first temperature is about 125° C. or higher and the second temperature is about 50° C. or lower.