VCMA MeRAM Array With On-Chip Electromagnet for AI Computing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional processing units struggle to meet the demands of large language models due to high power consumption and financial costs, and existing analog compute-in-memory solutions like ReRAM, PCM, and STT-MRAM are inefficient for AI computation.

Innovation Solution

A magnetoelectric device with an on-chip electromagnet is used to generate magnetic fields for programming VCMA MeRAM cells, enabling efficient vector-matrix multiplication for AI operations without external magnetic fields.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Use of energy by moving object

If conventional processing units are used for AI computation, then computational capability is provided, but power consumption is high and costs are high

Engineering Contradiction:

Improvepower consumptionVSAvoidcomputational capability

Core Design Contradiction:

Use of energy by moving objectVSProductivity

Solution Approach 1:

The patent replaces conventional electronic processing with magnetoelectric computing. VCMA MeRAM cells use voltage-controlled magnetic anisotropy to perform computations, substituting traditional electronic mechanisms with magnetoelectric effects. This enables AI workloads to be executed with significantly lower power consumption while maintaining computational capability through in-memory computing operations.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The magnetoelectric device serves multiple functions: it performs both storage and computation operations within the same memory array. The VCMA MeRAM cells can store data and simultaneously perform vector-matrix multiplication operations, eliminating the need for separate processing units and reducing overall system power consumption.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Productivity

If ReRAM, PCM, and STT-MRAM are used for analog compute-in-memory, then memory-based computation is enabled, but efficiency for AI computation is insufficient

Engineering Contradiction:

Improvecomputation efficiencyVSAvoidsuitability for AI computation

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent changes the fundamental operating parameters of memory-based computation by using voltage-controlled magnetic anisotropy instead of resistive or phase-change mechanisms. The VCMA effect enables precise control of magnetic anisotropy energy barriers through voltage, allowing for more efficient and accurate analog computations specifically optimized for AI workloads compared to traditional ReRAM, PCM, or STT-MRAM approaches.

Inventive Principle:

Principle #35Parameter changes

3Ease of manufacture

If external magnetic fields are used for programming VCMA MeRAM cells, then magnetic field programming is achieved, but device complexity and cost increase

Engineering Contradiction:

Improveprogramming capabilityVSAvoidsystem complexity

Core Design Contradiction:

Ease of manufactureVSDevice complexity

Solution Approach 1:

The patent merges the magnetic field generation function directly into the VCMA MeRAM cell structure by integrating on-chip electromagnets. This integration combines the programming mechanism with the memory cells themselves, eliminating the need for external magnetic field sources and reducing overall system complexity while maintaining full programming capability.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The electromagnets are nested within or integrated into the VCMA MeRAM array structure, with conductive lines positioned to generate magnetic fields precisely where needed. This nested configuration allows the programming mechanism to be embedded within the memory device footprint, reducing external requirements and simplifying the overall system architecture.

Inventive Principle:

Principle #7Nested doll (Nesting)

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

The on-chip electromagnet improves power efficiency and reduces costs for VCMA MeRAM arrays, enhancing their performance for AI computation.

Implementation Method 1

an electromagnet located over the substrate and configured to apply a magnetic field to the array of magnetoelectric unit cells

Methodology Applied

Scientific EffectElectromagnet: Electromagnet

Implementation Method 2

Voltage-controlled magnetic anisotropy (VCMA) refers to magnetic anisotropy that increase or decreases with application of an electric field across a magnetic tunnel junction

Methodology Applied

Scientific EffectVoltage-controlled magnetic anisotropy:

Data Source

PatentUS20250393218A1Voltage-controlled magnetic anisotropy device including an on-chip electromagnet

Publication Date: 2025.12.25 WESTERN DIGITAL TECHNOLOGIES INC

AI summary

A magnetoelectric device includes an array of magnetoelectric unit cells located over a substrate, and each of the magnetoelectric unit cells includes a first electrode, a second electrode, and a magnetic tunnel junction located between the first electrode and the second electrode, word lines contacting a respective row of first electrodes of the array of magnetoelectric unit cells, bit lines contacting a respective column of second electrodes of the array of magnetoelectric unit cells, and an electromagnet located over the substrate and configured to apply a magnetic field to the array of magnetoelectric unit cells.