Scaling is approaching its limits around 2nm, and the room for performance gains with conventional “silicon × CMOS” is shrinking.

For this reason, research and development around the world is heading in the following directions:

● Changing materials (Si → 2D materials, ferroelectric materials, etc.)

● Changing structures (FinFET → GAA → Beyond GAA)

● Changing operating principles (electrons → spin, photons, ferroelectrics)

This is the major trend known as “post-CMOS.”

A FeFET is a new type of transistor that uses a ferroelectric material.

Features of FeFETs

・Uses an HfO₂-based ferroelectric film as the gate insulating film

・Remanent polarization (PR) can be used as memory

・Ultra-low power consumption (lower than SRAM)

・Operates as non-volatile memory

・Likely to be manufacturable with the same process as logic (CMOS compatible)

Expected Applications

・Near-memory computing for AI inference

・Ultra-low-power memory (as an NV-SRAM alternative)

・Brain-inspired devices (neuromorphic)

Challenges

・Reliability (polarization fatigue)

・Rewrite endurance

・High-temperature stability

Intel, imec, Samsung, and others are accelerating their research.

One of the most closely watched candidates for what comes after silicon.

Strengths of CNTs

・Carrier mobility more than 10 times that of Si

・High current drive capability

・Strong gate control

・Capable of low-voltage operation

・Little degradation in characteristics even with scaling

・Can be realized as CMOS-compatible devices

Why They Are Difficult to Realize

・Separating metallic and semiconducting CNTs is difficult

・Variation in tube diameter

・Difficulty of alignment (making them face the same direction)

・Technology for arranging them uniformly over large areas is immature

IBM, Stanford, and TSMC are conducting joint research.

The idea of building devices just a few atomic layers thick.

Why Are 2D Materials Attracting Attention?

・Atomic-level thickness → channel scaling is easy

・Perfectly flat surfaces → small variation

・Extremely high electric field controllability

・Resistant to short-channel effects

・Small body effect

The representative examples are MoS₂ (a 2D semiconductor) and graphene (a high-mobility material).

Challenges

・Difficulty of large-area synthesis

・High metal contact resistance

・Uniformity at mass-production scale has not been established

They are seen as the leading candidate for “post-GAA.”

Technology that uses light instead of electrons for computing and communication.

Why Optical Devices Are Needed

Power consumption in data centers is approaching its limits. With electrical wiring, delay and power consumption increase as distance grows.

Light (photons) offers:

・Zero loss due to resistance

・Low-loss long-distance transmission

・High bandwidth and high speed

Strengths of Silicon Photonics

・Can be manufactured on CMOS lines

・Compatibility with optical communications

・Optical wiring within data centers is accelerating

Current Challenges

・Light sources (lasers) cannot be made from Si

・Process yield

・Temperature dependence must be compensated for

In the future, AI chips × optical interconnects will be the leading approach.

A new computing approach that uses the spin of electrons.

Features

・Low power consumption

・Non-volatility

・Ultra-fast switching

・Directly linked to faster MRAM

Applications

・STT-MRAM (already in volume production)

・Memory that requires no power

・Neuromorphic elements

Challenges

・Material stability

・Reducing switching current

・Scaling limits

It is also “a technology that blurs the boundary between memory and logic,” and is attracting attention as a structure well suited to AI processing.

Research around the world is progressing along the following three axes:

(A) Changing Materials

・CNTs

・2D materials

・Ferroelectrics

・New insulating films

・Optical materials

(B) Changing Structures

・GAA

・Nanosheets → nanowires

・CFET (stacking N-type and P-type vertically)

・3D logic

(C) Changing Operating Principles

・Spintronics

・Neuromorphic computing

・Optical computing

・Quantum devices

These are the breakthroughs for 2nm and beyond.

In particular, CFET (Complementary FET) is being researched in earnest by TSMC, Samsung, and imec, and is likely to come next after GAA.

・New materials and new structures are essential to break through the scaling barrier

・FeFETs are key to integrating logic and memory

・CNT/2D materials are post-CMOS candidates

・Silicon photonics is a data center revolution

・Spintronics is the leading technology for non-volatility and power savings

・The world is evolving along three axes: materials × structures × computing methods

1.Explain why FeFETs are attracting attention.

2.What is the biggest reason CNT-FETs are difficult to put into practical use?

3.Give the reasons why 2D materials such as MoS₂ are well suited to scaling.

 

Column Supervisor: Koji Kakumoto (President, OTIS Co., Ltd.)

After studying abroad and working in planning and development at a trading company, he joined OTIS Co., Ltd. in 2011. While primarily working in the Corporate Planning Department, he has also served concurrently in manufacturing and technical divisions, and since 2018 has served as President, working to drive business growth and strengthen the organization.

 

This article is a general technical explanation intended for educational purposes and does not refer to any specific company, product, or technology.