In the era of 2nm, GAA, multilayer wiring, and AI chips, it is said that reliability, rather than performance, is becoming the central design constraint.

The reasons are:

・Film thicknesses and wiring widths become extremely thin with scaling

・Current density and electric field strength rise sharply

・Materials are approaching their limits

・Operation in high-temperature environments (temperatures inside AI chips exceed 100°C)

・Growing demand for automotive and industrial applications

In particular, “no failures” and “small parameter variation” are essential requirements for end-product manufacturers (EVs / data centers / autonomous driving).

(1) TDDB (Time-Dependent Dielectric Breakdown)

A phenomenon in which the insulating film (SiO₂/High-k) gradually degrades over time, eventually leading to dielectric breakdown.

Characteristics:

・Accelerated at higher electric fields

・Its importance has increased with the adoption of High-k

・More serious at advanced nodes with thinner gate oxides

・Slow degradation → sudden breakdown at the end

Particular caution is needed in “gate-enclosed structures” such as GAA.

(2) BTI (Bias Temperature Instability)

A phenomenon in which the threshold voltage (Vth) shifts when voltage is continuously applied to the gate.

Types:

・NBTI: common in pMOS

・PBTI: occurs in nMOS

Effects:

・Long-term stress → Vth shift

・Increased circuit delay

・Fatal for the high-speed operation of AI chips

Causes:

・Defects in High-k films

・Trap capture and release

BTI is one of “the biggest enemies of the 2nm era.”

(3) HCI (Hot Carrier Injection)

A phenomenon in which high-energy electrons in the channel damage and degrade the gate insulating film.

Characteristics:

・Increases with high voltage and high-speed switching

・Serious in RF and AI applications

・Reduced transistor gm and Vth shift

It remains important in GAA as well.

(4) EM (Electromigration)

A phenomenon in which the enormous current flowing through metal wiring causes metal atoms to migrate, resulting in disconnections, voids, and hillocks.

Particularly in AI chips:

・Current density is very high

・Wiring widths are extremely narrow

・Temperature rise is large

→ The impact of EM is significant, and wiring materials are approaching their limits.

Wiring materials are shifting from Cu → Co / Ru.

(5) Reliability Requirements Specific to Automotive (AEC-Q100)

Because automotive devices operate under harsh conditions, the requirements are on a different level.

Examples:

・Temperature cycling from -40°C to 150°C

・High-humidity environments

・High current and noise

・Long service life (10–15 years)

For this reason, reliability testing is extremely demanding for:

・Power devices (SiC)

・ADAS chips

・Safety chips

TDDB Test

Applies high voltage until the insulating film breaks down, and measures the result.

BTI Test

Evaluates Vth shift under temperature and voltage stress.

EM Test

Measures the time until wiring opens at high temperature and high current.

HTOL (High Temperature Operating Life)

A continuous operation test at high temperature.

TCT (Temperature Cycling Test)

Evaluates degradation caused by temperature changes.

These require a systematic understanding of reliability engineering.

(1) Material Optimization

・Reducing defects in High-k films

・Work-function tuning of metal gates

・Adoption of Co/Ru wiring

・Reducing damage to ULK materials

(2) Process Control

・Reducing damage from deposition and etching

・Improving annealing conditions

・Reducing oxygen and moisture

(3) Device Design Optimization

・Electric-field-distributing structures

・Improved gate shapes

・EM-resistant wiring layouts

・Eliminating excessive stress concentration

(4) Circuit-Level Measures

・Appropriate margin design

・Digital correction to absorb errors

・High-temperature compensation

AI chips → high heat × high current × long operating hours → more EM/TDDB/BTI

Automotive chips → temperature cycling × long service life → material fatigue and interface degradation

Today, the design philosophy of prioritizing reliability over performance is growing stronger again.

The more scaling advances, the more reliability becomes an essential challenge.

● Redefining the fundamental properties of materials (renewing ULK, High-k, and metal materials)

● Degradation prediction using machine learning

● Integrated analysis of DFM × reliability

● The challenge of eliminating EM through backside power delivery

● Stronger thermal and stress models for 3D stacked devices

In particular, AI × reliability analysis has already begun to be introduced in volume production and is becoming key to process control.

・Reliability is a central challenge from 2nm onward

・Multiple degradation mechanisms exist, such as TDDB / BTI / HCI / EM

・Problems become pronounced with High-k and ultra-thin wiring

・AI chips face the greatest reliability load due to heat × current

・Automotive temperature and lifetime requirements are on a different level

・Comprehensive optimization of materials × process × design × circuits is essential

1.Why do reliability problems increase with scaling?

2.Explain why BTI becomes important in AI chips.

3.What is the mechanism by which EM (electromigration) occurs?

 

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.