Yield is “the percentage of chips produced that can be shipped as good products.”

Examples:

・95 good out of 100 → 95% yield

・60 good out of 100 → 60% yield

At advanced nodes, yield improves from:

・Initial volume production: 30–50%

・Stable volume production: 90% or more

Semiconductor defects are not simple.

The causes are compound:

・Equipment variation

・Material fluctuations

・Environmental changes

・Design dependencies

・Interactions between processes

They cannot be tracked by human intuition alone.

That is exactly why statistics and data science are needed.

The foundational technology of yield management.

It statistically monitors process data to detect abnormalities early.

Typical indicators:

・Film thickness

・CD dimensions

・Etching depth

・Implantation dose

Main control methods:

・Control charts

・Mean value control

・Variation monitoring

The aim is to stop problems before defects appear.

They quantify process stability.

・Cp: capability in terms of variation

・Cpk: capability including center offset

Guidelines:

・Cpk below 1.0 → unstable

・Cpk 1.33 or higher → volume production standard

・Cpk 1.67 or higher → high-quality process

They are monitored constantly in advanced fabs.

Yield is directly linked to defect density.

・Particles

・Scratches

・Residues

The lower the defect density, the higher the rate of good chips.

It is evaluated with wafer map analysis.

It visualizes the distribution of defects.

Examples:

・Concentrated in the center → deposition problem

・Concentrated at the edge → non-uniform etching

・Stripes → equipment-related

Causes are inferred from spatial distribution.

Technology for monitoring equipment data.

It analyzes equipment behavior in real time.

Examples of monitored items:

・Temperature waveforms

・Pressure fluctuations

・RF power

・Motor current

Signs of abnormalities are detected in advance.

Big data analysis of equipment.

・Data from thousands of sensors

・Long-term trends

・Comparisons between tools

Broader in scope than SPC.

It is also used for predictive maintenance of equipment.

Causes cannot be identified from a single parameter.

Multivariate analysis is used to extract correlations.

Examples:

・Temperature × pressure × time

・Film thickness × CD × etching conditions

Complex causal relationships are visualized.

The standard approach to process optimization.

Multiple conditions are varied systematically to derive the optimal conditions.

Examples:

・Deposition temperature

・Gas flow rate

・Pressure

The conditions that maximize yield are explored.

The main players in recent years.

Areas of use:

・Defect classification

・Yield prediction

・Anomaly detection

・Cause estimation

Correlations that humans cannot see are extracted.

The shape of the next-generation fab.

・In-line yield monitoring

・Automatic correction

・Feedforward control

Processes optimize themselves autonomously.

Yield is not determined by a single process.

Integration targets:

・Design data

・Equipment data

・Inspection data

・Material lots

Integrated analysis identifies the true cause.

・Fusion of AI × SPC

・Digital twin yield prediction

・Deep learning on defect images

・Autonomous process control

・Smart fabs

Yield improvement is shifting from experience-based engineering to data science.

・Yield is the core indicator of factory profit

・SPC is the foundation of quality control

・Defect density and yield are directly linked

・FDC detects equipment abnormalities early

・DOE derives optimal conditions

・AI analysis is the main player of the next generation

1.Explain the purpose of SPC in one sentence.

2.What can be learned from wafer map analysis?

3.Why is DOE effective for yield improvement?

4.What are the strengths of AI yield analysis?

 

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.