InsightFab
Knowledge Base/Defect Density Analysis: Poisson Model and Yield Prediction
Semiconductor Process6 min read

Defect Density Analysis: Poisson Model and Yield Prediction

This article addresses a critical scenario in semiconductor manufacturing: a sudden drop in yield, emphasizing why relying solely on final yield is insufficient. It demonstrates how defect density, utilizing the Poisson model, is crucial for identifying root causes and predicting future yield, offering vital insights for effective manufacturing yield management.

The Day Yield Dropped to 85%, Making the Boss's Face Paler Than the Cleanroom Walls

I still remember years ago, a new product batch had just entered mass production. We expected a yield of 95% or more, but when the first batch was processed, the next day's yield report showed a direct jump to 85%! All the senior engineers on the production line were stunned. The boss's face was paler than the cleanroom walls. He immediately called a morning meeting, asking, "Which part of the process went wrong? Isn't our process stable?" Everyone looked at each other, and there was a moment of silence. Finally, I spoke up: "Boss, let's not rush. Let's calculate the defect density. It might be that a machine parameter drifted, or a particular batch was exceptionally bad."

How Does Yield Relate to Defect Density?

You might wonder, yield is yield, defects are defects; what's the connection between the two? Simply put, yield is the proportion of your products that are "defect-free." If you produce 100 wafers today and 90 of them are good, then the yield is 90%. But the question is, how do you know if these "good" wafers truly didn't encounter any defects, or if the defects were just too small to be measured?

In a semiconductor factory, we don't just look at the final yield. By the time the yield drops, it's often already several days later. What we need to do is "predict" the yield. How do we predict it? By relying on defect density.

Imagine your wafer is like a table, and the dust on it represents defects. If the table is large and there's a lot of dust, the probability of getting a "clean" table decreases. The number and distribution of these defects can be described using something called the "Poisson Model."

So the key is: through defect density, we can predict future yield. This is like checking the weather forecast at home; while not 100% accurate, it at least provides a direction.

How Is It Calculated in Practice?

We typically start by finding a test wafer, or a wafer directly from the production line, and perform a defect scan. Let's say we scanned 10 wafers, each with an area of 100 square centimeters, and found a total of 500 defects.

  1. Calculate Defect Density (D0):

Total number of defects / Total area = 500 defects / (10 wafers * 100 cm²/wafer) = 0.5 defects/cm².

This D0 is your "defects per unit area."

  1. Predict Yield using the Poisson Model:

Yield = e^(-A * D0)

Where:

* e is the base of the natural logarithm, approximately 2.71828.

* A is your chip area.

* D0 is the defect density we just calculated.

For example, if your chip area is 2 cm² and D0 is 0.5 defects/cm².

Then the predicted yield = e^(-2 * 0.5) = e^(-1) ≈ 0.3678, or 36.78%.

In other words, when you find the defect density is 0.5, your chip yield is likely only around 30%. At this point, you need to quickly investigate which machine or process is having issues.

The Most Common Pitfall: Endless Arguments with the Boss

The most common pitfall I've encountered is when everyone only looks at overall yield and ignores defect density. Once, we had a product where the yield kept dropping, from 95% to 92%. The boss said, "It's only a 3% drop, isn't it just measurement variation?" I replied, "Boss, our defect density has already surged from 0.2 to 0.4, which means the number of defects has doubled, and the yield will inevitably drop further." The boss initially didn't believe me, thinking I was making a mountain out of a molehill.

Later, we took that problematic batch of wafers for rescanning and indeed found an explosive increase in the number of defects, many of which were concentrated in the same area. Only then did we trace it back and discover that there was build-up on the reaction chamber wall of a certain etching machine, causing many particles to fall. If we had looked at the defect density earlier, instead of waiting for the yield to drop before reacting, that batch of products would not have gone bad at all.

Honestly, many times, by the time the yield report comes out, it's already several days later. But defect density can be scanned and monitored daily. It's like driving a car: do you check the fuel gauge, or do you wait for the car to break down to realize you're out of gas?

One Thing You Can Do Today

Every day, open your defect monitoring report and check if the D0 value shows any abnormal fluctuations.

Want to try it yourself?

Every tool mentioned in this article is available on InsightFab — just upload a CSV to analyze.

Go to Tools

More in this category

Yield Engineering: From Die Yield to Wafer Yield

Ever wondered about wafer fab yield numbers but felt lost? This practical article introduces you to the world of wafer yield, using a vivid "pineapple field" analogy to clarify die yield, wafer yield, and what happens to defective chips. You'll learn how to pinpoint root causes from data, determining whether issues stem from equipment or processes, making crucial semiconductor knowledge accessible.

Yield Loss Classification: Random Defects vs. Systematic Defects

This article addresses the common factory challenge of unstable yield by differentiating between 'random defects' and 'systematic defects.' It provides a clear guide on how to identify the root cause of yield fluctuations, enabling more targeted problem-solving and significantly reducing time and cost spent on investigations.

Process Capability Assessment for Thin Film Deposition (PVD/CVD)

This article recounts a challenging experience where a newly promoted senior engineer faced intense pressure from management over a thin film deposition report showing a CPK of 1.08. It then provides an accessible introduction to thin film deposition, explaining common techniques like PVD and CVD. Readers will learn the critical importance of evaluating thin film process capability and the numerous intricacies involved in manufacturing these fundamental layers in chip production.