Summer00 production qc

Page 1

Lithography S

P

E

C

I

A

L

F

O

C

U

S

Production QC and Tool Monitoring Using an Automated Macro ADI Defect Inspection System by Iain Rutherford, Brian Haile and Tony DiBiase, KLA-Tencor Corporation This paper was presented at KLA-Tencor’s Yield Management Solutions Seminar during SEMICON/Europa in April 2000.

The major weakness of traditional, manual after-develop inspection (ADI) lies in the variability of the results. Defect capture rates are variable due to differences in the ability and experience of the inspection operators. Subsequent analysis of defects can also be inconsistent as some operators might flag a defect while others might pass it thinking it unimportant. Manual inspection is also one of the most tedious and unpopular jobs in the fab among the operators.

The ability to drive yield improvement from manual inspection results can be very poor. Data from manual inspections can be vague and subjective, making it difficult to archive or correlate with yield and parametric results. The bottom line is that manual ADI misses macro defects and costs money.

Manual vs. automated ADI

A solution to these issues is an automated, optical macro defect inspection system such as KLA-Tencor’s 2401. This system was installed for evaluation at NEC in Livingston, Scotland, and this paper reports the results from the evaluation at that site. The study was conducted in two parts: First, a comparison of manual versus automated inspection using the 2401 for ADI, and second, an investigation of the potential of the 2401 for evaluating and monitoring certain aspects of stepper performance.

Eight layers from one particular product were chosen. Lots were then randomly sampled from these layers. The layers represented a typical mix of front-end and back-end layers, both critical and non-critical.

The 2401 works by simultaneously scanning and capturing darkfield and brightfield images of a wafer. The inspected field is compared with two fields adjacent to it and any discrepancies between the two fields are flagged as a potential defect. The system has an 80 wafer-per-hour throughput and can capture defects greater than 50 µm. If needed, wafers can be reviewed using a variety of software tools on the machine. 32

Summer 2000

Yield Management Solutions

Over 5500 wafers were involved in the first part of the evaluation, and most of these (5019) were randomly selected for after-develop inspection. Some wafers with known problems were chosen (225), and some originated from engineering (322). The wafers were inspected manually, then taken to the 2401 for automated inspection.

Figure 1 demonstrates the overall sensitivity of the 2401 compared with manual inspection. While manual inspection found that 3 percent of lots had issues worth Visual Inspection Results

2401 Inspection Results

3%: Failed

97%: Passed

46%: Failed

54%: Passed

Figure 1. With 213 lots inspected at random, 6 lots failed visual inspection while 97 lots failed the 2401 inspection, demonstrating a 10X difference in capture rate when replacing manual with automated macro inspection.


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.