Magazine winter02 time to defect frames itergrated debate

Page 1

Cu/low S

P

E

C

I

A

L

F

O

Îş C

U

S

Time-to-Detect Frames the Integrated Debate Ralph Spicer, Dadi Gudmundsson, and Raman Nurani, KLA-Tencor Corporation

Is integrated defect inspection really the wave of the future? Analysis shows that a simple particle detection strategy is unlikely to be more cost effective than a comprehensive excursion inspection strategy, even if the particle detection can be integrated to a process tool.

The decision whether or not to integrate defect inspection onto process tools is one of the most important decisions facing 300 mm fab planners. This decision impacts everything from capital procurement strategies, to automation, to floor-planning, to data systems integration. And, once made, this decision is expensive, if not impossible, to change as the fab approaches first silicon. Therefore, it is important to understand the real variables behind this decision, moving beyond surface arguments that would appear to point strongly in favor of integration. This article discusses the relevant issues that must be analyzed when making decisions regarding the deployment of integrated versus non-integrated defect inspection in a new 300 mm fab.

significantly, making it vital to forecast such variables through several design rules before deciding on a strategy. When all of these considerations are taken into account, our analysis shows that the fact that particle detection can be integrated to a process tool does not necessarily make it the most cost effective strategy for 300 mm fabs. Trends affecting the decision

In order to determine whether or not an integrated inspection strategy makes sense for the fab, it is important to understand the variables that drive yield losses, and how technology trends are affecting these variables.

CYCLE TIME Cueue Times Automation

Layout

The decision to integrate

Recent arguments for integration cite equipment productivity as the driving variable upon which the decision should be based. While this is an important factor, it is also important to consider variables that drive yield, such as process tool excursion frequencies, defect kill probabilities, and the detection capability of the integrated and non-integrated systems being considered, as illustrated in Figure 1. Furthermore, while productivity-related variables remain relatively constant through design rule generations, the yield-related variables scale 42

Winter 2002

Yield Management Solutions

STEP YIELD

PROCESS DEVELOPMENT

Excursion Defect Frequencies Types Kill Probabilities

Yield Learning Support Design Rule Extendabilty

INSPECTION TECHNOLOGY

PROCESS TOOL OUTPUT Floor Orocess Tool Space Productivity Reliability

Inspection Strategy Decision

False Inspector Alarm Detection Capability Excursion Rate Time to Decision

Fastest Ramp Highest Yard Minimum Cycle Time

Figure 1. A multitude of factors must be considered when deciding on an inspection strategy for a 300 mm fab. These include productivity factors such as cycle time and process tool output, and yield factors such as step yield, process development, and the capabilities of the inspection technology.


S

The two primary yield-related variables are: Defect types:

Does the inspection strategy have a high probability of detection (pd) for the defect types that cause yield loss? Will it keep up with the (as yet unknown) defect types that will occur as design rules shrink and new processes such as copper dual damascene, low-κ dielectrics, DUV resists and SOI are introduced? Will it be able to adapt as yield learning in the fab proceeds?

Time to detection:

Does the approach minimize the time to detect the excursion? Can much of the benefit of integration be achieved through well-planned product flow and automation instead? With probability of detection and time to detection as the two main variables, a summary question arises: Is it better to detect a larger variety of excursions with some delay, or detect a subset of excursions with little delay? We studied this question as a part of a comprehensive study of integrated inspection technology. The results of this study led to the following key observations: • The excursion detection capability of a robust, comprehensive inspection strategy appears to outweigh the time-to-detect (td) benefits of integration.

• Significant time-to-detect (td) benefits may be achievable through optimized automation and fab layout without the loss of flexibility and added capital cost associated with integration. Integration per se has little incremental benefit once automation and fab layout are optimized. • The choice of inspection strategy must include provisions for future trends, such as new copper and low-κ defect types, and the growing importance of process margins and systematics as a source of excursions.

Metal Residue

Corrosion

Slurry Residue

Cu Flake

P

E

C

I

A

L

F

O

C

U

S

The industry has extensive experience in optimizing fab yield and productivity through the use of standalone inspectors. In the paragraphs that follow, we will review some of this experience, and focus on the issues that an integrated inspection approach can introduce to a fab. Defect types

A robust excursion control strategy must be capable of finding the defect types expected in today’s aggressive design rules, and must be adaptable to find the new types that are certain to occur in the future. Simple particle detectors are suited to find one particular defect type (large particles), but provide no capability in detecting other killer defects. True wafer inspection captures both process-induced and tool-induced defect types with a high probability of detection (pd). Examples of these defect types for Cu CMP are shown in Figure 2. Particles, of course, are captured as well, making additional particle detection inspections unnecessary. New processes and advanced design rules

At advanced design rules, defect densities must decrease to achieve viable yield1. As defect densities fall, the definition of an excursion becomes tighter, meaning that the inspection system must be able to detect smaller and smaller amplitude excursions without an undue increase in false alarms. The choice of appropriate inspection equipment typically includes a requirement for multiple-design-rule reuse; this is not supported by simple particle detection technologies. While it could be argued that particle detection technology is likely to improve in the future, such technology will remain behind the need as advanced design rules and new materials like copper and low-κ dielectrics are introduced. As design rules shrink, fewer excursions are caused by people and process tool contamination, making

Scratch

Particles

Micro Scratch Voids (from EP)

Figure 2. A comprehensive defect inspection strategy must detect the wide range of defect types that can cause yield loss. Many of these defect types, such as these Cu CMP defects, cannot be captured by simple particle detectors.

Winter 2002

Yield Management Solutions

43


E

C

I

A

F

L

O

C

U

S

particles a less important part of the picture relative to process-induced defects. This is due to the fact that environmental and equipment-induced particle sources are steadily being reduced and do not scale with design rule to the same extent as margins and systematics that are continually being pushed at each successive design rule generation, making such issues more likely at each successive design rule generation2. Inspection’s role in fast ramp

The ability of a fab to quickly ramp new processes is a major contributor to profitability. The faster the ramp, the faster the time to market, the higher the average device ASP.

unexpected problems tend to dominate. An inspection strategy that is to be effective during ramp must be able to capture defect types that cannot be predicted a priori. In contrast, once the fab reaches entitlement yield, it is less likely that new defect types will occur, making defect inspection much more predictable. This might lead to one asking whether the fab should invest in sensitive inspectors for ramp and development, and then switch to particle monitoring equipment during production, even though non-particle defects would still be present to some degree. Two important trends indicate that more sensitive tools are needed throughout the fab’s life, as shown in Figure 3. First, the yield “hurdle” that a process must pass prior to moving into ramp and production is moving ever higher, with a faster ramp. This means that there is less and less time to qualify lesser-capability inspection equipment to check that it finds the defect types identified during development. Secondly, the time period between process changes is decreasing, meaning that

The choice of inspection approach can have a major impact on the fab’s ability to ramp quickly. This is because the problems being solved during ramp are of a very different nature than those being solved during high volume production. During development and ramp, problems tend to be unpredictable; that is, new,

Trend: shorter time in high volume production

Production

3 Desi gn R

Desi gn R

Desi

Ramp

ule

ule

ule gn R

Yield %

Production 2

1

Production

Trend: higher yield required prior to ramp

Ramp

P

Ramp

S

Ramp

Dev

Dev

Dev

Dev Time

• Development and Ramp • Production

• Most problems UNPREDICTABLE

• Most problems PREDICTABLE

• New rocess-induced defect types

• Highest priority: YIELD LEARNING

• Defect types as determined during development/ramp

• High sensitivity/high capture • Accurate classification/quick sourcing

Production Ramp

• Highest priority: EXCURSION CONTROL • Fast time to detection • False alarm supression

Dev

Figure 3. Development and ramp drive fab profitability: the faster the ramp, the faster the fab can profit from higher ASPs. Most fabs are always in the development and ramp states, even after years of time, as new processes and shrinks replace older, lower ASP production products.

44

Winter 2002

Yield Management Solutions


S

E-test

CMP Bay

Litho Bay

E

C

A

L

F

O

C

U

S

Line Monitoring for Yield Correlation • Identify yield limiting defect types • Data customer = Yield Engineering

Etch Bay

Process Flow

High-Volume Inspections

I

DSA/Correlation

High-Sensitivity Inspections

Films Bay

P

Drives which excursion types to monitor

Tool Monitoring for Excursion Control • Decide process tool go/no-go based on SPC • Data Customer = Process Engineering • Must be able to detect newly identified types

Figure 4. A robust process tool monitoring approach includes the ability to feed back learning from more sensitive line monitoring inspectors. Without this feedback, it is difficult to control new excursion types introduced by new processes or design rule shrinks.

the typical fab is always in a development/ramp state. The net result is that today’s fab requires ever-morecapable excursion inspection capability to detect everchanging defect types. These observations imply that a robust tool monitor defect inspection approach must incorporate the learning from higher-sensitivity line monitor inspectors, as depicted in Figure 4. This allows the tool monitor inspectors to be tuned for new excursion types that are certain to be introduced by process changes or design rule shrinks. Experience suggests that without this vital feedback loop, it is impossible to sustain yield learning over time. While state-of-the-art standalone defect inspectors are able to adapt via this feedback, simple particle detectors cannot. Thus, a key element of excursion control would be eliminated were an integrated particle detector strategy to be chosen. Operational aspects of integration

The decision to integrate has major operational impacts to the fab: • One inspector is required for each process tool or cluster, meaning that there are more inspectors in the fab • Each inspector is tied to its process tool, meaning that a given inspector can only inspect wafers from its process tool.

Obviously, capital cost is a major consideration. Typical standalone implementations utilize a ratio of anywhere from 4:1 to 10:1 (that is, 4 to 10 process tools per inspector). Unless the integrated particle detectors are significantly less expensive than the standalone defect inspectors, the increased quantity of inspectors leads to a higher capital investment cost for the fab. Also, integrated inspectors add to each process tool’s footprint, reducing product output per unit area. False alarming is also a major consideration. Operators and engineers must respond to each report of an excursion as it occurs. As the simple technologies employed by particle detectors are readily confused by processinduced pattern variation on product wafers, they risk a higher incidence of false alarms than defect inspectors that incorporate technologies developed specifically to suppress such pattern variation. It is easy to see how a larger number of particle detectors, each false alarming more often, would lead to an intolerable distraction on process operations, or, worse yet, lead to process operations which either ignore alarms or “dumb down” the recipes to prevent them. These are situations that would completely negate the benefits of integration. Reliability is also a concern. With the particle detector dedicated to the process tool, a failure in the detector leads to a downed process tool (leading to lost productivity), or skipped inspections (leading to a possibility of an undetected excursion). Unless MTBFs and MTTRs of the particle detectors can be raised to levels many times better than that which is achievable today, Winter 2002

Yield Management Solutions

45


E

C

I

A

L

F

O

C

U

S

the sheer number of inspectors in the fab implies that at least one inspector will be down nearly all of the time, along with its associated process tool. A standalone approach, on the other hand, allows lots to be routed around downed inspectors in a way that is difficult with integration. Past studies of excursion control in 200 mm fabs indicate that the average time between lot processing and excursion detection in a typical fab can be eight hours or more3. This means that the average excursion affects yield across several lots of wafers before the source of the excursion is identified and action taken. One of the primary perceived benefits of integrated inspection is that this time is reduced significantly, since wafers can be inspected soon after processing. But, integration of the inspector to the process tool is only one way to reduce this time. For example, reducing queue and transport time from eight hours to two hours via efficient layout and inspector utilization would gain much of the benefit of integration while retaining the flexibility of a standalone inspection strategy. Such 300 mm automation concepts as multilevel transport and intrabay shuttles hold promise in making this feasible. We plan further modeling studies to quantify how these operational variables feed into the decision of an optimized excursion detection strategy. Modeling integrated versus standalone strategies

Given that there is significant industry effort toward integrating particle detectors, we set out to answer the question: which approach minimizes excursion losses in the fab: (1) a simple particle detector integrated to each process tool, or (2) a comprehensive defect inspection strategy implemented in a standalone fashion? Our preliminary results are that the second approach is the optimal strategy for the process steps we studied. Using KLA-Tencor’s Sample Planner™ software4, we were able to model the effects of defect type capture and inspection delay (such as that caused by transit, queueing, and inspection times) on the overall yield loss for various process steps. An example of the results of this study for Metal 3 etch are shown in Figure 5. The two curves show the value of inspection at Metal 3 etch for a 5000 wspw, 300 mm logic fab relative to performing no inspection at this step. As one would expect, the value increases as time to detection (td) decreases, since excursions are caught sooner. However, 46

Winter 2002

Yield Management Solutions

this increase in value is much more pronounced when a comprehensive defect inspection strategy is used, since the particle detection approach misses many excursions entirely, negating the benefit of decreased td. In fact, the cost of missed excursions is so substantial that, even with an eight-hour td, the standalone inspection approach provides a 30 percent yield benefit over an integrated particle detection approach ($1.3 m/year versus $1.0 m/year). The cost of the inspection tools has to be considered when comparing the value curves in Figure 5. When large scale integrated defect inspection was first conceptualized, it was clear from the start that each integrated inspector would have to cost considerably less than a standalone tool. This is clear from the fact that standalone tools are serving anywhere from four to ten process tools and an integrated inspector would serve only one tool. For this analysis, whether standalone or integrated inspection is used, it can be assumed that the overall inspection expenditure is the same. This arises from the observation that the price of integrated tools for our example are about a third of the standalone tool price. In addition, this particular example required slightly less than a third of a standalone’s tool capacity to serve one M3 etcher. That same etcher would require a single integrated inspector for itself, demonstrating the approximate cost equivalence. This follows the trend that to make integrated tools feasible they have to be cheaper; unfortunately, that can only be realized at the expense of detection capability. This analysis only quantifies the value of excursion control. The benefits of fast ramping due to accelerated $4.5

Value of Improved Yield

P

($m/year, relative to inspection)

S

Comprehensive Standalone (Optimized MH)

$4.0 $3.5

Large Particle Detector Comprehensive Inspector

$3.0 $2.5

Loss Due to Undetected Excursions

$2.0

Comprehensive Standalone (Unoptimized MH)

$1.5 $1.0 $0.5

Integrated Particle Detector

$0.0 0

1

2

3

4

5

6

7

8

9

10

Detection Delay td (Hours) Sum of material handing (transit + queueing) and inspection times

Figure 5. Particle detectors risk a substantial loss at M3 etch for a 5,000 wspw, 300 mm fab due to missed excursions, when comparing an integrated particle detector to a 2-hour-t d comprehensive standalone approach. Even with an unoptimized material handling time of eight hours, the comprehensive standalone approach shows a benefit over the integrated particle detector.


S

Yield Lea r

Yield %

E

C

I

A

L

F

O

C

U

S

reduction when using comprehensive standalone inspectors.

Baseline Yield

We now explain in more detail the key variables that drive excursion-related yield losses, and how these were analyzed to produce the values in the above figure.

ng i n

What is an excursion?

Excursions

An excursion is defined as an out of control condition at a single process step which impacts yield until the excursion is corrected. As shown in Figure 6, a fab’s baseline yield rises over time as yield learning is achieved. Excursions represent temporary dips in this yield, corresponding to a loss of profit. The goal of a comprehensive yield management program is one that raises baseline yields as high as possible as quickly as possible (hitting the “sweet spot” of high device ASPs), coupled with an effective strategy for preventing and minimizing the impact of excursions during volume production.

Time Figure 6. The baseline yield of the fab drives its baseline profitability. Excursions represent temporar y drops in yield, resulting in lost profit for the fab.

yield learning are harder to model, but experience shows that these benefits would further tilt the results in favor of comprehensive standalone inspectors.

In order to determine the best strategy for minimizing the financial losses from excursions, it is important to understand in greater detail where these losses come from. Figure 7 shows a more detailed timeline of an excursion and its losses for two situations: when the

These results indicate that substantial benefit can be derived through effective material handling time

Event Occurs

P

Chamber Down

td Good Product Scrap Product

Chamber Up

ttp

Good Product

Process Tool Productivity Loss (Ltp)

Cy

Yield Loss (Ltp) caught excursion

ed

XX

Chamber Down Scrap Product

Chamber Up

ttp

td

Good Product

n

io urs exc

Good Product

ss mi

Event Occurs

Ctp

Process Tool Productivity Loss (Ltp)

Cy

Ctp

Yield Loss (Ly) - missed excursion

Figure 7: The financial loss due to an excursion consists of two primar y elements: yield loss due to wafer scrap and/or die yield impact, and loss due to reduced productive tool time. An excursion that is not detected by the defect inspector increases yield loss dramatically.

Winter 2002

Yield Management Solutions

47


S

P

E

C

I

A

L

F

O

C

U

S

defect inspector detects the excursion (upper half of figure), and when it does not (lower half of figure). Three points in time define the timeline: (1) the time that the excursion event occurs, (2) the time that the chamber (e.g. deposition chamber or polishing head) is taken down, and (3) the time that the chamber is returned to production. Before the excursion event occurs, production lots are being processed normally. When the event occurs (for example, a problem with one etch chamber), product continues to move through that chamber, causing lots that will have to be partially or completely scrapped. This occurs until the event is detected and confirmed, and the chamber is taken down. Then, the tool is serviced, during which time no product is being processed by the chamber. Finally, the chamber is confirmed fixed, and processing resumes. The shaded regions of the figure depict the losses that occur. First is the yield loss (Ly). This is defined as the time to detect the excursion (td) times the cost of the yield loss (cy) per hour. Second is the process tool productivity loss (Ltp). This is the number of hours it takes to repair the chamber and bring it back into production (ttp) times the capital cost of the tool per hour. To give some sense of scale to the timeline, our experience in 200 mm fabs indicates that an average number for td is approximately eight hours, and for ttp 16 hours, in cases where the excursion is caught. The cost of yield loss (cy) is driven by two variables: excursion yield impact, and die ASP. Yield impact is the probability that a given excursion type will cause an electrical failure in a given die. For example, a micro excursion might cause sparse defects that kill, on average, 25 percent of the die on the wafer (yield impact = 0.25), whereas a macro excursion might impact every die on the wafer (yield impact = 1), leading to complete wafer scrap. For the case where the yield impact is less than 1, one might expect cy to scale with the yield impact; that is, a yield impact 0.5 excursion would lead to a cy of half of the yield impact 1 value. However, wafers will still be scrapped completely if the number of failed die exceeds a threshold beyond which the cost of continuing the wafers’ processing does not justify the smaller number of good die that will result. One immediate observation is that the cost of excursions is influenced heavily by product and yield-related 48

Winter 2002

Yield Management Solutions

variables. So, while it may appear that an obvious integrated strategy would be to set a capital cost target for process tools’ integrated inspector (say, 20 percent of the process tool’s cost), this is likely not the cost-minimizing approach. The cost-minimizing approach takes into account the cost of yield loss for the process step. For example, benchmarking studies indicate that the typical inspection investment is higher for early metal layers than for later metal layers, due the tighter design rules and higher defectivity encountered there, even though the exact same process tools are being used. One side effect of integrating inspection into the process tool is the absence of this kind of investment flexibility. For a complete discussion of this topic, see reference5. Another observation is that great leverage can be obtained by reducing td by reducing the time it takes to make a tool up/down decision. By removing the possibility of incurring queue time at standalone inspection, td can be reduced dramatically. This is the obvious attractiveness of integrated inspection. However, past experience indicates that a missed excursion is usually not detected until days later, when the lot is inspected by a more sensitive line monitor inspection, by the next step’s tool monitor inspection, or worse, by back end final electrical test. Most fabs have had the painful experience of an extensive “yield bust” because of a defect type that was not detected by the inspection equipment. A higher probability of excursion (pe), combined with a lower probability of detection (pd), make this undesirable scenario more likely. Another way of looking at the same issue is to use a weighted value for the time to detect, Td(effective) = td(caught) * pd + td(missed) * (1-pd) Given the potentially immense losses associated with missed excursions, it is dangerous to choose an inspection strategy without analyzing pd. To analyze pd, we utilized a historical root cause database and benchmarking data, combined with a survey of process experts to create an exhaustive list of the defect types and frequencies typically encountered at various process steps. We then assessed the capability of various inspection and particle detection approaches against these defect types. A subset of this assessment is shown in Figure 8. Here, the frequencies of excursion are multiplied by the pd for that excursion type to


0 = low pd, 1 = medium pd, 2 = high pd

Array

Logic

Darkfield Inspector Open

Brightfield Inspector Logic

Particle Detector

Array

Polymer buildup on the wafer from the process

Tool-Down Criteria

2

0

0

2

2

2

2

2

2

SPC based (2 out of 3 consecutive lots O.O.C.)

2

0

0

2

2

2

2

2

2

First Deep scratch in 0.5 events/month: occurance Increases with a line on wafer on particular MWBC and RF time map chamber

2

2

2

2

2

2

2

2

2

SPC based 0.25 events/month: (2 out of 3 Increases with consecutive MWBC and RF time lots O.O.C.)

2

0

0

2

2

2

2

2

2

First Aluminum balls 0.25 events/month: occurance shards on Increases with on particular wafer surface MWBC and RF time chamber

Polymer (organic and inorganic) deposit from Flakes/particles 2 events/month: on wafer Increases with electrode degradation, surface MWBC and RF time chamber walls and parts such as o-rings

Mechanical Mechanical failureScratches usually wafer handling

Residues

Frequency

Open

Particles

Deposits as a result of grounding failur l.e. arching

Appearance

Logic

Arcing

Root Cause

Array

Failure Mechanism

Open

Examples

Surface particles

Figure 8. Defectivity benchmark and defect root cause data were used to evaluate p d for various inspection technologies.

generate the results we showed earlier (see Figure 5). The comprehensive standalone approach indicated greater benefit than the integrated particle detection approach. Again, these results do not include the benefit of fast yield learning, a benefit that experience shows to be substantial.

We found that simple particle detectors missed a significant number of the excursion events at state-of-theart process rules, reducing pd considerably when compared to today’s standalone tool monitor inspectors. This, combined with the comparable effective throughput of darkfield inspectors and particle detectors (due to real-world factors such as wafer handling and alignment time), leads to a summary of average pd vs throughput for various inspection technologies such as that shown in Figure 9.

1

We then loaded the resulting probabilities of detection (pd) and throughputs into the Sample Planner™ model, which allowed us to more accurately model the interaction between the probabilities of detection, excursion frequency, sampling strategy, integration approach, and real-world issues such as transit and queue times.

0

Probability of Detection (pd)

obtain an event-weighted pd for the process step for various types of product wafers. These are then combined to obtain a die-weighted pd. Of course, this analysis is not static. As process tools mature and design rules shrink, we expect a decrease in the frequency of particle excursions, and an increase in the frequency of process-induced excursions.

E-beam Inspection Brightfield Inspection Darkfield Inspection Particle Detection Low High Throughput (Wafers/Hour)

Figure 9. The p d for particle detectors is significantly lower than

We modeled a range of td, from the historical average of eight hours of td, to an optimized standalone td of two hours, and on to an integrated td of 15 minutes to

currently available darkfield, brightfield, and e-beam inspectors, with effective throughputs comparable to darkfield inspectors. This is in direct relation to the lower cost of integrated inspectors.

Winter 2002

Yield Management Solutions

49


S

P

E

C

I

A

F

L

O

C

U

S

On the application front, metrology is applicable to advanced process control concepts such as feedforward (e.g. modifying the etcher recipe based on film thickness measurements) and feedback (e.g. modifying deposition tool parameters based on measured results from previous wafers). Defect inspection, on the other hand, does not lend itself to such applications: what would one adjust on an etcher were an excursion to be detected by the defect inspector?

Process control considerations

One obvious question that arises is: if the economics of integrated inspection do not appear favorable, why is there so much activity revolving around integrated technologies? The answer lies in the fact that three very different classes of integrated technologies are being pursued: integrated metrology, integrated particle detection, and integrated defect inspection. The differences among these applications are significant (as shown in Figure 10), and so it is vital to analyze each separately. It is important not to confuse the technology and applications of integrated metrology with those of defect inspection.

One proposed compromise solution for defect inspection is the particle detector, which performs a type of defect inspection, but is able to capture only one type of defect (large particles). As we showed earlier in this article, capturing large particles is insufficient to prevent costly yield busts due to lengthy out-of-control conditions. This is why particle detection does not lead to an optimal solution, even if it is integrated to the process tool.

On the technology front, the nature of the items to be measured (the targets) are known with certainty. Therefore, once a technology has shown an ability to measure targets with the required accuracy (for example, spectroscopic ellipsometry’s ability to measure film thickness on multilayer stacks to 0.5 percent) in a form factor which supports integration, there is higher confidence that integration may be an appropriate path.

Conclusions

Our conclusion is that a particle detection strategy is not likely to provide benefit over a comprehensive standalone inspection approach, even if particle detection can be integrated to process tools.

In contrast, defect inspection must take into account the added complexity of full-wafer scanning, which implies a highly variable background signal (the pattern) which reduces the signal-to-noise ratio of the inspection. The complex image acquisition and processing algorithms required to achieve a useful signal-to-noise ratio are not currently available in an integrated form factor. In fact, because this problem becomes more difficult as design rules shrink, it is a distinct possibility that adequate wafer inspection performance may never be available in an integrated form factor.

This conclusion was driven by these observations: • The excursion detection capability of a robust, comprehensive inspection strategy appears to outweigh the time-to-detect (td) benefits of integration. • Significant time-to-detect (td) benefits may be achievable through optimized automation and fab layout

Supported Fab Applications

Technology

Tool Go/No Go

Wafer Pass/ Scrap/ Rework

Tool Feedforward/ Feedback

Process Development Feedback

Target/ Area

Technology Maturity vs. Standalone

Metrology (CD, Film Thickness, Overlay)

Yes

Yes

Yes

Yes

Target

Medium

Defect Inspection

Yes

Yes

No

No

Full Wafer

Low

Particle Detection

Limited

Limited

No

No

Full Wafer

Low

Figure 10. The applicability of metrology to process control applications means that the decisions regarding the optimal approach for metrology are ver y different from those for defect inspection.

50

Winter 2002

Yield Management Solutions


S

without the loss of flexibility and added capital cost associated with integration. Integration per se has little incremental benefit once automation and fab layout are optimized. • The choice of inspection strategy must include provisions for future trends, such as new Cu and low-κ defect types, and the growing importance of process margins and systematics as a source of excursions. These observations suggest that integrated particle detection may not necessarily be the future trend that conventional wisdom might suggest. Integration of inspection will only become viable when integrated inspection technology is comparable to standalone technologies, and today’s candidate integrated particle detection approaches are not near this point. Even in the long run, the fact that inspection requirements scale with the design rule suggest that the crossover point at which integrated inspection becomes viable may be a long way off, if it appears at all.

P

E

C

I

A

L

F

O

C

U

S

The authors would like to acknowledge the contributions of Wayne McMillan, Anantha Sethuraman, Paul Marella, and Sanjay Tandon to the study. References 1. Stapper, C.H., Fact and Fiction in Yield Modeling. Microelectronics Journal, vol. 20, no. 1-2, 1989, p.129-151 2. Jensen, D. State of the Industry Address, 1995. 3. Esposito, T. et al. Automatic Defect Classification: A Productivity Improvement Tool. Conference proceedings IEEE/SEMI Advanced Semiconductor Manufacturing Conference and Workshop, p. 269-276. 4. Williams, R.R., Gudmundsson, D., Nurani, R.K., Stoller, M., Chatterjee, A., Seshadri, S., Shanthikumar, J.G. “Challenging the Paradigm of Monitor Reduction to Achieve Lower Product Costs”. The 10th Annual IEEE/SEMI Advanced Semiconductor Manufacturing Conference and Workshop, September 8 - 10, 1999. 5. Williams, R., Gudmundsson, D., Monahan, K., Nurani, R., Stoller, M., Shanthikumar, G. Optimized Sample Planning for Wafer Defect Inspection. IEEE International Symposium on Semiconductor Manufacturing, Santa Clara, California, October 11-13, 1999.

Read anything good lately?

Order your copy of Chris Mack’s Lithography Expert Booklet today! Log on to:

www.kla-tencor.com/litho


With the right adjustments, your 300 mm yield can be better than ever.

For more about how

When a major fab had to hit their 300 mm profitability goals as fast and efficiently

KLA-Tencor helped

as possible, they turned to us. That’s because they needed the most comprehensive,

a major fab accelerate

advanced suite of 300 mm-compatible process control tools available. A

300 mm yields, please visit

demonstrated track record of successful implementation. And an unwavering

www.kla-tencor.com/300mm.

commitment to faster yield ramps. As a result, the fab’s director identified our partnership as critical in helping reach 200 mm-equivalent yields on their very first 300 mm customer lots. Just another reason why more fabs depend on us to help make yield ramps – and ROI – look their very best. For more information, please visit www.kla-tencor.com/300mm, or call 1-800-450-5308. ©2001 KLA-Tencor Corporation

Accelerating Yield


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.