A faster route to better overlay

Product lifecycle management: added-value across the industry

Focus on a new star

ASML’s customer magazine | 2014

images | Colofon

Editorial Board

Lucas van Grinsven, Peter Jenkins

Managing Editor

Michael Pullen

Contributing Editor

Saskia Boeije

Contributing Writers

Paul Tuffy, Jan-Willem van der Horst,

Thomas van Wezel, Jules Tops,

Janneke van Heteren, Remi Pieternella,

Stuart Young

Circulation

Karen Lo, Michael Pullen, Saskia Boeije

For more information, please see:

www.asml.com/images

© 2014, ASML Holding BV

ASML, ASM Lithography, TWINSCAN, PAS 5500,

PAS 5000, SA 5200, ATHENA, QUASAR, IRIS, ILIAS,

FOCAL, Micralign, Micrascan, 3DAlign, 2DStitching,

3DMetrology, Brion Technologies, LithoServer,

LithoGuide, Scattering Bars, LithoCruiser, Tachyon

2.0, Tachyon RDI, Tachyon LMC, Tachyon OPC+,

LithoCool, AGILE, ImageTuner, EFESE, Feature Scan,

T-ReCS and the ASML logo are trademarks of ASML

Holding N.V. or of affiliate companies. The trademarks

may be used either alone or in combination with

a further product designation. Starlith, AERIAL,

and AERIAL II are trademarks of Carl Zeiss. TEL is

a trademark of Tokyo Electron Limited. Sun, Sun

Microsystems, the Sun Logo, iForce, Solaris, and the

Java logo are trademarks or registered trademarks of

Sun Microsystems, Inc. in the United States and other

countries. Bayon is a trademark of Kureha Chemical

Industry Co. Ltd. Nothing in this publication is intended

to make representations with regard to whether any

trademark is registered or to suggest that any sign

other than those mentioned should not be considered

to be a trademark of ASML or of any third party.

ASML lithography systems are Class 1 laser products.

6 10 164 A faster route to better overlay

6 EUV shows consistent performance

in the field

10 Product lifecycle management:

added-value across the industry

12 New nodes with your installed

TWINSCAN NXTs

16 Focus on a new star

2

As 2014 comes to a close and we look

back on our 30 year history of providing

lithography solutions for the ever growing

and changing semiconductor industry,

it is amazing to see the challenges

that we have overcome, and the

accomplishments that we have made,

together as an industry. As we look to

coming years, there are obvious hurdles

and roadblocks that we must overcome

to continue down the path set forth by

Moore’s Law.

In this issue of Images Magazine,

we will take a look at how we, at ASML,

are preparing to clear those hurdles and

continue down the path for another

30 plus years.

Going back nearly to our start in 1984,

the PAS 2500/5000 is still in use today

at several customers. While its End

of Service is approaching, the Mature

Products Service team is redefining

the life cycle of all ASML products and

mapping out four major lifecycle stages

that will increase transparency with

customers and help them improve their

own business planning in the future.

You will hear how ASML’s new Design

for Control (D4C) Overlay software

is tackling the current industry

requirements for overlay, focus control

and critical dimension uniformity (CDU)

and will meet the future demands by

allowing customers to design and

optimize metrology targets that can be

used to deliver significant improvements

to overlay performance.

Another hurdle you will read about is

the cost ineffectiveness of transitioning

from node to node and how ASML’s

System Node Extension Packages

(SNEPs) are breaking through this by

converting any TWINSCAN NXT system

to a newer model in the field, effectively

extending its capabilities another one or

two production nodes.

As you may have read recently, our EUV

tools are meeting and exceeding the 500

wafer per day barrier, with one machine

at IBM exposing 637 EUV wafers in a

single day! Additional advancements

and improvements are being made to

source power, availability and particle

contamination.

Last, but not least, you will read about

the latest YieldStar system, the 250D.

It is the first metrology system capable

of measuring overlay, focus and CD

in a production environment, helping

customers maximize their yields of good

wafers-per-day.

I hope you find this issue informative

and of value. Please feel free to

provide direct feedback to me at

michael.pullen@asml.com so that we

can continue to improve the magazine

and your experience.

Happy reading!

Mike

30 years down, 30 plus to go By Michael Pullen, Senior Communications Specialist

3

ASML Images, 2014

Editor’s note

A faster route to better overlayBy Paul Tuffy, Product Manager BRION Wafer Fab Applications

Abstract | ASML’s new Design for Control

Overlay software identifies the best

metrology target design for any given layer

combination and process in the shortest

time. It allows customers to design and

optimize metrology targets that deliver

the ideal balance of printability, detectability,

accuracy and device matching. These targets

can be used in an automated feedback

loop to deliver significant improvements to

on-product overlay performance.

Fast, precise and accurate, ASML’s

YieldStar diffraction-based metrology tool

has made it possible to continually monitor

on-product overlay performance and

provide faster feedback to the scanner.

This is done through a feedback loop

where overlay data from an in-track

YieldStar module is converted into

smart exposure corrections by system

enhancement packages like Litho Insight.

These corrections are fed back to the

scanner for subsequent wafers, resulting in

improved on-product overlay performance.

These overlay measurements are based

on grating targets included in the reticle

designs for the many overlay-sensitive

layer combinations in a product. The design

of these targets needs to fulfill certain

performance requirements. It must print

well across the process window, and

deliver an easily detectable diffraction

signal to ensure precision and short

measurement times. It also needs to

help ensure the measurements are

accurate and not affected by variability

in processing steps such as etching and

chemical mechanical planarization (CMP).

As design features shrink, more and more

overlay targets are needed to capture

the overlay fingerprint at the required

level of detail. At the same time, placing

targets within the actual device becomes

advantageous as it allows more accurate

and higher-order corrections. This in-die

placement is only feasible with very small

targets, making it important to optimize

the target design for a strong diffraction

signal. Moreover, the increased use of

opaque materials in the latest technology

process stacks – such as the sacrificial

layers used in spacer processes – make

target optimization for signal strength

doubly important.

Design for Control

ASML’s new Design for Control Overlay

(D4C Overlay) software package helps

semiconductor manufacturers identify

the best metrology target designs in the

shortest time. It does this by simulating

the lithography process and resulting

YieldStar measurements of candidate

targets including the full layer stack,

allowing the design parameters of the

target for a given layer to be optimized.

D4C Overlay transforms target

optimization from a lengthy trial-and-

error process to a quick and reliable

computational one. It avoids the need

to carry out repeated wafer experiments

on different test designs – significantly

speeding up process development.

And it allows users to try out thousands

more targets and fully explore the design

space – enabling the development of

Identify the best

target designs in the

shortest time

4

Design for Control Overlay optimizes YieldStar overlaytargets for best on-product overlay performance

• Target selection for robust and accurate overlay with lowest sensitivity to process induced asymmetry

D4C

Detectability

Accuracy: Process robustnessAccuracy: Device matchingdevice target

Printability

• Improved overlay due to improved aberration sensitivity matching (target to device)

• Mask optimization• Litho process window• Design rule compatibility

• Target selection to meet TMU/ MAM requirements• Including process variations for detectability robustness

∆SWA

D4C Overlay Target Design Flow

Input

Select targets basedon weighed KPIs

Target selection Filter out targets

with poor detectabilityand printability

YieldStar detectability

Simulatedetectability KPIs

@ nominal

Detectability robustness

Aberration matching

Overlay robustness

Simulate detectabilityrobustness KPIs

Simulate overlayrobustness KPIs

Simulate aberrationsensitivity KPIs

Simulateprintability KPIs

Litho Printability

Optional

(process flow)

Designinput

(Tachyon FEM+)

Lithomodel

(YS & target patterns)

Simultationsettings

target printability and detectability.

The full version including overlay

accuracy was launched in Q3 2014.

It is designed for use with ASML’s

TWINSCAN NXT immersion ArF and

NXE extreme ultraviolet scanners in

conjunction with YieldStar YS 200C or

YS 250D metrology tools and Litho Insight

overlay optimization software.

the final target design in a single reticle

tape out cycle.

By identifying targets that balance

precision and accuracy, D4C Overlay

helps deliver significantly better on-

product overlay performance. What’s

more, it can match the aberration

sensitivity of the target to that of the

critical device features being printed in

the specified layer, enabling further

overlay gains. (See Fig. 1)

Tailor-made targets

Design for Control Overlay is a complete

target optimization software package.

Its easy-to-use graphical user interface

guides the user through the target design

process step by step. This includes a

flexible interface for defining the process

stack in the same way as it is built in the

fab: adding etch, deposition, CMP and

patterning steps to build each layer.

This approach allows almost any stack

design to be simulated. (See Fig. 2)

Once the user has specified the process

design rules and constraints, the software

runs an initial simulation to identify target

candidates that meet the printability

and detectability specifications.

The detectability, overlay accuracy

and lens aberration performance of

these candidates are then tested in a

further round of simulations.

An extensive set of built-in analysis features

including 2D heat maps and 3D amplitudes

simplifies the selection of optimal targets.

Once the user has chosen a number of best

candidates for experimental validation,

the D4C Overlay software outputs the

complete reticle pattern for that layer in

the GDS format. (See Fig. 3)

In-house design and optimization

D4C Overlay runs on any fab server

cluster using the Tachyon Flex platform.

We offer extensive training and detailed

user manuals on how to get the best

from the package. This allows companies

to keep target design and optimization

completely in house.

D4C Overlay was initially released in

late 2013 in a version that focused on

Together these systems deliver a

holistic solution to the current industry

requirements for 5 nm overlay,

60 nm focus control and 1.6 nm CDU

(after etch). Planned enhancements to

all these products will support roadmaps

towards 2017 that demand 2.5 nm overlay,

50 nm focus control and 1.1 nm CDU

(after etch).

Fig. 2

Fig. 1

Fig. 3

5

ASML Images, 2014

EUV shows consistent performance in the fieldBy Jan-Willem van der Horst, Product Manager EUV

EUV lithography is making continued

progress towards maturity and production

insertion. Systems at customer sites are

delivering consistent performance fit for

development of 10 nm logic and sub-20 nm

DRAM products. Productivity levels are

up – leading to a record number of EUV

wafer exposures in a day.

The TWINSCAN NXE:3300B is our

third-generation EUV lithography system,

with a resolution specification of 22 nm.

Six of these systems have been qualified at

our facility in Veldhoven, the Netherlands

and shipped to customer sites. A further

five are currently going through the

qualification process.

Abstract | ASML has qualified and shipped

six TWINSCAN NXE:3300B extreme

ultraviolet (EUV) lithography scanners.

With multiple systems exposing wafers

at customer sites, the NXE:3300B

is demonstrating consistently good

performance. Meanwhile, enhancements

to source power and availability are driving

productivity gains. In addition, a joint

research program with customers and

materials suppliers is exploring a pellicle

concept to protect EUV masks from

particle contamination, and hence reduce

printed defect levels.

Excellence as standard

These installed systems demonstrate

that the NXE:3300B’s excellent imaging

performance is repeated across multiple

systems. In fact, the systems in the field

often achieve performance levels in excess

of specifications – for example printing

16 nm dense lines with large process

windows. Full wafer focus uniformity

better than 12 nm is common, as is critical

dimension uniformity (CDU) below 1.5 nm.

Excellent imaging

performance is repeated

across multiple systems

6

corrected using standard scanner model

Wafer – after std modelling

10

1.5

3

4.5

6

2 3 4 5 6 7 8

Mat

ched

mac

hine

ove

rlay

[nm

]

X - axisY - axis

Lot (3.2,3.0)

Logic 10nm Metal 1 layer: Wide Depth of Focus at 20 mJ/cm2

NXE:3300B, 10 nm logicMetal 1 layer, 45 nm pitch.

Clips courtesy of STMicroelectronics

Quasar illumination¬Dose ~20 mJ/cm2¬NXE OPC+

-80nm-60nm-40nm-20nm

0nm20nm40nm60nm80nmFO

CU

S:

DoF

120

nmFO

CU

S:

DoF

120

nm

Overlay performance too is consistently

excellent. The six installed systems exhibit

matched machine overlay (MMO) with

our immersion scanners of around 5 nm

or better. And full-wafer dedicated chuck

overlay (DCO) below 1.4 nm has been

achieved. (See Fig. 1)

Customers had already shown that the

NXE:3300B is capable of printing real

devices structures for the 10-nm logic

node with an extended depth of focus

(DoF). Now by combining the scanner’s

off-axis illumination (OAI) capabilities

with ASML-Brion’s advanced optical

proximity correction (OPC), they have

reproduced those results at much lower

doses. For example, customers have

printed 10-nm logic metal 1 layers with a

120 nm DoF at a dose of around 20 mJ/cm2

– both fi gures are within requirements for

high-volume production. (See Fig. 2)

Pushing up productivity

The six installed NXE:3300B systems

feature the new master oscillator

power amplifi er pre-pulse (MOPA-PP)

Systems feature the

new MOPA-PP source

confi guration running

at 40 W

Fig. 1: MMO performance for various machines

Fig. 2: 10-nm logic metal 1 layer

7

ASML Images, 2014

Start End

Start EndSubstrate radius, mm

Reflectivity Al-05 (unpolarized)

50 100 150 200 250 300 35005

10152025303540455055

Ref

lect

ivity

,%

After cleaningBrand new

Field collector cleaned in NXE:3300 source vessel test rig

Off-line cleaning using NXE:3300B source vessel with product configuration hardware

Reflectivity restored within 0.8% of original Cleaning in off-line MOPA Prepulse development vessel

In-situ collector cleaning¬Effectiveness of product configuration confirmed

MaskAbsorber pattern

Full-size EUV pellicle prototype manufactured

Pellicle transmission requirement: 90%

Current status: ~86%

Pellicle

µm-size particles do not aff ect imaging

Fig. 3

Fig. 4

8

source confi guration running at 40 W.

Complementing this increased source

power, we have developed a number of

features to improve source availability.

Among these is a system for cleaning the

collector mirror in-situ. The system can

restore the collector’s refl ectivity to almost

brand new, eliminating the need to take

the system off line to replace the collector.

We’ve also shown that the source can be

run in a fully automated mode with good

dose control. This increases availability

while maintaining a 99.9% die yield.

(See Fig. 3)

As a result, productivity from our

EUV scanners has been steadily rising –

resulting in the recent announcement by

IBM of a new 1-day EUV wafer exposure

world record. In a test of the source,

they exposed 637 EUV wafers in a single

day. This test to verify the source’s power

output and reliability was carried out with

the scanner linked to a track, and using

an exposure dose of 20 mJ/cm2 and

conventional illumination.

“The test was designed to check if the

newly installed source was working

correctly. Exposing so many wafers was

an unexpected bonus – due to the source

operating so well,” Dan Corliss, IBM’s EUV

development program manager.

In the meantime, several customers have

cleared the 500 wafers per day barrier.

These achievements show we are on track

to meet our previously stated end-of-year

goal of 500 wafers per day as an average

and our 2016 goal of 1500 wafers per day

to support volume production.

Addressing the mask defect challenge

With EUV lithography scanners becoming

more mature, ASML is also active in

addressing ancillary issues related to

EUV lithography wafer fabrication.

One example is mask front-side

defectivity. In parallel to improving the

system’s overall cleanliness, we are

working with customers and materials

suppliers in an ongoing joint research

program to explore pellicles for protecting

EUV masks from particle contamination.

Together these efforts aim to improve

printed defect levels. (See Fig. 4)

EUV lithography requires refl ective rather

than transmission masks. Hence pellicle

materials for EUV have higher single-pass

transmission requirements than for DUV

lithography, as the EUV light must pass

through the pellicle twice. The year-long

research collaboration has developed a

new higher-transmission pellicle. Tests on

half-size pellicles mounted on the reticle

have shown the new pellicle concept

has little or no impact on CD or line

roughness. The pellicles have also shown

good robustness, surviving multiple

exposures and extensive mechanical

testing – including unusually rough

handling – intact.

The fi rst full-size, free-standing pellicle

prototypes for EUV masks have been

manufactured. Exposure tests using these

prototypes are planned for later this year.

Enabling future shrink

Single-exposure EUV lithography enables

aggressive feature shrink to drive the

semiconductor industry forward over the

coming years. The imaging performance

reported from our NXE:3300B systems

already exceeds requirements for the

10-nm 2D logic node. And ASML’s

technology roadmap extends the NXE

platform to beyond the 7 nm node. As the

next step on that roadmap, integration

and qualifi cation of our fourth-generation

system – the TWINSCAN NXE:3350B –

has already begun.

“ Thank you for participating in our reader survey. We appreciate your input as it helps us improve the magazine. The Bose® SoundLink Mini Speaker is won by Sang Nyung Yoon of DuPont Electronics”

www.asml.com/ImagesSurvey

First full-size, free-standing

pellicle prototype for EUV

masks

9

ASML Images, 2014

Product lifecycle management: added-value across the industryBy Thomas van Wezel, MPS Product Manager Product Life Cycle

Abstract | ASML is introducing a more

formalized approach to product lifestyle

management. This involves mapping our

systems to four major lifecycle stages

defined by the level of support ASML offers.

Timelines for when each system type

transitions between stages are defined well

in advance, and reviewed annually based on

market demands and customer feedback.

By communicating this information in a

timely fashion, we will increase transparency

for customers and suppliers – helping them

improve their own business planning.

ASML is 30 years old this year.

In celebrating this anniversary, it is

remarkable to note that one of the very

first machine types we developed is still

in service today. Almost 30 years old

itself, the PAS 2500/5000 is still being

used in production at several customers.

Of course its role has changed: from the

cutting edge of semiconductor production

to More-than-Moore applications.

However, all good things come to an end.

And the PAS 2500/5000 will soon become

the first ASML scanner to officially come

to its End of Service (EoS), with ASML

no longer providing support. With this

milestone approaching, ASML is also

launching a more formalized approach

to product lifecycle management.

This involves clearly defining the various

stages of a product’s lifestyle and how

ASML supports machines in each of

these stages. Then by communicating

this information along with the relevant

timelines, including target dates for stage

transitions, we hope to create greater

transparency for both our customers and

our suppliers on exactly what they can

expect from ASML. This will in turn help

simplify business planning.

Defining the lifecycle

To start this process, we have divided

the product lifecycle into four stages

based on the level of service we offer.

We call these stages Regular Service,

Extended Service, Limited Service and

End of Service.

10

New

Refurbishment

Regular Service

7 yearsEconomic lifetime

extension for ASML and customer

ControlledExit (typically

3 years)

PossibleTransfer to3rd party

Extended Service

time

LimitedService

End ofService

In the Regular Service stage we provide

our highest level of support for new

and recently refurbished machines to

maximize system availability and enhance

performance. We offer full availability

of parts and a range of service level

options to suit different semiconductor

manufacturers’ needs.

Around seven years after we finish

manufacturing a machine type, it will

transition to the Extended Service stage.

Here again, we offer full parts availability

and a range of service level options.

But now the focus is more on extending

the economic lifetime of systems.

At some point after that, the system type

moves into the Limited Service stage

where we start to ramp down the support

we offer. Spare part availability isn’t

guaranteed and service is supplied on

a best effort basis.

Finally, the system type moves to End

of Service, where ASML stops providing

support altogether. However, systems

may continue to provide an economically

viable production facility – typically in

niche applications – and we will endeavor

to help customers find alternative service

solutions where possible. (See Fig. 1)

Communicating timelines

The timeline for when a system type

moves from stage to stage is not the same

for all types. Although the transition to

Extended Service is always seven years

after manufacturing stops, some system

types may be manufactured for longer

than others. This could depend on the

system’s popularity and intended role.

For example, a pre-production system like

the TWINSCAN NXE:3100 would come out

of manufacturing earlier than a production

workhorse like the TWINSCAN XT:1900i.

Similarly, customer demand for and usage

of refurbished systems plays a role in

deciding when a system type transitions

to Limited Service or EoS.

Hence, a key part of our product lifecycle

approach is deciding well in advance when

each system type will transition between

stages and communicating that in a timely

fashion to both customers and suppliers.

Of course, these timelines aren’t

something that we dictate and set in

stone. Transition dates are reviewed

based on feedback from customers and

suppliers. For example, we have already

decided to prolong the PAS 5500’s

Extended Service period by four years

to 2022. We will continue to review and

update our product lifecycle timelines

based on business potential and customer

feedback on a yearly basis.

Benefits throughout the chain

Product lifecycle management is

something we have always done within

ASML. But by formalizing it in this way,

we believe we can bring benefits for

everyone. For example, by considering

the lifecycle stages and end of life issues

more rigorously in our design process,

we can help extend the productive life

of new systems.

Meanwhile, by communicating the

transitions between lifecycle stages,

we make it easier for customers and

suppliers to plan their business and

investment in equipment. Our aim is to

share these timelines as early as possible

– particularly for the transition from

Extended to Limited service and from

Limited service to End of Service – and

at least one year ahead of the transition.

This will give everyone time to factor the

information into their business decisions.

Hence customers will be aware just how

long ASML will be supporting the systems

installed at their facility, and can decide

the best time to migrate to new tools

based on the service available, resale

potential, etc. If they are considering

buying a remarketed ASML system,

they will be able to confirm how long the

system will be supported before they

finalize the deal.

Suppliers will be better able to predict

the demand for their products as they

will know ahead of time when we will

be manufacturing, refurbishing

and supplying spare parts for each

system type. This allows them to plan

capacity, staffing levels and equipment

requirements with more confidence.

Given the long lifetimes of our systems,

this long-term planning of the entire

lifecycle will bring value for all parties

involved in the semiconductor industry

and associated businesses.

Details of product lifecycle stages and

transition timelines are available via

ASML’s Account Managers.

Fig. 1: Product lifestyle stages

Easier for customers and suppliers to plan their business

Extend the productive life

of new systems

11

ASML Images, 2014

New nodes with your installed TWINSCAN NXTsBy Jules Tops, SNEP Project Cluster Manager DUV, and Janneke van Heteren, SNEP Product Marketing Manager

Abstract | ASML’s System Node Extension

Packages (SNEPs) allow any TWINSCAN

NXT system to be converted into a newer

model in the fi eld. For example, the

SNEP:A2C and SNEP:B2C respectively

transform NXT:1950i and NXT:1960Bi

systems into an NXT:1970Ci. SNEPs allow

semiconductor manufacturers to extend

their installed NXT systems for volume

production at new nodes. This helps them

transition from node to node in a cost

effective manner, always having the latest

lithography system while managing their

capital expenditure.

12

New nodes with your installed TWINSCAN NXTsBy Jules Tops, SNEP Project Cluster Manager DUV, and Janneke van Heteren, SNEP Product Marketing Manager

Everyone likes getting value for their

money. And at ASML, we want to make

sure we deliver the maximum value for

our customers. That’s why our TWINSCAN

scanners have always been built using

a highly modular architecture that allows

systems to be upgraded in the fi eld with

new options to improve productivity and

performance. This allows semiconductor

manufacturers to buy new systems

in confi gurations tailored to their needs

at that time, safe in the knowledge

that the tool can be adapted as their

needs change.

Now with the TWINSCAN NXT platform,

we are taking that approach a step

further. Through System Node Extension

Packages (SNEPs), any installed NXT

system can effectively be converted to

a newer model in the fi eld – extending

its capabilities by one or two production

nodes. The transformed system meets the

full ATP specifi cation for the new model,

and comes complete with a standard

ASML warranty running from the date

of the upgrade. In effect, manufacturers

gain all the benefi ts of buying a brand

new system, without the hassle of

de-commissioning an installed tool.

Our SNEP strategy gives semiconductor

manufacturers the maximum fl exibility in

planning their investment in lithography

equipment. You can buy an NXT system

for high-volume manufacturing at one

node and then upgrade it when you are

ready to transition to the next node.

This avoids the need to invest in brand

new systems each time you move to a

new production node.

Any installed NXT system can be converted to a

newer model in the fi eld

13

ASML Images, 2014

NXT:1950i 1950/1951 lens

NXT:1970Ci

Node N Node N+1 Node N+2

NXT:1960Bi 1951 lens

NXT:1960Bi 1952 lens

NXT:1950i + PEP 1950/1951 lens

Per

form

ance

B2C path

SNEP: A2C (+ 2 nodes)

A2C path

Snep = System Node Extension Package

SNEP:B2C (+1 node)

NXT configurations in the field and transformation paths

Fig. 1: SNEP is a System Node Extension Package: New hardware and software installed during a field

transformation will bring the NXT:1950i or NXT:1960Bi to the newest node. (NXT:1970Ci specs)

Following this learning period, the first SNEP

upgrade projects at customer sites were

carried out earlier this year. These systems

are now operating in production and in spec,

allowing the customers to move into volume

production at a new node.

An ongoing roadmap for value

The SNEP:A2C and SNEP:B2C are just

the first steps in our roadmap for node

extension. As each new TWINSCAN NXT

system is released, we will also be making

available packages for transforming installed

NXT systems into the very latest model.

In this way, we aim to give semiconductor

manufacturers the ability to transition from

node to node in the most cost-effective way

possible – ensuring they always have the

capabilities needed for profitable volume

manufacturing of advanced products as

well as the freedom to control their capital

expenditure and investment in new equipment.

TWINSCAN NXT:1970Ci specifications

Full-wafer dedicated chuck overlay

2.0 nm

Full-wafer matched machine overlay

3.5 nm

Full-wafer focus uniformity

20 nm

Full-wafer CDU (isolated features)

1.3 nm

Full-field throughput (96 shots)

250 wph

Defects < 7 per wafer

A completely new system

Two SNEPs are available: the SNEP:A2C

and SNEP:B2C. These transform

NXT:1950i and NXT:1960Bi systems

respectively into our latest scanner,

the NXT:1970Ci. This represents a

one-node extension for the NXT:1960Bi and

a two-node extension for the NXT:1950i.

(See Fig. 1)

Upgrades are carried out by a dedicated

SNEP team, in cooperation with our

local customer support personnel.

Each upgrade is tailored to a specific

tool via the tool number. The exact

sub-systems to be upgraded are

determined based on the system’s

current configuration. But typically the

upgrade includes replacing the wafer

handler, stage and table, and installing

the new ultra-violet level sensor (UV-LS),

parallel ILIAS (PARIS) sensor and CO2

Extending its

capabilities by one or two

production nodes

Transform NXT:1950i and

NXT:1960Bi systems into

an NXT:1970Ci

immersion hood. Where necessary,

the projection lens may also be replaced

with a newer model.

Once all the necessary hardware and

software is upgraded, the team qualifies

all the modules and then the system

as whole before carrying out the ATP

specification. The final system has all

the features and performance of a

straight-from-the-factory NXT:1970Ci

(see box), backed up by our standard

12-month new system warranty.

A flexible process

Our dedicated SNEP team has already

carried out a number of system upgrades

at ASML facilities around the world.

This has allowed us to optimize the

upgrade procedure to maximize flexibility

and minimize downtime. For example,

by performing jobs in parallel wherever

possible, we reduce the time taken for

the upgrade. This also affords us some

freedom in the order in which tasks are

carried out to reduce the impact of any

unexpected delays. Moreover, the team

is able to adjust the upgrade process

according to the space available around the

system, reducing risk.

14

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 360,0

5,0

10,0

15,0

20,0

25,0

Latest data on the progression of the NXT:1970Ci’s availability and reliability performance

Ove

rlay

perfo

rman

ce [n

m]

Systems

Latest data on the NXT:1970Ci’s focus, overlay and defectivity performance, showing consistently within spec

0

0,5

1

1,5

2

2,5

3

3,5

4

4,5

5

Ove

rlay

perfo

rman

ce [n

m]

Systems

Matched Machine OverlaySingle Machine Overlay

The TWINSCAN NXT:1970Ci By Remi Pieternella, Senior Product Manager

The latest model in our TWINSCAN NXT

high-throughput, high-precision ArF

immersion lithography platform,

the NXT:1970Ci targets profi table

volume production at the 1x nodes.

Compared to previous systems,

the NXT:1970Ci includes a new

multifunction parallel ILIAS (PARIS)

sensor for measuring the effects of

lens and reticle heating. Together with

a multi-sector wafer table heater,

this enables better overlay performance.

In addition, a brand-new ultraviolet level

sensor (UV-LS) improves focus control

and reduces process dependency during

leveling. A new high bandwidth wafer

stage enables faster wafer exchange for

higher throughput, and features enhanced

thermal control to improve overlay and

imaging performance. Meanwhile, a new

immersion hood with a carbon dioxide

“gas knife” reduces defect levels and

allows scan speeds up to 800 mm/s.

Around fi fty NXT:1970Ci systems have

been shipped to customers and are

operating within specifi cations. A number

of these systems are being used in

high-volume production, where they

are exposing over 4000 wafers per day

(averaged over a week). The average

availability for these systems is around

95%, and mean time between failures

continues to rise thanks to continuous

software improvements.

15

ASML Images, 2014

Focuson a new starBy Stuart Young, Product Manager YieldStar

16

ASML has launched the latest member

of its YieldStar diffraction-based

metrology family. The YieldStar 250D

is a next-generation metrology tool,

targeting production at the 10-nm logic

and 1X-nm memory nodes. It offers even

better accuracy and precision than its

predecessors, and reduces the time taken

for an individual measurement by around

30%. This allows customers to measure

overlay, focus and CD on product wafers

– while maintaining the same high wafer

throughput as their TWINSCAN scanner.

Building on a solid platform

The new system builds on the success

of previous YieldStar family members.

YieldStar was the first metrology system

to deliver small-target diffraction-based

overlay (µDBO) measurements.

Together with ASML’s Design for Control

(D4C) metrology targets (see page 4),

this enabled a significant step forward

in accuracy, precision and speed for

FEOL, MOL and BEOL layers compared

to traditional image-based overlay (IBO)

metrology systems. (See Fig. 1)

YieldStar systems are available in

two configurations that are identical

in design: a traditional standalone tool

and system integrated into the wafer

track. The integrated systems use

less fab space and reduce lithography

cluster cycle time as the metrology can

be carried out in the lithography cluster

without overhead. They also allow real-

time metrology feedback to the scanner.

Hence for scanners with the right options

installed, data from each lot can be

Abstract | ASML has shipped several

of its new YieldStar 250D systems.

This next-generation metrology tool targets

the 10-nm logic and 1X-nm memory nodes.

It is the first metrology system capable

of performing simultaneous on-product

overlay, focus and CD measurements

at a speed and quality that allows

real-time feedback for scanner control.

This helps semiconductor manufacturers

maximize their yields of good wafers-

per-day. A roadmap of software-only

system enhancement packages will deliver

continued in-the-field enhancements to the

new system’s performance and capabilities.

X-overlay

Yiel

dsta

r µD

BO

mea

sure

men

t AD

I

Device overlay measured by CD-SEM after etch

y = 0.9178x - 0.8156R2 = 0.9393

Y-overlay

Yiel

dsta

r µD

BO

mea

sure

men

t AD

I

Device overlay measured by CD-SEM after etch

y = 0.9452x - 0.1708R2 = 0.969

Fig. 1: Yieldstar DBO overlay measurements show excellent correlation to real device overlay

First metrology system capable of measuring overlay,

focus and CD in a production environment

17

ASML Images, 2014

used to control the scanner and

optimize exposure for subsequent lots.

This improves both individual scanner

performance and machine matching.

These unique capabilities have seen

YieldStar become the metrology tool

of choice in high-volume manufacturing

(HVM) fabs around the world.

Following on from a very successful

year in 2013, we expect the number

of installed systems to double in 2014.

Over that time, shipments of integrated

systems have grown rapidly as more

manufacturers have seen the value of

real-time, data-based scanner control,

and have now become around half of the

total number of YieldStar shipments.

The installed YieldStar systems have

proven to be excellent performers

in fabs. They have, for example,

helped semiconductor manufacturers

successfully match on-product

overlay across their installed base of

scanners. YieldStar systems have also

demonstrated impressive reliability,

with availability above 99% and 13-week

mean time between interrupt (MTBI)

fi gures of over 1000 hours. (See Fig. 2)

Shedding more light on metrology

The YieldStar 250D is based on the

same fundamental platform as its

successful predecessor the YieldStar

200C. But it takes the YieldStar benefi ts

to the next level by dramatically

increasing the amount of light reaching

the wafer. It does this in two ways.

First, the xenon arc lamp used in the

YieldStar 200C has been replaced with

a new source that delivers substantially

more light and improved illumination

characteristics. Secondly, the optical path

from source to wafer has been improved

to increase transmission and extend the

range of transmitted wavelengths.

The more powerful source and improved

transmission increase the total amount of

light at the wafer level by up to 50 times.

As a result, the specifi ed move-acquire-

measure (MAM) time is just 0.35 seconds

(compared to 0.5 seconds for the

YieldStar 200C). Hence the YieldStar 250D

is capable of 1200-1600 measurements

per lot at the full production throughput of

ASML’s latest TWINSCAN NXT scanners.

Tests on product wafers from logic and

memory customers have shown that

the YieldStar 250D is consistently over

30% faster than the YieldStar 200C.

Improvements are even more pronounced

for more challenging layers – with gains

above 80% in some cases. (See Fig. 3)

Supporting advanced chip structures

Meanwhile, the extended wavelength

range allows the new system to support

3D chip features such as FINFETs and

3D-NAND structures. These structures

require thicker process stacks leading

to lower light transmission which makes

metrology more challenging. The YieldStar

250D uses sensing wavelengths up to

780 nm to increase transmission through

the stack. Hence, it can deliver fast,

accurate metrology on FINFET devices

at the 10 nm logic node as well as

3D-NAND structures.

More focus on focus

Prior to the YieldStar 250D, scanner control

based on metrology feedback focused

on optimizing on-product overlay

performance. Of course, overlay isn’t

the only performance metric that affects

wafer yield. The increased speed of

the YieldStar 250D means users can

now also measure on-product imaging

performance – specifi cally focus and CD.

This makes the YieldStar 250D the fi rst

metrology system capable of measuring

overlay, focus and CD in a production

environment.

YieldStar’s unique diffraction based-

focus (DBF) measurement directly and

accurately measures on-product focus

across the entire process window. Just as

for overlay, YieldStar focus measurements

can be used as the basis for calculating

and applying real-time corrections via

ASML’s Imaging Optimizer scanner option.

For focus critical features, these exposure

corrections have demonstrated on-product

CDU improvements of around 15%.

(See Fig. 4)

The YieldStar 250D also supports CD

measurements as an option.

CD measurement capabilities have

been extended, with a particular emphasis

on reducing the time required to create

recipes. This allows semiconductor

manufacturers to measure overlay,

focus and CD simultaneously with

no loss of productivity.

Facing the future

The YieldStar 250D is part of ASML’s

ongoing holistic lithography roadmap,

with future systems planned to keep

metrology performance on pace with

scanner capabilities. Just like our

scanner platforms, this roadmap is built

on a philosophy of continuous system

enhancement. All installed YieldStar 200C

systems can be fully upgraded to the new

YieldStar 250D in the fi eld. Several fi eld

upgrades have already been successfully

performed, and customer demand for

such upgrades remains strong. When the

time comes the YieldStar 250D will be

fi eld-upgradeable to subsequent models.

On-product CDU

improvements of

around 15%

18

Median Reliability

MTBI of Yieldstar machines

2013

1420

1316

2013

1820

1320

2013

2220

1324

2013

2620

1328

2013

3020

1332

2013

3420

1336

2013

3820

1340

2013

4220

1344

2013

4620

1348

2013

5020

1352

2014

0220

1404

2014

0620

1408

2014

1020

1412

2014

1420

1416

2014

1820

1420

2014

2220

1424

2014

2620

1428

2014

3020

1432

0

200

400

600

800

1000

1200

1400

MTB

I [ho

urs]

200C MAM250D MAM

MAM time data

Node

14nm

14nm

14nm

14nm

14nm

16nm

16nm

16nm

16nm

16nm

16nm

16nm

16nm

16nm

16nm

16nm

16nm

16nm

10nm

10nm

10nm

10nm

10nm

10nm

10nm

10nm

10nm D1x

D2x

D2x

D2x

D2x

D2x

D2x

D2x

D2x

D2x

14nm

14nm

14nm

0

0,20,1

0,30,40,50,60,70,80,9

1

Mov

e-A

cqui

re-M

easu

re ti

me

[sec

]

Slot #

Uncorrected focus uniformitySimulated improvement with 1) focus offset correction 2) focus offset and tilt correction

20

5

10

15

20

25

30

4 6 7 8 9 10 Ave

Focu

s 3σ

(nm

)

Interfield Focus 3σ (nm)Simulated Z per field correction potential 3σ (nm)Interfield Focus 3σ (nm)

26.3

25.1

22.0

19.2

26.2

21.8

19.5

24.1

20.2

18.0

26.9

23.3

21.0

25.6

23.2

20.7

25.7

22.2

20.2

25.7

22.0

19.821

.419

.6

Slot #1

0

5

10

15

20

25

30

2 3 4 5 6 8 97 Ave

Focu

s 3σ

(nm

)

Measured focus uniformity with focus offset & tilt corrections applied shows 20% improvements

Focus Uniformity (interfield) 3σ (nm)

20.2

20.2

20.4

20.8

21.2

20.9

21.221

21.5

21.2

Fig. 4: Yieldstar on product focus measurement together with Imaging Optimizer delivers 20% on-product focus uniformity improvement

Fig. 2

Fig. 3: The 250D delivers more than 30% increase in sampling

Improvements aren’t limited to just

developing new systems. Our YieldStar

roadmap also includes regular,

software-only system enhancement

packages to extend the capabilities and

economic lifetime of installed systems.

These packages allow us to make

signifi cant usability improvements for

both R&D engineers and HVM users,

and to introduce new functionality.

Furthermore, these enhancement

packages are used to deliver increased

sampling, accuracy and precision

performance. In this way, we help

ensure that all customers have the latest

metrology capabilities available in a

timely and robust fashion.

19

ASML Images, 2014

Corporate Headquarters

De Run 6501

5504 DR Veldhoven

The Netherlands

Phone +31 40 268 30 00

United States

2650 W. Geronimo Place

Chandler, AZ 85224 USA

Phone +1-480-696-2888

Asia

Suite 1702-3 17th Floor

100 Queen’s Road Central

Hong Kong, SAR

Phone +852 2295 1168

www.asml.com