In-Situ Metrology: the Path to Real-Time Advanced Process Control

3/27/03 12003 Int'l. Conf. Characterization & Metrology for ULSI TechnologyG. W. Rubloff © 2003

In-Situ Metrology: the Path to Real-Time Advanced Process Control

Gary W. RubloffProfessor, Materials Science and Engineering, Institute for Systems Research,

and Electrical and Computer EngineeringUniversity of Maryland

www.isr.umd.edu/gwrubloff, [email protected]

Department of MaterialsScience and Engineering

http://www.isr.umd.edu/gwrubloff


Synopsis• Advanced process control (APC) has become pervasive

• In-situ metrology is key to extending this to real-time APC

• In-situ chemical sensors provide viable quantitative real-time metrology– Multiple sensors deliver <1% precision– Real-time end point control demonstrated– Course correction as well as fault detection– Application to CVD, PECVD, etch, spin-cast, …

• New opportunities– Uniformity control spatially programmable reactor design– Precursor delivery control solid & low pvapor sources

ready for tech transfer & evaluation in manufacturing environment


Advanced Process Control (APC)

deposition pattern generation etching metrologymetrology

Advanced ProcessControl (APC)

Course correctionCompensate for variationsto maintain process targets

Fault managementIdentify and repair

equipment problemsSensor-drivenModel-based

run-to-runfeedback control

run-to-runfeedforward

control

Run-to-run controlFeedback & feedforward

Fault classification & responseOptimize management of faults

fault classification& response

Real-time controlEnd point and instantaneous

real-timecontrol

sensor

Real-time fault detectionKnown failure modes and signatures

real-time faultdetection

sensor


APC Hierarchy

supervisory

regulatory

Factory Control System

In-line & other metrologyMulti-step process controlMulti-scale time response

Run-to-runcourse correction-feedback-feedforward

Course correction

Tool maintenance- fault diagnosis- PM scheduling

Fault management

In-situ, real-time process sensorsUnit process control Real-time response

Real-timefault detection

Real-timecourse correction

quantitative metrologyin-situ, real-timeRealReal--Time APCTime APC

Real-time equipment sensorsEquipment component controlReal-time response

Regulatory tool control- equipment state

Tool Control System


In-Situ Sensors for Quantitative Process Metrology

REQUIREMENTS• In-situ, real-time• Quantitative precision (~1%)

– Required for course correction

• Process state• Wafer state• Preferably multi-use

– Indicators of process & wafer state

– Simultaneous application for fault detection

• Rich information– Chemically specific

• Robust, integratable

TECHNIQUES• Plasma optical emission

spectroscopy (OES)• Laser/optical interferometry• Mass spectrometry• Acoustic sensing• Fourier transform infrared

spectroscopy (FTIR)• Plasma impedance• Optical

thermometry/pyrometry• Ellipsometry• Optical scatterometry• …


Mass Spectrometry for Real-Time APC

PROCESS CHAMBER

Chemical vapor depositionchamber for tungsten metal

H2SiH4/WF6

BP

Reactor exhaust

Turbo pump50 l/s

QMS

Ion gauge

to drag stage

orificeorificeorifice

Pressurecontrolvalve

Baratron

PROCESSGAS FLOW

SENSORGAS FLOW

Ulvac multi-chamber“cluster” tool CHEMICAL SENSORS

InficonComposer

acoustic sensor

Inficon Transpectormass spec

chemical sensor

Pressure transductionto low pressure


Real-Time Mass Spec in W CVD

2 WF (g) + 3 SiH (g) 2 W (s) + 3 SiF (g) + 6 H (g)6 4 4 2WF6 (g) + SiH4 (g) W (s) + 2 SiHF3 (g) + 3 H2 (g)

2 WF6 (g) + 3 SiH4 (g) 2 W (s) + 3 SiF4 (g) + 6 H2 (g)

WF6 (g) + SiH4 (g) W (s) + 2 SiHF3• W CVD by SiH4reduction of WF6 in 0.5 torr thermal CVD

• Monitor process state as gas concentrations in reactor

• Product generation and reactant depletion reveal wafer state changes in real time

(g) + 3 H2 (g)


Real-Time Thickness Metrology

• Reasonable Conversion Rate of WF6 reactant (~20%)

• Metrology established from weight vs. integrated mass spec signal– Linear regression

standard deviation 1.09%

• Viable for manufacturing process control

SiH4 reduction of WF60.5 torr, 250°C

SiH4 reduction of WF60.5 torr, 250°C

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0H2 Integration Signal (arb unit)

W F

ilm W

eigh

t (g)

7/14/00, y = 89069 x + 0.0004, R2 = 0.99977/07/00, y = 90876 x - 0.0002, R2 = 0.99896/28/00, y = 90830 x – 0.0005, R2 = 0.99936/20/00, y = 87682 x + 0.0004, R2 = 0.9944


Real-Time Thickness Control

W F

ilm W

eigh

t (g)

H2

Inte

grat

ion

Sign

al

target H2 value

no control (experimental)

with control

Wafer Number Wafer Number

no control (experimental)

with control

0.0E+00

1.0E-07

2.0E-07

3.0E-07

4.0E-07

5.0E-07

0 1 2 3 4 5 6 7 8 9 100.000

0.010

0.020

0.030

0.040

0 1 2 3 4 5 6 7 8 9 10

Mass specMass spec Film thicknessFilm thickness

• Open-loop wafer-to-wafer thickness variation ~ 10%• Real-time end-point control of thickness to ~ 3%• Real-time course correction to compensate for BOTH:

– Random short-term variability– Systematic longer-term drift


Mass Spec Thickness Metrology

H reduction of WF2 6WF (g) + 3 H (g) W(s) + 6 HF(g)6 2

Fixed process condition: 10 torr, 500°C, 640 sec

H2 reduction of WF6WF6(g) + 3 H2(g) W(s) + 6 HF(g)


0.02.0x10-5 2.1x10-5 2.2x10-5 2.3x10-5 2.4x10-5 2.5x10-5 2.6x10-5 2.7x10-5 2.8x10-50

230

240

250

260

270

280

3

24

51

67

89

10

W F

ilm W

eigh

t (m

g)

Integrated HF Signal (Amps-sec)

0

15500

16000

16500

17000

17500

18000

Linear RegressionAverage Uncertainty +/- 0.56%

Standard Deviation 0.67%

0 0 0

4%

Estim

ated

Ave

rage

Film

Thi

ckne

ss (A

)

0 1 2 3 4 5 6 7 8 9 10 110

230

240

250

260

270

280

W F

ilm W

eigh

t (m

g)

Wafer (or Run) Number

0

15500

16000

16500

17000

17500

18000

10 wafer average drift +/- 1.18%(extreme case +/- 3.99%)

255.7

0 0 Estim

ated

Ave

rage

Film

Thi

ckne

ss (A

)

Run-to-run thickness driftAverage 1.18%Extreme 3.99%

Mass spec thickness metrologyAverage uncertainty 0.56%Standard deviation 0.67%


Mass Spec Thickness Metrology: Intentional Temperature Drift

Intentional Run-to-Run Temperature DriftFixed Deposition Time 618 sec

Intentional RunIntentional Run--toto--Run Temperature DriftRun Temperature DriftFixed Deposition Time 618 sec• Introduce significant

temperature drift to test robustness of metrology

• Substantial change in thickness (4X)

– Much larger than expected in manufacturing

0 5000 10000 15000 200000

1x10-10

2x10-10

3x10-10

4x10-10

400'C

WF6

Sig

nal (

Amps

)

0

2x10-8

4x10-8

6x10-8

18171615

14131211

10987654321

HF S

igna

l (Am

ps)

340350360370380390400

340'C350'C360'C

370'C380'C385'C390'C395'C

Tem

pera

ture

('C)

Time (sec)


Mass Spec Thickness Metrology: Intentional Temperature Drift

Moderate non-linearity over broad temperature rangeDeposition on showerhead, adsorption on chamber walls, …

Metrology precision ~ 0.6% near local process setpoint

0.01.0x10 -5 1.5x10 -5 2.0x10 -5 2.5x10 -5 3.0x10 -5 3.5x10 -5 4.0x10 -5

0

50

100

150

200

250

300

350

Wafer #2-#10 (400'C-380'C)Linear Regression

average uncertainty +/- 0.55%standard deviation 0.62%

Wafer #9-#18 (380'C-340'C)Linear Regression

average uncertainty +/- 0.57%standard deviation 0.69%

W fi

lm w

eigh

t (m

g)

Integrated HF signal (amp-sec)

0

5000

10000

15000

20000

Estim

ated

ave

rage

film

thic

knes

s (A

)


Mass Spec Thickness Metrology: Intentional Process Time Drift

0.0 1.0 2.0 3.0 4.0 5.0 6.00

50

100

150

200

250

300

350

400

W F

ilm W

eigh

t (m

g)

HF Integration Signal (10-5 Amps-sec)

0

5000

10000

15000

20000

25000

Estim

ated

Film

Thi

ckne

ss (A

)

H reduction of WF2 6WF6(g) + 3 H (g) W(s) + 6 HF(g)2

10 torr, 390°C

H2 reduction of WF6WF6(g) + 3 H2(g) W(s) + 6 HF(g)

10 torr, 390°C

• Introduce significant process time drift to test robustness of metrology

• Substantial change in thickness (4X)

– Much larger than expected in manufacturing

• Linear regression fit– Average uncertainty 1.19%– Standard deviation 1.59%

• Quadratic regression fit– Average uncertainty 0.48%– Standard deviation 0.57%


Acoustic Sensing for Real-Time APC

• Acoustic wave propagation and resonance

P > 50 torr

• Resonant frequency depends on average molecular weight, specific heat, and temperature of gas mixture

C = speed of sound

Gas compression(0.5 cfm diaphragm pump)

acousticsensor100 torr

Capacitancegauge

Pressure ctrl.valve

WF6 H2

4” wafer

Pressure controllerto process pumps

CVD process10 torr

Needle valve

Pressure transductionto higher pressure

Transmitter

InletAcoustic chamber Heater

Outlet Receiver

avg

avg

MRTγ

Cwith2LCF ==


Acoustic Sensor Thickness Metrology

H reduction of WF2 6WF (g) + 3 H (g) W(s) + 6 HF(g)6 2


H2 reduction of WF6WF6(g) + 3 H2

Run-to-run thickness driftAverage 4% over 10 runs

Acoustic sensor thickness metrology

0.5% average uncertainty from linear regression fit

(g) W(s) + 6 HF(g)Fixed process condition: 10 torr, 490°C, 640 sec

+/- 1 % weight error

01000 84000 88000 92000 960000.00

0.01

0.02

0.20

0.21

0.22

0.23

0.24

0.25

0.26

0.27

Wafer 203 to 212Linear Fit of Data2_B

W fi

lm w

eigh

t (g)

Frequency integrated signal (Hz.s)

0

1000

14000

15000

16000

17000

Estim

ated

film

thic

knes

s (A

)


FTIR Sensing for Real-Time APC• Implementation like acoustic sensor

P > 50 torr• Sense molecular vibrations (infrared) for product generation, reactant

depletion• WF6 product depletion thickness metrology precision ~0.5%

0.14 0.15 0.16 0.17 0.18 0.19 0.20 0.21580

600

620

640

660

680

410oC

410oC

405oC400oC400oC

395oC395oC

390oC390oC WF6 absorption at 712.5 cm-1

R=-0.99672

Inte

grat

ed W

F 6Si

gnal

Mass of deposited tungsten (g)

H2 reduction of WF6WF6(g) + 3 H2(g) W(s) + 6 HF(g)Intentional temperature drift: 10 torr, 390-450°C

H2 reduction of WF6WF6(g) + 3 H2(g) W(s) + 6 HF(g)Intentional temperature drift: 10 torr, 390-450°CWF6


Sensor Integration

real-time control

equipment stateLabView

sensorcontrol

process & wafer state metrology

sensor

Brooks

toolcontrol

Pumpsystem

W CVD Reactor

Centralwafer

handler

W CVD Reactor

Loadlock Ulvac ERA-1000

W CVD cluster tool

VisSim

dynsimul dynamic system

simulation


Interface and Thin Layer Sensitivity

SiF4 product of Si reduction reaction3 Si (s) + 2 WF6(g) 2 W(s) + 3 SiF4(g)

SiF4 product of Si reduction reaction3 Si (s) + 2 WF6(g) 2 W(s) + 3 SiF4(g)

1.0 1.5 2.0 2.5 3.00.0

1.0x10-12

2.0x10-12

3.0x10-12

4.0x10-12

5.0x10-12

6.0x10-12

Mas

s sp

ec s

igna

l (A

)

time (min)

“cold” waferHF signal

HF production “hot” waferHF signal

H reduction of WF :2 6WF6(g) + 3 H2(g) W(s) + 6 HF(g)

H2 reduction of WF6:WF6(g) + 3 H2

Initial reaction dominated by WF6 - Si nucleation

~30 nm W formedReadily observed

(g) W(s) + 6 HF(g)

Implications:Fault detection

assure oxide-free contacts)

Metrology for ultrathinCVD layers


Ready for Technology Transfer• In-situ sensors deliver metrology for real-time APC

– Quantitative precision for real-time course correction– Dual-use sensors to drive both course correction and fault

management (e.g., mass spec)• Research underpinnings in place

– Multiple sensors with metrology at 1% or better– Real-time end point control demonstrated– Sensor-tool integration

• Ready for implementation in manufacturing environment – Compatible with existing/installed real-time sensors for fault detection– UMD anxious to assist, collaborate, … –– Prediction: further improvement in metrology precisionPrediction: further improvement in metrology precision

• High throughput enhances sensor & tool conditioning


Across-Wafer Uniformity

• Key manufacturing metric for yield

• In-situ sensor capability to date– Spatially resolved optical (OES) –

process state– Full-wafer interferometry – wafer

state

• No mechanism for real-time uniformity adjustment

• Currently, process optimization involves tradeoff between material quality metrics and uniformity

Material quality

Uniformity

Recipe 1

Recipe 2

Recipe 3

Choose compromise as process design to balance uniformity and material quality for fixed reactor configuration


Programmable Uniformityto Optimize Quality & Manufacturability

Choose process design point for optimal material

and device performance

Choose process design point for optimal material


Uniformity unacceptable

Retune spatial conditions in programmable reactor

Retune spatial conditions in programmable reactor

Uniformity unacceptable

Produce high uniformity with optimal material


Produce high uniformity with optimal material



Spatially Programmable CVD Uniformity through a Smart Showerhead

Sensors - integrated into the showerhead

Spatially resolved, multizonewafer and process state measurements

Actuators - multizone, gas inlet

Gas flow rates and compositions controlled within each showerhead segment

Supplementary pumping through the showerhead

Reduced inter-segment gas mixing, precise composition control, gas sampling for chemical sensing

Simulation and reduced-order models

Support for process equipment design and control

Multipoint OpticalSensing

(e.g., full-waferinterferometry)

Heated SusceptorWafer

Gas Exhaust toProcess Pumps

Reactor

MultisectorGas Inlet

and Showerhead

IlluminationSource

Multipoint GasPumping & Sensing

(e.g., mass spec)


Experimental Testbed:Spatially Programmable CVD Showerhead

inlet

exhaust

3-segment prototypeExhaust and inlet in each segmentAdjustable spacings to wafer

intersegment vs. intrasegmentmixing

Extensive coupling to modeling

3-segment prototypeExhaust and inlet in each segmentAdjustable spacings to wafer

intersegment vs. intrasegmentmixing

Extensive coupling to modeling

individual segment :quasi-independent mini-showerhead incorporating gas inlet, exhaust, sensing, and model-based control of actuation


Experimental Testbed:Spatially Programmable CVD Showerhead

• Ulvac ERA-1000 W CVD cluster tool• W CVD process using WF6 + H2

Segmented showerhead

Gas management(inlets and exhaust)


Programmable Nonuniformityfor Rapid Materials & Process Development

One-wafer DOE process optimizationOne-wafer DOE process optimization

Combinatorial CVD new materials discovery and development Combinatorial CVD new materials discovery and development


Precursor Delivery Challenges

• Solid & low vapor pressure sources increasingly critical for new materials

• Precursor delivery control remains problematic

– Changing morphology with time and usage

– Adsorption on walls– Complex chemical precursors

• Options limited for both chemical precursor and delivery system design

Example: Cp2Mg temperature decrease 40 32°C reduces vapor pressure & composition 2X

Simulates “aging” effects


Real-Time Precursor Delivery Control

• Acoustic sensor for composition metrology

• Source and dilution gas flow control

Carrier gas flow set point

MFC

H2

H2 dilution

H2 carrier

P

Compositionmeasurement

Dilution flow set point

Composer controller

MFC

Composeracoustic

transducer50 g. Cp2Mgsolid source

precursordelivery

Source temperature varied from 4032oC

Σ (H2 flows) = 150 sccm, P = 300 torrCp2Mg target = 0.01 mol%

Cp2Mg composition controlled to 1% of target (0.0001 mol %)


MOCVD Precursor Delivery Line Control

T(source) = 40oC

60 ± 1.5oC

0 20 40 60

0.0092

0.0096

0.0100

0.0104

0.0108

Time (min)

20

40

60

80

100

120

Cp 2

Mg

com

posi

tion

(mol

%) Closed loop Open loop

H2

flow

rate

s (s

ccm

)0 20 40 60

0.0092

0.0096

0.0100

0.0104

0.0108

Time (min)

20

40

60

80

100

120Closed loop Open loop

Acoustic sensor reveals wall adsorption fluctuations optimize wall temperature

Control system suppresses delivery fluctuations

50 ± 1.5oCT (feed line) =

H2

flow

rate

s (s

ccm

)


• In-situ metrology is key to achieving real-time APC– Benefits in rapid feedback at unit process (tool) level– Implementation within hierarchical control framework

• In-situ chemical sensors provide quantitative real-time metrology– Multiple sensors with <1% precision– Real-time end point control demonstrated– Course correction synergistic with fault detection– Broad applications - CVD, PECVD, etch, spin-cast, …

• Ready for tech transfer, evaluation in manufacturing environment

• New opportunities– Uniformity control – Precursor delivery control

Conclusions


Acknowledgements• Research group

– L. Henn-Lecordier, J. N. Kidder, T. Gougousi, Y. Xu, S. Cho, R. A. Adomaitis, J. Choo, Y. Liu, R. Sreenivasan, L. Tedder, G.-Q. Lu, A. Singhal

• NIST– J. Whetstone, A. Lee, C. Tilford

• Inficon– R. Ellefson, L. Frees, C. Gogol, A. Wajid, J. Kushneir

• Other colleagues– Metrology TWG, AEC/APC, AVS MSTG

• Support

In-Situ Metrology: the Path to Real-Time Advanced Process Control

Documents

Transcript of In-Situ Metrology: the Path to Real-Time Advanced Process Control