Optical Emission Spectroscopy for Plasma Processing based ...adp-dresden.de/papers/PlasmaMPCA.pdfThe authors of this presentation would like to thank Siegfried Bernhard Lars Christoph

SEMATECH AEC/APC Symposium XIISept 24-28, 2000, Lake Tahoe, Nevada

24.05.2000Page 1

Jan ZimpelKnut VoigtländerFraunhofer InstitutIVI DresdenAndreas SteinbachInfineon DresdenDirk KnoblochInfineon München

Jan ZimpelKnut Voigtländer

Fraunhofer IVI DresdenAndreas Steinbach

Infineon Technologies DresdenDirk Knobloch

Infineon Technologies München

Using Multi Way PCA (MPCA) forAdvanced Monitoring and Diagnosis

for Plasma Processing based onOptical Emission Spectroscopy


24.05.2000Page 2


Acknowledgement

The authors of this presentationwould like to thank

Siegfried BernhardLars ChristophBarbara SchmidtInfineon Technologies Dresden


24.05.2000Page 3


Outline

� Introduction - APC in high volume production� Hardware integration and software structure� Data reduction by PCA� Experiments

– Contact etch at AMAT MxP+– Poly etch at AMAT DPS

� Summary and outlook


24.05.2000Page 4


APC - offline analysis andreal time process control including alarms

ToolParameters

ProcessParameters

ProductParameters

Data Base Offline AnalysisModel

Logistic Data(Wafer, Recipe)

Real TimeMonitoring Online Alarm

APC in high volume production


24.05.2000Page 5


Data reduction – an essential needfor APC in high volume production

� APC in high volume production creates large amounts of data� Data reduction is an essential need for off line analysis

and real time process monitoring� Methods for data reduction:

– Measurement techniques based on physical models– Calculation of statistical key numbers– Use of complex process parameters– Model based data analysis

APC in high volume production


24.05.2000Page 6


Features of Hamamatsu MPM spectrometer

� Spectral range: 200 - 950 nm� Resolution: < 2 nm� CCD line channels: 1024� Connection to Host PC

via TCP-IP, RS 232� Internal data processing for endpoint

detection; up to 100 endpoint scrip‘s areavailable

� Digital / analog port‘s for connection to tool

Hardware integration and software structure


24.05.2000Page 7


Integration of Hamamatsu MPM spectrometer

FAB-LAN

Plasma

Equipment

SECS SECSTICSpassthrough

server

EquipmentManager

start / stop spectrarecipe

process steplot number

winsockprotocol

application software

tool endpointinterface

� Tool interface for stand alone endpoint detection

� Interface for logistic datae.g., lot and wafer number, recipe, step number



24.05.2000Page 8


Software solution developed byFraunhofer Institut IVI Dresden

A p p l i c a t i o nprocess

database

measurementcomponent

singlespectrumanalysis

multiplespectrumanalysis

endpointsynthesis

tool

modeldatabase

on-linemonitoring

tool

FAB LAN

spectral data fromMPM

start /stop measurement

MatlabServer

�Database oriented spectrastorage and SQL- baseddata access for:

- Data visualization- Data analysis- Endpoint synthesis- Validation of endpoint

detection algorithms



24.05.2000Page 9


Data reduction by key number calculationwith PCA

� Simple key number extraction:mean, standard deviation, max, min, ...

� Extraction of key numbers using signal decomposition:– Tschebyscheff functions– Adjusted signal base (PCA)

Multivariate key number extraction -- Multi Way PCA� Adaptation of a nonlinear parametric signal model� Compromise between efficiency and effort / a-priori knowledge

Data reduction by PCA


24.05.2000Page 10


ntR ××λ∈Λ

Principle of PCA – Data cube

� Optical spectra visualized as a „Data cube“� Optical emission spectroscopy creates very large amounts of data !

wavelength λλλλwafer 1 ... n

Vertical and horizontal cutthrough Data cube

etch

tim

e t

wavelength λ

time

Data cubecontaining spectra



24.05.2000Page 11

Jan ZimpelKnut VoigtländerFraunhofer InstitutIVI DresdenAndreas SteinbachInfineon DresdenDirk KnoblochInfineon München 350 400 450 500 550 600 650 700 750

1

2

3

4

5

6

x 105

m11

m21

m31

m19

m29

m39

Principle of PCA – Matrix calculation

� Split of the original data matrix into orthogonal pattern ui andorthogonal scores mi:

� Scores represent the weight of the corresponding pattern in theoriginal data sample

⋅=⋅=i iuimTUMX T

350 400 450 500 550 600 650 7000

0.05

0.1

0.15

300 350 400 450 500 550 600 650 700-0.15

-0.1-0.05

00.05

300 350 400 450 500 550 600 650 700-0.2

0

0.2

wavelength

original data: X scores: mi base pattern: ui



24.05.2000Page 12


Application of PCA on DRAM contact etchat Applied Materials MXP+ chamber� Contact etch at Applied Materials MxP+ chamber� Standard oxide etch chemistry, CF4, CHF3, Ar� Observation of 5 wet clean cycles (WC), about 4000 wafers� Simple process mix, two different recipes for two high volume

DRAM products mainly

Step Product 1 Product 2 Descum -- N2 / O2 descum Main etch 1 BPSG etch BPSG etch Main etch 2 -- Nitride etch

Contact etch in AMAT MxP+


24.05.2000Page 13


3 0 0 3 5 0 4 0 0 4 5 0 5 0 0 5 5 0 6 0 0 6 5 0 7 0 0 7 5 0-0 .2 5

-0 .2

-0 .1 5

-0 .1

-0 .0 5

0

0 .0 5

0 .1

0 .1 5

w a ve le n g th [n m ]

norm

. Int

ensi

tät

W C 2

3 0 0 3 5 0 4 0 0 4 50 5 0 0 5 5 0 6 0 0 6 5 0 7 0 0 7 50-0 .1 5

-0 .1

-0 .0 5

0

0 .0 5

0 .1

0 .1 5

0 .2

0 .2 5


norm

. Int

ensi

tät

W C 2

PCA results obtained on DRAM contact etchat Applied Materials MXP+ chamber

300 400 500 600 700 800 9000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Matrix X: mean spectraof 4000 wafersof 5 wet clean cycles (WC)

3 0 0 3 5 0 4 0 0 4 5 0 5 0 0 5 5 0 6 0 0 6 5 0 7 0 0 7 5 00

0 .0 5

0 .1

0 .1 5

0 .2

0 .2 5


norm. Intensität

W C 2

PCA

1st order

3 0 0 3 5 0 4 0 0 4 5 0 5 0 0-6 0 0 0

-5 0 0 0

-4 0 0 0

-3 0 0 0

-2 0 0 0

-1 0 0 0

0

1 0 0 0

2 0 0 0

3 0 0 0

4 0 0 0W C 2W C 3W C 4W C 5

0 1 00 2 00 30 0 4 00 5 00 60 0 7 00 8 00 90 0-8

-6

-4

-2

0

2

4

6x 1 0

4

W C 2W C 3W C 4W C 5

wavelength [nm]

Inte

nsity

[arb

. uni

ts] stacked

mean spectraof 4000 wafers

3rd order

2nd order

0 2 0 0 4 0 0 6 0 0 8 0 0 1 0 0 0

2 .2

2 .4

2 .6

2 .8

3

3 .2

3 .4

x 1 05

W C 2W C 3W C 4W C 5

wavelength [nm]

Inte

nsity

[arb

. uni

ts]

wafer

scor

es [c

ount

s]

Patterns and scores 1st to 3rd order of WC 2 .. 5

1st order

2nd order

3rd order



24.05.2000Page 14


PCA results obtained on contact etch atApplied Materials MXP+ chamber, cont.� CF4, CHF3, Ar chemistry, two main DRAM products� Scores of 2st order of the first observed at wet clean cycle 1

� Product 1: high polymerizingProduct 2: low polymerizing

� Scores of 2st order decreaseduring WC1, caused by:

- Increasing light absorption atpolymer layer on the recessside window

- And real process drift causedby polymer on chamber wall

� Product dependent monitoringof chamber conditionrf hours [h]

scor

es [c

ount

s]

scores of 2st order vs. rf hours

product 1

product 2product 3

one point – one wafer



24.05.2000Page 15


Multi-Way Principle Component Analysis

N1,...,n for

(t)jv)(iui j

n)j,M(i,

=

′λ⋅

� “One way PCA”: use of one mean spectrum per wafer� Multi Way PCA: Calculation of orthogonal wave pattern ui

and orthogonal base time signals vi by unfolding the originaldata cube in time and wave direction

ntR ××λ∈Λ

wavelength λλλλwafer 1 ... net

ch ti

me t

Data reduction by Multi Way PCA


24.05.2000Page 16


50 100 150

0.12

0.121

0.122

50 100 150

-0.1

0

0.1

0.2

50 100 150

-0.2

0

0.2v1 v2v3

400 500 600 700-0.1

0

0.1

300 400 500 600 700-0.1

0

0.1

0.2

300 400 500 600 700-0.05

00.050.1

0.15

300 400 500 600 700-0.1

0

0.1

wavelength [nm ]

u2

u3

u4

u5

Some examples of key numbers obtained byMulti Way PCA on contact etch at AMAT MxP+

extra

cted

bas

ic wa

ve p

atte

rn u

i

resulting key numbers vs. wafer

extractedbasic time signals vj

wafer

scor

es o

f key

num

ber u

iv j [cou

nts]



24.05.2000Page 17


Interpretation of key numbers ui;vj

� Significant signatures up to10th…20th order of ui and vj, max. about 100...400 key numbers

� Significant key numbers limited by:– increasing order

decreasing information content– redundant signatures

� PCA = mathematical algorithm, nophysical or technological input

� Advantage: universal, applicationto any kind of data possible

� Disadvantage: no clear physicalmeaning of these key numbers

� Difficult interpretation


� Interpretation of key numbers with help of:– Physical, chemical, technological knowledge– Comparison to other measurement techniques,

delivering physical parameters


24.05.2000Page 18


Interpretation of optical key numberswith experience

� Key number u2;v1 showsreproducible long term driftbetween wet cleans.

� Experience possible reasons:– Light adsorption by polymer,

growing on recess side window– Drift of gas composition,

caused by polymer on thechamber walls

� No influence of power dissipationhere

wafer

key n

umbe

r [co

unts

]key number u2;v1 vs. wafer

one point –one wafer



24.05.2000Page 19


Reference: Plasma parameter measurementwith SEERS

� SEERS = Self Excited Electron PlasmaResonance Spectroscopy= „electrical“ plasma measurement technique

� Measurement of:– rf current– rf voltage

� Real time calculation of plasma parameters:– Electron collision rate [collisions per sec]– Electron density [electrons per cm³]– Bulk power [mW per cm²]– DC bias voltage [V]

� Plasma monitoring system HERCULES, basedon SEERS was used as reference system

FFT

rf currentrf voltage

Model SEERS

Electron collision rateElectron density

Bulk powerDC bias voltage



24.05.2000Page 20


0 50 100 150 200 250 300-1

0

1

2

x 104

50 100 150 200 250 300-1

0

1

2

3

x 104

wafer

key n

umbe

r [co

unts

]

wafer

key number ui;vj vs. wafer


Interpretation of optical key numberswith comparison to plasma parameters

� Short term drift indicated by:– Optical key numbers,

e.g., u4;v1, u5;v1– Electron collision rate

electron collision rate vs. rf hours

9,2

9,4

9,6

9,8

10,0

10,2

15 20 25 30

rf hours [h]co

llisi

on ra

te [1

0 7 s -1

]


� Possible reasons:– Temperature drift– Gas adsorption and desorption



24.05.2000Page 21


wafer

key n

umbe

r [co

unts

]


key number u4;v2 vs. wafer

Product 1

Product 2

Interpretation of optical key numbers withcomparison to plasma parameters, cont.

� Product indicated by:– Optical key number u4;v2

(no optical measurements available during the tool failure)

– Electron density

� Reason:– Different open area

of 2 products

electron density vs. rf hours

9

10

11

12

260 310 360 410

rf hours [h]el

ectr

on d

ensi

ty [1

08 /cm

3 ] WC3WC4WC5

Wetclean

one point – one lot

Tool failuredetected (Arcing)

Product 1Product 2



24.05.2000Page 22


2 0 4 0 6 0 8 0 1 0 0 1 2 0

-1 0 0 0

-8 0 0

-6 0 0

-4 0 0

-2 0 0

0

2 0 0

4 0 0

wafer

key n

umbe

r [co

unts

]key number u5;v6 vs. wafer

etch time

optic

al in

tens

ity [a

rb. u

nits

]

optical endpoint signal vs. etch time


•10 •20 •30 •40 •50 •60

•0.96

•0.97

•0.98

•0.99

•1

•1.01

•1.02

•1.03

•1.04

•1.05

•x 10•4

one curve –one wafer

Interpretation of optical key numbers withcomparison to endpoint signatures

� Key number u5;v6 corresponds with endpoint time� Superimposition of previous processes, depending on lot



24.05.2000Page 23


Topical Example: Chamber comparison atpoly recess etch in Applied Materials DPS

� Measurement atchamber B severalweeks later.

� Key numbers indicateother conditions atchamber B (see 3).

� Reasons not yetidentified.

Poly recess etch in AMAT DPS

wafer wafer

key n

umbe

r [co

unts

]key numbers vs. wafer

Chamber A Chamber B1

3

2

1

2

3one point – one wafer


24.05.2000Page 24


Summary and outlook

� PCA / MPCA is a universal mathematical method for data analysis and datareduction.

� Key numbers obtained by application of PCA / MPCA on optical spectra arecomplex process parameters, indicating tool and wafer impacts.

� Interpretation of key numbers is possible by use of:– extracted spectral wave pattern and basic time signals– physical, chemical, technological knowledge– comparison to other process parameters and tool parameters

� Actual evaluation / application status:– Endpoint detection demonstrated at contact etch processes– Application for optimization of endpoint signals and clean processes

� Use for real time process control in high volume production is a greatchallenge, due to large number of key numbers and complex interpretation.

Summary

Documents

Optical Emission Spectroscopy for Plasma Processing based ...adp-dresden.de/papers/PlasmaMPCA.pdfThe authors of this presentation would like to thank Siegfried Bernhard Lars Christoph