Read OES_MPCA.pdf text version

Supervision of Plasma Processes using Multi-Way Principal Component Analysis

Dirk Knobloch Ferdinand H. Bell Infineon Technologies AG Munich, Germany Jan Zimpel Knut Voigtländer Fraunhofer IVI Dresden

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

Page 1

Outline

Introduction - Supervision of plasma processes Experimental - Features of OES spectrometer Advanced data processing methods

­ Standard methods (mean value, PCA, etc.) ­ Multi-Way Principle Component Analysis (MPCA)

Multi Way PCA for contact hole etch

­ Interpretation of MPCA scores for process and tool phenomena

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

­ Monitoring of thermal and chemical chamber condition ­ Analysis of OES- Endpoint Signals

Conclusion

Page 2

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

Introduction

Equipment Data

int. & ext. sensors Data collection in high volume production creates large amount of data Data compression for key number extraction is essential

data processing

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

Information

relevant key numbers per wafer, recipe, tool

actions

Improvement

equipment and process stability, productivity, quality

Page 3

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

Introduction

Equipment Data

int. & ext. sensors Data collection in high volume production creates large amount of data Data compression for key number extraction is essential

data processing

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

Focus: Optical Emission Spectroscopy

Information

relevant key numbers per wafer, recipe, tool

actions

Improvement

equipment and process stability, productivity, quality

Page 4

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

Features of Hamamatsu Spectrometer MPM

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 scripts are available

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

digital/analog ports for connection to etch tool All experiments conducted on an AMAT MxP+ etch chamber (200mm)

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

Page 5

Optical spectra visualized as a ,,Data cube"

3 dimensional data cube containing OES spectra (oxide etch) Vertical and horizontal cut through data cube

etch time t

wavelength []

wafer 1 ... N wavelength

time [s]

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

Challenge: to obtain methods for extraction of key-numbers for longterm process monitoring, process diagnostics and fault detection

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

Page 6

Advanced data processing methods

Standard: univariate key number extraction 1. Simple key-numbers

­ standard deviation ­ time duration of process steps ­ mean value

multivariate key number extraction

2. Extraction of key-numbers using signal decomposition

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

2. Extraction of key-numbers using signal decomposition

­ application of Tschebyscheff functions ­ adjusted signal base decomposition (PCA)

­ Multi-Way PCA (MPCA)

3. Adaption for a nonlinear parametric signal model

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

Page 7

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

...

...

...

Multi-Way Principle Component Analysis (MPCA)

PCA:

etch time t

M(i, n) u ( ) i i

M(i, j,n) u ( )v (t) i j j i

for n = 1,..., N

MPCA:

wavelength wafer 1 ... N

"standard PCA": use of one mean spectrum per wafer

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

­ wavelength but no time dependency per wafer

Multi Way PCA: Calculation of orthogonal wave pattern ui and orthogonal base time signals vi by unfolding the original data cube in time and wave direction

­ wavelength and time dependency per wafer

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

Page 8

Multi-Way Principle Component Analysis (MPCA)

1st step: PCA of all time signals

­ calculation of the time signal pattern vj ­ in the calculation included 1024 (number of channels) x N* time traces ­ result: 1024 x N time signal patterns vj

*N = number of wafers

·0.122 ·0.1215

·0.121

Vx

·60 ·110 ·160

PCA

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

·0.1205

·0.12 ·10 ·0.25 ·0.2

...

wafer 1 ... N Different time traces

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

·0.15 ·0.1 ·0.05 ·0 ·-0.05 ·-0.1 ·-0.15 ·10

Vy

·60

·110

·160

Page 9

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

...

Multi-Way Principle Component Analysis (MPCA)

2nd step: PCA of all wavelength spectra

­ calculation of the wavelength pattern ui ­ in the calculation included Y (=number of spectra for each wafer) x N ­ result: Y x N wavelength patterns ui

Different spectra

wafer 1 ... N

·0.15

·0.2 ·0.15 ·0.1

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

...

PCA

·0.1

Ux

Uy

...

·0.05

·0

...

·405 ·505 ·605 ·705

·0.05 ·0 ·-0.05

...

·405 ·505 ·605 ·705

·-0.05

·-0.1 ·-0.15

·-0.1 ·305 ·305

Page 10

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

Multi-Way Principle Component Analysis (MPCA)

3rd step: calculation of the scores M(ui,vj,n), n=1...N

­ orthogonal patterns for time signal (vj) and wavelength (ui) ­ apply patterns on raw data of the optical emission spectra for each wafer ­ results in the scores M(ui,vj,n) ­ reconstruction of the raw data of optical emission spectra:

M(i, j,n) u ( )v (t) i j j i

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

for n = 1,..., N

­ number of useable patterns depends on information content of process data

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

Page 11

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

Example: Multi Way PCA for contact hole etch

intensity [a.u.]

·0.122 ·0.25 ·0.2 ·0.15 ·0.1 ·0.05 ·0 ·-0.05 ·-0.1 ·-0.15 ·10

extracted time signals vj

·0.1215

V2

·0.121

·0.1205

V1

·60 ·110 ·160

extracted wave pattern ui

·0.12 ·10

·60

·110

·160

time [s] scores of key number uivj [a.u.]

resulting MPCA scores vs. wafer

·0.15

·0.1

U2

·0.05

intensity [a.u.]

·0

·-0.05

·-0.1 ·305 ·405 ·505 ·605 ·705

M(U2,V1)

M(U2,V2) M(U5,V2)

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

·0.2 ·0.15 ·0.1 ·0.05 ·0 ·-0.05 ·-0.1 ·-0.15 ·305 ·405 ·505 ·605 ·705

U5

M(U5,V1)

wavelength [nm]

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

wafer no.

Page 12

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

MPCA scores - no process mix

scores of 1st and 2nd order of wet clean cycles (1..4) [time signal of 1st order]

6

3.4

x 10

4

wet clean 1 ·WC 2 wet clean 2 wet clean 3 wet clean 4

·WC 3 ·WC 4 ·WC 5

4

3.2

WC 2 WC 3 WC 4 WC 5

scores [a.u.]

2

3

0

2.8

2.6

-2

2.4

-4

2.2

match box change 1st order

·200 ·400 ·600 ·800 ·1000

-6

2nd order

-8 0 100 200 300 400 500 600 700 800 900

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

wafer no. Score correlations:

0

wafer no.

·1st order score: - match box change - decrease of scores through decreasing light transmission caused by polymer coating during wet clean ·2nd order score: - increase of scores caused by interaction between chamber wall and plasma

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

Page 13

MPCA scores - process mix

two products with high (product 1) and low (product 2) polymerizing gas chemistries scores of 2nd order [ time signal of 1st order] during 1 wet clean cycle scores of 2nd orders vs. rf hours

product 1 scores [a.u.]

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

scores of product 1 decrease significant more than scores of product 2 caused by different behaviour of chamber polymerization results in varying duration of wet clean cycle

product 2

one point ­ one wafer

rf hours [h]

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

scores are used to determine optimum wet clean cycles as a function of process mix

Page 14

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

MPCA scores - thermal and chemical variations caused by process mix

scores of 4th order [ time signal of 1st order] during 1 wet clean cycle

Process with different gas chemistries

different conditioning effects caused by product mix

­ (1) typical behavior for 1st wafer effect ­ (2) inverse 1st wafer effect caused by deconditioning of the chamber through different pre-process (3) scores are used to characterize memory effects

2

scores [a.u.]

3

1 3

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

wafer no.

Page 15

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

MPCA scores - correlation with variations in endpoint behavoiur

scores of 5th order [ time signal of 6th order] scores vs. wafer optical endpoint signal vs. etch time

·x 10 ·400 ·1.05 ·1.04

·4

·200

optical intensity [a.u.]

·1.03

scores [a.u.]

·0

·1.02

·-200

·1.01

·-400

·1

·0.99

·-600

·0.98

·-800

·-1000

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

one point ­ one wafer

·20

·0.97

·0.96

one curve ­ one wafer

·10 ·20 ·30 ·40 ·50 ·60

wafer no.

·40

·60

·80

·100

·120

etch time [s]

endpoint signal depends on previous process step variations (e.g. CMP, CVD, etc.) scores characterize variatons in endpoint signal behaviour scores are used to determine stability of previous process steps

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

Page 16

Summary of process and tool phenomena as characterized using Multi Way PCA

different mask and etch structures memory effects from previous recipes

influences from previous process steps

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

score [a.u.]

equipment influences

chamber condition

chamber cleanliness wafer no.

endpoint variations

Page 17

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

Conclusion

The supervision of plasma processes using Multi-Way Principal Component Analysis (MPCA) has been demonstrated for an oxide contact hole etch process The MPCA decomposition technique pinpoints significant process and equipment key numbers The MPCA analyses allows process and equipment monitoring, such as characterization of

Dirk Knobloch Ferdinand H. Bell Infineon Technologies Jan Zimpel Knut Voigtländer Fraunhofer Institut IVI Dresden

­ first wafer effect, chamber cleanliness, process mix

The MPCA is powerful in qualitative analysis of endpoint traces

­ robust endpoint analysis of endpoint traces as a function of previous processes

Page 18

IVC -15 Int'l Symposium Oct 02-06, 2000, Boston, MA

Supervision of Plasma Processes using Multi-Way Principal Component Analaysis

Information

18 pages

Find more like this

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate

949772