Plasma diagnostic in an inductively coupled plasma using chlorine

Workshop on Self Excited Electron Plasma Resonance Spectroscopy8. - 9. June 1999 at Infineon Technologies Dresden - 1 - P. Höhmann

Infineon Regensburg

Plasma diagnostic at LAM TCP 9600SETM

H. Steinmetz, J. Strobl, N. Rohn and T. Werner, Lam Research GmbH

M. Klick, W. Rehak, M. Kammeyer, and D. Suchland, Adolf-Slaby-Institute

S. Wurm, W. Preis and Ch. Koelbl, Infineon Technologies

Plasma diagnostic in an inductively coupled plasma using chlorine chemistry


Infineon Regensburg

Plasma diagnostics for production toolsComparison of complementary methods

Optical emission Self excited electron VI - probespectroscopy (OES) resonance spectroscopy (rf voltage and power)

(SEERS), Hercules______________________________________________________________________________________________________

Parameters - relative values - absolute values - absolute and relative values- line averaged - volume averaged - no well define averaging

- intensities of emission lines - plasma bulk power - peak voltage- identification of species - plasma density - real power in chamber- characterization of excitation - electron collision rate (not in plasma)

Measurement - optical - electrical, high frequency - electrical, radio frequency- passive - passive - passive- non-intrusive - non-intrusive - non-intrusive

Deposition,contamination - critical - no influence - no influence

Sensitivity - high - high - low/medium

Endpoint - yes - yes - (strongly) limited

Base lining, - strongy limeted - yes - limitedtool matching


Plasma diagnostics for production tools


Infineon Regensburg


Oil

13.56 million rps(generator)

Sheath(nonlinear part)

Sheath(linear part

- retarding electric field)

Inert massof plasmaelectrons

Collisionswith neutrals

Sinusoidal oscillation(simple sheath model)

Nonsinusoidal oscillation(realistic sheath model)

Mechanical analogon of nonlinear effect used for SEERS


Infineon Regensburg


rf current

HERCULES

Coaxial sensor and cable

TCP coil

Process data bank

SEERS

Fast ADC500 MHz, 1 GS/s

50 Ohms input

dielectricwindow

chamber

bottom power

peak voltage

top power

plasma

Experimental setup


Infineon Regensburg


TM®HERCULES - Sensor for LAM TCP 9600SE


Infineon Regensburg


TM®

Figure 1 shows window plate of the TCP-System in section

Detail figure: mounted sensor in the He leakage flange (principle)

He leakage flange

Contact spring or threadA

HERCULES - Sensor for LAM TCP 9600SE


Infineon Regensburg


0 0.5 1 1.5 2 2.5process time [min]

0

2.109

4.109

6.109

8.109

elec

tron

dens

ity [c

m-3

]

15 mTorr15 mTorr10 mTorr10 mTorrfirst wafer (cold machine)first wafer (cold machine)13 mTorr13 mTorr

, ,

The relative changes betweenprocessing a production waferon a warm and a cold machine as well as between running process at different total pressure are much more evident.

The overall curve shape does depend strongly on total pressure and on tool condition.

Pressure variation


Infineon Regensburg


0 100 200 300 400process time [s]

0

109

2.109

3.109

4.109

5.109

6.109

7.109

8.109

elec

tron

dens

ity [1

/cm

3 ], co

llisio

n ra

te [1

/s

60

90

120

150

peak

vol

tage

[V]

peak voltagecollision rateelectron density(filtered)

Process signaturesof the first three wafers

Al etching - first wafer effect


Infineon Regensburg


100 200 300 400 500 6002

46810

20

406080100

bulk

pow

er P

BP [m

W c

m-2

]

100 W100 W150 W150 W200 W200 W

100 200 300 400 500 600TCP - power PTCP [ W ]

109

2

468

1010

elec

tron

dens

ity n

e [cm

-3]

TCP power effects thedensity and collision rate ofelectrons and therefore theplasma impedance and thepower dissipation of thebottom power (capacitive).

Mainly dependent oncollision rate, the bulk power(bottom) decreases forincreasing TCP power(>250W). This is the reasonfor the plateau in the electrondensity.

Electron density and bulk power vs. TCP power


Infineon Regensburg


Product wafer - resist mask on Al (appr. 50%)

0 40 80 120 160process time [s]

3.109

4.109

5.109

6.109

7.109

8.109

elec

tron

dens

ity [1

/cm

3 ]

first waferfirst wafer

third waferthird wafer

secondsecond

main etchmain etch

The behavior of the mainetch for the first wafer ofa lot is quite different fromthe following ones.

Al etching in Cl2 - first wafer effect


Infineon Regensburg


Fault: no resist, faultless dashed (reference wafer)

20 40 60 80 100 120 140 160process time [s]

468

10

20

406080

100

200

peak

vol

tage

[V]

106

5

107

5

108

5

109

colli

sion

rate

[1/s]

100

500

1000

5000

10000

optic

al e

miss

ion

(EP)

The break-throughis not influenced(here separate step).

In case of the main etch thecollision rate decreases byone order of magnitude.

Wafer fault analysis


Infineon Regensburg


each curve averaged from five test wafers

0 20 40 60 80 100 120process time [s]

0

107

2.107

3.107

4.107

5.107

colli

sion

rate

[1/s]

with TiN (100 nm), Ti (15 nm)with TiN (100 nm), Ti (15 nm)

Ti layerTi layer

800 nm AlSiCu800 nm AlSiCu

break through (Al2O3)break through (Al2O3)

SiO2SiO2

Increase of the collision rate due tothe larger cross section of Ti (red curve).

TiN layer not visible.

Al etching-with/without barrier (TiN,Ti)


Infineon Regensburg


0 7 14 21 28 35 42Wafer No.

1.2.107

1.5.107

1.8.107

2.1.107

2.4.107

2.7.107

3.107

colli

sion

rate

[1/s] resist wafer (*0.7)resist wafer (*0.7)

Si wafer (*2)Si wafer (*2)product waferproduct wafer

First Si wafer (blue curve) after chambercleaning behaves different.

Conditioning is finished after processing10 resist wafers.

Deconditioning due to processing bare Siwafers.

Charactrization and calibration of machinestatus possible by measuring andcomparing absolute parameter values.

Characterization of chamber cleaning


Infineon Regensburg


one lot one point

After Cl2-MFC adjust-ments/substitution, thecollision rate returnedback to normal signallevels.

Shows the dramatic effectof the main clean on thecollision rate.

The endpoint level wasadjusted during cleaningprocedures.3640 3710 3780 3850 3920 3990 4060

Lot No.

3.2.109

4.109

4.8.109

5.6.109

6.4.109

7.2.109

8.109

elec

tron

dens

ity [1

/cm

3 ]

5.106

107

1.5.107

2.107

2.5.107

3.107

3.5.107

colls

ion

rate

[1/s]

optic

al em

issio

n (E

P) *

3000

40

60

80

100

120

140

etch

tim

e [s] quick cleanquick clean

Cl2-MFC drift/errorCl2-MFC drift/error

Cl2-MFC errorCl2-MFC error

main cleanmain clean

Al etching - trend analysis main etch


Infineon Regensburg


one lot one point

Trend analysis of amemory productcovering 77 lots.

Shows the effect of thequick clean on thecollision rate.

The collision rate isthe most sensitiveparameter, in particularforecasting MFC error.3640 3710 3780 3850 3920 3990 4060

Lot No.

3.2.109

4.109

4.8.109

5.6.109

6.4.109

7.2.109

8.109

elec

tron

dens

ity [1

/cm

3 ]

6.106

9.106

1.2.107

1.5.107

1.8.107

2.1.107

2.4.107

colls

ion

rate

[1/s]

optic

al e

miss

ion

(EP)

*50

0

0

10

20

30

40

etch

tim

e [s

]

Quick cleanQuick clean

Cl2-MFC drift/error Cl2-MFC drift/error Cl2-MFC errorCl2-MFC error


Al etching - trend analysis barrier etch


Infineon Regensburg


main etch (Al) and barrier etch (TiN, Ti)

One point represents one lot - stepwise analysis.

Shows the differentdependence of processsteps on events,e.g., cleaning procedures.

3640 3710 3780 3850 3920 3990 4060Lot No.

5.106

107

1.5.107

2.107

2.5.107

3.107

colli

sion

rate

[1/s]

main etch stepmain etch step

Cl2-MFC errorCl2-MFC error


barrier etch stepbarrier etch step

quick cleanquick clean

Al etching - trend analysis collision rate


Infineon Regensburg


K ey item s R esu lts

M onito r / contro l power coup ling in to the p lasm a? Y es

Too l m atch ing? Y es

Long term too l stab ility? Y es

D evelop ing / optim izing processes? Y es

C ontro lling cham ber c lean ing procedures? Y es

R eduction of test- and m onitor wafers? Y es

Layer reso lu tion? Y esdem onstra ted fo rTi,

S iO 2, A R C and photo res is t

S patia l reso lution? N o

D etection of too l fa ilu re? Y es(e .g . C l2-M FC fa ilu re)

Summary


Infineon Regensburg


increasing - overall equipment effectiveness- reliability and uptime- statistical analysis- MTBF- yield- throughput

decreasing - cost of ownership (equipment)- MTBR- time for equipment base lining- time for process development

TM

®Benefit from outfitting LAM TCP with

HERCULES /PL

Documents

Plasma diagnostic in an inductively coupled plasma using chlorine