Plasma etch control by means of physical plasma parameter measurement with HERCULES

1
Plasma etch control by means of physical plasma
parameter measurement with HERCULES
A. Steinbach S. Wurm F. Bell Ch. Koelbl D.
Knobloch D. Köhler
2
Contents

• Introduction - Motivation
• Plasma monitoring tool HERCULES
• Al etch on LAM TCP 9600 SE
• Contact etch on Applied Materials Centura MxP
• Summary

3
Our way of plasma processing today an
effective way ?
Process parameters powerpressureB fieldgas
flow......
Process results etch rateuniformityselectivityp
articles.......
Black Box called plasma processing

• Experience and statistical methods in process
  development
• Process Monitoring and Tool control by test
  wafers

4
Measuring Techniques for real time Plasma
Monitoring
5
Basic HERCULES Model

• High Frequency Electron Resonance Current Low
  Pressure Spectroscopy

6
HERCULES - Principle of measurement

• Passive electrical method,no influence on the
  plasma
• RF current measurement on ground potential at
  the chamber wall
• RF voltage measurement at the matchbox output
• Integral measurement
• Results of FFT and modelelectron collision
  rateelectron densitybulk powerbias voltage

7
Principle and experimental setup
Algorithm

• Passive electrical method,no influence on the
  plasma
• Integral measurement

8
SEERS provides reciprocally averaged parameters
Self Excited Electron Resonance Spectroscopy
9
Automatic data handling
10
TCP Al etch - trend analysis main etch

• Cl2 - MFC failure - Cleans

11
TCP Al etch - trend analysis barrier etch

• Cl2 - MFC failure - Cleans

12
TCP Al etch in Cl2 - first wafer effect

• First wafer effect in main etch

13
TCP Al etch - with / without barrier (TiN,Ti)

• Ti layer detected

14
LAM TCP 9600SE Al etch - electron density and
bulk power vs TCP power
TCP power effects the density and collision rate
of electrons and therefore the plasma impedance
and the power dissipation of the bottom power
(capacitive). Mainly dependent on collision
rate, the bulk power (bottom) decreases by
increasing TCP power (gt250W). This is the reason
for the plateau in the electron density.
15
AMAT MxP chamber - Hook up problems

• In the actual configuration no peak - to - peak
  rf voltage measurement possible? dark space
  thickness declared to be constant rough
  approximation for varying B field? systematic
  failure of electron density, bulk power and
  bias voltage
• Plasma oscillation caused the B field
  circulation? measurement is triggered by B
  field sensor? 1 measurement every 2 s only
  (minimum measurement period 0,2 s)

16
MxP CT etch Plasma parameters dependíng on
process parameters
Change of process chemistry ? strong nonlinear
correlation
17
MxP CT etch - Etch rate BPSG (blanket)
depending on plasma parameters
Obvious correlations between etch rate
and electron collision rateelectron densitybulk
power
18
MxP CT etch - Contact angle depending on
plasma parameters
Change of process chemistry ?no obvious
correlation between electron density and contact
angle
19
MxP Chamber monitoring of contact etch
processes on product wafers
Process mix in Applied Materials Centura MxP
chamber Oxide and Nitride etch with CF4 / CHF3
/ Ar / O2 chemistry Process 1 Process 2 Process
3 Descum --- N2 / O2 --- Step 1 BPSG BPSG Oxide S
tep 2 --- Nitride ---
20
MxP CT etch - Chamber monitoring of product
wafers electron collision rate

• Electron collision rate
• decreases with rf hours
• very sensitive to etch chemistry Pr1
  Pr2 !

One point - one wafer
21
MxP CT etch - Chamber monitoring on product
wafers electron density

• Electron density
• decreases with rf hours slightly
• sensitive to etch chemistry

One point - one wafer
22
MxP CT etch - Chamber monitoring on product
wafers bulk power

• Bulk power
• decreases with rf hours
• very sensitive to power input
• nearly not sensitive to etch chemistry

One point - one wafer
23
MxP CT etch - Chamber monitoring on blanket
BPSG wafers

• Electron collision rate correlates with
  uniformity.
• Electron density and bulk power too

24
MxP CT etch - Chamber monitoring on test
wafers Etch Rate and Uniformity

• Electron collision rate, electron density and
  bulk power correlate with change of etch rate at
  wafer edge, caused by growing polymer at chamber
  wall.

25
MxP CT etch - Chamber monitoring on test
wafers Particle density

• Electron collision rate, electron density and
  bulk power correlate slightly with change of
  particle density, caused by growing polymer at
  chamber wall.

26
MxP Conditioning after wet clean
Wetclean Stable chamber conditions after about
10 wafers.
27
MxP CT etch - short term chamber drift
depending on idle time
Electrical failure counts at Contact etch
Bad chamber
Wafer

• Collision rate shows dependence on chamber idle
  time.
• Constant chamber conditions after about 40 min.
• Change in electron collision rate corresponds to
  change in electrical failure counts.

28
MxP CT etch - short term chamber drift
Electrical failure counts at Contact etch
Wafer

• Electron density and failure counts increase by
  wafer number.

• One of four chambers causes high failure counts.

29
eMxP Arcing detection
Arcing between e - chuck and wafer
30
MxP CT etch - endpoint detection

• endpoint on patternd wafer
• endpoint on blanket wafer
• First test of endpoint detection on patterned CT
  test wafers showed lower sensitivity than AMAT
  HOT Pack.

31
Summary

• Al etch in LAM TCP 9600 SE, oxide and nitride
  etch in Applied Materials Centura MxP have been
  monitored with HERCULES.
• The measured parameters depend significantly on
  chamber conditions and etch results.
• The masured parameters are absolute values.
• No difficult modeling by the user is necessary,
  results are immediate.

32
Applications of the tool

• Development and optimizing processes yes
• Long and short term tool stability yes
• Tool matching yes
• Control of chamber cleaning yes
• Control of power coupling into plasma yes
• Endpoint detection possible
• Layer resolution possible
• Spatial resolution no
• Reduction of test- and monitor wafers yes
• Detection of tool failure yes
• Arcing detection yes