Molecular Dynamics Simulations of Plasma-Surface Interactions and Etching

1
Molecular Dynamics Simulations of Plasma-Surface
Interactions and Etching

• David Graves and David Humbird
• University of California at Berkeley
• FLCC Research Seminar
• February 23, 2004

• acknowledgements
• Gottlieb Oehrlein and group at U. Maryland
• Cam Abrams
• Harold Winters, John Coburn, Dave Fraser

2
Broader Issues in Plasma-Surface Interactions

• 1. Modifications of surfaces by plasma exposure a
  highly advanced empirical art for example...
• a.) etched structures controlled to within
  several nm over 300 mm wafer diameter
• b.) plasmas used to extend optical lithography
  (e.g. resist trim and spacer etch)
• 2. Models at plasma and feature scales require
  surface models
• 3. Complexity of plasma-surface processes and
  difficulties of direct observation (like all
  surface processes) challenges development of
  physically-based models

3
What are Key Mechanisms in Plasma-Surface
Interactions?
plasma creates positive ions and
neutral radicals, both of which hit surface
plasma
neutral species sputtered and desorbed
neutral radical at Tgas
positive ion impact at Vsheath
Near-surface region altered remains near Tgas
zone of energy release
4
Basic Challenge 1 Model Two Very Different Types
of Species
Ion bombardment and radical impact - ions impact
surface with 10s - 1000s eV - energy rapidly
released profound surface effects - ions have
chemical as well as physical effects - radicals
may adsorb/react at room temperature with no
barrier - coupled ion/radical effects at
surfaces thought to explain most observed
effects - experimental evidence for weakly bound
neutral surface species diffusion, reaction,
desorption dynamics
5
Basic Challenge 2 Model Effects with Sufficient
Accuracy
Major goals are to use model to a.
interpret/understand experiments - elucidate
mechanisms of etching b. develop surface rate
expressions that can be used in feature and
tool scale simulations
6
Basic Challenge 3 Model Processes with Range of
Time Scales
For example,
1. Ion bombardment results in collision cascade
that dissipates to heat in less than 10-12
s. 2. Radical diffusion/reaction/desorption may
take 10-12 - 10-3 s. 3. Experiments/processes
conducted for 102 s.
7
Basic Challenge 4 Model Surface Processes in
Which Surface Always Changes
The Plasma Alters the Surface!
The crucially important consequence simulations
must include enough impacts that the surface
achieves steady state composition and
structure. Methods that rely directly on
ab-initio electronic structure
calculations will be too slow for practical
purposes. Empirical potentials allow
realistic fluences, but are they accurate enough?
8
How to Model/Simulate Plasma-Surface Interactions?
Molecular dynamics simulations - classical,
semi-empirical potentials - resolves
vibrational timescales O(10-15 s)
1. Ion impact - crucially important energy input
10-13-10-12 s collision cascade - MD time
and length scales match physics of
interactions - weakly bound species after
collision cascade removed simple TST for
thermal desorption with Eb ? 0.8 eV. 2.
Radical-surface chemistry - accuracy of
interatomic potentials?? (cf. ab-initio) - time
and length scales adequate?? (cf. experiment)
9
Molecular Dynamics (MD) Simulation
Interatomic Potential
Interatomic Forces
10
MD Cell and Assumptions for Etch Simulation
Top exposed to ion neutral flux impact
location chosen randomly
Impact events followed for 1 ps excess energy
removed statistics collected new impact point
chosen repeat sequence 103 times for steady
state surface.
Surface composition and structure must reach
steady state.
Lateral boundaries periodic mimics semi-infinite
surface
Bottom boundary fixed new Si added here
2 nm
11
Mimic Experimental Time Scales by Accumulating
Effects of Many 10-12 s Impacts
Simulation strategy accumulate effects of
repeated impacts ignore time between impacts
except to remove weakly bound species
12
How to Simulate Radical-Surface Interactions?
Simplest problem simulate spontaneous etching
of un-doped room temperature silicon by F
impacting at room temperature. Previous studies
using Stillinger-Weber and related potentials
failed to predict ANY Si etching from F impact at
0.03 eV ( 300 K)!1 Recent results using
modified Brenner/Tersoff-style potential for Si-F
are more encouraging. 2
1 F.H. Stillinger and T.A. Weber, Phys. Rev.
Lett., 62, 2144, (1989) P.C. Wiekliem, C.J. Wu,
and E.A. Carter, Phys. Rev. Lett., 69, 200,
(1992) T. A. Schoolcraft and B. J. Garrison, J.
Am. Chem. Soc. 113, 8221, (1991). 2 C.F. Abrams
and D.B. Graves, J. Appl. Phys., 86, 5938,
(1999) D. Humbird and D.B. Graves, J. Chem.
Phys., in print, (2004).
13
Stillinger-Weber/ Carter Si-F Potentials
Spurious Barriers?

• Accurate representation of thermal Si-F chemistry
  is needed
• Stillinger-Weber/Carter Si-F potentials do not
  predict spontaneous etching
• Si-F potential of Cam Abrams does, but with
  product distribution in disagreement with
  experiment

H.F. Winters and J.W. Coburn, Surf. Sci. Rep.,
14 (4-6) 161-269, (1992)
14
Improving the Si-F potential
(Abrams)

• Added a correction function to Abrams Si-F that
  accounts for changes in bond energy as Si becomes
  more fluorinated
• Parameterized against DFT

(Humbird)
S. Walch, Surf. Sci. , 496, 271, (2002)
15
More Results Si-F Spontaneous Etch

• Results at 300 K are reasonable wrt reaction
  probability and product distribution
• Product shift at higher surface temperature
  matches experiment rate does not
• Etch kinetics above 400K dominated by
  spontaneous decomposition
• Simulation misses long-time scale events
  KMC/TST needed?

16
Spontaneous Etch Reaction Probabilities (Si
atoms etched per incident F atom)
Author Value Flamm et al.a 0.00672 Ninomiya et
al.b 0.025 Vasile and Steviec 0.064 H. F.
Wintersd 0.003250.0075 This work 0.03 a D. L.
Flamm, V. M. Donnelly, and J. A. Mucha, J. Appl.
Phys. 52, 3633 (1981). b K. Ninomiya, K. Suzuki,
S. Nishimatsu and O. Okada, J. Appl. Phys. 58,
1177 (1985). c M. J. Vasile and F. A. Stevie, J.
Appl. Phys. 52, 3799 (1982). d H. F. Winters,
private communication (2003).
Note Evidence that measured F coverage (7-10 ML
or more) may be due to roughness. This texture
question likely to become more important in
future as features/films become smaller. One
current issue is LER.
17
Ion-Assisted Etching Comparison to Experiment
200 eV Ar/Si F
18
Ion-Assisted Etching Weakly Bound Products
200 eV Ar/Si F
19
Fluorocarbon Plasma Etching of Si

• Important issue depositing species play role in
  selectivity and CD control
• FC plasmas readily etch SiOx Si and other
  materials etch more slowly
• Model case for studying mechanisms of etch
  selectivity
• Popular chemistry F-deficient (e.g. C4F8 C4F6
  C5F8, etc.) heavily diluted in Ar
• Model chemistry xCF2 yF Ar (20 eV and 200
  eV)
• Potentials Si-C-F (with recent Si-F revisions)

C.F. Abrams and D.B. Graves, J. Appl. Phys.,
86, 5938, (1999) J. Tanaka, C.F. Abrams and
D.B. Graves, JVST A 18(3), 938 , (2000)
20
Thermal CF2 / Ar 9/1 (Si Impacts)
Surface C, F, Si etch (ML) vs. CF2 Fluence
Si etch yield (Si/ion) vs. CF2 Fluence
Ar 20 eV
Uptake / Etch (ML)
(Steady Deposition)
CF2 Fluence (1015 cm-2)
Ar 200 eV
Uptake / Etch (ML)
Etch Yield Si/Ion
CF2 Fluence (1015 cm-2)
CF2 Fluence (1015 cm-2)
21
Thermal F CF2 / 200 eV Ar
Surface C, F, Si etch (ML) vs. CF2 Fluence
Si etch yield (Si/ion) vs. CF2 Fluence
CF2/F/Ar
Uptake / Etch (ML)
Etch Yield Si/Ion
8/1/1 (10 F)
CF2 Fluence (1015 cm-2)
CF2 Fluence (1015 cm-2)
Uptake / Etch (ML)
Etch Yield Si/Ion
7/2/1 (20 F)
CF2 Fluence (1015 cm-2)
CF2 Fluence (1015 cm-2)
22
Simulation Experiment Agreement
C(1s) XPS
Si(2p) XPS
Experiment (C4F8)
Simulated
Experiment
Simulated Ar/CF2
C4F8 / 90 Ar

• Increasing self-bias forms SiFx bonds
• Low energy passive deposition
• High energy CFx bonds reduced, Si-C, C-C, Si-F
  form Si etch observed

Measurements courtesy G.S. Oehrlein et al.
23
Stratified Layers
Close inspection reveals superficially
fluorinated Si-C network.
24
Silicon Transport and Surface Loss Mechanism
(as deduced from simulation results after steady
state reached)
Subsurface F attacks Si
Si joins Si-C mixing layer
Si appears on surface
Si stripped of F
Mixing
Mixing
Etching
F recycle
Deep mixing required
Shallow mixing required
25
Single-Impact Movies(available from website)
Deep impact From below Si-C layer
Shallow impact with product From overhead
Green Argon Blue Silicon Yellow Carbon White
Fluorine Red Fluorines of interest
0.6 ps duration
0.5 ps duration
26
Movies of Si Layer Etch F and Si Evolution
(available from website)

• Elapsed time 10 s (per mA/cm2 ion flux)
• 7 CF2, 2 F (each at 0.03 eV) per Ar (at 200 eV)
• Simultaneous images left F (red, green) right
  Si (blue)
• Note the front advance down into silicon

27
Ion Energy Deposition Through SiC Layer
top view
side view
0 F
10 F
20 F
28
Observations About FC/Ar Etching of Si
1. Remarkable layer segregation induced by strong
Ar bombardment. - SiCx and SiFx layers seen in
TEM images? (ASET) - leading front of SiF,
followed by SiC, then top layer F 2. Under
conditions examined (e.g. low Gn/G), SiC layer
plays key role in reduction of silicon etch
rate rather than CFx layer. 3. F adsorbed at
surface is transported to Si layer by ion
bombardment. - form of incident F less relevant
than total adsorbed C/F ratio - simulation shows
that F is most effective in thinning SiC
layer and in creating dynamic porosity in SiC
layer 4. Ion energy remaining at SiFx layer
appears to correlate to overall yield. 5. Etch
mechanism driven by ion mixing and assumes all
neutral species adsorb initially into strongly
bound states. 6. Recent results hint that
surface roughness could play role in FC
observed experimentally.
29
Concluding Remarks

• 1. Encouraging evidence that empirical potentials
  with parameters fit to DFT cluster calculations
  can capture spontaneous Si etching via F.
• - may be due to strongly exothermic,
  barrier-less process and prompt reactions
• - suggestion that etch at Tgt 400K requires
  KMC/TST
• 2. Simulations of Ar with FC radicals on Si
  shows general agreement with measurements and
  explains complex process.
• 3. Current simulation assumes that all processes
  driven by ion mixing and that thermal neutrals
  first chemisorb at surface. Ignores weakly bound
  reactants diffusing into sub-surface are these
  important?
• 4. Weakly bound etch products commonly observed
  (in agreement with experiment).
• 5. Field requires closer coupling between
  beam/plasma experiments and simulations to test
  assumptions and extend interpretations.