Eric Lifshin, Di Wu and Bob Geer

1
Task VI Modeling and Metrology
X-ray/E-beam Spectral Metrology Tools High
Spatial Resolution Compositional Imaging
Eric Lifshin, Di Wu and Bob Geer University at
Albany State University of New York Albany, NY
12222
2
Metrology Priority SUPPORT DRIVER DEMONSTRATION

• Goal
• Use electron beam X-ray microanalysis and
  scanning electron microscopy to characterize the
  composition, shape and dimensions of films,
  particles and fine structures to meet the
  challenge of future interconnect requirements

• Reduce current state of art 1000nm resolution to
  100nm or less
• for cross sectional analysis of thick samples!

• Metrology Innovation Requirements
• Small diameter intense electron beam at low
  voltage to minimize
• electron generation volume.
• Generate and detect statistically significant
  X-rays signal. Detector must have sufficient
  energy resolution to resolve closely spaced peaks

3

• There is a need to more fully
• characterize current and future
• devices and nanostructures
• ULSI (130 nm node and below)
• metal films
• dielectrics
• contaminants defects
• Photonic devices e.g.
• quantum wells and dots
• 3D interconnects

Electrons
X-rays
L. Larson in Characterization and Metrology for
ULSI Technology AIP 1998
4
Approach Combine Latest Schottky Electron
Sources, New X-ray Detector Technology and
Advanced Modeling
5
Experimentally Demonstrated the Value of the High
Energy Resolution of a Microcalorimeter X-ray
Detector in the Characterization of TaSiN Films

• Ta peaks still resolvable for 3 nm layer

Note Measurements were done with a NIST
microcalorimeter. SUNY microcalorimeter now
detects X-ray, but is undergoing further
optimization at the supplier.
6

• Z-axis Compositional Imaging mCal EDS
• Clearly resolves Ta M and Si K peaks
• Demonstrates sensitivity
• Resolves Ta signal from 3.5 nm film
• Demonstrates sufficient energy resolution for
  compositional metrology
• Probe depth exceeds sample thickness at 30 nm

TaSiN
TaSiN
TaSiN
TaSiN
Si
Si
Si
Si
7
Application of Parallel Beam WDS to AlGaAs/GaAs
Stack
5 KeV
8

• By lowering the electron beam voltage it appear
• that 100 nm resolution is possible!
• How low can you go?
• Modeling can provide both an answer to that
• question and criteria for choosing experimental
• conditions.

9

• The Ultimate Resolution Possible in the
  examination
• of thick samples viewed in cross-section depends
  on
• Sample Related Factors
• - shape, elements, concentration levels
• Instrumental Parameters
• - source, stability, spectrometer, KV, i
• Quality of the Results Needed
• - Accuracy (Correction Procedures, Input
  Parameters)
• - Precision (statistics, drift, repositioning)

10
Higher Spatial Resolution in Thick Samples
Requires Reducing the Beam Voltage
11
Monte Carlo Simulation Pure Al 1 nA 3 mm EDS
Detector 8 cm from sample 100 seconds
12
?-calorimeter EDS and Parallel Beam WDS
Conventional EDS and WDS
Low energy resolution of conventional
EDS (FWHM130eV at 6keV) and low X-ray
collection efficiency of conventional WDS make
them less suitable for low energy and high
spatial resolution electron beam micranalysis
13
Precision of Composition
Where
peak counts on sample
background counts on sample
peak counts on the standard
background counts on the standard
C the composition
a the parameter in the Ziebold-Ogilvie
Eqn.

14
Example of Calculation
15
95 Confidence Level for WDS Measurement of Al Ka
in AlGaAs
16
Relationship Between BeamVoltage, Diameter and
Current
17
Probe Current vs. Diameter
18
Monte Carlo Simulation of Electron Trajectories
Shows Improved Spatial Resolution for Aluminum
Analysis In AlGaAs Possible with Low KV High
Brightness Source
19
Comparison of 95 Confidence Limits for Different
X-ray Detectors
Current x time (nAsec)
20
Monte Carlo Simulation for Cross Sectional
Analysis of Banded Structure
AlGaAs film thickness (nm)
1.0E-04
0
100
200
300
400
500
600
1.0E-05
Al Ka
AlGaAs Band
X-Ray intensity (photon by electron by steradian)
Probe diameter d
1.0E-06
d 34 nm at 10 KV d 66 nm at 5 KV
Excitation Volume
1.0E-07
21
Comparison of 95 Confidence Interval is also
Possible When Peaks are not Resolved by using
Overlap Coefficients
Si
Si
TaSi2
Ta
22
Precision of Ta composition in TaSi2
23
Summary

• Monte Carlo modeling provides a way to predict
  both precision and spatial resolution for
  currently available scanning electron microscopes
  and spectrometer systems when used to examine
  nanostructures.
• Spatial resolution of 100 nm should be possible
  in the examination of thick samples providing
  stringent analytical conditions are met.
• Low voltage operation dramatically reduces
  precision particularly for situations where
  spectral overlap can occur.

24

• In the absence of peak overlaps EDS may provide
  the best precision. Accuracy may depend on the
  quality of the background subtraction process as
  well as the correction procedure.
• There is hope even in situations where serious
  overlaps are present providing the peak shapes do
  not vary between samples and standards. Having
  both high energy resolution combined with high
  count rate is still the most desired experimental
  situation.
• Counting statistics may not be the performance
  limiting factor if beam and stage drift, specimen
  damage and contamination effects are not
  minimized.

25
Acknowledgements

• Serge Oktyabrsky University at Albany
• Raynald Gauvin McGill
  University
• David Frey LEO Intruments
• Jeff Streger
• David Rhohde Thermo NORAN
• Audry Dow