1
Background
• In Situ Process Control Is Critical for theFuture Factory.
• Wafer State Monitors (CD, Film Thicknessand Profile) Are Important for the AdvancedSemiconductor Process Control.
• Physical Metrology Is Challenged by theShrinking Device Dimensions.
2
Overview• Specularly Reflected Light Measurements
(Ellipsometry, Reflectometry) Have ProvenAccuracy in Monitoring Blanket Thin FilmThicknesses.– Limited applications to patterned wafers
• Non-specular Scattering (Diffraction)Measurements Can Yield Detailed TopographyInformation From Patterned Wafers.– Scatterometry , Fourier Imaging– Both require multiple-angle view of wafer (hard to
implement for in situ monitoring)
3
Goals of Research
• Quantitative Analysis of Ellipsometry DataFrom Patterned Wafers Extract PatternTopography Information (Especially Depth)– Process control using product wafers
• Reduction/elimination of test wafers• Reduced offline metrology
– Increased knowledge of processes at real timelevel
4
Ellipsometry
Eip Erp
Eis Ers
Polarizer Analyser
LightSource Detector
Sample
)exp()tan(//
∆⋅=== iEEEE
RR
isrs
iprp
s
p ψρ
• Tan(Ψ ) and Cos(∆) are Measured by Ellipsometer.— Functions of λ
5
Scope of SE Applications
λ
4Smooth Surface (d ~ 0)– Thin film characterization d: feature size of uneveness
4Rough Surface (d << λ)– Surface/interface characterization
s Patterned Surface (d ~ λ)– Pattern profile characterization
This Work
6
Models For This Work
• Three Approaches for ModelingSpecular Reflection FromPatterned Structures– Scalar approach of Heimann
– Surface integral equation (SIE)
– Rigorous coupled-wave analysis(RCWA)In
crea
sing
Acc
urac
y an
dC
ompu
tatio
n Ti
me
SolvingMaxwell’s EquBoundary ValueProblem
7
Sample: Si Relief Grating
•Use Simple Structure to Minimize the Complexity in Initial Efforts•The Lineshapes can be Modeled as Trapezoidal.
— Described by 4 parameters : period, top linewidth, depth, wall angle•We Measure the Structure from SEM as: period=3.96 µm, top linewidth =2.2 µm, depth = 0.52 µm, and wall angle = 73.9º.
8
Configurations of SE Measurement
75º
•Align The Grooves Normal to the Plane of Incidence•Two Kinds of Measurements Conducted— Near normal (6º) incidence— Oblique (75º) incidence
9
region 1 region 2
2pR2sR,
1pR1sR,
Description of the Scalar Model
∆Ψ==ρ
+=+=
je)tan( RR
.Raf .Raf R .Raf .Raf R
s
p
s2s1sp2p1p 2121
af1 , af2 : Area Fractions Rp , Rs : Reflectances in pp and ss polarizations
• Nominally, af1 + af2 = 1. But, af1 and af2 may be free parameters to consider unmodeled effects, e.g., Reflection from side walls. • To obtain the reflectances, we add the corresponding complex fields, not their intensities.
10
4000 5000 6000 7000 80000
0.05
0.1
0.15
0.2
0.25
λ (Å)
|Rp|2|Rp|2
4000 5000 6000 7000 80000
0.05
0.1
0.15
0.2
0.25
|Rs|2|Rs|2
λ (Å)
Scalar Model Applied to the SiRelief Grating
• Region 1: 57.83% (4797.1 Å Si / Si )• Region 2: 25.11% (Si )
• Region 1: 57.83% (4797.1 Å Si / Si )• Region 2: 25.11% (Si )
MeasurementModel
11
SIE Model of Vector Diffraction
Patterned Wafer
Interference fromSecondary SourceIncident Wave
Secondary Source
• Based on the Surface Equivalence Theorem (Generalization of Huygens’ Principle )
12
SIE Simulation of Near-normalEllipsometry
Measured and SIE Fitted Ellipsometry Data at 6º for the Nominal 500 nmDeep, 4 µm Period Structure
0.6
0.8
1
1.2
1.4
0.4 0.5 0.6 0.7 0.8
tan(psi) Measured tan(psi) SIE
wavelength ( m)µ
Tan(
ψ)
13
Rigorous Coupled-Wave Analysis (RCWA)
• Numerical Eigen-matrix Solution for Maxwell’s Equation• Groove Is Sliced Into a Number of Thin Layers• Amplitudes of Different Diffraction Orders Are Obtained by Solving Coupled-wave Equations
ε 1
ε2
0-1 1
2-2Transmitted Waves
0
1-1Reflected Waves
14
RCWA Simulation of 75ºSpectroscopic Ellipsometry
Measured SE Data From the 500 nm Depth, 4 µm Period Sample. TheRCWA Simulation Yielded a Period of 4.0 µm, a Top Linewidth of 2.2µm, a Sidewall Angle of 72.95º, and a Depth of 480 nm.
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
-2
-1.5
-1
-0.5
0
0.5
1
0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8
TAN(psi)-RCWATAN(psi)-measured
COS(delta)-RCWACOS(delta)-measurede
Wavelength µ( m)
Tan(
ψ)C
os(∆)
15
Grating Analysis Approach• Successively Better Approximations
– Estimate Etch Depth From Near-normal IncidenceSpectral Reflectometry in p-polarized Mode (Fast)
– Extract the Grating Period From the DiffractionExperiment
– Refine Depth Estimate and Estimate Period,Linewidth, and Wall Angle Using SIE Analysis of s-and p-polarized Reflectances (~1 min/λ)
– Refine Topography Esitmates Using RCWA onSpectroscopic Ellipsometry Data (~5 min/λ)
16
Si Grating Etched to Different Depths
Si Relief Gratings with Nominal 2µm Lines and Spaces Etched to Depths ofApproximately 100, 200, 300, 400, and 600 nm.
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
0.4 0.5 0.6 0.7 0.8
tan(psi) etch1tan(psi) etch2tan(psi) etch3tan(psi) etch4tan(psi) etch5
Wavelength( m)µ
Tan(
ψ)
17
Simulation of Grating Etch
Approximate Simulation of Si Relief Grating Etch. Note That the Trends ofthe Spectroscopic Ellipsometry Data Are Captured.
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
0.4 0.5 0.6 0.7 0.8
tan(psi) simulation1tan(psi) simulation2tan(psi) simulation3tan(psi) simulation4tan(psi) simulation5
Wavelength( m)µ
Tan(
ψ)
18
Comparison With SEM
The Depths Extracted From the Simulation Are in Very Good AgreementWith Those Measured From SEM. The Non-uniformity Among Different DieMay Be Responsible for the Small Difference.
0
100
200
300
400
500
600
0 20 40 60 80 100
depth (simulation)depth (SEM)
Etching Time (sec)
Etc
hing
Dep
th (n
m)
19
Real Time Data Collection
• Part of SE Data During the Etch of Patterned PR/4800ÅPoly/300Å Sio2/si (6 sec in Total With 1 sec Intervals)
• The RTSE Possesses a Data Collection Capability of <0.5 secWith a Whole Visible Light Spectral Range.
0.2
0.3
0.4
0.5
0.6
0.7
1.2 1.6 2 2.4 2.8 3.2 3.6 4
Tan(psi)-40sTan(psi)-41sTan(psi)-42sTan(psi)-43sTan(psi)-44sTan(psi)-45sTan(psi)-46s
Energy (ev)
Tan(
ψ)
20
Deep Sub-micrometer Regime
• Simulations Shows That for 100nm Features, the Near-normalSE Curves Still Exhibits Strong Structures, As Opposed to EvenSmaller Structures and Blanket Wafers.
• The 100nm and 110nm Curves IS Different.• Conclusion: Can “See” 100nm and Resolute 10nm.
0.2
0.4
0.6
0.8
1
1.2
1.4
0.4 0.5 0.6 0.7 0.8
Tan(psi)-BareTan(psi)-100nmTan(psi)-110nm
Wavelength ( m)µ
Tan(
ψ)
21
Conclusions
• Spectroscopic Ellipsometry Can Give AccurateDepth and Topography Information FromPatterned Structures.
• Quantitative Analysis of Diffraction Effects IsComputationally Time-intensive.
• In Situ Rapid Data Acquisition Is Possible WithRTSE.
22
Future Efforts
• Improved Algorithms for Higher Speed VectorDiffraction Analysis
• Quantitative Analysis of Topography ParameterSensitivities and Optimal MeasurementConditions
• Non-linear Regression Method for TopographyParameter Extraction
23
Technology Transfer Possibilities
• Working With National SemiconductorMentors on Applications to Gate LinewidthControl
• Analysis Methods Easily Transferable/noNew Instrumentation Required– Exist SE’s for Blanket Measurements Can Be
Used for Patterned Wafer Metrology
24
Acknowledgements• This Work Was Supported in Part by
– Semiconductor Research Corporation (Contract 97-FC085)
– AFOSR/DARPA MURI Center for IntelligentElectronics Manufacturing (AFOSR F49620-95-1-0524)
– State of Michigan Center for Display Technology andManufacturing
• The Authors Would Also Like to Thank Dr. D. S.Grimard and Ms. M. Gulari for Assistance WithSample Fabrication