Control of Pad Damage Using Prober Operational Parameters

Control of Pad Damage
Using Prober Operational
Parameters
Deborah Miller
Micron Technology, Inc.
Jerry Broz, Ph.D.
International Test Solutions
Edward Robinson, Ph.D.
Hyphenated Systems, LLC.
June 3-6, 3
2007
San Diego, CA USA

Overview
• Introduction / Background
• Objectives
• Materials and Methods
• Results
• Analysis / Discussion
• Conclusions
June 3-6, 3
2007 IEEE SW Test Workshop 2

Introduction
• Probe card technologies have become advanced; however, the
basics of wafer sort really have not changed
• ALL probe technologies have a contact area substantially
harder than the pads or solder balls of the device
• “Contact and slide” is CRITICAL to break surface oxide(s), but
results in localized plastic deformation, i.e. a probe mark
• Volume of material displaced and/or transferred is a complex
function of dynamic contact mechanics, metallic interactions,
frictional effects, and other tribological properties
• Disclaimer: scrub mark photos, pad profiles, and pad structures shown
in this presentation are not considered representative of or meant to
infer anything about the process of record for Micron products.
June 3-6, 3
2007 IEEE SW Test Workshop 3

Background – Area Effects
Pad damage due to probe has been positively correlated
to bondability issues.
– Reduced ball shear strength and wire pull strength
– Increased NSOP (no stick on pad) and LBB (lifted ball bond)
Assembly Parameter vs. Probe Mark Area
Ball Shear (grams)
Wire Pull (grams)
Ball Shear
Wire Pull
Critical Value
Damage = 25%
% LBB
% NSOP
% AREA Pad Damage
% LBB Rejects
% NSOP Rejects
Sources …
Tran, et al., ECTC -2000
Tran, et al., SWTW-2000
Langlois, et al, SWTW-2001
Hotchkiss, et al., ECTC-2001
Hothckiss, et al., IRPS-2001
Among others …
June 3-6, 3
2007 IEEE SW Test Workshop 4

Area Effects Are Not Enough !
Blanket aluminum wafer from IMSI SEMATECH
A probe mark can have a
relatively small area of
damage, but exceed the
critical allowable depth.
– % Area Damage = 8.8%
which is within limits
– Depth = 10000Å which is
excessively deep
6000 Å aluminum + 5500 Å thermal oxide = 11000 Å
Probe Depth = 10000Å (punch-through)
June 3-6, 3
2007 IEEE SW Test Workshop 5

Background – Depth Effects
Excessively deep probe marks can cause …
– Underlying layer damage (low-k dielectric, circuitry under bond
pads, and aluminum capped copper pads)
– Bondability and long term reliability issues
Many steps are
needed to
assess cracks.
Images from Hartfield, et al, SWTW-2003
Sources …
Hartfield, et al, SWTW-2003
Martens, et. al., SWTW-2003
Hartfield, et al., SWTW-2004
Stillman, et al., SWTW-2005
Among others …
June 3-6, 3
2007 IEEE SW Test Workshop 6

Background – Height Effects
Pad material pile-up has also been correlated to
bondability issues.
– Reduced ball shear strength and wire pull strength
– Increased NSOP (no stick on pad) and LBB (lifted ball bond)
Assembly Parameter vs. Aluminum Pile-Up
Ball Shear (grams)
Wire Pull (grams)
Ball Shear
Wire Pull
Critical Value
“unknown”
% LBB
% NSOP
% LBB Rejects
% NSOP Rejects
Height of Pile - Up
Sources …
Langlois, et al, SWTW-2001
Among others …
June 3-6, 3
2007 IEEE SW Test Workshop 7

Controlling the Damage
• The depth of the probe mark can be controlled with
modifications to the probe card technology
– Low force probe cards (various manufacturers)
– Optimized probe to pad interactions
• Probers can effectively change the z-stage motion just
before contact and during overtravel to reduce damage
– Variable Speed Probing by Accretech ®
– Micro-Touch by Electroglas ®
– Micro-Force Probing by Tokyo Electron Limited ® (TEL)
June 3-6, 3
2007 IEEE SW Test Workshop 8

Assessing the Damage
Traditional depth, volume, and height measurements are
time consuming and can have long cycle times.
– Probing under different conditions
– Wafers must be scrapped
– Careful wafer sectioning
– Sample preparation and de-processing
– Electron-based microscopy
Probe Card + Wafer
• Touchdowns
• Variable Conditions
Manual Failure Analysis
• Sectioning
• Deprocessing
• Electron Microscopy
• Metrology / Correlation
Reporting
Damage Assessment
Feedback to Production
June 3-6, 3
2007 IEEE SW Test Workshop 9

Probe Mark 3D Cross Section
• From the wafer sort standpoint …
Displaced volume can be correlated to key sort parameters, e.g. z-stage
speed, overtravel, probe force, cracking, punch-through, etc.
• From a cleaning standpoint …
Displaced volume provides insight into accumulation rates and material
adherence.
PROBE MARK
Pad surface
-100nm
-200nm
-300nm
-400nm
-500nm
June 3-6, 3
2007 IEEE SW Test Workshop 10

Objectives
• Develop a multi-variable parametric DoE to identify
primary prober operational contributors to probe damage
• Perform a statistically valid and practical failure analysis
on damaged bond pads without cross-sectioning or deprocessing
the wafers
– 3D confocal, non-contact microscopy with a better than 50nm
resolution
– Wafers must be available for follow-on metrology
• Identify an optimized combination of prober operational
settings to reduce the overall area and volumetric probe
damage, i.e. disturbed pad area
Can reasonable steps be taken with existing technologies (e.g., an
existing probe card and a prober) to reduce pad damage in a costeffective
manner ?
June 3-6, 3
2007 IEEE SW Test Workshop 11

Factors (Prober Operational Settings)
• Number of Touchdowns
Single vs. Double
• Overtravel Magnitude
Low (50um) vs. Middle (63um) vs. High (75um)
• Undertravel Magnitude
Low (0um) vs. Middle (10um) vs. High (20um)
• Pin-Update Execution
– Abbreviated pin alignment to compensate for thermal movement
– On vs. Off
• Wafer Chuck Speed
Low (6000 um/sec) vs. High (18000 um/sec)
• Chuck Revise Execution
– Re-zero of the wafer chuck to compensate for thermal movement
– On vs. Off
June 3-6, 3
2007 IEEE SW Test Workshop 12

Responses (Probe Mark Features)
• Probe Mark Depth
• Pile-up Height
• Probe Mark
–Area
–Volume
• Pile-up
–Area
–Volume
June 3-6, 3
2007 IEEE SW Test Workshop 13

Design of Experiment (DoE(
DoE)
Touchdowns Undertravel Pin-Update Wafer Chuck Speed Chuck Revise Overtravel
ID
1 Single Off On High Off 63
Control
20 Single 10 Off High On 50
17 Single 20 On High On 50
9 Single Off Off Low On 50
21 Single 10 Off High Off 63
13 Single 20 Off Low Off 63
7 Single 10 On Low On 75
12 Single 20 Off Low Off 75
11 Single Off On Low Off 75
15 Double 10 On Low Off 50
14 Double 20 Off Low Off 50
23 Double Off On High Off 50
8 Double 10 On Low On 63
19 Double 20 On High On 63
10 Double Off Off Low On 63
24 Double 10 Off High Off 75
18 Double 20 On High On 75
16 Double Off Off High On 75
indicates a condition different than the CONTROL
June 3-6, 3
2007 IEEE SW Test Workshop 14

Materials and Methods
June 3-6, 3
2007 IEEE SW Test Workshop 15

Wafer Sort Tools
• Cantilevered probe card
Diagonal multi-site (X8) probe card representative of
production.
• 25-wafer LOT
Pad Lot wafer representative of pad metal layer.
• Production Tester + Prober combination
Test cell with a “known good” condition.
• One operator
Operator variability kept to a minimum.
June 3-6, 3
2007 IEEE SW Test Workshop 16

Probe Mark Inspection Layout
Location A
2
Two sites of interest
within each touchdown
1
Location B
1
2
2
Four pads of
interest in
each site
Location C
1
Three touchdowns
across 200 mm wafer
“Split Probe” marks were
observed AND included
in the dataset.
June 3-6, 3
2007 IEEE SW Test Workshop 17

Probe Mark Metrology
• Scrub mark volume and area
• Pile up volume and area
• Maximum depth
June 3-6, 3
2007 IEEE SW Test Workshop 18

Scrub Mark Measurement
Signal threshold reference level at the pad surface.
Region of
Interest
Threshold parameters set
to determine the area
falling more than 100nm
beneath average pad level.
June 3-6, 3
2007 IEEE SW Test Workshop 19

Pile-up Measurement
Signal threshold reference level at the pad surface.
Region of
Interest
Threshold parameters set
to determine the area
falling more than 200nm
above average pad level.
June 3-6, 3
2007 IEEE SW Test Workshop 20

Depth Measurement
A cross-section tool is used to measure the
probe depth.
June 3-6, 3
2007 IEEE SW Test Workshop 21

Test Results
June 3-6, 3
2007 IEEE SW Test Workshop 22

Pad Damage Pareto
350
Sorted by Pad Volume Damage
microns 2
300
250
200
150
100
50
microns 3
0
-50
-100
W16
W18
W24
W10
W08
W19
W14
W23
W15
W12
W07
W11
W21
W22
W13
W01
W09
W20
W17
Area Damage
Volumetric Damage
June 3-6, 3
2007 IEEE SW Test Workshop 23

Pile-up Pareto
350
Sorted by Pad Volume Damage
microns 2
300
250
200
150
100
50
microns 3
0
-50
-100
W16
W18
W24
W10
W08
W19
W14
W23
W15
W12
W07
W11
W21
W22
W13
W01
W09
W20
W17
Area Damage
Volumetric Damage
June 3-6, 3
2007 IEEE SW Test Workshop 24

Maximum Scrub Volume
One Way Analysis Variance of MAXIMUM Scrub Volume By Wafer
Scrub Volume
100
90
80
70
60
50
40
30
20
10
1 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Wafer
Each Pair
Student's t
0.05
W16
W18
W24
Statistically
Significant
Max Damage
Touchdowns Undertravel Pin-Update Wafer Chuck Speed Chuck Revise Overtravel
24 Double 10 Off High Off 75
18 Double 20 On High On 75
16 Double Off Off High On 75
June 3-6, 3
2007 IEEE SW Test Workshop 25

Minimum Scrub Volume
One Way Analysis of Variance of MINIMUM Scrub Volume By Wafer
Scrub Volume
100
90
80
70
60
50
40
30
20
10
1 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Wafer
Each Pair
Student's t
0.05
Borderline
Statistically
Significant
Min Damage
W09
W17
W20
Touchdowns Undertravel Pin-Update Wafer Chuck Speed Chuck Revise Overtravel
20 Single 10 Off High On 50
17 Single 20 On High On 50
9 Single Off Off Low On 50
June 3-6, 3
2007 IEEE SW Test Workshop 26

Maximum Scrub Area
One way Analysis of MAXIMUM Scrub Area By Wafer
350
Scrub Area
300
250
200
150
W024
Statistically
Significant
Max Damage
100
50
1 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Wafer
Each Pair
Student's t
0.05
Touchdowns Undertravel Pin-Update Wafer Chuck Speed Chuck Revise Overtravel
24 Double 10 Off High Off 75
June 3-6, 3
2007 IEEE SW Test Workshop 27

Minimum Scrub Area
One Way Analysis of MINIMUM Scrub Area By Wafer
350
Scrub Area
300
250
200
150
Borderline
Statistically
Significant
Min Damage
W09
100
W17
50
1 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Wafer
Each Pair
Student's t
0.05
W20
Touchdowns Undertravel Pin-Update Wafer Chuck Speed Chuck Revise Overtravel
20 Single 10 Off High On 50
17 Single 20 On High On 50
9 Single Off Off Low On 50
June 3-6, 3
2007 IEEE SW Test Workshop 28

Analysis / Discussion
June 3-6, 3
2007 IEEE SW Test Workshop 29

Scrub Mark Data Modeled in JMP
Scrub Mark Data Modeled in JMP
Actual vs. Predicted Results
Response Mean (Scrub Volume)
Actual by Predicted Plot
Mean(Scrub
Volume) Actual
80
70
60
50
40
30
20
10
10 20 30 40 50 60 70 80
Mean(Scrub Volume) Predicted
P

Significant Factor Estimates
Probe Mark Volume
Primary Responses
td[Double]
ot[50]
13.3180
-12.4100
Single vs. Double Touchdown
Minimum vs. Maximum Overtravel
speed[Low]*ot[63]
ot[63]
speed[Low]*ot[50]
speed[Low]
pu[Off]
speed[Low]*cr[Off]
cr[Off]
ut[10]
ut[20]
2.0472
-1.9320
1.6657
-1.1663
1.0323
-0.9465
-0.7554
-0.6623
-0.0079
Secondary Responses
No clear wafer chuck speed
dependency was surprising.
0 2 4 6 8 10 12 14 16
Scaled Estimates
June 3-6, 3
2007 IEEE SW Test Workshop 31

Significant Factor Estimates
Probe Mark Area
Primary Responses
td[Double]
ot[50]
39.9630
-38.75266
Single vs. Double Touchdown
Minimum vs. Maximum Overtravel
speed[Low]
ot[63]
speed[Low]*ot[50]
speed[Low]*ot[63]
speed[Low]*cr[Off]
pu[Off]
cr[Off]
ut[20]
ut[10]
-9.0581
-7.1279
5.4209
5.0350
-4.8920
2.7873
2.0336
1.6605
-1.5891
Secondary Responses
A reduced dataset analysis
showed that speed was the
third largest factor
contributing to the
probe mark area response.
0 5 10 15 20 25 30 35 40 45
Scaled Estimates
June 3-6, 3
2007 IEEE SW Test Workshop 32

Correlation Between Responses
Scatterplot Matrix
80
Mean(Scrub
70
Volume)
60
50
40
30
td
Double
Single
• As expected, the probe mark
scrub area and scrub volume
showed statistically high
correlations to each other
20
300
250
200
150
Mean(Scrub
Area)
• The correlation between
probe mark depth and
volume was not statistically
significant
100
0.5
0.45
0.4
20 30 40 50 60 70 80
100 150 200 250 300
Mean(Scrub
Depth)
.4 .45 .5
• Possible reasons for lack of
correlation to probe depth
– Small sample size effects
– Operator-induced variability
– Probe tip diameter
– Probe gram force
June 3-6, 3
2007 IEEE SW Test Workshop 33

Pile-up Data Modeled in JMP
Actual vs. Predicted Results
Response Mean (Pile Up Volume)
Response Mean (Pile Up Area)
Actual by Predicted Plot
Actual by Predicted Plot
60
120
Mean(Divot
Volume) Actual
50
40
30
20
10
Mean(Divot
Area) Actual
100
80
60
40
0
0 10 20 30 40 50 60
Mean(Divot Volume) Predicted
P=0.0012 RSq=0.95 RMSE=4.6836
20
20 40 60 80 100 120
Mean(Divot Area) Predicted
P=0.0009 RSq=0.96 RMSE=8.0552
Summary of Fit
RSquare 0.954192
RSquare Adj 0.882209
Root Mean Square Error 4.683565
Mean of Response 26.51567
Observations (or Sum Wgts) 19
Summary of Fit
RSquare 0.957954
RSquare Adj 0.891882
Root Mean Square Error 8.055216
Mean of Response 62.74199
Observations (or Sum Wgts) 19
June 3-6, 3
2007 IEEE SW Test Workshop 34

Significant Factor Estimates
Pile-up Volume
Primary Responses
td[Double]
ot[50]
13.3180
-12.4100
Single vs. Double Touchdown
Minimum vs. Maximum Overtravel
speed[Low]*ot[63]
ot[63]
speed[Low]*ot[50]
speed[Low]
pu[Off]
speed[Low]*cr[Off]
cr[Off]
ut[10]
ut[20]
2.0472
-1.9320
1.6657
-1.1663
1.0323
-0.9465
-0.7554
-0.6623
-0.0079
Secondary Responses
Additional dataset analysis
showed that speed was
a contributing factor for
pile-up volume
0 2 4 6 8 10 12 14 16
Scaled Estimates
June 3-6, 3
2007 IEEE SW Test Workshop 35

Significant Factor Estimates
Pile Up Area
Primary Responses
td[Double]
ot[50]
-11.6636
20.6415
Single vs. Double Touchdown
Minimum vs. Maximum Overtravel
ut[20]
speed[Low]*cr[Off]
speed[Low]
pu[Off]
cr[Off]
ut[10]
ot[63]
speed[Low]*ot[63]
speed[Low]*ot[50]
-4.4643
2.5774
1.9346
1.9195
1.8716
1.5082
-1.4899
1.3116
-0.4748
Secondary Responses
No clear wafer chuck speed
dependency was surprising.
Undertravel was a more significant
contributing factor than chuck
speed.
0 5 10 15 20 25
Scaled Estimates
June 3-6, 3
2007 IEEE SW Test Workshop 36

Correlation Between Responses
Scatterplot Matrix
70
50
30
20
300
250
Mean(Scrub
Volume)
• Statistically significant
correlation between the
primary responses (area
and volume) was
observed.
200
150
100
60
50
40
30
20
10
Mean(Scrub
Area)
Mean(Divot
Volume)
• Modeled data can be
used to investigate
optimal conditions.
120
100
80
60
40
Mean(Divot
Area)
20 30 40 50 60 70 80
100 150 200 250 300
10 20 30 40 50 60
40 60 80 100 120
June 3-6, 3
2007 IEEE SW Test Workshop 37

Best Case Combinations
• Modeled response data can be used to investigate the effects of
changing one parameter and keeping the other constant.
– Slopes of the lines can give some indication of sensitivity to the change.
Mean(Divot
Volume)
8.732232
±8.432413
60
40
20
0
Double
Single
td
Single
Low
Hig h
speed
High
Off
Off On
cr
On
50
63
50
ot
75
10
20
20
ut
off
Off
On
pu
On
Mean(Scrub
Area)
64.05595
±19.72275
300
250
200
150
100
50
Double
Single
Low
High
Off
On
50
63
75
10
20
off
Off
On
Single
Low
Off
50
10
On
td
speed
cr
ot
ut
pu
June 3-6, 3
2007 IEEE SW Test Workshop 38

Summary / Conclusions
• Reasonable steps can be taken with “existing” hardware to
reduce pad damage in a cost-effective manner.
• Volumetric probe damage assessment can be used to
optimize probe/pad interaction and reduce yield fallout.
– Non-contact methods are critical for statistically valid and practical failure
analysis without de-processing and/or cross-sectioning.
• Even with the small sample size, the statistical power was
adequate to give an indication for response sensitivity to
primary process factors.
• The influence of second order factors for fine-tuning the
operational parameters can be performed using modeled
response data.
June 3-6, 3
2007 IEEE SW Test Workshop 39

Follow-On Work
• Investigate improved height and depth measurements
– Larger sample size
– Improved automated methods to reduce operator induced
variability
• Consider probe card parameters
– Probe tip diameter
– Probe gram force
• Validate test results using a secondary metrology
evaluation
• Further assessment of the “optimized” operational
parameter combination
– CRES performance evaluation
– Bondability testing
June 3-6, 3
2007 IEEE SW Test Workshop 40

Acknowledgements
• Chuck Jensen, Micron Technology, Inc.
• Brett Crump, Micron Technology, Inc.
• John Little, Hyphenated Systems, LLC
June 3-6, 3
2007 IEEE SW Test Workshop 41