Control of Pad Damage Using Prober Operational Parameters

Control of Pad Damage Control of Pad Damage Using Prober Operational Using Prober Operational

Parameters Parameters

June 3June 3--6, 20076, 2007San Diego, CA USASan Diego, CA USA

Deborah MillerMicron Technology, Inc.

Jerry Broz, Ph.D.International Test Solutions

Edward Robinson, Ph.D.Hyphenated Systems, LLC.

June 3-6, 2007June 3June 3--6, 20076, 2007 IEEE SW Test WorkshopIEEE SW Test WorkshopIEEE SW Test Workshop 222

OverviewOverview• Introduction / Background• Objectives • Materials and Methods• Results• Analysis / Discussion• Conclusions


IntroductionIntroduction• 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.


Background Background –– Area EffectsArea EffectsPad 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

% L

BB

Rej

ects

% N

SO

P R

ejec

ts

% AREA Pad Damage

Bal

l She

ar (g

ram

s)W

ire P

ull (

gram

s)

Sources …Tran, et al., ECTC -2000Tran, et al., SWTW-2000Langlois, et al, SWTW-2001Hotchkiss, et al., ECTC-2001Hothckiss, et al., IRPS-2001Among others …

Critical ValueDamage = 25%Ball Shear

Wire Pull

% LBB% NSOP


Area Effects Are Not Enough !Area Effects Are Not Enough !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)

Blanket aluminum wafer from IMSI SEMATECH


Background Background –– Depth EffectsDepth EffectsExcessively 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

Sources …Hartfield, et al, SWTW-2003Martens, et. al., SWTW-2003Hartfield, et al., SWTW-2004Stillman, et al., SWTW-2005Among others …

Many steps are needed to

assess cracks.

Images from Hartfield, et al, SWTW-2003


Background Background –– Height EffectsHeight EffectsPad 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

% L

BB

Rej

ects

% N

SO

P R

ejec

ts

Height of Pile - Up

Bal

l She

ar (g

ram

s)W

ire P

ull (

gram

s)

Sources …Langlois, et al, SWTW-2001Among others …

Ball ShearWire Pull

% LBB% NSOP

Critical Value“unknown”


Controlling the DamageControlling 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)


Assessing the DamageAssessing the DamageTraditional 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

Damage Assessment

Feedback to Production

Reporting


Probe Mark 3D Cross SectionProbe 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.

Pad surface-100nm

-200nm-300nm

-400nm-500nm

PROBE MARK


ObjectivesObjectives• 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 de-processing 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 cost-effective manner ?


Factors Factors (Prober Operational Settings)(Prober Operational Settings)• Number of Touchdowns

Single vs. Double

• Overtravel MagnitudeLow (50um) vs. Middle (63um) vs. High (75um)

• Undertravel MagnitudeLow (0um) vs. Middle (10um) vs. High (20um)

• Pin-Update Execution– Abbreviated pin alignment to compensate for thermal movement – On vs. Off

• Wafer Chuck SpeedLow (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


Responses Responses (Probe Mark Features)(Probe Mark Features)

• Probe Mark Depth

• Pile-up Height

• Probe Mark – Area – Volume

• Pile-up– Area– Volume


Touchdowns Undertravel Pin-Update Wafer Chuck Speed Chuck Revise OvertravelID1 Single Off On High Off 63

20 Single 10 Off High On 5017 Single 20 On High On 509 Single Off Off Low On 50

21 Single 10 Off High Off 6313 Single 20 Off Low Off 63

7 Single 10 On Low On 7512 Single 20 Off Low Off 7511 Single Off On Low Off 75

15 Double 10 On Low Off 5014 Double 20 Off Low Off 5023 Double Off On High Off 50

8 Double 10 On Low On 6319 Double 20 On High On 6310 Double Off Off Low On 63

24 Double 10 Off High Off 7518 Double 20 On High On 7516 Double Off Off High On 75

Design of Experiment (Design of Experiment (DoEDoE))Control

indicates a condition different than the CONTROL


Materials and MethodsMaterials and Methods


Wafer Sort ToolsWafer Sort Tools• Cantilevered probe card

Diagonal multi-site (X8) probe card representative of production.

• 25-wafer LOTPad Lot wafer representative of pad metal layer.

• Production Tester + Prober combinationTest cell with a “known good” condition.

• One operatorOperator variability kept to a minimum.


Probe Mark Inspection LayoutProbe Mark Inspection Layout

Location A

Location B

Location C

Three touchdownsacross 200 mm wafer

Four pads of interest ineach site

Two sites of interestwithin each touchdown

“Split Probe” marks wereobserved AND included

in the dataset.

2

12

1

2

1


Probe Mark MetrologyProbe Mark Metrology• Scrub mark volume and area• Pile up volume and area• Maximum depth


Scrub Mark MeasurementScrub Mark Measurement

Threshold parameters set to determine the area

falling more than 100nm beneath average pad level.

Region ofInterest

Signal threshold reference level at the pad surface.


PilePile--up Measurementup MeasurementSignal threshold reference level at the pad surface.

Region ofInterest

Threshold parameters set to determine the area

falling more than 200nm above average pad level.


Depth MeasurementDepth MeasurementA cross-section tool is used to measure the probe depth.


Test ResultsTest Results


-100

-50

0

50

100

150

200

250

300

350

W16 W18 W24 W10 W08 W19 W14 W23 W15 W12 W07 W11 W21 W22 W13 W01 W09 W20 W17

Area Damage Volumetric Damage

Sorted by Pad Volume Damagem

icro

ns2

mic

rons

3

Pad Damage ParetoPad Damage Pareto


-100

-50

0

50

100

150

200

250

300

350

W16 W18 W24 W10 W08 W19 W14 W23 W15 W12 W07 W11 W21 W22 W13 W01 W09 W20 W17

Area Damage Volumetric Damage

Sorted by Pad Volume Damagem

icro

ns2

mic

rons

3

PilePile--up Paretoup Pareto


Maximum Scrub VolumeMaximum Scrub VolumeOne Way Analysis Variance of MAXIMUM Scrub Volume By Wafer

10

20

30

40

50

60

70

80

90

100

Scru

b Vo

lum

e

1 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Wafer

Each PairStudent's t0.05

Touchdowns Undertravel Pin-Update Wafer Chuck Speed Chuck Revise Overtravel24 Double 10 Off High Off 7518 Double 20 On High On 7516 Double Off Off High On 75

W16

W18

W24

StatisticallySignificant

Max Damage


Touchdowns Undertravel Pin-Update Wafer Chuck Speed Chuck Revise Overtravel20 Single 10 Off High On 5017 Single 20 On High On 509 Single Off Off Low On 50

One Way Analysis of Variance of MINIMUM Scrub Volume By Wafer

10

20

30

40

50

60

70

80

90

100

Scr

ub V

olum

e

1 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Wafer


Minimum Scrub VolumeMinimum Scrub Volume

W09

W17

W20

BorderlineStatisticallySignificant

Min Damage


One way Analysis of MAXIMUM Scrub Area By Wafer

50

100

150

200

250

300

350

Scru

b Ar

ea

1 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Wafer


Maximum Scrub AreaMaximum Scrub Area

Touchdowns Undertravel Pin-Update Wafer Chuck Speed Chuck Revise Overtravel24 Double 10 Off High Off 75

W024

StatisticallySignificant

Max Damage


One Way Analysis of MINIMUM Scrub Area By Wafer

50

100

150

200

250

300

350

Scru

b Ar

ea

1 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Wafer


Minimum Scrub AreaMinimum Scrub Area

Touchdowns Undertravel Pin-Update Wafer Chuck Speed Chuck Revise Overtravel20 Single 10 Off High On 5017 Single 20 On High On 509 Single Off Off Low On 50

W09

W17

W20

BorderlineStatisticallySignificant

Min Damage


Analysis / DiscussionAnalysis / Discussion


Response Mean (Scrub Area) Actual by Predicted Plot

50

100

150

200

250

300

Mea

n(S

crub

Area

) Act

ual

50 100 150 200 250 300Mean(Scrub Area) Predicted

P<.0001 RSq=0.99 RMSE=7.9659

Summary of Fit RSquare 0.992907RSquare Adj 0.981761Root Mean Square Error 7.965932Mean of Response 151.0761Observations (or Sum Wgts) 19

Response Mean (Scrub Volume) Actual by Predicted Plot

10

20

30

40

50

60

70

80

Mea

n(S

crub

Volu

me)

Act

ual

10 20 30 40 50 60 70 80Mean(Scrub Volume) Predicted

P<.0001 RSq=1.00 RMSE=1.3295


Scrub Mark Data Modeled in JMPScrub Mark Data Modeled in JMPActual vs. Predicted Results


13.3180

1.6657

-1.1663

-0.9465

-0.7554

-0.6623

2.0472

-1.9320

1.0323

-0.0079

-12.4100

0 2 4 6 8 10 12 14 16

ut[20]

ut[10]

cr[Off]

speed[Low]*cr[Off]

pu[Off]

speed[Low]

speed[Low]*ot[50]

ot[63]

speed[Low]*ot[63]

ot[50]

td[Double]

Significant Factor EstimatesSignificant Factor EstimatesProbe Mark Volume

Single vs. Double TouchdownMinimum vs. Maximum Overtravel

Primary Responses

Secondary Responses

No clear wafer chuck speed dependency was surprising.

Scaled Estimates


1.6605

2.0336

2.7873

5.0350

5.4209

39.9630

-38.75266

-1.5891

-4.8920

-7.1279

-9.0581

0 5 10 15 20 25 30 35 40 45

ut[10]

ut[20]

cr[Off]

pu[Off]

speed[Low]*cr[Off]

speed[Low]*ot[63]

speed[Low]*ot[50]

ot[63]

speed[Low]

ot[50]

td[Double]

Significant Factor EstimatesSignificant Factor EstimatesProbe Mark Area


Primary Responses

Secondary ResponsesA reduced dataset analysis showed that speed was the

third largest factor contributing to the

probe mark area response.

Scaled Estimates


Correlation Between ResponsesCorrelation Between Responses

20

30

40

50

60

70

80

100

150

200

250

300

0.4

0.45

0.5

Mean(ScrubVolume)

20 30 40 50 60 70 80

Mean(ScrubArea)

100 150 200 250 300

Mean(ScrubDepth)

.4 .45 .5

DoubleSingle

td

Scatterplot Matrix

• As expected, the probe mark scrub area and scrub volume showed statistically high correlations to each other

• The correlation between probe mark depth and volume was not statistically significant

• Possible reasons for lack of correlation to probe depth

– Small sample size effects– Operator-induced variability– Probe tip diameter– Probe gram force


Response Mean (Pile Up Area) Actual by Predicted Plot

20

40

60

80

100

120

Mea

n(D

ivot

Area

) Act

ual

20 40 60 80 100 120Mean(Divot Area) Predicted

P=0.0009 RSq=0.96 RMSE=8.0552


Response Mean (Pile Up Volume) Actual by Predicted Plot

0

10

20

30

40

50

60

Mea

n(D

ivot

Volu

me)

Act

ual

0 10 20 30 40 50 60Mean(Divot Volume) Predicted

P=0.0012 RSq=0.95 RMSE=4.6836


PilePile--up Data Modeled in JMPup Data Modeled in JMPActual vs. Predicted Results


13.3180

1.6657

-1.1663

-0.9465

-0.7554

-0.6623

2.0472

-1.9320

1.0323

-0.0079

-12.4100

0 2 4 6 8 10 12 14 16

ut[20]

ut[10]

cr[Off]

speed[Low]*cr[Off]

pu[Off]

speed[Low]

speed[Low]*ot[50]

ot[63]

speed[Low]*ot[63]

ot[50]

td[Double]

Significant Factor EstimatesSignificant Factor EstimatesPile-up Volume


Primary Responses

Secondary ResponsesAdditional dataset analysis

showed that speed wasa contributing factor for

pile-up volume

Scaled Estimates


1.3116

1.5082

1.8716

1.9195

1.9346

2.5774

20.6415

-4.4643

-11.6636

-1.4899

-0.4748

0 5 10 15 20 25

speed[Low]*ot[50]

speed[Low]*ot[63]

ot[63]

ut[10]

cr[Off]

pu[Off]

speed[Low]

speed[Low]*cr[Off]

ut[20]

ot[50]

td[Double]

Significant Factor EstimatesSignificant Factor EstimatesPile Up Area


Primary Responses

Secondary Responses

No clear wafer chuck speed dependency was surprising.

Undertravel was a more significantcontributing factor than chuck speed.

Scaled Estimates


Correlation Between ResponsesCorrelation Between Responses• Statistically significant

correlation between the primary responses (area and volume) was observed.

• Modeled data can be used to investigate optimal conditions.

2030

50

70

100

150

200

250

300

102030405060

40

60

80

100

120

Mean(ScrubVolume)

20 30 40 50 60 70 80

Mean(ScrubArea)

100 150 200 250 300

Mean(DivotVolume)

10 20 30 40 50 60

Mean(DivotArea)

40 60 80 100 120

Scatterplot Matrix


0

20

40

60

Mea

n(D

ivot

Vol

ume)

8.73

2232

±8.4

3241

3

Dou

ble

Sing

le

Singletd

Low

Hig

h

Highspeed

Off

On

Oncr

50 63 75

50ot

10 20 off

20ut

Off

On

Onpu

Off

50100150200250300

Mea

n(S

crub

Are

a)64

.055

95

±19.

7227

5

Dou

ble

Sin

gle

Singletd

Low

Hig

h

Lowspeed

Off

On

Offcr

50 63 7550ot

10 20 off

10ut

Off

On

Onpu

Best Case CombinationsBest 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.


Summary / ConclusionsSummary / 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.


FollowFollow--On WorkOn 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


AcknowledgementsAcknowledgements• Chuck Jensen, Micron Technology, Inc.• Brett Crump, Micron Technology, Inc.• John Little, Hyphenated Systems, LLC

Date post:	16-Jan-2022
Category:	Documents
Upload:	others
View:	3 times
Download:	0 times

Control of Pad Damage Using Prober Operational Parameters

Documents