Eric Schulte 1 , Matthew Lueck 2 , Alan Huffman 2 , Chris Gregory 2 , - - PowerPoint PPT Presentation

▶
eric schulte 1 matthew lueck 2 alan huffman 2 chris
SMART_READER_LITE
LIVE PREVIEW

Eric Schulte 1 , Matthew Lueck 2 , Alan Huffman 2 , Chris Gregory 2 , - - PowerPoint PPT Presentation

Feb 19, 2024 39 likes •400 views

INTERCONNECT STRUCTURE FOR ROOM TEMPERATURE 3D-IC STACKING EMPLOYING BINARY ALLOYING FOR HIGH TEMPERATURE STABILITY Eric Schulte 1 , Matthew Lueck 2 , Alan Huffman 2 , Chris Gregory 2 , Keith Cooper 1 , Dorota Temple 2 1 SET North America,

slide-1
SLIDE 1

INTERCONNECT STRUCTURE FOR ROOM TEMPERATURE 3D-IC STACKING EMPLOYING BINARY ALLOYING FOR HIGH TEMPERATURE STABILITY

Eric Schulte1, Matthew Lueck2, Alan Huffman2, Chris Gregory2, Keith Cooper1, Dorota Temple2

1SET North America, Chester, NH, USA 2RTI International CMET, Research Triangle Park, NC, USA

eschulte@set-na.com

11/6/2013 1

slide-2
SLIDE 2

Outline

  • Intro: Current Methods of 3D Assembly
  • Proposed Solution: Room Temp Bonding
  • Methodology and Characterization
  • Evaluation of Experimental Results
  • Conclusions and Next Steps

11/6/2013 2

slide-3
SLIDE 3

3D Promise / 3D Issues

Promise:

  • High speed
  • Low power
  • High density

Issues:

  • Bonding Registration Issues
  • Serial Yield Issues
  • Operability/Reliability Issues

11/6/2013 3

slide-4
SLIDE 4

Assessment of Conventional Reflow and Thermocompression Bonding for 3D-IC

11/6/2013 4

slide-5
SLIDE 5

Conventional Reflow and Thermocompression Bonding REFLOW

( e.g. SnAg/Cu)

THERMOCOMPRESSION

(e.g. Cu/Cu)

Fast Slow Solder Bridges Confined, Stable Low force High Force Lateral instability Laterally Stable Solder Compliance Ultra Flatness Required Unstable during stacking Thermally Stable CTE Mismatch CTE Mismatch Controlled Atmosphere Controlled Atmosphere

Too Much Red Here!!

11/6/2013 5

slide-6
SLIDE 6

Ideal 3D Metallurgy and Bond Process Would Have the Following Characteristics

  • High speed bond cycle.

– Room temperature bond at low force. – Air ambient.

  • Fine pitch capability (<10µ) without bridging.
  • Compliant metallurgy to give flatness margin.
  • Unlimited wafer level chip stacking.

– Mechanical stability during (1+n) bonds. – No concerns for oxidation buildup.

  • Immune to “next-higher-assembly” reflow.

11/6/2013 6

slide-7
SLIDE 7

Proposal: A Novel Metallurgy and Bond Process for Room Temperature 3D Multi-Chip Stacking

11/6/2013 7

slide-8
SLIDE 8

Proposed Solution: InAg Binary

In Ag Ag In In/Ag Alloy

11/6/2013 8

slide-9
SLIDE 9

Advantages of InAg Binary

  • Deoxidized Ag and In bond instantly at RT.
  • Compliant Indium allows flatness tolerance.
  • Indium has easily controlled squeeze-out.
  • Low bonding force: < 0.1 gram per bump at

atmospheric ambient.

  • Mechanical stability during subsequent bonds.
  • InAg alloy anneal is performed at 120-140C

(solid state), then stable to >600C.

11/6/2013 9

slide-10
SLIDE 10

InAg Binary Bonding- Engineering Details

11/6/2013 10

slide-11
SLIDE 11

Detail: Surface Prep

  • De-oxidized Indium and Silver will

cold-weld instantly at room temp.

  • Could wet etch oxide, but thruput is

slow and oxide re-grows, making the process time-dependent.

  • Atmospheric plasma quickly removes
  • xide and passivates die for bonding.
  • Passivation enables long queue

lifetime (hours).

11/6/2013 11

slide-12
SLIDE 12

Detail: In-Situ Probing

  • Room temp bonding and no

confinement enable in-situ probing during bonding.

Lower vacuum tool Upper vacuum tool

i.e. Processor i.e. Stacked Memory

RT bond, no confinement

11/6/2013 12

slide-13
SLIDE 13

Detail: In-Situ Probing

  • Room temp bonding and no

confinement enable in-situ probing during bonding.

  • Operability of each bond

can be checked during the stacking operation.

Lower vacuum tool Upper vacuum tool RT bond, no confinement

11/6/2013 13 i.e. Stacked Memory i.e. Processor

slide-14
SLIDE 14

Detail: In/Ag Alloy Anneal

  • Indium and Silver interdiffuse rapidly, even below the

melting point of Indium. (~135C)

  • Since the bonded connections remain in the solid

phase, no compression force is needed during anneal. Die flatness/bowing issues are avoided.

  • Ideal volume ratio of Ag to In is 2:1 to form Ag2In with

a melting point of ~600C.

  • Diffusion kinetics depend on metal purity, time,

volume, and temperature.

  • Cross-section + EDS provide interdiffusion data.

11/6/2013 14

slide-15
SLIDE 15

Experimental

11/6/2013 15

slide-16
SLIDE 16

Test Chips

Chip (In bump):

  • Silicon chip.
  • 256 Copper daisy chain

continuity channels.

  • 1280 bumps each.
  • Bumps are 4µ dia, 4µ tall
  • 10 µ centers.
  • Copper pillars (plated).
  • Nickel barrier (plated).
  • In cap (plated).
  • No CMP.

Substrate (Ag bump):

  • Silicon substrate.
  • 256 Copper daisy chain

continuity channels.

  • 1280 bumps each.
  • Bumps are 4µ dia, 4µ tall
  • 10µ centers.
  • Copper pillars (plated).
  • Nickel barrier (plated).
  • Ag cap (plated).
  • No CMP.

4µ 4µ 4µ 4µ

11/6/2013 16

slide-17
SLIDE 17

Wet Etch Surface Preparation

  • Pre-bond wet etch option:

– Dilute HCL to remove oxidation from Ag and In. – Extreme care required to avoid over-etching. – Bond parts within 10 minutes to avoid re-oxidation.

11/6/2013 17

slide-18
SLIDE 18

Atmospheric Plasma Surface Prep

  • Reducing chemistry converts

bump oxide back to native metal.

  • Passivating chemistry ties up

metal dangling bonds.

  • Process takes less than 1
  • minute. Atomic passivation

inhibits re-oxidation for hours, is bond-able.

  • Activates chip surfaces for

enhanced underfill wicking.

11/6/2013 18

slide-19
SLIDE 19

Room Temperature Bonding

  • 27° C substrate and chip.
  • Compression bond at <0.1 gram per

bump (32Kg total force on 640x512 bumps).

  • Maintain 1 µ alignment accuracy thru

bonding.

  • Confining gas not required.
  • Multiple-chip automatic placement

available but not used for these experiments.

11/6/2013 19

slide-20
SLIDE 20

Post-Bond Alloy Anneal

  • Alloy anneals performed in room air.
  • Programmed ramp, temperature, and time.
  • RT-140C alloy anneal temperature.
  • 0-32 Kg compression force applied during anneal.
  • 0-30 minutes alloy anneal time.
  • Can be performed simultaneously with underfill

cure.

11/6/2013 20

slide-21
SLIDE 21

Experimental Results

11/6/2013 21

slide-22
SLIDE 22

Atmospheric Plasma Cleanup, RT Bond, 200C 10 min Alloy Anneal (no force)

  • Strong adhesion of In/Ag

as evidenced by tensile rupture.

  • Ag2In alloy is ductile, not

fragile

  • Capable of removing

alloyed In/Ag bump from its Ni pad.

Indium-bumped chip Silver-bumped substrate

11/6/2013 22

slide-23
SLIDE 23

RT Bond, A.P., 200C 30 min Anneal Cross-section and EDS

A B C D E F Ag 95.7% 67.6% 58.8% 25.1% 33.0% 0.0% In 4.3% 32.4% 37.9% 65.1% 59.0% 0.0% Ni

  • 100.0%

Cu 0.0% 0.0% 3.3% 9.8% 7.2% 0.0%

11/6/2013 23

  • No pure Indium remaining.
  • Region B is ideal Ag2In alloy ratio.
  • Region C, D & E some Cu, so less

Indium available for Ag alloying.

  • Cu is probably a remnant of seed

layer removal by sputtering. Wet etch next time!

  • Nickel barrier (F) shows no diffusion
  • f In, Ag, or Cu.
  • Region A is still 96% Ag, indicating a

depletion of In for alloying.

  • Take-aways:

– Indium prefers Cu to Ag for alloying. – Cu ties up Indium efficiently – must eliminate from bonding region. – Cu/In intermetallic is reported as fragile – may explain signs of voiding/cracking at original bond interface.

slide-24
SLIDE 24

RT Bond, A.P., 200C 30 min Anneal Cross-section and EDS

A B C D E F Ag 95.7% 67.6% 58.8% 25.1% 33.0% 0.0% In 4.3% 32.4% 37.9% 65.1% 59.0% 0.0% Ni

  • 100.0%

Cu 0.0% 0.0% 3.3% 9.8% 7.2% 0.0%

11/6/2013 24

  • No pure Indium remaining.
  • Region B is ideal Ag2In alloy ratio.
  • Region C, D & E some Cu, so less

Indium available for Ag alloying.

  • Cu is probably a remnant of seed

layer removal by sputtering. Wet etch next time!

  • Nickel barrier (F) shows no diffusion
  • f In, Ag, or Cu.
  • Region A is still 96% Ag, indicating a

depletion of In for alloying.

  • Take-aways:

– Indium prefers Cu to Ag for alloying. – Cu ties up Indium efficiently – must eliminate from bonding region. – Cu/In intermetallic is reported as fragile – may explain signs of voiding/cracking at original bond interface.

slide-25
SLIDE 25

RT Bond, A.P., 200C 30 min Anneal Cross-section and EDS

  • No pure Indium remaining.
  • Region B is ideal Ag2In alloy ratio.
  • Region C, D & E some Cu, so less

Indium available for Ag alloying.

  • Cu is probably a remnant of seed

layer removal by sputtering. Wet etch next time!

  • Nickel barrier (F) shows no diffusion
  • f In, Ag, or Cu.
  • Region A is still 96% Ag, indicating a

depletion of In for alloying.

  • Take-aways:

– Indium prefers Cu to Ag for alloying. – Cu ties up Indium efficiently – must eliminate from bonding region. – Cu/In intermetallic is reported as fragile – may explain signs of voiding/cracking at original bond interface.

A B C D E F Ag 95.7% 67.6% 58.8% 25.1% 33.0% 0.0% In 4.3% 32.4% 37.9% 65.1% 59.0% 0.0% Ni

  • 100.0%

Cu 0.0% 0.0% 3.3% 9.8% 7.2% 0.0%

11/6/2013 25

slide-26
SLIDE 26

RT Bond, A.P., 200C 30 min Anneal Cross-section and EDS

A B C D E F Ag 95.7% 67.6% 58.8% 25.1% 33.0% 0.0% In 4.3% 32.4% 37.9% 65.1% 59.0% 0.0% Ni

  • 100.0%

Cu 0.0% 0.0% 3.3% 9.8% 7.2% 0.0%

11/6/2013 26

  • No pure Indium remaining.
  • Region B is ideal Ag2In alloy ratio.
  • Region C, D & E some Cu, so less

Indium available for Ag alloying.

  • Cu is probably a remnant of seed

layer removal by sputtering. Wet etch next time!

  • Nickel barrier (F) shows no diffusion
  • f In, Ag, or Cu.
  • Region A is still 96% Ag, indicating a

depletion of In for alloying.

  • Take-aways:

– Indium prefers Cu to Ag for alloying. – Cu ties up Indium efficiently – must eliminate from bonding region. – Cu/In intermetallic is reported as fragile – may explain signs of voiding/cracking at original bond interface.

slide-27
SLIDE 27

RT Bond, A.P., 200C 30 min Anneal Cross-section and EDS

A B C D E F Ag 95.7% 67.6% 58.8% 25.1% 33.0% 0.0% In 4.3% 32.4% 37.9% 65.1% 59.0% 0.0% Ni

  • 100.0%

Cu 0.0% 0.0% 3.3% 9.8% 7.2% 0.0%

11/6/2013 27

  • No pure Indium remaining.
  • Region B is ideal Ag2In alloy ratio.
  • Region C, D & E some Cu, so less

Indium available for Ag alloying.

  • Cu is probably a remnant of seed

layer removal by sputtering. Wet etch next time!

  • Nickel barrier (F) shows no diffusion
  • f In, Ag, or Cu.
  • Region A is still 96% Ag, indicating a

depletion of In for alloying.

  • Take-aways:

– Indium prefers Cu to Ag for alloying. – Cu ties up Indium efficiently – must eliminate from bonding region. – Cu/In intermetallic is reported as fragile – may explain signs of voiding/cracking at original bond interface.

slide-28
SLIDE 28

RT Bond, A.P., 200C 30 min Anneal Cross-section and EDS

A B C D E F Ag 95.7% 67.6% 58.8% 25.1% 33.0% 0.0% In 4.3% 32.4% 37.9% 65.1% 59.0% 0.0% Ni

  • 100.0%

Cu 0.0% 0.0% 3.3% 9.8% 7.2% 0.0%

11/6/2013 28

  • No pure Indium remaining.
  • Region B is ideal Ag2In alloy ratio.
  • Region C, D & E some Cu, so less

Indium available for Ag alloying.

  • Cu is probably a remnant of seed

layer removal by sputtering. Wet etch next time!

  • Nickel barrier (F) shows no diffusion
  • f In, Ag, or Cu.
  • Region A is still 96% Ag, indicating a

depletion of In for alloying.

  • Take-aways:

– Indium prefers Cu to Ag for alloying. – Cu ties up Indium efficiently – must eliminate from bonding region. – Cu/In intermetallic is reported as fragile – may explain signs of voiding/cracking at original bond interface.

slide-29
SLIDE 29

Makeshift Structure To Avoid Cu Contamination - Replace Ag-Bumped Sub With Ag Planar Coupon

Si Cu Ni In Si Ni Ag Planar Silver coupon Cu base Ni Barrier Plated Ag Bond In-Bumped Chip To Ag coupon Si Ni In Cu contam. Sputtered Cu “Jacket”

11/6/2013 29

slide-30
SLIDE 30

Indium Chip To Silver Coupon; AP Prep RT Bond; Anneal (no force) 30 min/135

High force shear

11/6/2013 30

InAg alloy separated in bulk

slide-31
SLIDE 31

Electrical Continuity Testing

  • 256 daisy chain strings per chip.
  • 1260 bumps in each string.
  • Samples potentially compromised by Cu

contamination. Anneal Temp Ramp up time Hold Time Avg. Ω/bump Yield to

  • pens

Yield to shorts Shear/ mil spec 135C 20 sec 600 Sec 0.248 98% 98% 242% 190C 60 sec 90 sec 0.108 93% 96% 172% 190C 240 sec 90 sec 0.084 100% 98% no test

  • Increased anneal time/temp appears to improve bump conductance.
  • Anneal above Indium melt temp does not seem to affect opens or shorts.
  • Limited data suggests capability for low resistance, high yield contact.

11/6/2013 31

slide-32
SLIDE 32

Shear Testing

  • Bonded pairs were shear-tested in accordance with MIL-STD-883 which

specifies die shear strength for this size die as 5.0 kg.

  • Although shear data is limited, shear strengths on all samples

measured did easily exceed the MIL-STD requirement.

  • Shear strength is expected to improve when Cu is kept out of bond

zone.

  • The current data suggest that this bond scheme is capable of robust

mechanical performance.

Anneal Temp Ramp up time Hold Time Shear Strength (Kg) Shear/ MIL-STD 135C 20 sec 600 Sec 12.1 242% 190C 60 sec 90 sec 8.6 172%

11/6/2013 32

slide-33
SLIDE 33

Surface Activation for Capillary Underfill

Die surfaces are not naturally wetting. Contact angle ~50-70 De-oxidizing Atmospheric Plasma also activates die surfaces for enhanced CUF. Contact angle <10

11/6/2013 33

slide-34
SLIDE 34

Conclusions

  • AgIn system is capable of high speed, low force, room

temperature bonding.

  • 3DIC stacking at room temperature has significant benefits.
  • Metallurgy is capable of MIL-STD mechanical stability following

solid-state alloy anneal.

  • Copper participates aggressively in Indium metallurgy – keep

isolated.

  • Nickel appears to be a suitable barrier layer to isolate Cu from

Ag and In.

  • Atmospheric Plasma enables fluxless instant RT bonding of In-

to-Ag bumps and enhanced wicking of capillary underfills.

  • These preliminary results for InAg binary bonding are very

encouraging, and warrant further investigation.

11/6/2013 34

slide-35
SLIDE 35

Future Plans

  • Fabricate new test chips confining the Cu to the

interconnect layer.

  • Characterize the interdiffusion mechanisms of the

Ag/In binary system for small bump volumes.

  • Characterize series resistance, shear, and high-

temperature stability of the Ag/In binary system.

  • Demonstrate multi-chip 3D stacking and subsequent

underfill and reflow with the Ag/In binary system.

  • Cultivate industrial partnerships to develop and

implement this technology.

11/6/2013 35

slide-36
SLIDE 36

Acknowledgements

  • The authors would like to thank Professor C.C.

Lee and his students at U.C. Irvine. Their pioneering work in Ag/In metallurgy and processes provided inspiration for this development project.

  • Fabrication, testing, and analytical work was

performed and supported by RTI International CMET, internal funding.

  • Surface preparation, bonding, and assembly

process development was performed and supported by SETNA Corp.

11/6/2013 36