PROCESS AND EQUIPMENT ENHANCEMENTS FOR C2W BONDING IN A 3D INTEGRATION SCHEME Keith A. Cooper, Michael D. Stead SET- North America Daniel Pascual, Sematech Gilbert Lecarpentier, Jean-Stephane Mottet SET SAS

Outline 

  



Introduction and Statement of Challenges Hybrid Bonding with Collapsible Polymer Particulate Improvement Study Removal of Oxides prior to Bonding Conclusions and Further Work

3D Definitions  3D-Packaging: Traditional

 

packaging interconnect technologies, including package stacking, wire bonding 3D-WLP: Wafer-level packaging, where interconnects processed post-IC passivation 3D-IC: IC technology, where 3D interconnects processed at the local level* * Adapted from Huyghebaert, Soussan, et al. IMEC. ECTC 2010

Via “middle” Cu process W-via or contact plug

Start: After W CMP 1. TSV litho (I-line)

2. TSV etch, strip & clean

3. O3-TEOS liner

After TSV: BEOL or passivation + sintering

4. Ta/Cu Cu seed

5. Cu electroplating/anneal

6. Cu/barrier/ liner CMP

Adapted from Huyghebaert, Soussan, et al. IMEC. ECTC 2010

Attractions of Cu TSV Fill 

  



Familiarity of Processing Mechanical Strength Electrical Integrity Scalability of Copper Cost

Die-to-Wafer (D2W) Bonding DIE-TO-WAFER  Lower Throughput  Single Chip Placement  Long Bond Processes  High Yield  Known Good Die  Good Overlay  Flexibility  Component and wafer sizes  Different Technologies

Heterogeneity!

Challenges with Cu–Cu bonding 

 

Bond requires high temp, long process Flat, particle-free surfaces Oxides 



Cu oxidizes at STP, oxidizes rapidly at elevated temperatures Metal oxides inhibit mechanical and electrical integrity

Outline     

Introduction and Statement of Challenges Hybrid Bonding with Collapsible Polymer Particle Improvement Study Removal of Oxides prior to Bonding Conclusions and Further Work

Collective Hybrid Bonding Cost-effective processing by segmentation of 3D assembly into D2W + Collective Bonding 

Pick-and-place tool Patterned dielectric glue

Wafer-level bonding tool TSV-die

Landing wafer

Die pick and place



Landing wafer

Collective bonding

Combines:  High Yield and flexibility of D2W  High Speed and efficiency of parallel process

Temp.

 Landing wafer sees only one

temperature cycle  Accuracy depends upon several process steps

Collective bonding @ wafer level Bonding, polymer cure

Wafer population @ wafer level

Polymer Reflow

 Higher throughput

time

die n

Collective D2W bonding

Metal bonding Die n

controlled by the bonder  Time consuming  Landing wafer sees several bonding T-cycles

Metal bonding Die 2

Temp.

die 2

 High Accuracy capability,

Pick & place: die 1

Sequential D2W bonding

Metal bonding Die 1

In-Situ vs. Collective Bonding Temperature Profile

LT

time

2-Step Cu-Cu Direct Bond* 

Advantages  Low temp and force attachment process  Strong initial bond maintains alignment for collective bond step



Challenges  Very planar, clean, smooth surfaces  Long diffusion process  Very clean bonding environment

Bond evolution with annealing Diffusion cones

T-Shape Triple junctions

Triple junctions at equilibrium

Direct Metallic Bond after annealing (2h @ 400C)

*Source: CNRS-CEMES and CEA-LETI

Tack/Collective Bond Overview 2.

1.

Heat + Force

Bonding Plate

N2 Environment

4.

3.

1. 2. 3. 4.

TSV wafer with bond and probe pads Spin coat thin layer of sacrificial adhesive Tack dice individually using die bonder tool Apply heat/force to decompose the adhesive and bond all dice in parallel using wafer bonding tool

Die Tacking Results



Tack dice onto wafer 



 

Align each die to bond site on 300 mm wafer Place die onto wafer and apply force at low temperature (~135 C) Repeat tack process to populate wafer 2.5 μm average placement accuracy observed

Source: Sematech

Alignment Shift From Collective Bonding 11 10 9 8 7 6 5 4 3 2 1 0 -11

-10

-9

-8

-7

-6

-5

-4

-3

-2

-1

0 -1

1

2

3

4

5

6

7

8

9

10

11

Misalignment vector map 300 mm wafer 1unit = 1μm

-2 -3 -4 -5 -6 -7 -8 -9 -10 -11

• Optimized Tooling and Process • Alignment improved to 2 μm (average = 0.8 um) • No damage to tooling

FIB-SEM Sectional Image

• Diffusion of Cu across bonding interface

Particles at Cu-Cu interface were major source of yield loss

Outline     

Introduction and Statement of Challenges Hybrid Bonding with Collapsible Polymer Particle Improvement Study Removal of Oxides prior to Bonding Conclusions and Further Work

Schematic of Cu-Cu Bonding

Areas of Particulate Contributors

Particle Reductions

Performed in the framework of PROCEED project funded by French authorities and by European authorities (FEDER). PROCEED partners are: ALES, CEA LETI, STMicroelectronics, CNRS-CEMES and SET.

Sample Modifications  Stages and guides of low-

particulate materials  Teflon Cable channels  Enclosures around specific assemblies to exhaust any particles generated locally

TEFLON Cable channels

Particle collection

After Particulate Improvements 

 

Particle counts reduced by 2-3 orders of magnitude Alignment improved to ± 1μm Tight distribution of daisy chain contact resistance

Outline     

Introduction and Statement of Challenges Hybrid Bonding with Collapsible Polymer Particle Improvement Study Removal of Oxides Conclusions and Further Work

Requirements of Oxide Removal Process 

  

 

Rapid and effective Inert to surrounding materials Minimal or no residue EHS Compliant Long-lasting Low-cost

Historical Methods of Reducing Oxides 

 

Wet acid dips, e.g. HCl, citric acid Liquid or paste fluxes Vacuum plasma treatments

In-situ Removal of oxides 



Description Schematic of In-situ reduction

Reduction Chamber Hardware 



2 versions – D2D and D2W Photos of micro-chamber D2D version:

View of Chuck

View of Bonding Arm

Proposal: Novel Ex-situ Removal of Oxides 

  

Dry process at atmospheric ambient Non-toxic, non-corrosive chemistry Rapid turnaround (< 1 minute) Reduces oxide from metal surfaces and passivates surface against re-oxidation

Ellipsometry   

Change in polarization defined by Δ = phase change of reflected light Δ indicates morphology or composition

Ellipsometry of In ► Ellipsometry confirms oxide removal

SETNA/SET Proprietary

Results with Indium Bumps 

Untreated Indium  



No adhesion Bumps were coined

Treated Indium  

Good adhesion Good “taffy pull”

Process Validated for Indium Validated for: 



    

Indium-to-Indium Indium-to-metal contact pads

Room temp bonding process Strong bump-to-bump adhesion Perfect tensile rupture with pull test Demonstrated for Indium-to-Nickel Demonstrated for Indium-to-Titanium

SETNA/SET Proprietary

Process Validated for In alloys Validated for Indium alloy-to-metal contact pads

    

Room temp and elevated temp bonding Strong bump-to-pad adhesion MP > In, depending on composition Demonstrated for In alloy to Ni or Ti Projected to work on Sn and Ag solders

Protection from Re-oxidation Passivated Indium surface remains stable after 50 hours

SETNA/SET Proprietary

Application-Specific Metallurgy Surface prep process shows promise for a broad range of metals and alloys:         

Indium* Indium alloys* Titanium* Nickel* Copper** Silver ** Tin** Aluminum** SnAg**

*Demonstrated with bonding tests **Ellipsometry results are promising, no bonding tests yet

Summary 



3D-IC Integration opportunity is expanding, good process flow options Technical hurdles addressed:   

Throughput – Hybrid Polymer Bonding Yield – Particulate Reduction Materials – Oxide Removal Options

Areas for further study 

Further Work foreseen:  Characterization of ex-situ oxide reduction process  Further exploration of Collective Hybrid Bonding

Thanks for your attention

For further info, please contact:   

www.set-sas.fr

[email protected] [email protected] +33 4 50 35 83 92