High throughput cross sectioning using plasma FIB milling and laser ablation 3D-IC workshop CNES, Toulouse November 28-29, 2011 L. Kwakman, G. Franz, M. Straw
Confidential
Outline • Introduction to Plasma-FIB technology • Application results • Vion Plasma-FIB product launch • Expanding the application space: combined laser ablation & FIB milling workflows • Conclusions 2
Confidential
3D IC integration and packaging
Courtesy of FhG EMFT, Sintef, Infineon
“More than Moore” : smart, integrated & heterogeneous systems! Delivered in smart, advanced and reliable packages….
3
Confidential
3D IC integration and packaging Fabrication of stacked and vertically interconnected device layers
Au wires
TSV’s
μ-Bumps
4
Confidential
3D device analysis challenges • For “Chip Access” large volume material removal is needed within reasonable time and at precise locations… • The zone of interest has a Field of View of ~ 100 x 100 μm, and has to be inspected with nm resolution…. We need to think big while keeping an eye for details!
More Moore: Typical volumes: 10 x 10 x 10 μm3 Standard FIB mill rate < 5 μm3 /s Total Mill time ~ 3 - 10 minutes 5
Confidential
More than Moore: Typical volumes: 100 x 100 x 100 μm3 Standard FIB mill rate < 5 μm3 /s Total Mill time ~ 50 hrs!
3D IC interconnect technologies
6
Confidential
Inductively Coupled Plasma Ion Source Broad source with high angular intensity Angular Intensity = 50 mA/sr Virtual source = 15 μm Brightness ~1x104 Am-2sr-1V-1
LMIS FIB: ~ 60 nA 7
Confidential
Plasma-FIB: ~ 1 μA
Advantage over LMIS Ga Source
8
Confidential
Xe instead of Ga, high sputter yield
High volume milling / high beam current
High brightness, low energy spread
Ga-FIB loses size advantage above 60 nA (spherical aberrations)
Plasma-FIB system performance
9
Confidential
FIB imaging resolution: ~ 30 nm at 20 pA
FIB milling rate @ 1.0 μA: ~ 300 μm3/s
FEI’s plasma-FIB activities so far
• Plasma-FIB source & prototype system developed in 2009-2010 • Plasma-FIB Beta tool installed at FhG Munich in Q3 2010 • Plasma-FIB evaluated in ENIAC JEMSIP-3D project context • Plasma-FIB demonstrated on many customer samples (demo’s) 10
Confidential
Plasma-FIB in action:
APPLICATION RESULTS
11
Confidential
Reduced curtaining: automated sample rocking
After regular FIB mill
After sample rocking based FIB mill
FIB induced curtaining has been successfully suppressed by an automated routine in which the sample is rocked during milling. 12
Confidential
Stacked device: TSV and SLID bond technology courtesy of JEMSIP3D project partner FhG-EMFT
200 μm top layer face down
50 μm middle layer double sided, TSV
720 μm bottom layer Face up
~ 150 min PFIB mill
Total view of 3D-stack (without TSV) 13
Confidential
Detailed view of TSV
Detailed view of SLID metallurgy
Stacked device: TSV and SLID bond technology courtesy of JEMSIP3D project partner FhG-EMFT
Third pad layer (bottom of 2nd layer)
Reversed engineering: successful deprocessing combining plasma-FIB and XeF2 chemistry: bond overlay errors, free-standing TSV’s etc.
14
Confidential
Conductive adhesive for WtW bonding courtesy of JEMSIP3D project partner SINTEF
Metal coated polymer spheres
Electrical contacts
~ 30 min PFIB mill
Bulk mill after cleavage ~ 200x50x600 μm3 15
Confidential
Curtain-free fine mill of w2w bonding region
Zoomed-in PFIB images of conductive spheres
BGA package solder bump, MEMS device
BGA solder bump: ~300 μm cross section ~60 min @ 1.3 μA (No sample rocking used to reduce curtaining)
MEMS access: 7.1×105 μm3 volume of Si was removed to create an access hole. ~ 18 minutes @ 1.5 μA
16
Confidential
Some customer demo results courtesy of ST-Ericsson Physical Analysis Lab Grenoble
Copper wire analysis on 2-device stack • PFIB @ 74 nA/27 nA for 12 min. • ST-E standard: polishing + Ga FIB mill: 6-8 hrs • important curtaining observed as no beam rocking routine has been applied. Silicon trenching for FIB Edit in FCBGA • PFIB @ 1.3 μA for 5 hrs. • 600x300x40 μm3 Î 500 μm3/s • Flat bottom reveals non-optimized process (OL?) 17
Confidential
Some customer demo results courtesy of ST-Ericsson Physical Analysis Lab Grenoble
PFIB X-section of CSP Ball • PFIB @ 1.3 μA, 350x150 μm2 for 53 min. • CCS PFIB @ 570 – 230 - 74 nA, ~350 x 20 μm2 for total 35 min. • Away from CSP Ball the surface appears damaged and curtaining exists Requires protective metal layer and/or sample rocking 18
Confidential
Some customer demo results courtesy of ST-Ericsson Physical Analysis Lab Grenoble
RDL Cu lines PFIB edit on FOWLP • PFIB navigation + marker box at 7nA • PFIB @ 1.3 μA, 20x680 μm2 for 29 min. • Real time monitor nor stage current showed any “endpointing” • important damage of top polymer layer observed • FIB/SEM X-section shows significant overmill Î device =dead FIB based navigation creates damage of sensitive layers… FIB edit needs more application development: mill rates/endpointing methods 19
Confidential
Some customer demo results courtesy of Infineon Central Lab, Munich Cu pillar
20
t = 24 min
t = 32 min
t = 32 min
t = 60 min
t = 100 min
t = 100 min
Confidential
Some customer demo results courtesy of Infineon Central Lab, Munich
t = 30 min
t = 35 min
t = 45 min
Backside metalization
Void
t = 12 min 21
Confidential
t = 35 min
t = 45 min
The Plasma-FIB product launch:
VION PLASMA-FIB SPECIFICATIONS
22
Confidential
Official Vion Plasma-FIB product launch
Market introduction in June 2011
23
Confidential
Technology Deployment to Plasma FIB V600 • General Platform • 6” stage • IR camera (option)
FIB/DB Developments • Tomahawk base column •iFAST automation (option)
New Technologies • Productized Plasma Source
24
Confidential
2006
2009
2010
System Beta Q4‐2010 Mfg Q2‐2011
Vion System Specifications Plasma ion column
Milling
Stage
Landing Energy Range Max Beam Current Resolution at 30kV at Eucentric WD on Graphite (from Image.exe 35/65) Primary Milling Species Maximum FOV Number of aperture positions Aperture Lifetime Working Distance Si mill speed (top down milling, no gas enhancement)
2-30 keV >1 µA 20 pA < 30 nm 1 µA < 2.3 µm
Type
6” Piezo stage: 5-axis, eucentric tilt
Maximum Sample Size
150 mm diameter with full rotation (larger samples possible with limited rotation)
Maximum Weight Options Compatible with existing SDB/FIB options Beam chemistry GIS Support 25
Confidential
Xe 750 um square 24 >1 month 16.5 mm >20x faster than current Ga columns. 3
Xe: 25,000 µm /min at 1 µA 3 Ga equivalent is 1000 µm /min at 65 nA
500 g including sample holder e.g. IR camera, NavCam, CadNav,… Compatible w/standard FEI GIS
Expanding the application space
LASERS & PLASMA-FIB
26
Confidential
Customer demonstrations • Several demonstrations have been carried out for Asian, US and EU customers, a lot of market interest has developed over last months, demonstrations cover: • TSV analysis • Bond analysis (Wire bonds, BGA, u-bumps, Cu pillars) • (Buried) Bond analysis in stacked devices • PCB copper wire analysis, PCB level circuit edit
customers appreciate: significant analysis time gains, good Xsection and image quality Plasma-FIB is considered very valuable for 3D IC technology analysis and for certain packaging related analyses but..Plasma-FIB is not fast enough to do large area chip access, a combined workflow of laser ablation + plasma-FIB milling can reduce the bulk material removal time.
27
Confidential
Practical limitations of Plasma-FIB
The large hole 1.5 mm deep:
Circuit layer
• 7 μA Xe FIB: ~15 hours
Use case = analysis of bumps in packaged flip chip, target sample prep < 3 hrs: • Over-Mould = 200-400 μm, Silicon = 50-100 μm thick, Bump diameters 40-100 μm • Best case scenario = 300 μm deep mill Î minimum = 300 x 300 x 100 μm3 • Worst case scenario = 600 μm deep mill Î minimum = 600 x 600 x 100 μm3 • Assume average mill rate @ 1.3 μA: ~ 300 μm3/s
Access to bumps takes ~ 8 hrs to 65 hrs 28
Confidential
A combined laser & plasma-FIB workflow • Laser ablation rates are > 105 μm3/s Î very fast removal! • ns-lasers generate damage (HAZ) Î unpractical for FA! • Removal of HAZ (several μm) takes too long for standard Ga-FIB Plasma-FIB offers the unique capability to remove the laser induced damage zone very fast…
• Possible scenario 1: ns-laser + Plasma-FIB clean-up + X-section • Possible scenario 2: a-thermal fs-laser + Plasma-FIB X-section
29
Confidential
A combined laser & plasma-FIB workflow Laser ablation of large hole (blue is HAZ) Plasma FIB box mill + S&V Mould Silicon active
PCB Laser ablation: ~600 x 600 x 400 μm3 @ 105 μm3/s Î ~ 25 minutes PFIB X-sectioning: ~ 100 x 100 x 100 μm3 @ 300 μm3/s Î ~ 55 minutes LMIS Ga FIB X- sectioning: ~ 100 x 100 x 100 μm3 @ 15 μm3/s Î ~ 18 hrs 30
Confidential
Example of laser + plasma-FIB analysis
350 μm SiC substrate Au-Sn SLID bonded to 2nd substrate • Laser ablation of top ~ 300 μm of SiC substrate • Plasma-FIB cross section ~ 100x100x100 μm3 Remaining thickness of Substrate 1 after laser ablation Bond region
Substrate 2
31
Confidential
Example of fs-laser Slice & View Proof of concept: Integrated fs-laser-SEM system
fs-Laser ablation with 200 nm slice steps @ 1 Khz, ~ 3000 μm3 /s
• • 32
~ 6 hrs total, ~ 2 hrs for laser ablation (25 .106 μm3 removed)
Confidential
Example of fs-laser Slice & View Proof of concept: Integrated fs-laser-SEM system
fs-Laser ablation with 200 nm slice steps @ 1 Khz, ~ 3000 μm3 /s
• • 33
~ 6 hrs total, ~ 2 hrs for laser ablation (25 .106 μm3 removed)
Confidential
CONCLUSIONS
34
Confidential
Conclusions • After >6 months of extensive beta system testing FEI has launched a new product, the Vion plasma-FIB system • The plasma-FIB system significantly speeds-up sample prep and provides good resolution images • The system includes standard options such as GIS units, IR camera, optical microscope, CAD navigation and a new Software package, i-Fast, for automated recipe execution • For even more demanding applications such as depackaging, FEI is exploring workflows combining ultra short pulse lasers and plasma-FIB • Application development is ongoing to arrive at BKM’s and optimized recipes 35
Confidential