WAFER-BOND MICRO-VOID PHYSICAL ANALYSIS
C. Cassidy1, H. Plank2, T. Ganner2, L.G.W. Tvedt3, C. Gspan2, M. Dienstleder2, J. Wagner2, M. Krause4, C. Patzig4, S. Brand4, B. Böttge4, J. Siegert1
1austriamicrosystems
AG 2FELMI-ZFE (Graz, Austria) 3SINTEF ICT, Microsystems and Nanotechnology (Oslo, Norway) 4Fraunhofer-IWM (Halle, Germany)
Purpose • Share a 3D-integration analysis problem • Highlight real-life practical problems & difficulties • Get feedback & ideas…
2
Outline 1. Introduction to bonding & “micro-voids” 2. PFA Successes 3. PFA challenges & problems 4. Summary, current status 5. Workshop question
3
Technology overview Solder balls
TSVs
- CMOS processing - Wafer bonding - TSV (unfilled “via last”) - Finish CMOS/bumping
Wafer bond interface
Wachmann et al., ICMAT 2011
4
Wafer Bonding • Bond process:
EVG Gemini toolset
– Plasma-activation – Direct bonding (Si/SiO2) – Low temperature anneal – CSAM Pre-bonding
• Bond surfaces – 1) Polished Si wafer – 2) Deposited oxide layer, CMP finish (lower) Post-bonding Kraft et al., ECTC 2011
5
What is a “micro-void”? Wafer-level CSAM detects circular features in the bond interface. Specific size distribution (diameter ~100µm) CSAM
Micro-void
TSVs
“upper wafer”
“Micro-void”
TSV
Bond oxide
Wafer-level Scanning acoustic microscopy
“lower” wafer
Bond Interface
6
Micro-void questions CSAM
– What is the root cause? – Reliability risk or cosmetic defect?
Fig. A
– Do they move or grow? – Interrupted TSV connection? – Fracture/delam initiation point?
TSVs Fig. B
Fig. C 7
Micro-void physical analysis
8
Physical analysis The interface is not easy to access
Bond interface
Accessing the bond interface 9
Sample preparation approach 1. Si Etch removal, oxide underside as etch stop Lower Si removed using wet/dry etch
PRO: Defect is visible and accessible to PFA techniques CON: Defect might be changed by etch/release process
“upper wafer” Bond oxide
Lower wafer selectively etched away
Bond interface
Optical microscopy
10
Optical microscopy (after Si etch removal)
CSAM
Delamination/ voids/ bond defects
“upper wafer” Bond oxide
3
Optical microscopy
2
1
11
Optical microscopy; 3 distinct zones can be discerned 11
Cross-section & in situ optical microscopy Defects appear to consist of… - A. outer debonded area - B. sealing ring (bonded/ contaminant) - C. internal debonded area
1.
1. Defect is undisturbed
2.
3.
2. Outer delaminated region is disturbed. Trapped gas escapes, unbonded oxide at the polishing front fractures off, following the contour of the ring 12
3. As 2. Polishing front is now right “in” the ring region, but has not compromised it yet.
A
4.
4. The ring is “opened” by the polishing front, allowing the gas inside to escape.
B
C
5.
5. None of the oxide at the polishing front in the vicinity of the defect is now bonded, and fractures and rips off easily during polishing
12
Is there a foreign contaminant? - ToF-SIMS depth profile (F-IWM), through bond interface - No foreign elements/compounds could be detected
Defect 1
Before ToF depth profile
After ToF depth profile 13
Void collapse example (OM)
Pre-collapse
Post-collapse
- Voids collapse after e.g. ultrasonic cleaning, mechanical, pressure, vacuum exposure (SEM/FIB), time… - Complicates PFA attempts 14
What topograpy is associated with the defect? AFM on the oxide surface (ZFE), after collapse
Post-collapse (defocus) 8nm ridge
10nm recess
Ultrasonic agitation to collapse the void
15
TEM & analysis (bond oxide still in place) What are the recess and ridge composed of?
Bond oxide
Si substrate
Successful microvoid TEM lamella preparation
16
TEM Sample prep Pt
2.8 nm
Thick (5-8nm) oxide layer? 5
Si
n m
Bond oxide
Bond oxide
EFTEM shows Si and O
EFTEM shows Pt EFTEM shows Si and O
5
Si
n m
Reference area
Si 5
n
m
Defect area
Typically 1-2nm native oxide expected. e.g. Suni et al, J. ECS, 2002
17
Sample preparation approach 2. (Assisted) Debonding
A
B PRO: Bond interface is exposed for surface analysis CON: Unreliable debonding, possible material transfer, defect navigation, contamination
C 18
White light interferometry (debonded Si surface) WLI (Sintef) identified a slight recess (~-10nm, blue in the image below), with an enclosing raised ring (~+15nm, red in the image below).
19
AFM (debonded surfaces, ZFE) Topographical surface analysis on both surfaces… Polished backside Si surface
Polished oxide surface AFM artifact
No significant topography
Problem stems from topography on polished Si surface.
Polished oxide surface is OK!
Summary of useful PFA results • Micro-voids are caused by existing surface topography. • This topography is uniquely associated with the polished Si wafer. • ToF-SIMS indicates that no foreign species are present. • TEM identifies a suspiciously thick oxide layer (5-8nm) in the defect region 21
Pending • Further defect TEMs • Reference, pre-bonding TEMs (surface polishing, cleans, plasma activation) – Characterization of surface, particularly oxidation
• Elemental analysis on debonded defects 22
PFA Problems
23
1. Targeting/navigation - Buried, featureless surfaces, disappearing voids
Optical
Laser marks
24
2. FIB/SEM problems • Cross-sections: Nothing at the interface – A) Void collapsed? – B) Insufficient resolution/lack of contrast mechanism
Micro-void
Oxide
Deposite d FIB markers
Si
Bond interface
25
- Lamella fracture, owing to unbonded/weakly bonded surfaces
Oxide
Fracture at micro-void area
1
Bond interface
µ m
3.TEM problems
Si
26
3.TEM problems “reweld”
after 19 seconds (in the electron beam)
2 0
2 0
n m
1
µ m
n m
Bright Field 27
4. ToF-SIMS problems • Spikes (+ve, -ve) in ToF depth profiles • Assumption is instability of defect, opening of void, “explosions” of trapped gas +ve spikes 5
10
4
4
10
10
3
3
10
10
2
10
2
1
10
0
10
10
OHIntensity (counts)
Intensity (counts)
C-
5
10
FSiO2ClSiOBr-
1
10
0
10
200
400
600
800 Time (s)
All counts fall to zero as bond interface is exposed
28
Process improvement/reliability assessment
29
Strategy 1: Elimination • No progress to date – Physical root cause – Identification of processing step(s) – Process improvement
30
Strategy 2: Reliability assessment
(A)
• Accelerated stresses – Micro-voids are stable (A)
• Micro-manipulation
(B)
– Trapped bubbles can’t be “micro-maneuvered” beyond their original boundaries (B)
• Bond strength tests – Micro-void presence has no detectable effect on fracture mode or fracture stress values (C)
(C) 31
Conclusions • Micro-void origin – Micro-voids are caused by nm-scale local topography on the backside of the polished bond surface, perhaps caused by local oxide thickness variations. – The root cause not yet found. – Reliability data: Do not pose a reliability risk
• PFA of wafer-bond interface defects is challenging – – – –
Buried defects, navigation/targeting Instability of defects during FA (trapped bubbles) Aspect ratio (defect, defect to enclosing material) High bond strength
32
Thank you for your attention Workshop question…
33
Workshop question Until now, we always had to bond the wafers to find the defect in CSAM. How could we find the flaws on the surface prior to bonding? 34
Backup
35
What topograpy is associated with the defect? AFM on the oxide surface (ZFE)
Micro-void
40nm air bubble
36
AFM (oxide surface)
Pre-collapse
40nm air bubble
Post-collapse (defocus) 8nm ridge
10nm recess 37
Strategies
1. Prove that there is no reliability risk
2. Eliminate microvoids
a) Understand the mechanism b) Gather reliability data c) Convince customers
a) Understand the mechanism b) Process improvement
Both approaches require physical analysis to determine the root cause.
38
Probe needle
Exposed oxide: micro-manipulation • Defect oxide cap is very elastic – Voids disappear with gentle pressure – Re-appear when pressure removed
Probe needle
• Collapse – With extensive force and scratching, voids can be permanently collapsed
• Movement – The debonded area/bubble cannot be moved, manipulated or increased beyond its original area
TSV
39
4. ToF-SIMS problems • Si etch process (sample prep) shows some nonuniformity and consumption of oxide (A) • Large sputter volume, thin defect layer - 0.9-2.2µm oxide sputter, 15nm defect • Interpretation of depth profiles was difficult (B)
Sputtering direction ILD0 Bond oxide
(A)
(B)
40
1. Targeting/navigation • If debonding the surfaces, no surface defect can be seen afterwards. • No navigation guides.
Intact oxide
Oxide etched (HF) Debonded surface, oxide side
Debonded surface, Si side
The defect cannot be seen 41
1. Targeting/navigation • Invisible in electron/ion microscopy. • Laser marking is possible, but often void disappears
Laser marks
Void has disappeared
42
5. White light interferometry • Excellent for Si surface • Not possible for transparent BO surface => Large area AFM a better option
WLI: Si surface 43