electromigration studies on 6 solid cu tsv (via …...menglu li1, prakash periasamy2, k. n. tu1, and...
TRANSCRIPT
Electromigration Studies on 6 �� Solid Cu TSV (Via last) in 32 nm SOI Technology
Prakash Periasamy1, Michael Iwatake2, Menglu Li3, Joyce Liu2, Troy Graves-Abe2, Thuy Tran Quinn2, Subramanian S. Iyer3
1 GLOBALFOUNDRIES INC, Malta, NY12020, 2 GLOBALFOUNDRIES INC, Hopewell Junction, NY12533 3 Center for Heterogeneous Integration and Performance Scaling, UCLA, Los Angeles, CA90024.
h Peeeeeeeeeeeeeeerrrrrrrrrrrrrrrrriiiiiiiiiiiiiiiiaaaaaaaaaaaaaaaaaasssssssssssssssssaaaaaaaaaaaaaaaaammmmmmmmmmmmmmmmmyyyyyyyyyyyyyyyyy111,,,,,,,,,,,,,,, aaaaaaaaaaaaaaeeeeeeeeeeeeeeeelllllllllllll IIIIIIIIIIIIIIIwwwwwwwwwwwwwaaaaaaaaaaaaaatttttttttttttakeMMMMMMMMMMMMMMMMiiiiiiiiiiiicccccccccccccccccchhhhhhhhhhhhhhhhhh 2, MeTrrrrrrrrrrrrrrrrroooooooooooooooooyyyyyyyyyyyyyyyyyy GGGGGGGGGGGGGGGGGrrrrrrrrrrrrrrrrraaaaaaaaaaaaaaaaavvvvvvvvvvvvvvvvvveeeeeeeeeeeeeeeeeeessssssssssssssssss----------AAAAAAAAAAAAAAAAAAAbbbbbbbbbbbbbbbbeeeeeeeeeeeeeeeeee222222222222222222,,,,,,,,, TTTTTTTTTTTTTTTTThhhhhhhhhhhhhhhhuy Tran Q
NNNNNNNNNNNNDDDDDDDDDDDDDDDDDDDRRRRRRRRRRRRRRRRRIIIIIIIIIIIIIIEEEEEEEEEEEEEEEESSSSSSSSSSSSSSSSS aaaaaaaaaaaaa,,,,,,,, IIIIIIIIIIIIIIIIINNNNNNNNNNNNNNNNNNCCCCCCCCCCCCCCCCC,,,,,,,, MMMMMMMMMMMMMMMMMaaaaaaaaaaaaaaaaallllllllllllllllttttttttttttttttttaaaaaaaaaaaaaa NNNNNNNNNNNNNY12DDDDRRRRRRRRRRRRRRRRRIIIIIIIIIIIIIIEEEEEEEEEEEEEEEEESSSSSSSSSSSSSSSSSSSSSDDDDDDDDDDDDRRRRRRRRRRRRIIIIIIIEEESSSSSSSSSSSSSSSSS eeeeeeeeeeeeeewwwweIIIIIIIIIIIIINNNNNNNNNNNNNNNNNCCCCCCCCCCCCCCCCC, HHHHHHHHHHHHHHHHHHoooooooooooooooooppppppppppppppppppeeeeeeeeeeeeeeeeIIIIIIIINNNNNNNNCCCCCCCCCCCCCCCC HHHHHHHHHHHHH
gennnnnnnnnnnneeeeeeeeeeeeeeeeeeooooooooooooooooouuuuuuuuuuuuuuuussssssssssssssssss IIIIIIIInnnnnnnnnnnnnnnntttttttttttttttttteeeeeeeeeg
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Outline
2
Background – Via Last TSV Integration
Problem Statement (What, Why and How?)
1
2
Experimental Details3
Results & Discussions4
Concluding Remarks5
ation
(What, Whhhhhyyyyy aaaaannnnnnnddddddd HHHHHHHHoooowwwwww????????))))))))
meeeeeennnnnnnnttttttaaaaaaaalllll Detaaaaaaaaiiiiiillllllsssss
Resssssssuuuuuuuullllllllttttttttssssssss &&&&&&&& DDDDDDDDDiiiiiisssssssssccccccccuuuuuuuusssssssssssssssssiooooooonnnns
udddddddiiiiiiiinnnnnnnggggggg RRRRRRRR
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Background: via-last TSV IntegrationNeed for via-last TSV integration
Elements/links of interest in this study
� TSV/Capture Level
� TSV itself
� TSV/Redistribution layer (RDL)
This study
6 µm solid Cu via-last TSV
What is the electromigration limited min/max. current for each of these links?
Iyer, Subramanian S. "Three-dimensional integration: An industry perspective." MRS Bulletin 40.03 (2015): 225-232
s/linkkkkkkkkkkksssssssssss ooooooooooofffffffffff iiiiiiiiiiinnnnnnnnnnnttttttttttteeeeeeeeerrrrrrrrrrreeeeeeeeeeesssssssssssttttttttttt iiiiiiiiiiinnnnnnnnnnn ttttttttttthhhhhhhhhhhiiiiiiiiiiiissssssssssss studddddddddyyyyyyyyyy
ptttttttttttuuuuuuuuuuurrrrrrrrrreeeeeeeeeee LLLLLLLLLLLeeeeeeeeeeevvvvvvvvvvveeeeeeeeeeellllllllll
bbuuuuuuuuutttttttttttiiiiiiiiiiiiooooooooooonnnnnnnnnnn lllllllllaaaaaaaaaaayyyyyyyyyyyeeeeeeeeeerrrrrrrr ((((((((((RRRRRRRRR
This
6 µm soliiiiidddddd Cu
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Problem StatementWhat?
Why?
How?
� What is the electromigration limited min/max. current for the links comprising of TSV, TSV/Capture Level and TSV/GrindsideRDL?
� Maximize the current loading (within the existing design) per TSV and thereby reduce the number of TSVs per chip if possible
� Aid designers and package engineers with electromigration related guidelines
� By making each of the elements (TSV, Capture Level, GrindsideRDL) in the link as the weakest in properly designed electromigration test structure to gain failure distribution
min/max. cuuuuuuuuuurreure Level aaaaaaaaaannnnnndddddd TTTTTTTTTSSSSSSSSSSV/
guuuuuuuuuuuiiiiiiiiiiidddddddddddeeeeeeeeeeelllllllllllliiiiiiiiiiinnnnnnnnnnneeeeeeeeeeeesssssssssss
ach o
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
6 µm solid Cu TSV (not drawn to scale)
25 µm wide capture level landing pad
redistribution level landing pad25 µm wide
Experimental Details – Test Structures (TSV/Capture Level)(a) single finger
(b) five fingers
Link of interest: TSV/Capture level. All other links are design robust
Two variations: (a) Single capture level finger vs (b) Five fingers
6 µm solid Cu u u u uu TSTSTSTSTTSTTTT V V V V V V VVV (nototototototot ddd dddddrawnwnwnnwnnnnwn t o oooooooo scscscscscscalalallalalalalalale)
redistribbbububbbbbbb titititititiiiioooonoo levelllll l l l l l l annnnnnnnndid ng padadadadaadaa255555555555 µµmµmµmµmmµmµmµmµmµmµm wwwwwwww wwididididididididdideeeeeeeeee
((bbbbbbbbbbb) fiv
rrrrrreeeeeeeeeeessssssssssssttttttttttt::::::::: TTTTTTTTTTTSSSSSSSSSSSVVVVVVVVVVV///////////CCCCCCCCCCCaaaaaaaaaaappppppppppptttttttttttuuuuuuuuuuurrrrrrrrrrrreeeeeeeeeeee llllllllllleeeeeeeeevel.kkkkkkkkkkksssssssssss aaaaaaaaaaarrrrrrrrrrreeeeeeeeeee dddddddddddeeeeeeeeeesssssssssssiiiiiiiiiiiggggggggggnnnnnnnnnnn rrrrrrrrrrrooooooooooobbbbbbbbbbbust
s: (((((((((aaaaaaaaaaaa))))))))))) SSSSSSSSSSSiiiiiiiiiiinnnnnnnnnnngggggggggggglllllllllleeeeeees (b) Fi
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Experimental Details – High Current Electromigration Oven
1
2
3
4
Current range: ± 2 µA to 4 ATemperature: 60°C to 400°C
ge: ±±±±±±±±±± 222222222 µµµµµµµµµµµAAAAAAAAAAAAA tttttttttttooooooooooo 44444444444 AAAAAAAAAAAure: 600000000000°°°°°°°CCCCCCC tttttttttttooooooooooo 44444444444000000000000000000000°CCCCCCCCCCC
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Experimental Details – TSV RIE and Fill Process
TSV post Metallization
Cu seed
TaN/Ta
TSV post etch
ost MMMMMMMMMeeeeeeeettttttttaaaaaaaaallllllllllllllllliiiiiiizzzzzzzzzaaaaaaaaattttttttioooooooonnnnnnnn
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Results & Discussions – TSV/Capture Level
Popup failures (rapid rise in resistance) �� five parallel capture level fingers connecting two TSVs
10% failure criteria -> chip that exhibits 10% increase in resistance than the baseline resistance at stress temperature is classified as failed chip.
hipchcreaincan ththssssssiiiiiisssssssssstanrrreeeeeemmmmmmppppppppppeeeeeeeeeeerratutttttteeeeee
s fffffffffaaaaaaaaaiiiiiiiiilllllllllleddddddddddd chaaaaaasssss
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Results & Discussions – Electromigration Performance (TSV/Capture Level)
� Single finger exhibits 10x times better EM performance than five fingers capture level
� Higher joule heating in five finger indicates a design influenced current crowding (intrinsic integrity of the link is robust)
� At lower stress current as that of single finger, five finger lasted with no faill till 2700 hrs
Joule heating extracted by TCR (temperature co-efficient of resistance) measurements
10ºC 2ºC
Stress current is 2.4x times higher for five fingers (300 vs 125 mA)
xxxxxxxhhhhhhhhhhhiiiiiiiiiibbbbbbbbbbbbiiiiiiiiiiittttttttttttsssssssssss 111111111111000000000000xxxxxxxxxxx ttttttttttttiiiiiiiiiimmmmmmmmmmmeeeeeeeeeees
ing in fivof th
Joule hby TCR (effiiiiiicccccciiiiiieeeeeeennnnnt ommmmmmeeeeeeaaaaaassssssuuuuuuuuuurrrrrrrrrreeeme
2ºC
Stttttttrrrrrrressssssssssssssssssss cccccccccccuuuuuuuuuuurrent istimmmmmmmmmmmeeeeeeeeeeesssssssss hhhhhhhhhhhiiiiiiiiiiggggggggggher fofffffffffffiiiiiiiiiiinnnnnnnnnnngggggggggggggeeeeeeeeeeerrrrrrrrrrrssssssss (300
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Results & Discussions – Cross sectional Image (Single finger versus Five fingers)
Five Fingers
Single Finger
No significant degradation. Few voids at capture pad and TSV
No significant degradation. Voids at capture pad and capture level
single finger
SSSSSSSSSSSSSiiiiiiiiiiiiiiinnnnnnnnnnnngl
gnificant deggggggrrrrrraaaaaaddddddaaaaaaaaaationvoids at capturrrrrrrrrrreeeeee ppppppaaaaaaddddddd aaaaaaaaaannnnnnnnnnnd T
gnifiiiiiiiiicaaaaaaaaaannnnnnnnnnntttttttttt ddddddddddddture p
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Experimental Details – Test Structures (TSV/RDL)
Link of interest: TSV/RDL. All other links are design robustoffffffff iiiiiiiiiinnnnnnnnnnnnttttttttttteeeeeeeeeeerrrrrrrrrrreeeeeeeeeeessssssssssstttttttttt::::::: TTTTTTTTTTTSSSSSS
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Results & Discussion– Electromigration Performance (TSV/RDL)
� TSV/RDL structure EM performance equivalent to TSV/five finger capture level
� The failure criteria is aggressive and it only takes 0.05 ohms resistance increase to classify as a failure for this huge structure. The structure didn’t result in hard open
� Test done at lower current didn’t tail till 2700 hrs of stress
10ºC 2ºC Stress current range (300 to 750 mA)
turrrrrrrrrrreeeeeeeeeee EEEEEEEEEEEEMMMMMMMMMMM ppppppppppppeeeeeeeeeeerrrrrrrrrrfffffffffffffooooooooooorrrrrrrrrrrmmmmmmmmmmm
a is age f
2ºC SSSSSSSSSSStreraaaaaaaaaaangemmmmmmmAAAAAAAAAAA)))))))))))
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Results & Discussion– Cross-sectional Images (TSV/RDL)
TSV/RDL structure post EM performance is still intact. No significant degradation was observed
Showing robust TSV/Capture level
interface
p
ShowingTTTTTTTTTTSV/C
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Experimental Details – Test Structures (TSV itself)
Link of interest: TSV. All other links are design robust
RDL only
Capture level only
k
CCCCCCCCCCaaaaaaaaaapppppppppppttttttttttture leve
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Experimental Details – Test Structures (TSV itself)
Good candidate to test in package form with C4s. Need a robust design to maintain TSV as the weakest link and uniform current distribution
(a) Wirebonds are not robust enough to test the TSV itself structure
(b) 1.5 mil diameter wirebonds fuse beyond 1 A for 5 parallel wires (250 mA/wire)
onnnnnnnnnnnddddddddddddsssssssssss aaaaaaaaaaarrrrrrrrrrreeeeeeeeeeee nnnnnnnnnnnoooooooooooottttttttttt rrrrrrrrrrro
r wireb
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Take Away Message/Concluding Remarks
Link of Interest
Experimental Condition CurrentLoading
Capability*
Stress Current
(mA)
Stress Temperature
(ºC)
Joule Heating
(ºC)
Electrical C/S Area
(µm2) TSV/Capture Level (Single Finger) 125 300 2 ~5 1A
TSV/Capture Level (Five Fingers) 300 300 10 ~12 >1A
TSV/Grindside RDL 300 300 Negligible ~20 >1A
TSV element by itself 540-1000 300 Negligible ~20 >>1A
* By passing electromigration targets comfortably. Parameters used in this calculation: 10KPOH, Activation Energy: 0.9 eV, Current Exponent=1.1, Operation condition = 100ºC, 1A, Lifetime=11.4 years
What is the electromigration limited min/max. current for the links comprising of TSV, Capture Level and Grindside RDL?
Integration Scheme is robust enough to handle 1A* and above
eating (ºC)
ElectC/S Ar
((((((µµµµµµmmmmmmmmmm2)
300 2222222222 ~55555555555
0 300 111111111100000 ~~~~~~~~~~1111111111122222222222
33333333333000000000000 30000000000 NNNNNNNNNNNNeeeeeeeeeeegggggggggggllllllllllliiiiiiiiigggggggggggiiiiiiiiiiibbbbbbbbbbbllllllleeeeeeeee
ent bbbbbbbbbbbyyyyyyyyyy itttttttttttssssssssssseeeeeeeeeellllllllllfffffffffff 5555555555554444444400000000000----------111111111110000000000000000000000000000000000 333333333300000000000000000000 Neg
eeeeeeeeeeellleeeeeeeeeeeccccccccccccttttttttttrrrrrrrrroooooooooommmmmmmmmmmiiiiiiiiiiigggggggggggrrrrrrrrrrraaaaaaaaaaattttttttttiiiiiiiiiiooooooooooonnnnnnnnnnn tttttttttttaaaaaaaaaaarrrrrrrrrrrgggggggggggggeeeeeeeeeeeetttttttttttssssssssss comfortabyyyyyyyyyyy: 00000000000........99999999999 eeeeeeeeeeVVVVVVVVVVV,,,,,,,,,, CCCCCCCCCCCuuuurrrrrrrrrrrrreeeeeeeeeeennnnnnnnnnttttttttttt EEEEEEEEEEExxxxxxxxxxxpppppppppppooooooooooonent=1
e eeeeeeeeeeellllllllllleeeeeeeeeeeeccccccccccctttttttttttrrrrrrrrrrrooooooooooommmmmmmmmmmiiiiiiiiiiigising o
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Concluding Remarks� In the 6 µm solid Cu via-last integration scheme, current loading as high
as 1 A is demonstrated successfully
� Results indicate strong interaction of current crowding. EM intelligent design to route the TSVs in the device side and grind side RDL will result in current loading higher than 1A comfortably
� Improved structures and testing methodology needs to be derived to test the limitation of the TSV element itself by making other elements/links robust enough. >1 A stress current to 5 parallel wire bond wires had fused
17
Optimized Power Delivery for 3D IC Technology Using Grind Side
Redistribution Layers
Menglu Li1, Prakash Periasamy2, K. N. Tu1, and Subramanian S. Iyer1
1Center for Heterogeneous Integration and Performance Scaling, Henry Samueli School of Engineering and Applied ScienceUCLA, Los Angeles, CA 90024 2GLOBALFOUNDRIES INC, Malta, NY-12020Friday, June 3, 2016. Session 41: Student Interactive Presentations8:30 AM - 10:30 AM
en
urrent crowddddddddddiiiiiinnnnnngggggg. EEEEEEEEEEMMMMMMMMMMM invice side andddddddd ggggggggggggrrrrrrrrrrriiiiiiiinnnnnnnnnndddddddddddd sssssiiiiiiiidddddddddeeeeeeeeeee RRRRRRRRRRRDL
A comfortabbbbbbbllllllyyyyyyyyy
truuucccccccccccctttttttttttuuuuuuuuuuurrrrrrrrrrreeeeeeeeeeesssssssssss aaaaaaaanof the
Reddddddddddddisssssssssstttttttributiiiiiiiiiooooooooonnnnnnnnn Layyyyyyyyyeeeeeeeeerrrrrrrrrrrrssssssssssss
ngluuuuuuuuu LLLLLLLLLLiiiiiiii1,,,,,,,,, PPPPPPPPPPPrrrrrrrrraaaaaaaaaaakkkkkkkkkkkaaaaaaaaaassssssssssshhhhhhhhhhh PPPPPPPPPPPeeeeeeeeeerrrrrrrrrrrriiiiiiiiiiiaaaaaaaaaaassssssssssaaaaaaaaaammmmmmmmmmmyyyyyyyyyyyy2, KKKKKKK. NNNNNNNNNNN......... TTTTTTTTTTuuuuuuuuuuu11111111111,,,,,,,,, aaaaaaaaand Sterogggegeggggggg nnnnennnnn ouououououououououoouousssssss InInInInInInInInInInntetetetetetetetetetetet grgrggggggggg atatatatatatatatatatatioioioioioioiooioiionnnnnnnnnnn anananananananananaa ddddddddddd PePePePePePePePePePeP rfrfrfrfrfrfrfrfrfrforororororororrororrmmamamamamamammammam nccccccccccceeeeeeeeee ScScScScScScScScScSS alaalalalaaala ing,,,,,,,,,, HeHeHeHeHHHeHeHeHeHenry Selessss,s,ssssss CACACACACACACACACACACA 909090909090909009090020202020202020202020244444444444 2GLGLGLGLGLGLGLGLGLGG OBOBOBOBOBOBOBOBOBOBOOBBALALALALALALALALALALFOFOFOFOFOFOFOFOFOFOFOFOUNUNUNUUNUUNUNUNUNUNDRDRDRDRDDRDRDRDDRDD IES INC, Ma201111111166666666666.... SSSSSSSSSSSeeeeeeeeeeessssssssssssssssssssiiiiiiiiooooooooooonnnnnnnnnnn 444444444441111111111: SSSSSSSSSSSttttttuuuuuuuuuuuudddddddddddeeeeeeeeeennnnnnnnnnnntt IIIIIIIIIIInnnnnnnnnnntttttttttterrrrrrrrrractive PreseAMMMMMMMMMMM
IEEE 66th ECTC – Las Vegas, NV USA May 31 – June 3, 2016 Session 29. Paper 5. Prakash Periasamy
Acknowledgements
� One of the authors (Prakash P) sincerely thank Tim Sullivan (IBM, Retired) for his mentorship and structure concepts used in this study
� Authors also thank Carole Grass, Dinesh Badami, Henry Nye and Gary St Onge for their managerial support during this work
� Thanks to Daniel Berger for financially supporting the wire bonding activity
Thank You18
� Thanks to Hally McHugh and Chad Burke (GLOBALFOUNDRIES,BTV) for preparing the EM samples
� Thanks to GLOBALFOUNDRIES management and Technical Council for the opportunity to present this work
Sullivs used in t
nesh Badammmmmmmmmmiiiiiiii, HHHHHHHHHHHeeeeeeeeeeeennnnnnnrrrrryyyyyyyyy NNNNNNNNNNNyyyyyyyyyeeeeeeeeeee apport duringgggggggg tttttthhhhhhiiiiiissssss wwwwwwoooooooooooorkkkkkkkkkkk
to HHHHHHHHHHHaaaaaaaaaalllllllllllllllllllyyyyyyyyyyyy MMMMMMMMMMMccccccccccccHHHHHHHHHHHuuuuuuuuuuuuggggggggggghhhhhhhhhhh aaaaaaaaaaaannnnnnnnnnndddddddddd CCCCCCCCCChhhhhhhhhhhhaaaaaaaaaaaadddddddd BBBBBBBBBBBuuuuuuuuuuurrrrrrrrrrrkkkkkkkkkkkeeeeeeeeeee (((((((((((GGGGGGGGGGGLLLLLLLLLLLOBAing ttttttttttthhhhhhhhhhheeeeeeeee EEEEEEEEEEEMMMMMMMMMMM sssssssssssaaaaaaaaaaammmmmmmmmmmmppppppppppplllllllllllleeeeeeeeeeesssssssssss
oooo GGGGGGGGGGGGLLLLLLLLLLLOOOOOOOOOOOBBBBBBBBBBBBAAAAAAAAAAALLLLLLLLLLLFFFFFFFFFFFOOOOOOOOOOOOUUUUUUUUUUUNNNNNNNNNNNNNDDDDDDDDDDDRRRRRRRRRRIES muuunnnnnnnnnniiiiiiiiiitttttttttttyyyyyyyyyyy ttttttttttooooooooooo pppppppppppprrrrrrrrrreeeeeeeeeeessssssssssseeeeeeeeeeennnnnnnnnnnnnttttttttttt ttttttttttthhhhhhhhhhhis w