Slides - Agenda INFN
Download
Report
Transcript Slides - Agenda INFN
Hybridization,
interconnection advances
Massimo Manghisoni
Università degli Studi di Bergamo
INFN Sezione di Pavia
December 17, 2015
Motivations
Experiments at the future high luminosity colliders set challenging
requirements on the design of semiconductor pixel detectors
Designers are currently considering two different approaches:
• moving to higher density 2D technology nodes
• moving to 3D technologies with vertical integration techniques
3D integration is a technology that makes it possible to devise pixels with
•
•
•
•
advanced architectures
smaller form factor
less material and dead areas
separation of sensing, analog and digital function
Features of 3D integration can be applied to
• Particle tracking: wide interest in the high energy physics (HEP) community in
view of the design of novel pixel detector systems with advanced readout
electronics
• X-ray imaging: development of advanced X-ray imaging instrumentation for
applications at the free electron laser (FEL) facilities
2 of 25
Outline
Focus of the talk: R&D activity on advanced interconnections
The AIDA 2020 Collaboration (WP4)
Pixel modules for tracking at HL-LHC
High-resolution imagers for photon science application
Medipix - CERN
VIPIC - Fermilab
PixFEL Project - INFN
3
AIDA 2020
WP4 - Task 4.4 Interconnections and TSVs
CERN, INFN (GE, PV, PG), CNRS (CPPM, LAL)
MPG-MPP, UBONN, UNIGLA
Coordination: INFN
Bonding and TSV technologies
Qualification of Industrial bonding techniques: requirements of future HEP
experiments addressed in terms of
pixel readout cell pitch (order of 20 μm) and geometry
interconnection density
Etching TSV in fully processed 65 nm CMOS wafers for peripheral backside
interconnections: the baseline TSV technology will be “via last”
Explore processes for relatively fine pitch TSVs in thinned CMOS wafers,
which will also include chip backside processing steps
4
LHC Hybrid pixel modules
Flex
FEIx
Sensor
Flex
FEIx
Sensor
5
Wire-bond
Stave surface
Wire-bond
Stave surface
Standard pixel modules
Wire bonds are the weak link
Oscillation in magnetic field
Protection required or desired
Limit on envelopes
Require flex to be glued very precisely
Access to bond area on FEIx limits the
sensor active area (not 4-side
buttable)
Modules without wire-bonds
Reverse sensor and FEIx order
Better controlled envelopes
Sensor can be larger than FEIx size and
is 4-side buttable
No fragile wire bonds needed
Sensor cooled directly by stave
Enabling Technologies
TSV-last + RDL
For Flex to FEIx connection
Connect chip M1 from front to
back of chip
RDL distributes all FEIx
connections over full chip surface
• Do not need fine-pitch connections
• Power can be brought to chip at
several places , not just on the edge
Direct laser soldering
for flex to chip connection
Thin 2-layer Al flex
No glue layer needed
Connections are solder 1-by-1,
module stays at RT
Reworkable
Laser%
Solder%balls%
Flex
Wire-bond
FPC%
Chip%
leti
CEA-LETI
FEIx
6
Sensor
Stave surface
Direct laser soldering
Flex is thin 2-layer Al on Kapton flex with Solder ball (200mm) inside which is
melted by laser
No module heat up & No thermal stress on module
No glue is needed between flex and bare module
Solder holds flex mechanically well in place
P. Riedler, A. Di Mauro, A. Junique
(PH-AID ALICE)
Coverlay
Metal
Polyimide
Metal
Coverlay
Chip
27 cm long
7
9 MAPS solder to flex (50 µm thick chips, 15x30mm2 each chip)
7
LETI TSV+RDL
Medipix & LETI developed TSV on Medipix
IBM chip to allow for BGA assembly for
buttable X-ray/particle detectors
Second run processed 6 wafers of
Medipix3rx : TSV yield ~ 70% to 80%
M.Campbell /CERN-PH – G.Pares / CEA-LETI
Medipix specifications
Design
Process Flow
8
Wafer view
Test structures
Single chip
Wafer diameter: 200mm
Wafer thickness: ~725um
IC Technology: 130 nm / IBM
Top Surface: Al + Nitride
Chip size : 14100 x 17300 µm
TSV per chip: ~100
TSV aspect ratio :
120:60 µm (MEDIPIX RX)
50: 40 µm (timepix3)
Activity
Collaboration: CERN, Glasgow, LAL & MPI
Flip chipping at LETI
FEI4 B + Sensor, Satisfactory results (prototype irradiated and Beam tests ..)
Start With FEI4 , TSV + RDL
The Future: Build a demonstrator:
TSVs in a 65 nm front-end chip (from RD53 engineering run)
The Ultimate Goal
100 μm
150 μm
or less
back side
TSVs module
front end chip (65 nm)
front side
Bonding …
New (LETI Proprietary)
Bump Bonding
sensor
HV
9
Direct Wafer-to-Wafer bonding
Bond sensor wafer directly to FE wafer
• Remove single die flip-chip process
Cu-Cu / SiO2-SiO2 and hybrid bonding
• Direct bond formed between 2 wafers
bond Cu & Cu / SiO2 & SiO2
• Requires smooth and activated wafer
surfaces (Roughness ~ 0.5 nm)
• Room temperature and room pressure
bond process
Requires
Post processing at wafer level to
produce smooth surfaces
Matching wafer size
Excellent alignment
TSVs to bring front signal to backside
Wafer-Wafer assembly
Thin ROC wafer is “fused” to
sensor wafer
Allows
Thin devices (Process as a single
wafer after bonding)
TSVs to be added after the bond
Minimizes bond cap/inductance
CEA-LETI Demonstrated direct bonding Cu-Cu, SiO2-SiO2 and more complex Hybrid bonding
10
Bonn/CPPM TSV project
! "##$%&&' () *+(, -" ./01
Goal of the project: develop modules for ATLAS pixel detector at the HL-LHC
using
a&'via
last
TSV
process
! " #$%
($&)$
*" #$
+,&-#.
*$/0$*&$1#2#(&+$3 &14(#0$)&,$5! 657$+/8#($1#*#. *&,$' *$*" #$9 6:
; 40/<%$' $2/' $(' 0*$
! 7=$+,&. #00
69Post-processing
technology
applicable on existing FE electronics
>&0*:+,&. #00/<%$*#. " <&(&%?$' ++(/. ' @(#$&<$#8/0*/<%$AB$#(#. *,&</. 0
Dead
area
periphery
can
be reduced
C#' 1$
' ,#' at
$' *$the
*" #$.chip
" /+$+#,
/+" #, ?$. ' <$
@#$,#14.
#1
Compact, low mass hybrid pixel modules with minimal modification to the FE
; &3 +'and
. *D$(&E
$3 ' standard
00$" ?@,/1$+/8#($
3 &14(#0$
E /*" $3 /</3 ' ($3 &1/)/. ' */&<$*&$*" #$AB$(' ?&4*$' <1$
layout
using
CMOS
technology
40/<%$0*' <1' ,1$; F G7$*#. " <&(&%?
no<&$
wire
bonds
needed/)$if. &3
combined
flex
hybrid
interconnection
E /,#$
@&<10$<##1#1$
@/<#1$E /*"with
$<#E $new
)(#? " ?@
, /1$
/<*#,. &<<#.
*/&<$3 #*" &10
>&*#<*/' ($
)&,$4
H$0/1#$
@4**' @(# 3 &14(#0$
40/<%
$1#1/. 'dedicated
*#1$0#<0&,$('sensor
?&4* layout
Potential
for
side' abuttable
modules
using
Modules
with
be
used)&,$
for*"the
outermost
detector
layers
at9 6:69
the ;HLF &14(#0$
E /*" $TSV
! 7=$.can
' <$@
#$40#1$
#$&4*#,3
&0*$1#*#.
*&,$(' ?#,0$
' *$*" #$
$*&$
LHC
to provide
full*&,$
detector
coverage
+,&2/1#$
)4(($1#*#.
. &2#,' %#$
&2#,$*" #$('over
,%#$'the
,#' large area
5! 657$IJ 6$3 &14(#
B8' 3 +(#$&)$! 7=$3 &14(#
11
" 4#%%/<%U +"?0/NV4</:@&<<V1#
5IC5:KLKL$M/. N:&))$F ##*/<%D$; BOP $: LHQLRQKLST
W
Via last TSV at IZM
! "#$%#&' $( ) ! $#' $*+,
Process: IZM via last tapered TSV
Demonstrator
single chip modules built with:
! "#$%&&'()*+ (, -. (/. &0(0. 1%"%2(345
6 %7FE-I2/3
ATLAS
pixel
chip-0:90μm
thin, with tapered profile TSV and RDL
#8&0". 0#"(&
-89/%($:
-1(7 readout
#2; /%&(<; -/0(=
'
Planar
n-in-n sensor
>?@
)ABC(D3ED4(1-F%
/("%. 2#; 0($: -1(GHI 7 (0: -8J(= -0: (0. 1%"%2(1"#K-/%(345(. 82(L 6 E(
! /. 8. "(8@
-8@
%8&#"
Standard
flip8(&
chip
process (no handle wafer for thin chip handling used,
unconnected
expected
perimeter)
40. 82. "2(K/-1($:pixels
-1(1"#$%
&&(M
8#(: along
. 82/%(=the
. K%chip
"(K#"(0:
-8($: -1(: . 82/-89(; &%2J(
; 8$#88%
$0%2(1-F%/&(%F1%
$0%2(.
/#89(0: %($:
-1(1%"-7TSV
%0%"N
No loss
in performance
wrt
modules
without
O#(/#&&(-8(1%"K#"7 . 8$%(= "07 #2; /%&(= -0: #; 0(345
D7 @
APQ(&#; "$%
+ #2; /%(#8(<#. "2
: 001'BB-#1&$-%8$%R-#1R#"9BQSPT@
HAAQBSBHTB! HTHHTB
12
: ; %99-89Z 1: [ &-VR; 8-@
<#88R2%
D)6D@
AHAH(U-$V@
#KK(+ %%0-89J(W?L O(@HPBHXBAHQY
T
U! ()U"+
*"1)
1) "%0-+
"<$.+,12-+
V"W
X "5 "G$-"2(*0"&.0<$/ -+0, "R1) .="<+KK) .)
#$%&'
**"+
, -).="2--.2/
./ 0, , ) /-+
-""'
0,
Y+, 2("V02(I".) &(2/ ) "G$%&) <"K(+&' / J+&"P+-J"U! U"QB45 "+, -) ./ 0, , ) / -"-0
C2C
SLID (Solid
Inter-Diffusion)
HG1+
."&+-/ J" "K+Liquid
, ) ."&+L) ("K0."J=G.+
<"<) -) / -0.*"
" 0$*I"*%2(()
() **"%2-) .+2("M", 0"G$%&*NN"
Z-+(+[ ) "Y(+
&',UJ+
&") S$+&%) , -"K0."QB45 "
Bump-less
interconnect
O
J) "%0*-"&.0%+
) , -"P2=I"QB45 "RQ0(+<"B+S$+<"4, -) .<+KK$*+0, T""
" 2."%0*-(="8
Q
0"K
7"0."U!
8 "P+
' / 0%&0*+
-+0, "0K"/ J+&*"0, "*$&&0.-"P2K) ."
smaller
pitch ->! 8finer
pixel
for-J".)
hybrid
detectors
U!
*"1) .="2--.2/
-+1)bumps
"<$.+, V"WX 5 "G$-"2(*0"&.0<$/ -+0, "R1) .="<+KK) .) , -"/ J+&*??T"
"U"+material
less
– no
Y+, 2("V02(I".) &(2/ ) "G$%&) <"K(+&' / J+&"P+-J"U! U"QB45 "+, -) ./ 0, , ) / -"-0"-J+, "3Q4U*"
The
mostV="K
prominent
Q-.2-)
0."WX 5way:
I" SLID (Solid Liquid Inter-Diffusion)
"
So far mostly
or C2W
re-composition
5 ) *+V, W2W,
"2, <"&.0<$/
) "with
-) *-"1)
J+/ () *"R/ J+&*T"of chips on support wafer
Z-+(+[ ) "Y(+&' UJ+&") S$+&%) , -"K0."QB45 "
very attractive
during R&D but also production (very different chips..)
K="&.0/ ) **"
"C2C is S$2(+
replace
bumped
flip-chip
with C2C SLID interconnect to thin ASICs
-.2, *K
) ."-0"\.)
2(]"/ J+&*"
"
"Final goal:
Q-.2-) V="K0."WX 5 I"
"
5 ) *+V, "2, <"&.0<$/ ) "-) *-"1)CuJ+
/ J+&*T"
- / () *"R
TiW:N
Interdiffusion
S$2(+K="
Cu&.0/ ) **"
Sn,
-.2, *K)Sn."-0"\ .) 2(]"/ J+&*"" liquid
Cu3Sn
eutectic
alloy
"
Cu Interdiffusion
TiW:N
Contact under Pressure
andSn,
Heat
Through MaskCu
ElectroplatingSn
~ 5 bar, 260 –liquid
300 °C (Sn-melt)
from
under Pressure
6! 67"8 9:
"; R&D
+<=07">activity
01?"!Contact
6@A"
Through Mask
Electroplating
13
and Heat
~ 5 bar, 260 – 300 °C (Sn-melt)
"! 6! 67"8 9: "; +<=07"> 01?"! 6@A"
Formation of
CuEutectic
3Sn
Alloy;
eutectic
Tmelt > 600 °C
alloy
! " #$%
MPP/EMFT
!"
Formation of
Eutectic Alloy;
Tmelt > 600 °C
!"
B2<+*(21
B2<+*(21"3, <.+/ ) C7"D 9E"F2(G() +-) .(2G0."
3D integration in Photon science
Medipix3 chip
256 X 256 pixels on a pitch of 55μm
130 nm CMOS technology
I/O wire bonding pads compatible with TSV technology
TSV=connections between front and back sides of the ASIC
CERN-LETI collaboration (TSV Last developed by LETI and applied to pixel
detector read- out chip designed and supplied by CERN)
VIPIC1 Demonstrator
64 X 64 pixels on a pitch of 80 μm
130 nm CMOS technology
PixFEL Project
64 X 64 pixels on a pitch of 100 μm
65 nm CMOS technology
14
VIPIC1 Vertically Integrated Photon Imaging Chip
W2W tier stacking
Cu-DBI (oxide-oxide fusion bonding) used for bonding
tiers of 3D VIPIC
8” bonded wafer pair with top wafer thinned to expose
6mm TSVs (6mm of silicon left of the top wafer)
D2W ASIC-sensor LTD-bonding
Ni-DBI (oxide-oxide fusion bonding)
15
Detector
Si d=500 mm, pitch 80×80 mm2,
soft 8keV X-rays
Application
XPCS (X-ray Photon Correlation
Spectroscopy)
Electrically testable using
wire bonded connections
bump-bonded connections
Tests in configuration with
Sn-Pb bump-bonded sensor
Fusion-bonded sensor
(Low-Temp. Direct bonding)
3D technology providers
Tezzaron
Ziptronix.
Collaborating institutions
Fermilab, AGH-UST, BNL
LTD-bonded vs b-bonded
Bump-bonded VIPIC1 pitch 100mm vs. 80mm for fusion-bonded, nevertheless …
competitive
to MAPS!
32×38 pixels bonded, 2880 pixels floating
bump-bonded: ENC=69.6 e- ± 5.1 mV/elarger input capacitance = larger noise,
lower gain and more dispersions
LARGE FEEDBACK RESISTANCE
ENC=36.2 e- ± 2.6 mV/esymmetrical noise distribution with
<3.4 % of pixels outside of ±3 s range
ENC on fusion bonded device is close to that measured for floating inputs!
ENC=40e- Cin<20fF, ENC=70e- Cin>80fF
16
From VIPIC1 to VIPIC-Large: 1Mpixel XCS Detector
X-ray back-side illumination
Features
1Mpixel = 3 shingles of 6×2 or straight 7×7 VIPICs-L LTD-bonded (D2W) to a sensor wafer
No dead edges, no peripheral circuitry on the ASIC
65 mm - pitch square pixel, 36,864 pixels/chip grouped in indiv. readout subchips of 1024 pixels
1 FPGA per VIPIC-L for on the fly data processing (up to 0.7 Tbps of raw data produced)
Multi-layer (>20 routing layer LTCC) supports b-bonded detector structure
Flat back-side - friendly for mounting of a cooling plate
17
B-TSV
TSVs are critical component for
VIPIC-Large
small diameter ~1mm TSVs
wire-bond-less
no-dead-zone
Middle
TEOS SiN
1470A
469A
Via-Last TSVs (inserted in postprocessing after thinning of a 3D bonded
wafer pair)
B-TSV (requirement: identical DRC as
GF TSVs)
Work with Tezzaron/Novati on B-TSVs
Bosch etching used for B-TSV cavities
and internal metal (M1) is Cu
- DRC rules = GF TSV rules
- 2 3D-integrated wafer pairs tested
18
Bottom
Top
Oxide
SiN
B-TSVs – capacitance and resistance
B-TSV to substrate capacitance
B-TSV resistance
1TSV
2TSVs
TSVs to substrate breakdown
voltage measurement:
W#5 VBREAK=38.5V ± 9V
W#6 VBREAK=36.5V ± 5V
19
Single B-TSV resistance is a difference
between 1TSV/connection and
2TSV/connection 0.343W.
Wafer #6 showed more defective BTSV in the chains
VIPIC-L Design Highlights
New approach for entirely edgless design – enabled by 2 tier 3D integrations
array of pixels on the analog tier
whole digital tier P&R-ed at once without respecting pixel boundaries
32×32 VIPIC-L analog pixels forming single
analog tier subchip
(6×6 subchips form one full analog tier chip)
20
Digital tier chip designed in 130nm process;
next design is clearly 65nm or beyond
32×32 pixels VIPIC-L digital tier subchip
with no explicit boundaries of pixels
The PixFEL Project
Long term goal: Develop a four-side buttable module for the assembly of large
area detectors with no or minimum dead area to be used at FEL experiments
Collaboration: INFN (PV,PI,TN) UniPV, UniPi, UniBI, UniTN
Multilayer device: active edge thick pixel sensor, two tiers 65 nm CMOS readout
chip (analog+digital/memory) with low/high density TSV, pixel pitch of 100 mm
21
Demonstrator
slim edge sensor
high density
interconnects
bump
bonding
TSVs
front-end
& ADC
readout active/slim
chip
edge sensor
hybrid board
memory and
digital readout
Interconnection and vertical integration
bond
pads
Main features
32x32 array of hybrid pixels
three-tier structure
slim edge, fully depleted silicon
sensor
two-tier readout chip, in 65 nm
CMOS technology
1 layer with analog front-end
and ADC
1 layer devoted to digital
memory
• Sensing layer - front-end chip: relatively large pitch of 100um easily achieved
through bump-bonding techniques
• Through silicon vias (TSVs) to access the circuits from the back side and including
substrate thinning
• 3D integration to interconnect the analog front-end and ADC on one tier to the
memories on the other tier
22
Summary
Pixel detector requirements for next
generation experiments at high luminosity
colliders and X-ray sources are extremely
challenging
•
•
•
•
high granularity → small room for electronic
high hit rate → high speed, on-chip memory
data reduction → hit discrimination capability
radiation hardness
More functionalities need to be built into the
readout chip to satisfy the specifications
• front-end performance improvement
• data selection and hit discrimination
(bandwidth reduction)
3D integration technologies may provide several advantages in the design
of detectors for vertexing and imaging applications
23