Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Future ASIC technologies in HEP Experiments
Michael Campbell and Federico Faccio
Microelectronics Section
ESE Group, EP Department, CERN
Acknowledgements
• All those who signed up to the WP mailing list
• Those who presented as first meeting on 5th March
• Members (and one retired member) of ESE Group who provided important inputs. In particular Philippe Farthouat, Erik Heijne, Kostas Kloukinas, Sandro Marchioro and Walter Snoeys
• External experts: Alvin Loke from Qualcomm, Michael P. King from Sandia
Resources:
• Michael P. King : FinFET technologies for Digital Systems with Radiation Requirements: TID, SEE, Basic Mechanisms and Lessons Learnt https://indico.cern.ch/event/666568/
• Alvin Loke: Analog/Mixed-Signal Design in finFET Technologies https://indico.cern.ch/event/662048/
• Erik Heijne: Report on IEEE-ISSCC 2018: micro-nano-pico, we must go on https://indico.cern.ch/event/709523/
2
Outline
• Context
• ASIC’s used during LHC Runs 1 and 2 and lessons learned
• Transistor scaling: opportunities and challenges
• Chip stacking
• Radiation tolerance
• First work package meeting and preliminary conclusions
• One (optimistic) long term scenario
• The way ahead for this work package
3
Context
• The EP Department asks us to prepare a document by the end of 2018 outlining plans for R and D in the period beyond 2020
• A major decision concerning the next large project (expected ~2020) will have a large impact on which direction to take
• Let’s look briefly at ASICs from the first generation LHC detectors
• Let’s see where we are now with respect to industry
• What can we learn from our past and from what’s happening outside
4
pixeldetector
beam pipe
Si strip tracker
e-CAL
TRT
h-CAL
muon chambers•
Some of the ASICs in ATLAS
FE-I3 pix det
28 000 chips
80 M segments
1.7 m2 Si sensor
ABCD Si det
50 000 chips
6 M segments
60 m2 Si sensor
ASDBLR TRT det38 000 chips
DTMROC TRT det
19 000 chips
ASD muon det148 000 chips
Total ATLAS100 million sensor cellsappr. 800 000 chipsmajority ASICs
Chips to scale 1 cm
Slide by E. HeijneISSCC 2014 5
pixeldetector
Total CMSappr. 1 million chipsof which 700 000 ASICs
MAD muon det181 000 chips25 000 m2 gas-filled
APV25 Si det
110 000 chips
9.3 M segments
198 m2 Si sensorPSI46 pix det
16 800 chips
66 M segments
1 m2 Si sensor
QIE8 calorimeter220 400 chips
beam pipe
Si strip tracker
e-CALh-CAL
muon chambers
Chips to scale 1 cm
Some of the ASICs in CMS
Slide by E. HeijneISSCC 2014
6
List of main functions performed by on-detector ASICs
• Front-end ASICSignal amplificationNoise filteringPipelining (analogue, binary or digital)Discrimination / A to D ConversionL1 selectionData transmission off chip
• Module controllerSlow control interfaceMonitoring of FE ASICsTemperature monitoring
• Data transmissionData de/serialisation on/off moduleOptical receiver and transmitters
• Voltage regulators7
Radiation levels foreseen in ATLAS
Sub-detectorNIEL
[1MeV n.cm-2.year-1]
TID
[Rad.year-1]
SEU tolerance
[>20 MeV hadron.cm-2.year-1]
Pixel B-layer 1.62 1014 1.1 107 2.30 1014
Pixel 7.58 1013 4.98 106 8.98 1013
SCT 1.59 1013 7.83 105 1.22 1013
TRT 7.23 1012 1.56 105 2.86 1012
Lar 1.85 1011 4.77 102 3.78 1010
Tile 2.74 1010 37.8 6.74 109
CSC 8.34 1011 23.0 102 1.59 1011
RPC 5.71 109 11.5 1.04 109
TGC 6.53 109 16.5 1.75 109
MDT 4.80 1010 74.1 8.25 109
Slide by Philippe Farthouat 8
Example of ATLAS tracking ASICs
Tracking detectors ASIC Technology Functionality Quantity & remarks
Pixel
FEI 0.25 µ CMOS Front-end ~28000
Module Controller 0.25 µ CMOS Control ~1800
VDC 0.25 µ CMOS VCSEL driver ~500
DORIC 0.25 µ CMOS Timing and control receiver ~400
Silicon Strips
ABCD DMILL Front-end ~50000BiCMOS design
DORIC 0.35µ BiCMOS Timing and control ~4100
VDC 0.35µ BiCMOS VCSEL driver ~8200
TRTASDBLR DMILL Amplifier-shaper-
discriminator~38000
Bipolar design
DTMROC 0.25 µ CMOS Digitiser ~19000
Slide by Philippe Farthouat 9
Example of ATLAS MUON spectrometer ASICsASIC Technology Functionality Quantity & remarks
MDTASD 0.5 µ CMOS Amplifier-shaper ~5000
AMT 0.3 µ CMOS Time to digital conv. ~15000
CSC
CSC
ASM1 0.5 µ CMOS Preamplifier ~1300
ASM2 0.5 µ CMOS Multiplexor ~1300
Clock driver 0.5 µ CMOS Clock driver ~200
HAMAC-SCA DMILL Analogue memory ~2600Common with LAr
RPC
RPCASD GaAs Amplifier-shaper ~47000
CMA 0.18 µ CMOS Coincidence matrix ~3300
TGC
TGC
ASD Bipolar Amplifier-shaper ~81000
HpT 0.35 µ CMOS Trigger ~800
PP 0.35 µ CMOS Trigger ~15000
SLB 0.35 µ CMOS Trigger ~3000
JRC 0.35 µ CMOS JTAG controller ~1400
Slide by Philippe Farthouat 10
Additional functions in on-going upgrades
• Enhanced radiation hardness in trackers
• New architectures – with highly complex digital circuitry - needed to deal with higher pile up
• Fast timing layers also for pile up mitigation
• Serial powering or DCDC convertors
11
10um
3um 1.5um
1um
0.8um
0.35um
0.25um
180nm 65nm130nm
Moore’s uncertain future
Alice SPD chip 1999
RD-53 chip2017
12
Si CMOS technology nodes in papers ISSCC
2018
65 nm
14 nm
~300 papers acceptance <40%
'SiGe'
Slide by E. HeijneFrom ISSCC 2018
28 nm
13
Looking ahead – scaling
• As a community we have accumulated (at least) 10 years delay since 1999
• With the 65nm process we cannot increase IO speed beyond 10Gbps
• FPGA chips (which we rely on off-detector) are pulling away from us
• We cannot stand still but going forward requires significant resources
• Below 28nm FinFETs become the workhorse
14
Journey to FinFETs• 16/14nm complexity accumulated from scaling innovations
introduced earlier across multiple earlier nodes
TechnologyInnovation
FoundryDebut
ReasonRequired
Mechanical stressors 40nmMobility boost for more FETdrive & higher Ion/Ioff
HKMG replacement gateintegration
28nm (HK-first)20nm (HK-last)
Higher Cox for more FET drive & channel control
Multiple-patterning 20nmSub-80nm pitch lithography without EUV (13.5nm mostly l= 193 nm)
Complexmiddle-end-of-line
20nmContact FET diffusion & gate with tighter CPP
Slide courtesy of Alvin Loke, Qualcomm 15
Planar NMOS vs FinFET
Planar
p-substrate
channel
p-well
gaten+ drain
n+ source
STI
p-well tie
FinFETfully-depleted
body
p-substrate
n+ drain
p-well
STI
NMOSn+ source
p-well tie
Slide courtesy of Alvin Loke, Qualcomm16
A closer look
• Most commercial fabs have migrated to FinFETs below 20-nm gate length feature sizes
• FinFETs exhibit improved electrostatic control of the channel and improved reliability compared to equivalent scaled planar CMOS
T. Hook, FDSOI Conference, Taiwan, 2013
Courtesy of Michael P. King, Sandia17
Mechanical Stressors• Mobility depends on channel lattice strain (piezoresistivity)• Grow stressors to induce channel strain along L• Tensile for NMOS, compressive for PMOS• Techniques: S/D epitaxy, stress memorization, gate stress
• Anisotropic mobility & stress response• L vs. W direction, (100) fin top vs. (110) fin sidewall
NMOS PMOS
Garcia Bardon et al., IMEC [5]Liu et al., Globalfoundries [6]
Slide courtesy of Alvin Loke, Qualcomm18
Bulk FinFET Processing Technology
19
A. Yagishita (Toshiba), SOI Short Course (2009)
• Increasing processing complexity
• More challenging lithography• Quad patterning
• Soon EUV
• Line edge roughness
• Isolation steps• STI
• CSD/SSRW
Courtesy of Michael P. King, Sandia
FinFET evolution at Intel
Zheng Guo et al. Intel. paper 11.1 Slide by E. HeijneFrom ISSCC 2018 20
Stress-Related Layout Effects• Stressors are stronger in 16/14nm for more FET drive, so layout
effects can be more severe schematic/layout Δ
• Stress build-up in longer active, ID/fin not constant vs. # fins
• Interaction with stress of surrounding isolation & ILD
• NMOS/PMOS stress mutually weaken each other
• New effects being discovered, e.g., gate-cut stress effect
NMOS
Faricelli, AMD [7]Lee et al., Samsung [8]
Sato et al., IBM [9]
PMOS
Slide courtesy of Alvin Loke, Qualcomm 21
Advantages / Challenges of FinFETs
M. G. Bardon (IMEC) ICICDT (2015)
Gate length shrinkPerformance scaling
FET is on edgeDual gateReduces Ioff
Courtesy of Michael P. King, Sandia22
Designing with FinFET• More drive current for given footprint
draincontact
well
spacer
gatesourcecontact
Sheu, TSMC [18]Hsueh et al., TSMC [19]
• Quantized channel width
– Challenge for logic & SRAM
– OK for analog, enough gm granularity
• Less DIBL better rout, 3 intrinsic gain
• Essentially no body effect (ΔVT < 10mV)
• Higher Rs & Rd spreading resistance
• Lower Cj but higher Cgd & Cgs coupling
• Higher Rwell (Rdiode, latch-up)
• Mismatch depends on fin geometry, MG grains, gate density, stress, less on RDF
Slide courtesy of Alvin Loke, Qualcomm23
I/O Voltage Not Scaling With Core Supply
• Many I/Os still use 1.8V signaling despite core VDD reduction• Many peripheral ICs remain at lower cost nodes• Backward compatibility is key constraint for some I/Os
• Increasingly tough to keep 1.8V thick-oxide devices• Thick-oxide HKMG ALD fill not easy for tighter fin pitch• More complex level shifters to deal with wider voltage gap• Some standards no longer support legacy modes in favor of
higher link rate & lower power (e.g., LPDDR5)
• Need ecosystem consensus• Industry has migrated from 5.0V to 3.3V to 2.5V to 1.8V• Obvious power & area benefit to migrate to say 1.2V• 1.8V remains an industry-wide issue until next transition
Wei et al., Globalfoundries [22]Slide courtesy of Alvin Loke, Qualcomm
24
Considerations for HEP Application
• Significant logic area scaling migrating to finFET• Though not as good as 4x reduction from 28nm to 14nm (marketing)
• Mature 14nm & 10nm process
• No model corner uncertainty less overdesign & perf. compromise• Hf-based HK gate dielectric reliability
• May be prone to hysteretic (ferroelectric) polarization & worse BTI at high radiation levels, causing undesirable VT shift
• HK polarization issues resolved for “typical” CMOS usage• Latch-up prevention
• High fin resistance enforces stricter well-tie spacing & guard ring DRCs
• Beta ratio (NMOS-to-PMOS drive strength) 1• Mechanical stressors & (110) fin sidewall much more effective to boost hole vs. electron
mobility strong PMOS• Device mismatch
• Fully-depleted structure intrinsically superior (less/no RDF)• Benefit reduced by new mismatch sources (fin dimensional control, MG grain
orientation, LDE sensitivities)
Slide courtesy of Alvin Loke, Qualcomm25
Looking ahead – scaling
• As a community we have accumulated (at least) 10 years delay since 1999
• With the 65nm process we cannot increase IO speed beyond 10Gbps
• We cannot stand still but going forward requires significant resources
• Below 28nm FinFETs become the workhorse• Increased gain, faster switching, lower leakage current• Increased layout restrictions (already at planar 28nm) • Radiation tolerance largely unknown (although initial indications not promising)
26
Looking ahead - chip stacking
Po-Sheng Chou et al.TSMC; paper 5.5
1.1 µm pixelsmotion detection
Oichi Kumagai et Sony; paper 5.4
1.5 µm pixelsevent driven10b ADC
TSMC example
27
Slide by E. HeijneISSCC 2014
Looking ahead - chip stacking
Wafer level stacking @ pixel pitch 6.9 mm
SONY example
28
Slide by E. HeijneISSCC 2014
Radiation tolerance requirement is strongly machine dependent
Here reference TID levels are in the innermost tracker layer (note that in some cases the forward calorimeters might be exposed to larger doses, such as in FCC-hh).For a lepton FCC machine, no
simulation of the detectors’ environment is still available.
Slide from Federico Faccio
29
TID leakage current can depend strongly on manufacturer
1.E-12
1.E-11
1.E-10
1.E-09
1.E-08
1.E-07
1.E-06
1.E+04 1.E+05 1.E+06 1.E+07 1.E+08 1.E+09
Vgs (V)
Ileak (A
)
1/0.08
0.5/0.08
pre-rad annealing
1.E-12
1.E-11
1.E-10
1.E-09
1.E-08
1.E-07
1.E-06
1.E+04 1.E+05 1.E+06 1.E+07 1.E+08 1.E+09
TID (rad)
Ileak (A
)
1/0.12
1/0.24
1/0.5
1/1.6
1/3
1/5
pre-rad annealing
Figure 2: Leakage current evolution with TID for different transistor sizes: W array (left), L array (right).
A peak can be observed also in the evolution of the threshold voltage shift with TID
(Figure 3). It is interesting to observe that the dose at which the threshold voltage shift
peaks is the same as for the leakage current, i.e. 1Mrad. Thus, we can conclude that as
well as for the leakage current, the degradation of the threshold voltage is due to the
balance between positive charge trapped in the bulk of the STI and interface states
creation at the STI-Si interface and not to radiation effects in the gate oxide. This effect,
already observed in the 130nm CMOS process, is called Radiation-Induced Narrow
Channel Effect (RINCE) [1]. In narrow channel transistors (W≤ 1µm) radiation induced
defects in the STI not only determine the working conditions of the lateral parasitic
transistors but also influence the electric field of the main transistor. This gives origin to a
shift of the threshold voltage inversely proportional to the W. As we can observe in figure
3 (left) where the threshold voltage shift evolution as a function of the TID is shown for a
W array of transistors, the transistor with W=0.5µm has a bigger threshold voltage shift
than the transistor with W=1µm. It is interesting to observe (Figure 3, right) that given
the same W there is a small variation of the maximum threshold voltage shift value: the
longer the transistor the bigger the threshold voltage shift. For all the transistors sizes the
value of the threshold voltage shift is anyway negligible being of only a few mV.
-0.006
-0.005
-0.004
-0.003
-0.002
-0.001
0.000
1.E+05 1.E+06 1.E+07 1.E+08 1.E+09
Vgs (V)
ΔV
th (V
)
1/0.08
0.5/0.08
annealing
-0.006
-0.005
-0.004
-0.003
-0.002
-0.001
0.000
1.E+05 1.E+06 1.E+07 1.E+08 1.E+09
TID (rad)
ΔV
th (V
)
1/0.12
1/0.24
1/0.5
1/1.6
1/3
1/5
annealing
Figure 3: Threshold voltage shift evolution with TID for different transistor sizes: W array (left), L array
(right). The threshold voltage shift has been extracted by linearly fitting the square root of the Ids vs Vgs
curve in saturation and finding the intercept with Ids=0A.
Figure 4 shows immediately that PMOS transistors are less affected by radiation than
NMOS. In the case of PMOS transistors both fixed charge and interface states are
positively charged and contribute to increase the threshold voltage shift of the parasitic
lateral transistors in such a way that no leakage current degradation is observed. The
small deformation of the Ids vs Vgs curve at Vgs<0V is due to a phenomenon called Gate
after L.Gonella et al., “Total Ionizing Dose effects in 130-nm commercial CMOS technologies for HEP experiments”, NIM A 582 (2007)
130nm bulk CMOS
90 nm bulk CMOS
Manufacturer A
Manufacturer B
30
TID leakage current can depend strongly on a particular plantFA C C IO et al.:R IS C E A N D N A R R O W C H A N N E L (R IN C E ) E F F E C T S 2 93 7
F ig.8. Ion decreas e w ith T ID for N M O S tran s is tors at differen t bias in the
6 5 n m n ode (room T ).T he bias effect is large an d can be obs erved even on
un m atched tran s is tors (on differen t chip s ).
T he m agn itude of the n arro w chan n el effect (R IN C E ) is
s m aller for the N M O S ,but the obs ervation of the degra-
dation of tran s is tors w ith differen t W [F ig. 6 (b)] reveals
in teres tin g details of the dyn am ic evolution of the defects in
the S T I oxide.A t m oderate T ID ,hole trappin g in the bulk of
the oxide dom in ates , lo w erin g the equivalen t of n arro w
chan n el tran s is tors [7],[16 ] an d in creas in g their .A t higher
do s es ,m o s t of the hole trap s are already occupied an d room -T
an n ealin g is active, s uch that additio n al do s e is les s efficien t
in accum ulatin g po s itive charge in the S T I.A t the s am e tim e,
laten t in terface trap buildup occurs w here electron s can be
trapped an d com pen s ate for the fiel d gen erated by the trapped
holes [16 ].A s a res ult, the differen ce betw een n arro w an d
w ide chan n els dis appears at aroun d a T ID of M rad .
A t higher do s es , the in terplay betw een R IN C E an d R IS C E
determ in es the n et degradation of the tran s is tors : the on ly
con tributio n of R IS C E is s een in the E L T tran s is tor of F ig.6 (b),
w hich has m in im um degradatio n . H o w ever, its com paris o n
w ith the lin ear tran s is tor w ith m in dicates that,as for
the PM O S ,R IN C E s till affects tran s is tors w ith this gate w idth.
T herefore, the approxim ate 5 % degradation of the larges t
available tran s is tor [ m ,in F ig.6 (a)] is traceable
to R IN C E an d n ot to effects in the thin gate oxide.
a) Effect of the Bias: T he electric field applied durin g the
irradiation has a s tron g in flue n ce on the phen om en a des cribed
before.In particular,the m agn itude of the R IS C E s eem s to be
s tron gly depen den t on the app lication of durin g irradiation .
In F ig.8, w e report the decreas e for iden tical tran s is tors
bias ed w ith a (1.2 V ) but w ith or w ithout : the abs en ce
of drain -s ource field m akes the dam age con s iderably m ilder.
N ot s ho w n are the an d S ubS in creas es that are als o m uch
s m aller.
a) Effect of the Temperature: A s for the PM O S ,radiation -
in duced dam age in creas es w ith tem perature in the explored T
ran ge ( 3 0 to C ).A ls o,for the N M O S ,the effect of the
applied bias is relevan t at room tem perature an d at C ,
w hile it is n egligible at C .
2) 130 nm Node: S am ples available in this techn o logy w ere
m an ufactured in tw o differen t productio n plan ts : w hile the ra-
diation -in duced leakage w as n egligible in on e cas e,it w as very
F ig.9. E volution of the s ource-drain leakage curren t (m eas ured for
0 V an d 1.2 V ) for m in im um -s ize N M O S tran s is tors in the 13 0 n m
n ode ( 0.15 /0.13 m ).T he irradiatio n took place at room T an d w ith a
1.2 V bias .
s ign ific
a
n t in the other (F ig.9).T his is a clear exam ple of ho w
s en s itive the radiation res pon s e can be to the proces s in g details
of C M O S m an ufacturin g: fres h tran s is to rs from the tw o plan ts
have iden tical electrical perform an ce,an d the m an ufacturer can
us e both plan ts for an y product,depen din g on the available pro-
duction capacity.H o w ever,the radiation toleran ce of the prod-
ucts — in the pres en t cas e— can be very differen t.
C on cern in g T ID effects o n the other tran s is tor param eters ,
N M O S devices appear to be very res ilien t up to the m axim um
do s e that w e reached ( M rad ).T he degradation is
lim ited to belo w 5 % ,w ith a com parable decreas e in an d
s hifts belo w 5 0 m V .W ith the degradation bein g s o s m all,it
is n o t po s s ible to dis tin guis h the pres en ce of R IS C E ,n either the
effect of tem perature an d bias .H o w ever,s om e m an ifes tation of
the R IN C E is eviden t for m in im um W tran s is tors as an in creas e
of the at m oderate do s es for n arro w chan n el tran s is tors .T his
effect is s im ilar to,but s tro n ger than the on e s ho w n in F ig.6 (b)
for the 6 5 n m n ode belo w M rad .
A lthough the radiation -in duced leakage curren t w as very dif-
feren t for s am ples from the tw o production plan ts ,this w as n o t
the cas e for all other characteris tics : T ID effects w ere iden tical
for all other param eters .
IV . D IS C U S S IO N
F rom our res ults o n the PM O S an d N M O S tran s is tors ,w e
can con clude that the thin gate oxide is affected little by T ID .
R adiation effects determ in in g the tran s is tor’s res pon s e occur in
thicker “paras itic” oxides .
A. RINCE in PMOS and NMOS Transistors
D efects in the lateral S T I oxide determ in e the perform an ce
degradation of n arro w chan n el tran s is tors (R IN C E ).F ig.10
com pares the evolutio n of P- an d N -chan n el tran s is tors
w ith m in im um W (0.12 m ) an d m oderately lon g L (1 m ).
C om pared to the data s ho w n in F ig.2 (a) an d 6 (b),w here all
tran s is tors in the W array had m in im um gate len gth an d w ere
hen ce heavily affected by R IS C E ,this figure en ables to better
s ort out the phen om en a behin d the n arro w chan n el effect alon e.
W hile the tw o types of defects in the S T I oxide (oxide trapped
after F.Faccio et al., “Radiation-Induced Short Channel (RISCE) and Narrow Channel (RINCE) Effects in 65 and 130 nm MOSFETs”, IEEE TNS 62, n.6, Dec.2015
31
Some devices in some bulk processes are ok for HL-LHC
65nm bulk CMOS 28nm bulk CMOS
100Mrad
14 nm14 nm
32
But the device physics is complicated
STI = Shallow Trench Isolation oxide
Edge on view of gate Edge on view of channel Real edge on view of channel
Spacer Oxide and nitride(above source/drain LDD diffusion)
33
Radiation Induced Short Channel Effects (RISCE) and radiation induced Narrow Channel effects (RINCE)
paper ID# 1297
6
degradation of narrow channel transistors (RINCE). Fig. 10
compares the Ion evolution of P- and N-channel transistors
with minimum W (0.12 µm) and moderately long L (1 µm).
Compared to the data shown in Fig. 2.a and 6.b, where all
transistors in the W array had minimum gate length and were
hence heavily affected by RISCE, this figure better allows
sorting out the phenomena behind the narrow channel effect
alone. While the two types of defects in the STI oxide (oxide
trapped charge and interface traps) tend to compensate in the
NMOS, as discussed in III.B.1.a, they add up in the case of the
PMOS. For the minimum W transistor in Fig. 10, this positive
trapped charge modifies the electric field in the full transistor
transforming its characteristics: the transistor almost end up
behaving as a large resistor at doses above about
400 Mrad(SiO2).
B. Possible causes for the RISCE
We can exclude that the damage responsible for the RISCE
occurs in the STI oxide edgeless (ELT) transistors (as reported
in Fig. 2.b and 6.a) are insensitive to STI charging. Instead, it
may be traced to the region of the spacers, thick insulator
islands around the gate used in CMOS processing for
source/drain engineering – Fig. 11 shows a schematic view of
a transistor. Note that this is an example of a textbook process
flow, as we do not have any information on the construction of
the specific transistors used in our study. During the
manufacturing process, a moderate doping implant covering
the full drain/source area (extension module) is first applied to
produce a shallower and more lightly doped region. In the
successive spacer module, the spacer insulator is built with the
role of preventing the high doping implant of the source/drain
regions to reach the proximity of the channel. This gives
origin to the Lightly-Doped Drain (LDD) construction that is
efficient in reducing the electric field responsible for hot
carrier degradation. The construction of the spacer normally
starts with the deposition of a relatively thin oxide (about
15 nm) replacing the thermal oxide that has been damaged by
the source/drain extension implant: this deposited oxide is
likely to have a much worse quality, in terms of defects in the
bulk and interface, than the thermal gate oxide. On top of this
thermal oxide, a thick layer of nitride is deposited, which
normally contains hydrogen in some percentage.
Radiation can induce significant damage in this deposited
oxide and/or at its interface with the silicon (lightly doped
source/drain region), with significant charge trapping that can
be responsible for an electric field affecting the surface
potential in the LDD at high levels of TID. The consequences
on the electrical characteristics of the transistors differ
according to their polarity.
In PMOS transistors the LDD regions are lightly doped with
acceptors (p-). The accumulation of positive charge in the
spacer oxide at high TID can build an electric field effectively
lowering the net effective doping of the LDD region, or in
parts of it. As a result these regions would present a larger
resistance to current flow. However, this resistance is not an
access resistance independent from the applied gate voltage:
such a model does not match the experimental data. On the
contrary, it is reasonable to believe that the fringing field lines
at the edges of the polysilicon gate (closing on the
source/drain LDD regions) have an influence on this
resistance. Overall the transistor would be able to carry less
current (the effective mobility would be smaller) and some
characteristics dependent on the effective gate length might be
affected – all in agreement with our observations.
In NMOS transistors the LDD regions are lightly doped
with donors (n-). In this case the accumulation of positive
charge in the spacer oxide would tend to increase the net
effective doping of the LDD region without impacting the
current in the transistor (the native resistance of the LDD
region is already smaller than the channel resistance so that its
further decrease would be irrelevant). On the other hand the
effective doping of the LDD is an important parameter in
determining the response of the device to Hot Carriers
Injection (HCI). The reason to mention HCI here is that some
of our observations resemble damage mechanisms typical of
hot carrier stresses:
- Only the shortest channel transistors (L = 60 nm), with the
highest source-drain field, are significantly damaged.
- Interface traps are created, as suggested by a sizeable
increase of the subthreshold swing (more than
30 mV/decade for the shortest channel transistors, Fig. 7).
- The damage is strongly bias dependent, being very large
only in transistors biased with a large drain-source voltage
Fig. 10. Ion decrease with TID for minimum W transistors with moderately
long channel (1 µm). In NMOS, the compensation of the effects of oxide
trapped charge and interface traps makes the radiation performance much
better than in PMOS, where the two effects add up.
Fig. 11. Qualitative view of the construction of a MOSFET, showing the
spacers used for the Low-Doped Drain (LDD) engineering.
were either very large (W = 20 μm) or Enclosed Layout Tran-
sistors (ELTs, where the central diffusion is completely sur-
rounded by the poly-crystalline Si gate [11]). Results of these
different transistors were very comparable, confirming that
RINCE becomes negligible for very wide transistors (above
about 4 μm in our experience). Groups of identical transistors
with separate terminals were used to study the influence of the
bias on the radiation effects. Large ELT transistors (W =
10 μm) with different L and without ESD protection were used
for charge pumping and noise measurements, performed in a
restricted set of conditions (bias, temperature, transistor type).
Comparison of the results with those obtained for the complete
study of the arrays, where terminals were ESD protected, con-
firmed that ESD structures did not influence the results. We
focused our study on standard-Vth devices; however, results on
low- and high-Vth transistor arrays, not reported in this paper,
yielded comparable results.
B. Radiation sources and exposure conditions
The CERN X-ray irradiation system (Seifert RP149) was
used for most of the current/voltage measurements and for the
charge pumping study. This system uses a 3 kW tube with a
tungsten target and a 150 μm Al filter to generate ~ 10-keV X-
rays, a typical configuration for radiation-effects studies on
semiconductor devices [12]. Bare Si chips were kept under
bias and measured with the help of a wafer prober inside the
irradiation cabinet. A thermal chuck controlled the tempera-
ture of the samples between –30 and 100 °C. DC bias was
applied during irradiation and post-exposure annealing, most
often in a “diode” bias configuration where |Vgs| = |Vds| =
1.2 V, which was observed to generally produce the largest
degradation [1]. Unirradiated control devices (data not shown)
show less than 2% bias-temperature instability for the bias,
times, and temperatures used in these irradiation and annealing
studies. Other bias configurations were the “off” (all shorted)
and “on” (where |Vgs| = 1.2 V and Vds = 0 V) conditions.
The Pelletron at Vanderbilt University was used for irradia-
tion with 1.8 MeV protons of devices used for the noise study.
Here samples were mounted in ceramic packages and exposed
in the “diode” bias configuration at room T. Both facilities
allow for a very high dose rate, which is necessary to reach
TID levels up to 1 Grad(SiO2). Typical dose rates were ~ 9
Mrad/h for the X-ray tests, with similar, effective dose rates
for proton irradiation.
C. Measurement details
Static transistor measurements, mainly output and current
transfer characteristics, were taken with a semiconductor pa-
rameter analyzer (Keithley 4200A or HP 4155 for the X-ray
study, HP 4156 for the proton tests). The main parameters we
extracted from the raw data are the maximum current Ion (drain
current for |Vgs| = |Vds| = 1.2 V), the maximum transconduct-
ance in the linear regime (Gmmax), the subthreshold swing
(SSW), and the threshold voltage (Vth). The absence of radia-
tion-induced leakage current in NMOS FETs in the studied
technology allowed the precise extraction of all parameters,
and in particular Vth and SSW, in these devices.
Charge Pumping (CP) measurements were performed at the
University of Padova. This technique can be used to determine
the average interface-trap density of MOS transistors, as well
as the spatial distribution and energies of these traps [13],
[14]. The low-frequency 1/f noise of ELT PMOS transistors
with channel lengths of 60 nm to 480 nm was measured at
room temperature at frequencies f of 3 Hz to 390 Hz using the
apparatus and procedures shown in [15],[16], before and after
1.8-MeV proton irradiation, at Vanderbilt University. During
1/f noise measurement, devices were operated in the linear
mode of operation, with Vds = –30 mV and 0.1 ≤ |Vgs – Vth| ≤
0.5 V. At least two devices of each type were evaluated for all
experimental conditions, with typical results shown below.
III. RESULTS AND DISCUSSION
Transistors in the studied 65 nm technology exhibit two dis-
tinct radiation-induced degradation mechanisms. During irra-
diation, the drain current Ion decreases in both P and NMOS
transistors with inverse proportionality to the gate length
(Fig. 1), mainly because of the increase of the parasitic series
resistance. In addition, short-channel transistors also exhibit a
threshold voltage shift ΔVth that depends on bias and tempera-
ture. In PMOS FETs exposed at room temperature this mainly
takes place during post-irradiation high-T annealing (Fig. 2).
This behavior makes it possible to separately study the series
resistance and the threshold voltage shift: in this paper we will
hence focus our attention mainly on PMOS transistors.
A. Drain current decrease during irradiation
A relevant Ion decrease is observed during irradiation of
both N and PMOS FETs. This performance degradation gets
milder during cold exposures (-30 °C), is consistently larger
for shorter length devices, and almost negligible for gate
lengths above 1 μm (Fig. 1). In degraded transistors the most
relevant parameter change is the transconductance Gmmax,
whose evolution closely follows changes in current drive. A
detailed observation of the Gm = f(Vgs) curve at different doses
(not shown here) displays comparable decreases for the full
range of Vgs, suggesting that mobility degradation is not the
dominant mechanism – or the degradation would be smaller at
Fig. 1. Radiation-induced degradation of the current in strong inversion
(|Vgs| = 1.2 V) and in the linear regime (|Vds| = 20 mV) for PMOS and NMOS
ELT transistors “diode”-biased during exposure at room temperature up to
400 Mrad(SiO2).
PreRad 107
108
TID [rad]
-40
-35
-30
-25
-20
-15
-10
-5
0
5
I ON
[%
]
pMOS (ELT)
W=1.32 m;L=0.06 m
W=1.42 m;L=0.12 m
W=1.63 m;L=0.24 m
W=1.84 m;L=0.36 m
W=2.43 m;L=0.6 m
W=3.24 m;L=1 m
W=8.95 m;L=4 m
PreRad 107
108
TID [rad]
nMOS (ELT)
Vgs
=1.2V; Vds
= 20mV
after F.Faccio et al., “Radiation-Induced Short Channel (RISCE) and Narrow Channel (RINCE) Effects in 65 and 130 nm MOSFETs”, IEEE TNS 62, n.6, Dec.2015
after F.Faccio et al., “Influence of LDD spacers and H+ transport on the total-ionizing-dose response of 65 nm MOSFETs irradiated to ultra-high doses”, presented at NSREC 2017, to be published in IEEE TNS Jan.2018
34
2642 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 64, NO. 10, OCTOBER 2017
Fig. 3. Transfer characteristics of n (a), (b) and p (c), (d) types ofMOSFETs in saturation mode (|VDS| = 1.1 V) with respect to TID.(a) and (c) are for DUT1 with the largest W/ L ratio (3 µm/30 nm).(b) is for nDUT10 with the smallest W/ L ratio (100 nm/1 µm) and (d) isfor pDUT9 with 100 nm/1 µm. ID0,of f is the off-current (VGS = 0) before
irradiation.
a higher amount of oxide trapped charges. This decreases the
threshold voltage, increases the on-current, and degrades the
subthreshold swing. The negative threshold voltage shift also
leads to a higher off-current for short-channel nMOSFETs.
Long-channel nMOSFETs present a lower off-current, which
could be explained by the compensation of the interface
charge trapping. The bias loss between 940 and 963 Mrad for
pMOSFETs explains the slight fluctuation of the extracted
parameters for the last two TID steps.
B. TID Effects on Transfer Characteristics
Fig. 3 plots the normalized drain current of nDUT1,
nDUT10, pDUT1, and pDUT9 with respect to TID. The
on-current (ID,on) is defined at |VGS| = 1.1 V, and the
off-current (ID,off) at VGS = 0. nDUT1 has a negative
threshold voltage shift and a monotonic on-current increase,
as shown in Fig. 3(a). Nevertheless, as seen in Fig. 3(b),
the threshold voltage of nDUT10 first decreases and then
increases, leading to the corresponding on-current evolution.
Both irradiated nDUT1 and nDUT10 have a significant
off-current increase. In contrast, the threshold voltage of both
pDUT1 [Fig. 3(c)] and pDUT9 [Fig. 3(d)] decreases with TID,
leading to a continuous on-current loss. Note that when the
threshold voltage of a pMOSFET decreases, it becomes more
negative and its absolute value increases. The off-current of
both pDUT1 and pDUT9 is not sensitive to TID and has a
slight off-current change.
C. On-current
The relative on-current variation is plotted as a function
of TID in Fig. 4. All nMOSFETs present an on-current
improvement at a lower TID (∼ 10 Mrad) in Fig. 4(a). Then
Fig. 4. TID-induced on-current (ID,on ) variation of (a) n and (b) p typesof MOSFETs in saturation mode (|VDS| = 1.1 V). ID,on is obtainedat |VGS| = 1.1 V.
Fig. 5. Drain current (ID) versus the overdrive voltage (VG − VT0) with
respect to TID for (a) and (b) n and (c) and (d) p types of MOSFETs insaturation mode (|VDS| = 1.1 V).
for nDUT1-5, the on-current increase continues until reaching
a ∼ 15% increase. However, the on-current of nDUT6-10 starts
to decrease at a higher TID. nDUT10 with the smallest W/ L
ratio (100 nm/1µm) evidences as the most dynamic case with
a maximum on-current loss of ∼ 25%. Fig. 4(b) shows that
pMOSFETs are grouped into two distinct categories: pDUT1-6
and pDUT7-9. In each group, the narrower the channel,
the higher the on-current loss. Eventually, most of pMOSFETs
show a less than 30% degradation in the on-current. Narrow-
channel pMOSFETs with a smaller W/ L (pDUT6 and
pDUT8-9) present a more than 50% on-current loss.
Nevertheless, these results are much better compared with
the commercial 65-nm bulk CMOS technology from the same
foundry [9].
Both the threshold voltage (VT0) shift and the free carrier
mobility (µ ) reduction influence the on-current. To extract
VT0 and µ ,√
ID–VG curves are extrapolated linearly at the
maximum slope. The intercept at the VG axis is defined as VT0.
The slope of the linear extrapolation provides insights into µ .
after C.Zhang et al., “Characterization of GigaRad Total Ionizing Dose and Annealing Effects on 28 nm Bulk MOSFETs”, IEEE TNS 64, n.10, Oct.2017.
PMOSNMOS
Short channel
Narrow channel
35
28nm shows pronounced RINCE but little RISCE
Results from Scaltech28 (INFN) and GigaRadMOST (SNSF)
CHATTERJEE et al.: BIAS DEPENDENCE OF TOTAL-DOSE EFFECTS IN BULK FinFETs 4477
Fig. 1. 3D schematic illustration of a triple-well bulk FinFET.
Fig. 2. (a) Top-down SEM image of the structure of the FinFETs. (b) Cross-
sectional TEM image of an nMOS FinFET showing the fin structure [15].
III. EXPERIMENTAL DETAILS
Transistors were irradiated at room temperature with
10-keV X-rays using an ARACOR Model 4100 irradiator at
a dose rate of 31.5 krad SiO min to a cumulative dose of
300 krad SiO . The bias conditions during irradiation corre-
spond to 1) on-state (ON) and 2) off-state (OFF) for inverter
and 3) transmission gate (TG) operation. Other bias conditions
were also tested, with the source, drain, and gate either at 0 V
(ALL-0), and/or at negative gate bias, as shown in Table I. The
body and triple-well were held at 0 V and 0.7 V respectively
during all the irradiation experiments.
Current-voltage ( ) characteristics were measured with
an Agilent 4156 semiconductor parameter analyzer on unpack-
aged wafers. During measurements, both the source and the
Fig. 3. characteristics as a function of dose for irradiation at a dose
rate of 31.5 krad SiO min. The drain was biased at 0.7 V, the fin width is
13 nm, and the devices comprise 4 fins.
substrate were grounded, and a 50 mV bias was applied to the
drain. The gate voltage was varied from 0.3 V to 0.7 V. The
body terminals were grounded for all the bias conditions. The
triple-well (deep n-well) terminal was held at . Annealing
experiments were performed at room temperature, with all ter-
minals grounded, to understand the subthreshold slope modula-
tion in irradiated FinFETs.
IV. EXPERIMENTAL RESULTS AND DISCUSSION
A. Off-state Leakage Current
Total-ionizing-dose irradiation induces net positive trapped
charge in oxides and interface traps at silicon/oxide interfaces.
The extremely small increase in post-irradiation gate leakage
observed in these transistors suggests that there were no leakage
paths created in the gate insulator stack (Fig. 3). There is little
shift in the threshold voltage following irradiation, indicating
that the trapped charge in the gate insulator is quite small. How-
ever, similar to trends observed in planar CMOS technologies,
the radiation-induced charge trapping in the STI oxide still leads
to macroscopic effects, such as the drain-to-source leakage cur-
rent, and ultimately limits the radiation tolerance of CMOS cir-
cuits [16]. Thus, the post-irradiation response of these bulk Fin-
FETs is dominated by buildup of charge in the isolation oxides
(the shallow trench isolation) around the transistors. If an elec-
tric field exists across an insulator during total dose irradiations,
electrons and holes in the insulator will immediately begin to
transport in opposite directions. Electrons are extremely mobile
in the silicon oxides and are normally swept out of it in picosec-
onds [17], [18]. Holes generated in the silicon oxides transport
much slower than electrons and a substantial fraction may be
trapped. As a result, hole trapping usually determines the tran-
sistor response after irradiation. As the electric field increases,
the probability that a hole will recombine with an electron de-
creases and the fraction of un-recombined holes increases [19].
4478 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 60, NO. 6, DECEMBER 2013
Fig. 4. Schematic showing the trapped charges in the isolation oxide in a
multi-fin FET (representative figure only, not to scale). [20].
The electric field extends into the trench region and plays a piv-
otal role in both the initial separation of electron–hole (e–h)
pairs and the charge migration. Fig. 4 shows an illustration of
the trapped charges in the isolation oxides.
In contrast, for SOI FinFET devices, positive trapped charge
in the BOX is a greater concern. In FD SOI devices, charge trap-
ping in the BOX can affect the device degradation through a
direct coupling effect between the front and back interfaces. In
non-planar multiple-gate devices, the coupling behavior is dif-
ferent and complex, especially as fin widths and gate lengths
decrease below 130 nm. A lateral coupling effect induced by
the lateral gates appears in addition to the vertical coupling ef-
fect of single-gate devices [6], [8]. These complex electrostatic
coupling effects are strongly geometry dependent.
Fig. 5 shows a schematic diagram of the trapped charges in
the buried oxides. A wide-fin FinFET is similar to a “pseudo”
single-gate FD SOI transistor. Its electrostatic behavior is dom-
inated by the front and back gates. The electrostatic control of
the lateral gates over the potential in the active silicon film and
in the BOX under the silicon-film/BOX interface is weak [11].
Ionizing radiation exposure induces a positive charge buildup in
the BOX of SOI devices, which increases the back-gate surface
potential [21]–[23]. In single-gate FD SOI devices, the charge
trapped in the BOX acts as a positive back-gate bias. Because
of the strong vertical electrostatic coupling effects, the front
surface potential increases and induces a negative front-gate
threshold voltage shift. For narrow fin devices, on the other
hand, the primary effect is the screening of the trapped charges
into the BOX due to the strong electrostatic control of the lateral
gates when they are close to each other. The electrostatic poten-
tial in the silicon body and in the BOX under the Si fin/BOX
interface is dominated by the lateral gates, thereby limiting the
amount of radiation-induced hole charge trapped in the middle
of the channel [6]–[8].
The bias applied to the transistor terminals during exposure
to radiation is a critical parameter influencing charge trapping.
For fully depleted SOI transistors, the worst-case bias is fab-
rication process dependent. For partially-depleted SOI tran-
sistors, the worst-case bias conditions are when source and
drain are at with other terminals grounded (the trans-
mission-gate configuration) [24]. For some fully-depleted SOI
Fig. 5. Schematic showing the trapped charges in the BOX in a SOI
omega-FET (representative figure only, not to scale). [7].
technologies, the worst-case bias condition is the same as that
for partially-depleted SOI transistors, the Transmission-Gate
configuration, causing the most radiation induced charge trap-
ping in the buried oxide [25]. However, for other technologies,
the worst-case bias condition was determined to be the ON
bias configuration (gate at , other contacts grounded) [26],
[27].
Fig. 6 shows the pre- and post-irradiation character-
istics for the five bias conditions under consideration. The
highest increase in off-state leakage is observed for the OFF-
state bias condition, for all doses considered. The pre-rad off-
state leakage is 1 pA and increases to 25 nA after a cu-
mulative dose of 300 krad SiO . The smallest shift is ob-
served in the case of the negative gate and ALL-0 bias condi-
tion. The off-state leakage for these bias conditions increased
from 1 pA (pre-rad) to 100 pA 300 krad SiO . Fig. 7
shows vs. cumulative dose for the bias conditions under
consideration. At low dose, the high electric fields in the cor-
ners of the shallow trench isolation are partly responsible for
the increased transistor leakage current [16], [25]. The simu-
lations detailed in Section V show that the electric field at the
trench corners is highest for the OFF state bias. There are a
large number of oxygen vacancies in the STI close to the sil-
icon/oxide interface due to the out diffusion of oxygen near
the oxide and the lattice mismatch at the surface [16], [19].
These oxygen vacancies can act as trapping centers. Some
fraction of the holes will be trapped there, with the fraction
strongly related to the electric field in the oxide during irra-
diation [16], [26]. The overall response of bulk FinFETs is
similar to planar bulk MOSFETs in terms of the parasitic tran-
sistor dominating the radiation response of the devices.
B. Parasitic Transistor Threshold Voltage Shift
At a high dose, the leakage current becomes relatively inde-
pendent of gate voltage, which means that the parasitic tran-
sistors play an important role for the leakage current. The ac-
tual threshold voltage shift is calculated after subtracting the
leakage current from the measured characteristics. The in-
significant shift in threshold voltage shift observed is because
of a relatively small amount of charge trapping in the gate insu-
lator stack. As the fin width is less than 15 nm, the electrostatic
CHATTERJEE et al.: BIAS DEPENDENCE OF TOTAL-DOSE EFFECTS IN BULK FinFETs 4479
Fig. 6. characteristics as a function of dose for irradiation at a dose rate of krad SiO min for the various bias conditions under consideration.
The DUT is 4-fin 90 nm FinFET. (a) ON state. (b) OFF state. (c) ALL-0 state. (d) Negative gate. (e) Transmission gate (TG).
potential is predominantly controlled by the lateral gates, fully
depleting the silicon fin in the ON state. The longitudinal pen-
etration of the fringing electric field from the source and drain
to the channel, e.g., the drain-induced virtual substrate biasing
(DIVSB) effect [8], is prevented. Fig. 8 shows the threshold
voltage shifts in the DUTs after subtracting the leakage current
from the measured characteristics.
Fig. 9 shows the subthreshold swing ( )
of the 90-nm-gatelength FinFET as a function of dose for the
worst case and best case bias condition. The swing was mea-
In FinFETs source-drain leakage due to oxide charging occurs
36
A Tale of Two Commercial Processes
• Typically comes about when they fix a leakage problem
• Impossible to say if TID resilience remains a permanent feature of the technology going forward
-0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.210-6
10-5
10-4
10-3
10-2
10-1
100
101
102
103
104
Medium-Vth Device
On-state
Vgs = 1.0 V
Dra
in C
urr
en
t (m
A/m
m)
Gate Voltage (V)
Initial
100 krad(SiO2)
200 krad(SiO2)
300 krad(SiO2)
500 krad(SiO2)
1000 krad(SiO2)
Two snapshots of a commercial 14/16-nm FinFET technology show very different TID results
-0.2 0.0 0.2 0.4 0.6 0.810-8
10-7
10-6
10-5
10-4
10-3
10-2
10-1
100
LVT Dev 1
I ds (
mA
)
Vgs (V)
pre
100 krad
200 krad
300 krad
400 krad
500 krad
1 Mrad
Vg = 0.8 V
Vd = Vs = Vb = 0 V
W = 0.192 mm
L = 0.014 mm
Courtesy of Michael P. King, Sandia 37
Approach to R and D for Future IC Technologies
1. List of topics were drawn up:
• Pixel detector readout (hybrid) – monolithic is in Si WP• Strip detector readout• Calorimeter readout• Muon Detector readout• Control and monitoring
Slow control ADC, DACs etc• Trigger generation• Fast timing• Power supply distribution
DCDCLinear regulators
• Serial powering
• NB This list covers the applications of the Technologies. They must inform the technology choice but this R and D focusses on the technologies themselves, not the particular circuit solutions
38
Approach to R and D for Future IC Technologies
2. Some key players were identified and contacted
3. A Vidyo meeting was held on 5th
March to gather inputs:
4. Each presentation was followed by a discussion
/indico.cern.ch/event/704624
39
Some preliminary observations/conclusions
• Future experiments will rely on ASICs
• ASIC interconnect technology (TSV, bump bonding) should be included in the work package as it depends on or is inseparable from the ASIC Technology chosen
• As a community we would benefit from accessing newer technologies but large projects ($$$$) are needed to justify the effort to access such processes
• The provision of tools via Europractice has been vital for our community
• Access to technologies and the appropriate tailored design kits is also essential
• Many of our designs are still done in the ‘old fashioned’ analog way. This is already changing e.g. RD-53, MPA, SSA, Velopix etc
• Qualification of devices for operation sometimes falls between the cracks of the design team and the users
• Could we benefit from closer collaboration with other fields (ITER, synchrotrons..)?
40
Preliminary list of needs for IC Technology development
• CERN Microelectronics should continue to study and select (along with the design community) appropriate technologies to act as a common platform for developments.
• A deep understanding of radiation effects is essential
• Equally important is the provision of tailor made design kits for our community
• Designs can only be done using the latest tools
• Europractice needs our strongest support!
• Training in the new tools is essential to make the most of new opportunities
• More young (digital) engineers are needed
• We must learn to work in large collaborative efforts. 41
A few unpleasant practical facts - inconvenient truths?
• Our community represents a tiny fraction of the world market in microelectronics. The only leverage we have is good will.
• NDA’s take an age to negotiate but are an absolute necessity to enable collaborative work. They must also be scrupulously respected.• Confidential foundry information • Use restrictions• Export controls
• Accessing such technologies will not be cheap, but not going there could be even more costly
42
One optimistic scenario:
• FCC or CLIC gets approved and 1 000m2 of tracker needs to be constructed.
• We agree to use the same technology for all 1 000m2
• This represents 28 000 300mm wafers @ $1k per wafer (assuming 50% yield)
• Because of the extreme radiation these should be replaced every 5 year or so.
• We suddenly become an interesting client for a leading edge company with wafer stacking.
“You may say I’m a dreamer…”
“…but I’m not the only one…”
Jacques Dubochet, Nobel Prize for Chemistry 2017 after John Lennon.
https://www.nobelprize.org/nobel_prizes/chemistry/laureates/2017/dubochet-lecture.html
43
So what’s next?
• Complete work on understanding radiation tolerance of current processes and define qualification protocol
• Start to test 28nm and FinFET devices to understand the physics
• Put in the place the necessary infrastructure to allow us access to these processes• NDA’s• Technology files and support• Foundry access • Training
• Continue to work as a community to restructure our efforts to permit the design of future complex ASICs
• More fledged out plan to be presented at second meeting44
Thank you for our attention!
45