Transcript Presentazione standard di PowerPoint

Accelerated Testing
• If the required failure rate is ~ 100 FIT or less, then the time required
to observe one failure in 100 devices is approximately 100,000 hr
(11.4 yrs)
• Thus it is impossible to test the required reliability under normal
operating conditions
• This necessitates means to accelerate the mechanisms that cause
devices to fail
1
• 5 common stresses are used to accelerate device failure
– temperature
– voltage
– current
– temperature cycling to accelerate mechanical failure of chips and
assembly package
• In such studies, different failure mechanisms may be accelerated by
different level of stress even for the same type of stress
• A device may fail at normal operating conditions because of 2 completely
different mechanisms
• Under the applied stress, one of these failure modes may be accelerated
much more than the other
• Thus, we only see 1 failure mode in those tests
• After successfully eliminating the mode we may only the failure rate by
small factor under normal operating condition
• Adequate studies must be done under normal operating conditions to
satisfy that no failure mechanism remain that were not accelerated by
the applied stress
2
Temperature Acceleration Arrhenius Model
• Many failure mechanisms involve chemical or physical processes that can
be accelerated by raising the temperature
R  R0 exp(  Ea / kBT )
• Ea is the activation energy
• If some parameter of IC changes as a function of time, and if the IC fails
when the parameter exceeds certain value.
• The Rates at two different temperatures are related as
E
t1 R2

 exp  a
t 2 R1
 kB
 1 1 
  
 T1 T2 
• The IC would fail when the destruction reaction proceeds to some value
equal to the failure criterion
• Rtf = constant
t f  t f 0 exp( Ea / k BT )
• Thus a plot of ln(MTTF) versus 1/T, the slope will correspond to the
activation energy
3
Voltage and Current Acceleration
• Voltage and current are effective acceleration stresses
• Voltage stress cause failure in devices due to:
– dielectric breakdown
– interface charge accumulation
– charge injection
– corrosion
• Most studies indicate that the reaction rate, Rx, of the failure mechanism
is proportional to a power of the applied voltage
R(T ,V )  R0 (T )V  (T )
•
•
•
•
where R0(T) if thermally activated
 varies between 1 to 4.5
For dielectric breakdown, a different type of acceleration occurs
For a given field, a certain fraction of devices fail in a very short time, of
the order of seconds
• Very few additional failure occurs as the field is maintained
• If the applied field is increased, additional fraction failures occur
4
• In such cases, operation at an increased voltage is more in the nature
of screening rather than accelerating the failure mechanism
• Increased current level is used to accelerate failures caused by
electromigration in metallic conductors
R  R0 J  (T )
• 1<<4
Stress-Dependent Activation Energy
• (T) describes the temperature dependence of the reaction rate
• For failure accelerated by voltage or current stressing, the activation
energy is dependent on the applied bias
• The Eyring model states that
 a(T ) S 
 Q 

R  sinh 
 exp 
k
T
k
T
 B 
 B 
• where S = applied stress
a(T )  k BT (T )  k BT ( 0 
1
k BT
)
5
• Q is related to the Arrhenius activation energy
Q  Ea 0  a(T )S B
• SB is the breakdown stress, the value of the applied stress where
failure of the device occurs essentially instantaneously
• The reaction rate can also be expressed as
 E  S 
R  sinh[  (T ) S ] exp   a 0 1 B  exp(  0 S B )
k BT


• under conditions of low stress, R reduces to
 E  S 
R   (T ) S exp   a 0 1 B  exp(  0 S B )
k BT


• At high stresses
 E  S  S 
R  exp[ 0 ( S  S B )] exp   a 0 1 B 1 
k BT


• From the above expressions Reaction Rate is a function of the applied
stress, the effective activation energy will decrease with increased
rate
6
Humidity-Temperature Acceleration
• Presence of water vapor in the chip environment introduces a new
variety of possible failure mechanism
• Water vapor quickly permeates plastic packaging material
– 1. water vapor transports contaminants from surface of package
through the plastic
– The chip is then exposed to water vapor and various contaminants
– 2. Diffusion of the contaminated water vapor through the passivation
layer of the chip
– This step can be speeded up if the passivation contains defects or
cracks
– The penetration of water vapor through the passivation layer
determined the reaction rate
– 3. Once water reaches the metallization level, electro-chemical
corrosion process can occur
– The ions needed for this corrosion process can arise from the
contaminants which diffused through the passivation layer.
7
– If the intermediate dielectric of the chip is a phosphorous-doped
glass, the water vapor can extract the phosphorus from the
dielectric
– Electrochemical corrosion is a rapid process leading to
metallization failure
– This failure mechanism can be accelerated by increasing the
partial pressure of water vapor in the environment
8
Burn-in
• For mature products, the initial reliability studies would have
identified and eliminated failure processes so that steady-state failure
rate meets or exceeds design goal
• However, the manufactured devices still show existence of continuing
early failure
• Generally, manufacturing defects cause the infant mortality failures
e.g. pinholes, photoresist or etching defects resulting in near-opens or
shorts
• Contamination on the chip of the package, scratches, weak chips or
wire bonds, partially cracked chips or packages
• The purpose of the burn-in procedure is to operate the devices for
some time during which most of the devices that are subject to
• Infant mortality failure actually fail
• The conditions during burn-in presumably accelerate the failure
mechanisms that contribute to infant mortality failure
• Studies of infant mortality under increased T conditions show that
infant mortality have an activation energy of 0.37 to 0.42 eV
9
Properties of Metal-Oxide Silicon (MOS) System
• To understand the MOS system the step is to derive the energy band
diagram
• We note that at thermal equilibrium the Fermi level is constant
• The energy band diagram for a separated system is shown below:
E0
E
C
q
= 4.1 eV
q = 4.1 eV
M
8 eV
E
C
E
Ei
fm
E fs
EV
EV
10
• When connected the Fermi
level will be constant. The
Fermi level in Si depends on
doping level of the Si
• For Ei - Ef = 0.29 eV the band
diagram is shown
Flat-band Condition
Accumulation
E fm
EC
Ei
E
EV fs
E fm
EC
Ei
Efs
EV
11
Weak Inversion
Depletion
EC
EC
Ei
Efs
EV
E fm
Ei
Efs
EV
E fm
Strong Inversion
q p
E fm
q s
EC
Ei
Efs
EV
12
• Strong inversion condition stipulates that qP = qs
• Thus to turn an MOS into strong inversion -- a condition for the
formation of a conduction channel or inversion layer under the gate, a
minimum band bending of 2 qP is required
• The threshold voltage VT is then represented by
1
VT  VFB  2 |  p | 
Cox
4 s qN A |  p |
• We can gain further insight into the MOS by realizing that it is
basically a capacitor
• The charges in the Si of the MOS system can be expressed as
Q  Cox [(VG  VFB )  ( s   p )]
• The terms in the square brackets constitute the voltage drop across
the oxide. Since part of the charges are associated with the dopants
in the depletion layer, therefore free carrier concentration in the
inversion layer is
Qn  Cox (VG  VFB  2 |  p |)  4 S N A |  P |
13
• The second term is the charge in the depletion layer VG here is the
gate bias required to produce a band bending of 2|P| and is
therefore equivalent to the threshold voltage
• The proof for the expression of charge in Si is
 qN A xd

s
14
• Where E is the electric field in Si

• The total band bending in Si is | s |  E0 xd
• From Poisson’s equation we have
|  s |
qN A xd
2 s
• For xd = xdmax we have s = 2p and xd max 
• Charge is the Si layer is qNAxdmax, thus
4 p
qN A
Qd  4qN A p
• The total charging voltage is V  Vox  Vsi
• where V    
si
s
p
VG  VFB  Vox  ( s   p )

Vox [(VG  VFB )  ( s   p )]
Eox 

xox
xox

Es 

 ox Eox
s
Cox' 
 ox
xox
15

'
• Substituting into the previous equation s Es  Cox
[(VG  VFB )  (s   p )
'
'
'
• From Poisson’s equation we obtain   s E  Qs  Qn  Qd
• Concentration of free carriers in the inversion layer is Qn'  Cox (VG  VT )
Oxide and Interface Charges
• Charges at Si-SiO2 interface and the oxide may influence the threshold
voltage through the modifications of the flat-band voltage. If the density
of charge at x=x1. It induces an equal and opposite charges divided
between silicon of the metal gate. The closer is x1 to xox (the Si-SiO2
interface), the greater will the fraction of induced charges at Si-SiO2
interface and the oxide may influence the threshold voltage through the
modifications of the flat-band voltage. If the density of charge at x=x1. It
induces an equal and opposite charges divided between silicon of the
metal gate. The closer is x1 to xox (the Si-SiO2 interface), the greater will
the fraction of induced charge within the Si.
• The induced charge changes the charge stored in the Si at equilibrium
therefore it alters the flatband voltage.
• The size of VFB can be found using Gauss’ law to obtain the value of the
gate voltage that causes all of the oxide charge Qoxto be mirrored in the
gate electrode.
16
• Under this condition the field is constant between x = 0 and x =x1 and
0 for x > x1.
• For 0 < x < x1 we have the following relationship

Qox'
Eox  
, 0  x  x1
 ox
VFB
x1 Qox'

xox Cox'
• The result can be generalized to account for shift in VFB using an
arbitrary charge distribution
xox
VFB
1
 '
Cox
x
0 xox  ( x)dx
• where (x) is the volume charge density at x
Origins of Oxide Charges
• There are at least 4 distinct types of charges in the oxide-silicon
system
– Qf ’ -- fixed interface charge density
– Qot’ -- oxide trapped charge density
– Qit’ -- interface trapped charge density
– Qm’ -- mobile charge density
17
• Qf ’ is positive and is located within a very thin (1 - 2nm) laryer fo nonstoichiometric silicon oxide (SiOx)
• Qot’ can be both positive and negative, typically predominantly
negative. Located in traps distributed throughout the oxide layer.
Distortion in the C-V curve is due to unstable charges at the interface.
• The trapping sites Nit (cm-2) are located at the Si-SiO2 interface and
have energy levels within the bandgap with density Dit cm-2eV-1
• To relate behavior of these traps to the distorted C-VG, as shown in
the following figure, consider an oxide-Si interface characterized by
interface trapping levels at energy Es. The gate voltage causes the
Fermi level at the surface to cross Es, the charge state of these levels
will change. This introduces a voltage dependent term, Q’it/Cox’, into
the equation for VFB above making both flatband and threshold
voltages vary with VG that led to the distortion the C-V curve.
• For Nit > 1010cm-2 is generally unacceptable for reliable device design.
Using modern MOS technology, Nit is reduced by annealing device in
forming gas (90% N2 and 10% H2)
18
Mobile charge results from alkali-metal ions particularly sodium.
They induce VFB.
• The alkali ions have sufficient mobility when relatively low gate biases
are applied. The mobility increases with temperature and thus
magnifies the problem of VFB instability at high temperatures. The
ions are positively charged, therefore -VG draws the ions to the metalSiO2 intrface where their effect is minimal. +VG pushes them to the SiSiO2 interface where their effects are most significant.
• For voltage stability of 0.1V, less than 2x1010cm-2 mobile ions can be
tolerated. Mobile ions can be avoided by careful processing and
oxidation in HCl that immobilizes alkali ions.
Hot Electron Degradation
• The assessment and improvement of reliability on the CKT level
should be based on both failure mode analysis and the basic
understanding of the physical failure mechanisms.
• Processes such as electromigration and electrostatic discharge cause
catastrophic changes in device characteristics
19
• Other mechanisms such as hot-electron effects cause noncatastrophic failures which develop gradually over time and change
CKT performance
Scattering of Channel Hot-Electrons into the Oxide
• In order for electrons to obtain enough kinetic energy to be injected
into the oxide, it has to
– gain K.E.
– its momentum redirected towards the oxide through a quasielastic collision
– following the collision, the electron must travel to S-SiO2 without
further collision
• These processes are statistically independent, the injection
probability is obtained as the product of the probabilities of each
event
• MOSFET gate current is made up of electrons injected into the gate
oxide by quasi-elastic scattering
• It consists of electrons that overcome the image potential well in the
oxide and reach the gate electrode
20
Hot Carrier Effects
• Advances in VLSI is achieved through down scaling of device
dimension such as channel length, junction depth and gate oxide
thickness without proportional scaling of power supply voltage
• Decrease in device dimensions results in significant increase of the
horizontal and vertical electric fields in the channel region
• Electrons or holes with high K.E. may be injected into the gate oxide,
degrading I-V characteristics of MOSFETs.
• This is one of the important factors that limits the maximum
achievable device densities in VLSI circuits
• Hot-carrier damage results in:
– trapping of carriers on defect sites in the oxide
– creation of interface states at Si-SiO2 interface leads to
degradation in transconductance, shifts in threshold voltage and
decrease in drain current capability.
21
Oxide degradation Mechanism in MOS system
• Cause by injection of high-energy electrons and holes into the gate
oxide near the drain
• The damage is in the form of localized oxide charge-trapping and
interface trap generation
• Recent experimental evidence shows that hot-carrier related
degradation can occur in deep-submicron devices with Leff = 0.15m
at drain voltage as low as 1.8V. Therefore hot electron degradation
may occur even with significant reduction in drain voltage.
• The continuing technology thrust must therefore accompanied by
some limitations to ensure hot-electron reliability
• Hot-carrier injection causes degradation in the transconductance,
shift in threshold and decrease in sub-threshold drain current
• There are many disagreement concerning the physical degradation
mechanisms due to the lack of a reliable and sensitive techniuqe to
evaluate hot-carrier damage at the interface. Moreover, hot-carrier
induced oxide damage is very localized, the interpretation of the
analysis is complex
22
Injection of Hot-Carriers into Gate Oxide
• Hot-carriers are electrons and holes that have a much higher K.E. than
average carrier population
• E’s in S.C. at equilibrium mostly have energy about kBT above EC. At
equilibrium, K.E. of carriers that encounter large E may gain
significant K.E. in a short distance. Thus E- EC = kBTe where Te is the
effective electron temperature. There are 2 distinct modes of
electrons injection in nMOS
– Substrate hot-electron effect (SHE)
– Channel hot-electron effect (CHE)
• Substrate hot-electrons are derived from leakage current. Electrons
generated in the channel depletion region or diffusing from the bulk
neutral region of the substrate drift toward the Si-SiO2 interface and
gain K.E. from the high field in the surface depletion region
• The energetic electrons may overcome surface energy barrier and
inject into the gate oxide.
• Some of the injected electrons are trapped in the oxide, resulting in a
relatively uniform oxide charge accumulation that shifts the threshold
voltage over time.
23
• The SHE is observed mainly in long-channel MOSFETs. As channel
length decreases, SHE decreases since a large fraction of the hotelectrons generated in the substrate region are swept into the source
and drain regions instead of the device surface
• CHE is more pronounced at large VDS. Electrons reaching Si-SiO2
interface with large K.E. may surmount the energy barrier. Electrons
and holes generated by impact ionization also contribute to charge
injection into the oxide. Hot-electron current and oxide degradation
occurs mainly at the drain end. From the figures, the increased
density of equipotential lines leads to larger horizontal field. Hotelectrons and hot-holes can be injected into the oxide interface with
the aid of the vertical field or with their K.E. energy alone.
• Injected current density is
J e ( x)   n( x, y ) pinj ( x, y )dy
y
• Where n(x,y) is the local electron concentration at (x,y)
• pinj(x,y) is the spatial distribution of injection probability
• pinj(x,y) depends on several events that provide the electron with a
momentum directed towards the oxide interface and with a K.E.
sufficient to overcome interface potential barrier.
24
• Injection of hot carriers occurs mainly in a narrow injection zone at
the drain end of the device where lateral field reaches maximum.
• In log channel MOSFETs, the spatial extent of injection region and
magnitude of electric fields near the drain are largely independent of
the channel length, L.
• For short channel devices, the heavier dopin or shallower junctions
increase the electric fields in the drain region. Due to short channel
effects the channel current entering the drain increases more rapidly
than 1/L.
• Thus, devices with smaller geometries will be more sensitive to hotcarrier related degradation.
• Oxide degradation in the form of charge trapping which occurs in a
short distance of about 0.1 m. However, a large percentage of the
electrons entering the oxide are either scattered in the oxide and/or
returned to Si substrate by the opposing field.
• The charges that do not reach the gate electrode degrades the oxide
by charge trapping and interface trap generation.
25
Impact Ionization
by Hot-Electrons

• In saturation region a high E exists in the channel depletion region.
Electrons will, therefore, be accelerated by the field. Some move
horizontally and creates electron hole pairs (EHP) by impact ionization
near the drain.
• Impact ionization process creates an avalanche plasma consisting of
generated EHP in the pinch-off region
• The holes created are collected by the substrate constituting the drift
component of the substrate current
• The drain current that contributes to impact ionization substrate
current is a function of lateral electric field in pinch off region and VGS
and L
Impact
Ionization
26
• Some electrons and holes in the avalanche plasma can gain sufficient
K.E. to surmount Si-SiO2 potential barrier and become injected into
SiO2. Majority of the holes generated constitute substrate current of
MOSFET. Therefore, substrate current is considered a reliable and
convenient monitor of the amount of hot-carrier degradation in nMOSFETs
• To create EHP, hot-electrons
must have K.E. > (impact
ionization 
i

energy). Thus i / qEmis the distance the electrons must travel in E  Em
to gain energy i.
• The probability of an electron travelling a distance to gain the
required K.E. or more is
i
Pi  exp[ 
qEm
]
where  is hot-electron mean-free path. Since IDS is the total electron
flow in the channel. Rate of supply of hot-electrons with K.E. > i is

I sub  C1 I DS exp(  i )
qEm
where C1 is a weak function of the max. channel field Em.
27
Oxide Traps and Charge Trapping
• Concentration of oxide trapped charges and interface trapped charges
are changed by capture of excess electrons or holes by existing traps
in the oxide.
• The oxide charge distribution can also be changed by impact release
of the trapped electron or hole by a hot-carrier. The electron or hole
traps in the gate oxide are mainly Si “dangling bonds”. The dangling
bonds give rise to the electron and hole traps.
Interface Trap Concentration
• Interface trapped charge arises from a.) structural, oxidation-induced
defects; b.) metal impurities; c.) defects caused by radiation or hotcarriers
• Unlike other trapped charges, interface trapped charges are in
electrical communication with the underlying Si. Thus, influence of
interface trapped charge on electrical characteristics of MOSFETs
depends on its bias conditions.
• Generation of new interface trap is the primary cause of degradation
of MOSFETs. New traps generated by hot-electrons and hot-holes
through breaking:
28
– Si-Si and Si-O bonds
– breaking of H bonds at interface, releasing H and leaving dangling
Si- or O- bonds
– H released by hot-carrier impact migrates towards Si-SiO2
interface and is then trapped by proton traps.
Bias Dependence of Degradation Mechanism
• Oxide degradation takes place by carrier trapping in oxide due to hotelectron carrier injection and interface trap creation
• These are the two most significant degradation mechanism, but there
is no clear consensus on their relative contribution Hot-carrier related
device degradation reaches maximum when VGS = VDS/2. This
coincides with max. of substrate current, thus it is often linked to
impact ionization.
• When VTH shift is plotted as a function of gate voltage, the
degradation exhibits 2 local maxima:
– First peak: electron injection into SiO2 is max. resulting in charge
trapping
– Second peak: at VGS = VDS/2, corresponds to impact ionization of
electrons and holes.
29
• For VTH = VDS/2, VTH(t) = Atn where 0.5 < n < 0.7
• A depends on ISUB, IDS and processing parameters
Effects of Hot-Carrier Damage on Device Characteristics
• Trapped charges in gate oxide influence surface potential and thus
local flat band voltage
• C-V measurement can be used to measure total amount of trapped
charge in SiO2. Accumulating negative charge shifts the local VFB to
positive direction.
• Influence of interface traps generated by hot-carriers depends on the
instantaneous bias conditions. Since interface traps are in electrical
communication with the underlying Si substrate, occupation of the
traps depends on the Fermi level at the Si-SiO2 intface and energy
distribution of interface traps and the physical nature of NiT (whether
the trap is acceptor or donor type).
• In n-MOS most generated NiT are acceptor type, mostly located at the
drain end. Traps will start to be charged by electrons from substrate
as the surface is biased from accumulation into weak inversion, ant
then into strong inversion
30
• Once all traps are filled, their influence is similar to fixed oxide
charge, which is the case because for all practical purposed the
device are in strong inversion
• Surface mobility is decreased due to increased surface scattering
• Significant reduction in ID in linear region
• Less effect on ID in saturation region because once in saturation ID
is governed by the channel region between the source and pinch off
point
• Asymmetry between forward and reverse I-V curves also due to
localization of oxide damage near the drain end
– as n decreases gm decreases
– ID/ID0 decreases
– VT increases
Radiation Induced Interface Traps
• The major effect of ionization radiation on MOS devices is the
generation of positive oxide charge resulting from hole-trapping at SiSiO2 interface.
31
• Radiation consists of high energy particle such as electrons, neutrons,
protons and energetic x-ray and gamma ray
• Photons with E > EG of SiO2 can generate EHP. Some of the generated
carriers recombine. Most are driven toward the electrode by the
oxide field. Electrons rapidly drift toward the positive electrode and
flow out of the circuit. Holes drift much more slowly towards
negative electrode.
• Once holes reach Si-SiO2 interface a fraction becomes trapped
constituting the radiation induced positive oxide charge
32
Latchup in CMOS Circuits
• CMOS (Complementary MOS) is a very important class of circuits in
VLSI technology
• Advantages of CMOS circuits includes low power, high speed logic
circuits
VDD
VDD
PMOS
Vo
Vi
Vo
NMOS
Vib
Via
• In bulk CMOS CKTs both n- and p-channel MOSFETs exist side by side.
This is achieved by starting with a Si wafer of one type and creating in
it regions of the opposite type. In so doing, FETs are not the only
structures fabricated, pnpn devices consisting of parasitic bipolar
transistors are also created.
33
• Under normal operation the circuit performs as an inverter and the
bipolar portion can be ignored. However, if the bipolar circuit
switches from its normally high impedance state to its low impedance
state, the power supply sees a low impedance path to ground.
• If the current from the supply is not limited somehow, there might be
irreversible damages done to the circuit.
• However, even if the circuit is protected, the pnpn’s low impedance
state can still cause the circuit to malfunction.
34
Switching Mechanism (Streetman p. 401)
• The operation of a CMOS
circuit can be understood using
a two-transistor analogy
i
1
E
p
B
n1
C
p
1
J 1 i b 1 = ic 2
J2
2
J2
ic1 = i b 2
J3
n1
C
p
2
B
n2
E
2
Thus the parasitic bipolar transistors in a CMOS inverter can be viewed
has two interconnected BJTs
Each BJT can be modeled by Ebers Moll model
S PE
Pn
 N S PE
Pn
E
C
 ICS PC
Pn
B
CS PC
Pn
35
• The Ebers-Moll model uses the diode equations for the emitter and
collector plus extra terms that provide coupling between the emitter
and collector.
• For a forward biased diode, the excess carriers is
pE  pn (e qV / k BT  1)
• Thus the emitter current is
pC I ES
pE
  I I CS

(pE   N pC )
pn
pn
pn
pC I CS
pE
I C   N I ES
 I CS

( I pE  pC )
pn
pn
pn
I E  I ES
• Thus, IC and IE can be expressed as
I C   N I E  (1   N I ) I CS (e qVCB / k BT  1)
I C   N I E  I C 0 (e qVCB / k BT  1)
I E   I I C  (1   N I ) I ES (e qVEB / k BT  1)
I E   I I C  I E 0 (e qVEB / k BT  1)
36
Two-Transistor Analogy
• The analysis of the 2-transistor analogy is given below:
ic1  1i  I C 01  ib 2
ic 2   2i  I C 02  ib1
• where 1,2 are the emitter-to-collector current transfer ratios for the
transistors
• However, the sum of ic1 and ic2 is the total current through the device.
Thus, ic1  ic 2  i
i (1   2 )  I C 01  I C 02  i
I C 01  I C 02
i
1  (1   2 )
• As indicated in the above equations, as long as the sum, 1 + 2, is small
compared to unity, the current i is small at approximately the combined
collector saturation currents of the 2 equivalent transistors. As 1 + 2
approaches unity the current i increases rapidly. At this state both
transistors become saturated and will remain in the low resistance state
after the switching.
37
Variation of  with Injection
• Since the 2-transistor analogy implies that switching involves an
increase in the 1 and 2 to the point that 1 + 2 becomes unity, it is
important to understand how 1 and 2 depend on injection for a
transistor.
•  is the product of the emitter injection efficiency, , and the base
transport factor, B where
I En
I

;
B C
I En  I Bp
I En
• At very low currents,  is dominated by recombination in the
transition region of the emitter junction
• As the current increases, injection across the junction begins to
dominate over the recombination. This leads to increase in B due to
the saturation of recombination centers as excess carrier
concentration becomes large.
38
Forward Blocking State
• At this state the applied voltage is mainly across the reverse biased
junction j2
• j1 and j3 are forward biased but current remain small because holes are
injected from p1 into n1.
• If a hole recombines with an electron in n1 the electron must be
replenished to maintain space charge neutrality.
• The supply of electrons is very restricted because j2 is reverse biased.
• As a result current through j1 is approximately the same as the reverse
saturation current of j2. Similar arguments hold for current through j3.
39
Conducting State
• As 1 + 2
1, many holes injected at J1 survive to be swept across j2
into p2. Similarly, electrons injected at j3 are collected at j2. This is
regenerative because more electrons injected into n1 induces more holes
injection across j1 to maintain neutrality.
• When this happens depletion at j2 begins to shrink. Finally, j2 becomes
forward biased.
• The overall voltage across the devices approximates the potential drop
across 1 forward biased pn junction.
• The triggering of the junction is often caused by the breakdown of j2
40
Avoiding Latchup
• Latchup can be avoided by incorporating guard structures in the circuits.
• There are two types of guard structures
– minority carrier guards
– majority carrier guards
• The guard structures are used to decouple the parasitic bipolar
transistors from each other
Minority Guard
Minority carrier guards are used to collect injected minority carriers
before they can cause a problem.
The minority carriers injected into the substrate could be collected by a
reversed-biased well/substrate junction and flow through the well as
majority carriers.
Majority Guard
The basic mechanism of a N+ majority guard ring in the well is to steer
current away from the parasitic emitter.
41
Silicon on Insulator (SOI)
• Devices fabricated on SOI substrates are fabricated by dielectric
separation
• At this point the most mature technology for manufacturing SOI
substrates is Separation by IMplantation of OXygen (SIMOX)
• In this technology heavy dosage of oxygen is implanted into Si wafer
to obtain SiO2 with a layer of single crystalline Si on top
• Separation of devices can therefore be accomplished by etching the Si
layer through
• Therefore there will not be any parasitic bipolar transistors
42