Review LOW TEMPERATURE ELECTRONICS

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Solid-State Electronics Vol. 37, No. 12, pp.

1967-1975, 1994
~ Pergamon Copyright © 1994 ElsevierScienceLtd
0038-1101(94)E0024-9 Printed in Great Britain. All rights reserved
0038-1101/94 $7.00+ 0.00

BRIEF REVIEW OF THE MOS DEVICE PHYSICS F O R


LOW T E M P E R A T U R E ELECTRONICS

F. BALESTRA and G. GHIBAUDO


Laboratoire de Physique des Composants ~i Semiconducteurs, URA CNRS, ENSERG/INPG,
23, rue des martyrs, B.P. 257, 38016 Grenoble, France

(Received 3 August 1993; in rev&ed form 18 January 1994)

Abstract--A brief review of the main physical phenomena involved in the cryogenic operation of CMOS
silicon devices down to liquid helium temperature is given. Going from solid state physics towards
electrical engineering point of views, several aspects such as the quantification of the inversion layer, the
electronic transport in the 2D electron or hole gases, the scattering mechanisms, the impurity freeze-out
in the substrate or in the lightly doped source and drain regions, the field-assisted impurity and impact
ionization phenomena, the influence of series resistance and other parasitic effects (kink effect, hysteresis,
transient, ...) which alter the device characteristics will be discussed. The short channel effects such as
drain induced barrier lowering, punch through, velocity overshoot will also be addressed.

1. INTRODUCTION the two-dimensional channel, the impurity freeze-out


in the bulk substrate[9,10] or in the lightly doped
The operation of C M O S devices at low and very low source and drain regions, the field assisted impurity
temperature has received much attention during the and impact ionization phenomena occurring near the
past years[l-15]. The device operation at low tem- drain region in saturation regime[3,10], the influence
perature is of prime importance since it allows sub- of series resistance and other parasitic effects (kink
stantial increase of the circuit performances by effect[9,11,12], hysteresis[10,15], transient[3,4] . . . . ) on
comparison to room temperature operation. More the device performances. Furthermore, the character-
specifically, this results in higher carrier mobility and istics related to the scaling down of the device such
saturation velocity, better turn-on capabilities, i.e. drain induced barrier lowering (DIBL)[5,6], punch
subthreshold slope, latch-up immunity, improved through, velocity overshoot[14] will also be ad-
reliability due to activated degradation processes, dressed.
reduced power consumption, decrease of leakage
currents, lowering of interconnection resistance, in-
2. ELECTRONICTRANSPORTIN THE INVERSIONLAYER
creased thermal conductivity, reduction of thermal
noise . . . . Nevertheless, some disadvantages, which 2.1. Inversion layer quantum subbands
derive from the operation of C M O S devices at low The energy quantization in the inversion layer is
temperature, still need to be understood in order to clue to the confinement of the carriers in the potential
overcome the problems due to a cryogenic operation. well caused by the band bending at the silicon-dioxide
For example, impurity freeze-out, kink effect, series interface[16,17]. The discretization of the energy in
resistance effects, hysteresis behavior, anomalies in the direction perpendicular to the interface results in
transient response . . . may strongly affect the func- the formation of 2D subbands. In the triangular
tionality of the circuits. Other problems due to low potential well approximation, the energy levels E k
temperature are related to the major changes in the which define the subband edges with respect to the
fundamental mechanisms associated with the elec- conduction band E c are given by[16,17]:
tronic transport, the time response and/or the fluctu-
ations in the device. E t : Ak(qhF,)z3/(2 m*) 1'3, (1)
The aim of this paper is to give a brief review of where the AkS are the zeros of the Airy function, F~
the main physical phenomena involved in the oper- is the surface electric field, q is the elementary charge,
ation of C M O S silicon devices under cryogenic con- m* is the effective mass depending on the silicon
ditions. F r o m solid state physics towards electrical surface orientation. Self-consistent Poisson's and
engineering, several issues will be discussed such as Schrodinger's calculations have shown that the tri-
the quantum mechanical approach of the inversion angular potential approximation is good for electric
layer[7], the ohmic[7] and non-ohmic[13] transport in field up to 1 MV/cm[17,18].
the 2D gas constituted by the quantum subbands, the The inversion charge of the channel Q~ can then be
scattering mechanisms[8] which alter the mobility in obtained by adding all the contributions of each

1967
1968 F. BALESTRAand G. GHIBAUDO

subband assuming a constant 2D density of states /~k(E) is experimentally inferred by the fact that. at
and using the Fermi-Dirac statistics[7,19]. It should liquid helium temperature, where the inversion layer
be mentioned that an experimental verification of the is degenerate and where the fundamental subband
inversion layer quantization has been demonstrated prevails,/t.fr also exhibits a bell-shaped function of Q,
by capacitance measurements on MOS transis- of the form[25]
tors[19]. Furthermore, a well-known proof for the
existence of 2D subbands in a MOS transistor is the kleff= 21%/(Qi/Qm + Qm/Qi). (3)
manifestation of the quantum Hall effect[20]. The validity of eqn (3) has been confirmed exper-
In practice, the quantization of the inversion layer imentally by studying the effects of stress-induced
produces a shift of the charge threshold voltage interface charge and substrate voltage on the vari-
resulting from the leveling of the fundamental sub- ation of the maximum mobility #m and the corre-
band edge AV~ "-- Eo/q[21 ]. This leads for example to sponding inversion charge Qm[26],
typical AV, of the order of 0.1 V for doping level The emergence of excited subbands and of the non
10JT/cm3.
degeneracy at higher temperatures leads to more
The electronic transport coefficients in a 2D system complex behaviors which can only be approached
can be evaluated using the Kubo-Greenwood inte- empirically. Several works have been made in order
gral by summing, at first order, the parallel contri- to provide a universal description of the mobility as
bution of each subbands while neglecting intervalley a function of Fourand temperature[8,27 30]. In par-
scattering[7], such as ticular, an empirical mobility law which is valid from
T = 4.2 K up to 300 K has been demonstrated, gener-
a ( r ) = y G~(T) alizing the very low temperature mobility law of
k
eqn (3) and the usual room temperature law[31]. In
with fact, this generalized mobility law reads
fZ ~c
ak(T) = ek(E)(-~f/~E)dE, (2) Pen-=#g(Qi/Qc)"-:/[ 1 +(QUQc)" '], (4)
k
where ak(E) is the energy conductivity function where Qc is a critical inversion charge and the expo-
which is proportional to the energy mobility function nent n varies from 3 to 2 as the temperature varies
Pk(E) and f is the Fermi-Dirac distribution. These from very low to room temperature. Figure 1 illus-
formulae can be applied whatever the statistical trates the variations of the exponent n with tempera-
degeneracy, and, in extended as well as in localized ture for various MOS technologies. It should be
states[7]. The effective mobility of the channel can noted that the simplicity of this mobility law enables
then be obtained as po~,= cr(T)/Qi. an accurate modeling of the transfer characteristics of
Si MOSFETs to be achieved in a wide range of
2.2. Principal scattering processes and carrier mobilio, temperature[31].
The mobility /~k(E) at a given energy in each
4 --
subband can be determined as the inverse of the sum
of the scattering rates corresponding to all relevant
scattering modes (Mathiessen rule). The main scatter- WIL = Q 25/25
ing mechanisms which control the inversion layer o 100/100
mobility are the Coulomb, the so-called surface • 2/50
roughness and the phonon scattering modes[16]. The ~- ~ ( ~ l x m ~ 50/5
Coulomb and the surface roughness processes prevail e- • • 50/5
O
e~
at low temperatures, whereas the latter one is effective
mainly at room temperature. For Coulomb scatter-
ing, ,tik(E ) is inversely proportional to the charge
density and has a linear variation with energy[22].
For phonon scattering, pk(E) may vary as o 2 "°~o • o . o ~ o"0~ .-
e~
F~n~~T ~[23],while, for surface roughness scattering,
the mobility should depend primarily on electric field
as F~2[16], Fee being the effective electric field in the
inversion layer.
Following Stern[24], one might consider in a sim- 1 I I I
plified approach that Pk (E) may have a bell-shaped 0 100 200 300

behavior with energy resulting from combining Cou- Temperature (K)


lomb and surface roughness scatterings. This enables
us to provide a consistent picture of the effective Fig. 1. Variation of the mobility law exponent n with
temperature for various technologies. N channel devices
mobility calculated using eqn (2) both as a function from: O, SGS-Thomson; O, LETI-LIR; I , Plessey; A,
of inversion charge and temperature (4.2 K up to CNET; A, CNET with oxynitride gate dielectrics; after
300 K)[7]. In practice, the bell-shaped variation of Ref.[31].
MOS device physics for low temperature electronics 1969

nearly equal to the scattering relaxation time. This


3 [-- Exp.
phenomena results in velocity overshoot and nearly
. / ..... ballistic transport effects[36]. For silicon, velocity
overshoot appears as the momentum relaxation time
>_. 2 I-I? Theo,. becomes smaller than the energy relaxation time. As
a result, the carrier velocity passes through a maxi-
mum as a function of time. The maximum velocity
~= II ' r = l l0K can therefore overcome the velocity saturation by
~ i[-[, w/L=,OO/lOO 50-100%[35], leading to a significant excess of
::~ ] : t°x=120nm
transconductance and drain current in short channel
MOS devices ( - 0.1 p m), especially at liquid nitrogen
It I I I temperature[14,37] (see Fig. 3).
0 5 10 15

Vg (v)
3. I M P U R I T Y F R E E Z E - O U T - R E L A T E D PHENOMENA
Fig. 2. Experimental and theoretical variations of the field
effect mobility /~fe with gate voltage Vg exemplifying a Impurity freeze-out becomes important for tem-
double-subband effect (after Ref.[34]).
perature lower than 150 K for shallow-energy-level
dopants. At liquid nitrogen temperature, weak freeze-
Around the liquid nitrogen temperature, the field out takes place, which is mainly annoying for LDD-
effect mobility variations with gate voltage in n type devices and depletion-mode MOSFETs. For
channel MOS transistors show a pronounced spike temperatures below 30 K, strong freeze-out occurs (at
near threshold[32,33]. This phenomenon can be prop- liquid helium temperature, practically no carriers
erly interpreted as a consequence of the contribution remain in the bands if no field is applied), which
of the first excited subband for (100) or (110) surface affects all the devices and leads to a number of specific
orientations[34]. The amplitude of this behavior phenomena. It is worth mentioning that strong impu-
which results from a higher density of states and a rity freeze-out may occur at higher temperature for
higher mobility in the second subband, is reduced by other dopant species e.g. at 77 K for tellurium doped
increasing the depletion charge via the bulk doping MOS transistors[38].
and/or the substrate bias[34] (see Fig. 2).
3.1. Depletion-mode devices
2.3. High .field transport
Around liquid nitrogen temperature, an unusual
As the longitudinal electric field increases, non-
behavior is observed in the subthreshold region of
ohmic conduction resulting from hot carrier effects
the transfer characteristics of depletion-mode
arise[16]. In inversion layers, according to a displaced
device (Fig. 4). Since the energy band structure
Maxwellian theory which involves elastic and acous-
responds to the gate potential change, the impurity
tic phonon scatterings[16], the mobility should vary
ionization is changed. For P-channel devices, the
as the squared electric field at very low temperature
increase of (negative) gate voltage produces a shift
and low electric field (< 10V/cm). The electronic
of the valence band towards the Fermi level. The
temperature may therefore change with respect to the
resulting band bending causes acceptors to become
lattice temperature quadratically with the electric
field. Experimental results obtained on Si MOS tran-
sistors show only a qualitative agreement with this ]-~---- C h a n n e l m,[
prediction[13]. At high electric field ( > 103 104 V/cm),
velocity saturation due to intervalley and/or optical 4 -- ___ At S i - S i O 2 i n t e r f a c e VGS = 0 " 7 V
--'-- 20 .~ below interface VDS = 0.6 V
phonon scatterings results in a strong decrease of the E
mobility whatever the temperature[13,16]. The satu- - - <,:x> v
/t . .\ 0.043 lam
ration velocity ranges between 6 × 106 and 107 cm/s as
7 ' ,k.~ ~ 77K
the temperature is changed form 300 K down to ~, Vsat = 1.2 x 10 cm/s /j. ,x~.
4.2 K[13]. A simple empirical formula of the form

/~en=/~0/{ 1 + [F/(vs, t/#o)]2} ':z (5) e- . . . . . . . . . . . . . . . . . . .

is generally sufficient for the modeling of the mobility


!
reduction at high electric field F[13,16,35].
The above results have been obtained in the case of
long channel devices. As the devices are scaled down, 0 0.02 0,04 0.06 0.08 0.10
new features arise from the non stationary transport x-Directed distance (p_m)
which takes place in the channel[35-37]. In this
Fig. 3. Variation of electron velocity along the channel
situation, the mean free path becomes comparable to illustrating the velocityovershoot phenomenon in very short
the channel length, and, in turn, the transit time is channel devices (after Ref.[14]).
1970 F. BALESTRAand G. GmBAUDO

10-3 device, the lateral electric field necessary to induce an


efficient impurity ionization is only large enough in
the drain region, which is insufficient to alter signifi-
cantly the device performances.
10-5 f - - - ~ " ~ " X ~ ~ \ ~ 300
I \\\ \\T-----,,o 3.2. Influence qf LDD regions
For degenerately doped source and drain regions,

9.2
\li" no freeze-out takes place until liquid helium tempera-
ture. However, for the lightly doped source and drain
regions (LDDs), freeze-out becomes important even
at liquid nitrogen temperature and therefore has a
strong impact on the electrical properties of the
lO 11 devices. Indeed, important series resistance effect
s arises when decreasing the temperature in the LDDs,
leading to large degradation of the MOSFET per-
10-13 ~ formance. For low applied voltages, the electrical
-3 -2 1 0 1 properties are optimum at intermediate temperature
and are lowered for very low temperature. Neverthe-
vo~v~ less, for high applied gate and drain voltages, an
Fig. 4. Evolution with temperature of the ld(Vg) character- interesting feature is observed, which is the increase
istics of a depletion-mode MOS device (P channel device; of the drain current until liquid helium temperature
after Ref.[42]). (see Fig. 5). This phenomenon is related to a re-
duction of the parasitic LDD resistance due to the
frozen-out[39-41]. This effect has a strong impact on partial ionization of the frozen-out impurities in the
the threshold voltage of the device as well. LDDs caused by Poole-Frenkel effect or impact
For temperature above 150 K, freeze-out is very ionization[44]. Therefore, high performance and re-
weak as the gate voltage is increased, since the Fermi liable Si MOSFETs may be designed for low tem-
level is well above the impurity level. When the perature operation while using LDD technologies.
temperature decreases, the Fermi level approaches Nevertheless, the use of LDDs for improving the
the valence band and therefore boron ions start to be hot-carrier-induced degradation immunity has not
deactivated at lower Vg values. On the other hand, at been proved to be necessary[45].
very low temperature (9 K) the Fermi level ranges
between the acceptor level and the valence band, so 3.3. Kink and hysteresis eff&'ts
that these impurities are already frozen-out even at At very low temperature (T < 30 K), enhancement-
Vg = 0. Any threshold voltage change can be expected mode bulk Si MOS transistors exhibit a change in the
by increasing the gate voltage, for boron atoms are output conductance at high drain voltages in satu-
never ionized (Fig. 4)[42]. ration which is called the kink effect (Fig. 6119,t2].
Besides, another interesting feature, which is ob-
served for depletion-mode device operating at very
L = 5Bm LDD-1013/cm2 3(I
low temperature, is the suppression of the kink effect
in saturation even at liquid helium temperature[42]. S
This phenomenon, which appears by a change in the
output conductance for high drain voltage, is gener- - - ~ \ V = 3V
ally observed for enhancement-mode devices for tem-
6 2(I
perature lower than 30 K, as will be shown below.
It has been shown lately that the influence of
freeze-out on the operation of depletion-mode MOS-
FETs at liquid nitrogen temperature is reduced in 4
short-channel devices for high drain voltages[43]. " I (1
I
This phenomenon has been attributed to the
Poole-Frenkel effect which increases as Vd increases. -- it V =2V ~
/ I g
In fact, the impurity ionization is enhanced for
! /
short-channel devices by the lowering of the effective iiii
barrier height due to the high lateral electric field. In - I J
the case of a sub micrometer device, for not too small 0 100 200 300
Vd, the lateral electric field is quite large on the main T (K)
part of the channel which leads to significant impurity Fig. 5. Variations with temperature of the saturation (solid)
ionization, resulting in an improvement of the drain and linear (dashed) drain current illustrating the impact of
current characteristics. In the case of long-channel the LDDs on the electrical performances (after Ref.[44]).
MOS device physics for low temperature electronics 1971

43I ZIL = 50/10S


avalanche multiplication near the drain is related to
the drain current and the drain voltage by a classical
expression[47]. This model leads to a satisfactory
modeling of the kink effect in MOS transistors for
very low temperature operation[46]. Moreover, this
analysis has also been successfully applied to the case
of short-channel MOSFETs where a reduction of the
normalized kink effect Ald/los,t is obtained due to
combined effects of charge sharing and velocity satu-
ration via the lowering of the depletion capacitance
2V and the normalized saturation transconductance[11].
I
The critical drain voltage for which the kink effect
appears increases as the temperature is increased
(Fig. 6). This effect can be interpreted by the decrease
of the substrate current which flows from the drain to
0 4 8 12 16
the source and the decrease in the substrate resist-
Vd(V) ance, both effects leading to a lower internal potential
Fig. 6. Typical Id( Vd) output characteristics of a MOSFET for the same applied voltages[9,12].
illustrating the onset of kink effect at very low temperature Hysteresis phenomena have also been observed in
(after Ref.[9]). the Id(Vd) characteristics[10,15,48], and are at-
tributed to transient effects which will be discussed in
the following section.
This behavior is quite similar to the one observed in
silicon on insulator devices. The kink effect results 3.4. Transient behat~ior
from a shift in the threshold voltage. Indeed, at very At very low temperature (T ~ 30 K), in the strong
low temperature, the equilibrium situation, reached freeze-out regime, as a carrier is trapped in a shallow
after a certain time which depends on temperature level, it will remain in the trap for a certain time
and applied biases (see Section 3.4), can be described before picking up enough thermal energy to escape.
by a depletion region with ionized impurities and a However, in the liquid helium temperature range, the
frozen-out Si substrate. The impact ionization ma- substrate of a MOSFET could remain frozen-out
jority carrier current, generated in the drain space indefinitely if thermal energy provides the only
charge region, can flow through the source junction means to release the carriers from the shallow
and through the we1119,12]. The avalanche-generated traps. The application of a field of large enough
current flowing from the drain to the source creates strength can induce ionization of trapped carriers and
in turn an internal substrate voltage at the source depletion region formation. The transient time for
substrate junction. The other part of the avalanche which an excess drain current is observed (the current
current may constitute a space charge limited current decreases when the depletion region is created)
flowing through the well, which also forward biases depends on temperature and applied field. Three
the substrate, both effects leading to a decrease in the types of field dependent ionization are generally
threshold voltage. For large enough substrate possible[3,4,49-53].
voltage, the source-substrate junction becomes con- A first mechanism is the Poole-Frenkel ionization,
ducting and causes the major part of the impact for which the barrier for thermal emission is lowered
ionization current to be collected by the source. For by the application of a field. The change in the
the sake of simplicity, one may consider that the magnitude of the barrier depends entirely on the field
avalanche current flows directly to the source, what- F such as AE=q(qF/rcEsi) ~'2. As the field or the
ever the internal substrate voltage. temperature increase, the ionization time decreases,
The excess drain current due to the parasitic sub- which leads to shorter transients. At temperature
strate biasing is, at first order, proportional to the above 30 K, a field larger than 103 V/cm leads to short
internal substrate bias change Vb near the source and transient times lower than 1/~s. However, in order to
to the body transconductance gb = dlJdVb such as have transient times at 4.2 K of the same amplitude,
Aid = gb Vb[46]. The body transconductance is related one has to apply a field above 3 × 103V/cm[52].
to the gate transconductance gm=dld/dVg as A second possible mechanism of field dependent
gb = (Cd/Cox)gm, where Cd and Cox are the depletion ionization is by tunneling. At sufficient field strength,
and gate oxide capacitances, respectively. the carrier is able to tunnel through the barrier into
The internal substrate bias change with respect to the band. This phenomenon is thus weakly tempera-
the source is related to the substrate current /sub ture dependent contrary to the Poole-Frenkel field-
flowing in the forward-biased source-substrate assisted thermal ionization, but is an extremely
junction through a classical diode formula or an sensitive function of the electric field. Therefore,
equation adapted for the very low temperature when the temperature decreases (T < 30 K) and for
case[46]. The substrate current generated by large enough field ( > 2 x 103V/cm), tunneling be-
1972 F. BALESTRAand G. GHIBAUDO

comes the dominant ionization process[52]. For lower 80 --

fields, the temperature for which tunneling is the main


ionization mechanism decreases. Tunneling may also
be phonon-assisted, but this process is rather unlikely
in the very low temperature range. 60
The third important mechanism is the shallow-level
impact ionization. When majority carriers are gener-
ated by avalanche multiplication for sufficient field
near the drain, these carriers flow through the sub-
strate and can induce the ionization of dopant impu-
:L:\
rities by impact ionization (the so-called "forced
depletion layer formation"). The transient response
in the drain current of a M O S F E T originates from
the change with time of the forced depletion region
width, which depends on the substrate current and on
20
\\'<L. P MOS •
the rate of impact ionization and recombination by
thermal capture[10,50,51,53]. This kind of ionization
seems to be dominant at very low temperature I L .I
( < 10 K) for intermediate electric field. A typical 0 100 200 300

transient in the drain current of a M O S F E T is T(K)


displayed in Fig. 7 as a function of temperature[4].
Fig. 8. Variation of the maximum substrate current l~,~......
The transient time generally increases as the tempera- with temperature T for N and P MOS devices
ture is lowered and decreases as the applied gate and (W/'L = 50/10, Vd = 3 and -4.5 V: after Ref.[58]).
drain voltages are increased. Similar transients have
also been observed for MOS capacitors, diodes and
in the substrate current of M O S F E T s for very low degradation. A detailed study of the substrate
temperature operation in the strong freeze-out current in MOS devices is therefore mandatory for
regime[3,49,52]. the qualification of a C M O S technology for cryogenic
Other types of transient effects can be due to device application. Several models have been reported in
self-heating. This feature produces a change in the the past for the substrate current but they were
surface band structure and in the channel mobility, restricted to room and/or liquid nitrogen tempera-
both phenomena giving rise to a transient in drain tures[56,57]. In this section is presented the tempera-
current[54,55]. However, this effect seems to be less ture dependence of the substrate current in MOS
important than the previous ones in the liquid helium devices operated between liquid helium and room
temperature range. temperatures.
In Fig. 8 are reported the variation with tempera-
ture of the maximum substrate current /,t,b..... for
4. I M P A C T IONIZATION PHENOMENON VERSUS constant drain voltage. The substrate current is found
TEMPERATURE
to be much larger in N channel devices than in P
The impact ionization phenomenon from which MOS transistors (almost 3 order of magnitude). This
originates the substrate current is a key feature as is justified by the higher mobility and higher impact
regard to the kink effect and the hot-carrier-induced ionization coefficient for electrons. As the tempera-
ture is reduced, the maximum substrate current is
enhanced by about a factor of 3 and 6 for N and P
channel devices, respectively.
1.2 -- The variation of l~ub/I d and of (l, uh/ l,l )/( V d - Vd~,, ),
, j 22.5 K Vd~t being the saturation drain voltage, have becn
1.0 --
plotted as a function of 1,'(Vd--~ / d,,,) for various
~ 0.8 -- temperatures in order to check the validity of the
classical substrate current law[47]. It is worth noting
0.6 - -

that the plots Lub/Id and (l, ub/Id)/(Va--Vd~,,,) VS


0.4 -- @ 11.8 1/(Vd--Va~,) have been found to be practically
linear (better than 99% over 2-3 decades) over
0.2
the whole temperature range 20-300 K. for both
0 -- plots, independently of the gate voltage and device
I J I I L I I I I types[581. This is in fact due to the dominating
0 0.5 | .0 1.5 2.0 2.5 3,0 3.5 4.0 role of the exponential term in the substrate current
Time (ms) law.
Fig. 7. Typical drain current transient as observed at low Therefore, a satisfactory modeling of the substrate
temperature (after Ref.[4]). to drain current ratio can be obtained over the whole
MOS device physics for low temperature electronics 1973

temperature range 20-300 K by using the simple The charge sharing effect, which characterizes the
substrate current 1aw[58,59]: control loss of the gate on the depletion charge, is not
very much altered by the temperature reduction[5,61].
l~ub = A i d e x p [ - B / ( V d - Vds,t)], (6a)
This is due to the fact that the depletion layer width
or depends weakly on temperature[61].
Unlike charge sharing, the punch-through current
Is, b = A l d ( V d - Vd~,t)exp[--B/(V d- Vd~,t)], (6b) which flows between source and drain via the bulk is
where A and B are empirical constants related to the strongly modified by the cooling down. This is due to
impact ionization law coefficients as A = a~l and the activated behavior of the punch through with
B = b~l with 1 being the effective ionization length temperature such as[61],
near drain[47].
In, = I0 exp(-¢b/kT), (7)
The constants A and B have been extracted from
where ¢I) is the source-to-bulk barrier height. Thus, as
the y axis intercept and slope of the plots l s . b / l d vs
the temperature is reduced, the punch-through cur-
1/( Vd - Vd~,t),respectively. Figure 9 displays the evol-
utions with temperature of these parameters as ob- rent, which is mainly efficient in the weak inversion
tained for electrons and holes. region of the MOSFET operation, is strongly reduced
For both N and P channel transistors, the co- (see Fig. 10). The punch-through current may also
efficient B is almost constant with temperature (31-35 strongly modify the subthreshold swing and the
and 46-50 V), while A is a monotonically decreasing threshold voltage[61]. Therefore, the operation of
function of temperature. If l is assumed to be con- scaled down MOS devices at low temperature con-
stant with temperature, it implies that b~ and a~ have siderably improves the device characteristics.
the same temperature dependence as B and A. The DIBL which measures the modification of the
source surface potential by the electrostatic influence
5. SMALLGEOMETRYEFFECTSVERSUSTEMPERATURE of the drain voltage, is also reduced by the tempera-
ture lowering[5,6].
As the devices are scaled down, new physical
phenomena which may considerably alter the opera- 6. SUMMARYANDCONCLUSION
tionality of the device evolve. These limitations are
generally referred to as small geometry or short A brief review of the main physical phenomena
channel effects. The series resistance, the DIBL, the involved in the cryogenic operation of CMOS silicon
punch-through or the charge sharing effects represent devices down to liquid helium temperature has been
the major causes to these effects. given. Going from solid state physics towards electri-
The series resistance effects are considerably in- cal engineering point of views, several aspects such as
creased by the device cooling down. This is mainly the quantification of the inversion layer, the elec-
due to the higher carrier mobilities at low tempera- tronic transport in the 2D electron or hole gases, the
ture, although the contact and interconnection resist- scattering mechanisms, the impurity freeze-out in the
ances are reduced at low temperature[60]. substrate or in the lightly doped source and drain

100 - -- 0.4 80--

80
\ 0.3 60

60 "\ ° ""~" ° ~° "'~ ° • •


P MOS
e-

"V "\ 40
o~°'O-o~o~o~ N MOS
40

20
20
~ M~OSi ~ ~ t 0.1

I I I I I
100 200 300 0 100 200 300
T(K) T(K)
Fig. 9. Variations of the impact ionization law coefficients A and B with temperature T as obtained for
P and N MOS devices (after Ref.[58]).
1974 F. BALESTRA and G. GHIBAUDO
10-1

,o-r
1o . 5

~a~ 10-7

Vale 2.5V

10 I1 i / L=0gBm
/
10 -13 / I I I
-I 0 I 2 3
V (v)

Fig. I0. Weak inversion ld(V~) characteristics showing the suppression of punch-through vs temperature
(P channel device).

regions, the field-assisted impurity a n d i m p a c t ioniz- 24. F. Stern, Phys. Rev. Lett. 44, 1469 (1980).
a t i o n p h e n o m e n a , the influence o f series resistance 25. G. Ghibaudo and F. Balestra, Solid-St. Eleetron. 31,
a n d o t h e r parasitic effects (kink effect, hysteresis, 105 (1988).
26. A. Emrani, G. Ghibaudo and F. Balestra, Proe. qf Low
transient . . . . ) w h i c h alter the device characteristics Temperature Electronic Det'ive Operation Symposium
have been discussed. T o s o m e extent, s h o r t c h a n n e l (Edited by D. Foty, N. Saks, S. Raider and G. Oleszek),
effects such as D I B L , p u n c h t h r o u g h , velocity over- Electrochem. Soc.. Vol. 91-14, p. 61 (1991).
s h o o t have also been addressed. 27. N. Arora and G. Gildenblat, IEEE Trans. Electron
Devices ED-34, 89 (1987).
28. V. M. Agostinelli, H. Sin and A. Tasch, IEEE Trans.
Electron Deviees ED-38, 151 ( 1991 ).
REFERENCES 29. K. Masaki, K. Taniguchi, C. Hamagichi and M. lwasea.
J. appl. Phys. 30, 2734 (1991).
I. Special issue of IEEE Trans. Electron Devices 36 30. A. Rothwarf. IEEE Eleetron Deviee Lett. EDL-8, 499
(1989). (1987).
2. R. K. Kirschman, Cryogenics 25, 115 (1985). 31. A. Emrani, F. Balestra and G. Ghibaudo, IEEE Trans.
3. N. S. Saks and A. Nordbryn, J. appl. Phys. 50, 6962 Eleetron Devices ED-40, 564 (1993).
(1979). 32. F. Fang and A. Fowler, Phys. Rer. B 3, 619
4. S. Tewksbury, J. appl. Phys. 53, 3865 (1982). (1968).
5. J. C. Woo and J. D. Plummer, 1EEE Trans. Eh, ctron 33. T. Sato, Y. Takeishi and H. Hara, Phys. Rev. B 4, 1950
Devices ED-33, 1012 (1986). (1971).
6. Z. Yan and J. Deen, Co,ogenies 30, 1160 (1990). 34. A. Emrani, G. Ghibaudo and F. Balestra, to be pub-
7. G. Ghibaudo, J. Phys. C 19, 767 (1985). lished: see also A. Emrani, Ph.D. thesis, INP Grenoble
8. C. Huang and S. Gildenblat, IEEE Trans. Electron ( 1992).
Devices ED-37, 1289 (1990). 35. T. Kobayashi and K. Saito, IEEE Trans. Eh>etron
9. F. Balestra, L. Audaire and C. Lucas, Solid-St. Elec- Devices ED-32, 788 (1985).
tron. 30, 321 (1987) and Proc. o [ ' E S S D E R C 86, Cam- 36. S. L. Teitel and J. W. Wilkins, IEEE Trans. Eleetron
bridge, U.K., p. 31 (1986), Devices ED-30, 150 (I983).
10. B. Dierickx, L. Warmerdam, E. Simoen and C. Claeys, 37. G. G. Shahidi, D. Antoniadis and H. Smith, IEEE
1EEE Trans. Eleetron Deviees ED-35, 1120 (1988). Electron Device Lett. EDL-9, 94 (1988).
11. I. M. Hafez, G. Ghibaudo and F. Balestra, IEEE 38. S. T. Tewksbury, M. R. Biazzo and T. L. Lindstrom,
Electron Device Lett. EDL-11, 120 (1990). IEEE Trans. Electron Devices ED-32, 67 (1985).
12. E. Simoen, B. Dierickx and C. Claeys, Solid-St. Elee- 39. F. H. Gaensslen and R. C. Jaeger, Solid-St. Electron.
tron. 33, 445 (1990). 22, 423 (1979).
13. F. Fang and A. Fowler, J. appl. Phys. 41, 1825 (1970). 40. R. C. Jaeger and F. H. Gaensslen, IEEE Trans. Eleetron
14. S. Laux and M. Fischetti, IEEE Electron Derice Left. Devices ED-27, 914 (1980).
EDL-9, 467 (1988). 41. K. A. Wilson, P. Tuxbury and R. L. Anderson, IEEE
15. E. Rocofyllou, A. Nassiopoulos, D. Tsamakis and F. Trans. Electron Devices ED-33, 1731 (1986).
Balestra, Solid-St. Electron. 32, 603 (1989). 42. T. Elewa, F. Balestra, S. Cristoloveanu, 1. M. Hafez,
16. T. Ando, A. Fowler and F. Stern, Rev. mod. Phys, 54, J. P. Colinge, A.-J. Auberton-Herve and J. R. Davis,
437 (1982). IEEE Trans. Electron Devices ED-37, 1007 ([990).
17. F. Stern, CRC Critical Rev. Solid-St. Sci. 5, 499 (1974). 43. J. Gautier, G. Guegan, M. Guerin and M. Lerme, 1EEE
18. C. Moglestue, J. appl. Phys. 59, 3175 (1986). Trans. Electron Devices ED-38, 1832 (1991).
19. J. Pals, Phys. Rev. B 7, 754 (1972). 44. A. Emrani, K. Rais, I. M. Hafez, F. Balestra, G.
20. K. Von Klitzing, G. Dorda and M. Pepper, Phys. Rev. Ghibaudo and M. Haond, Proe. of Symposium on Low
Lett. 45, 494 (1980). Temperature Eh, etronies and High Temperature Super-
21. M. J. Van Dort, P. H. Woerle, A. J. Walker, C. A. H. conductivity, Honolulu, Hawaii (1993).
Juffermans and H. Lifka, Microelectron. Engng 15, 551 45. M. S. Song, J. S. Cable, J. C. S. Woo and K. P.
(1991). MacWilliams, IEEE Trans. Eleetron Deviees ED-12, 375
22. C. T. Sah, T. Ning and L. Tschopp, S u r f Sci. 32, 561 (1991).
(1972). 46. I. M. Hafez, G. Ghibaudo and F. Balestra, IEEE Trans.
23. S. Kawaji, J. Phys. Soc. Japan 27, 906 (1969). Eleetron Devices ED-37, 818 (1990).
MOS device physics for low temperature electronics 1975

47. T. Y. Chan, P. K. Ko and C. Hu, IEEE Electron Devices 55. D. P. Foty, IEEE Trans. Electron Devices EDo36, 1542
Lett. EDL-5, 505 (1984). (1989).
48. A. Nassiopoulos, D. Tsamakis and E. Rocofyllou, 56. N. D. Arora and M. S. Sharma, IEEE Trans. Electron
J. appl. Phys. 68, 1896 (1990). Devices 38, 1392 (1991).
49. E. Rosencher, V. Mosser and G. Vincent, Solid State 57. A. K. Henning, N. N. Chan, J. T. Watt and J. D.
Commun. 45, 629 (1983). PlummeL IEEE Trans. Electron Devices ED-34, 64
50. E. Simoen, B. Dierickx, L. Warmerdam, J. Vermeiren (1987).
and C. Claeys, IEEE Trans. Electron Devices ED-36, 58. K. Rais, G. Ghibaudo and F. Balestra, Electron. Lett.
1155 (1989). 29, 778 (1993).
51. E. Simoen, B. Dierickx, M. Gao and C. Claeys, Proc. 59. E. Simoen and C. Claeys, Solid State Commun. 82, 341
Syrup. on Low Temperature Electronic Device Operation, (1992).
(Edited by D. Foty et al.), The Electrochemical Soc., 60. I. M. Hafez, F. Balestra and G. Ghibaudo, J, appL
Vol. 91 14, p. 107 (1991). Phys. 68, 3694 (1990).
52. D. P. Foty, Cryogenics 30, 1056 (1990). 61. Y. W. Yi, K. Masu, K. Tsubouchi and N. Mikoshiba,
53. E. Simoen, B. Dierickx, U Deferm, C. Claeys and G. Proc. Low Temperature Electronic Device Operation
Declerk, J. appl. Phys. 68, 4091 (1990). Symposium (Edited by D. Foty, N. Saks, S. Raider
54. D. P. Foty and S. L. Titcomb, IEEE Trans. Electron and G. Oleszek), Electrochem. Soc., Vol. 91-14, p. 99
Devices ED-34, 107 (1987). (1991).

SSE 37/12--F

You might also like