Professional Documents
Culture Documents
Solid-State Circuit Conj., Dig. Tech. Papers, p. 225, E. J. Boleliy and J. E. Meyer, High-performance low-power CMOS memories using silicon-on-sapphire technology, IEEE J. Solid-State Circuits (Special Issue on Micropower Electronics), vol. SC-7, pp. 135-145, Apr. 1972. R. W. Bower, H. G. Dill, K. G. Aubuchon, and S. A. Thompson, [MOS field effect transistors by gate masked ion implantation, IEEE !trams. Electron Devices, vol. ED-15, pp. 757-761, Oct. 1968. J. Tihanyi, Complementary ESFI MOS devices with gate self adjustment by ion implantation, in Proc. 5,th Iwt. Conj. Microelectronics in Munich, Nov. 2729, 1972. MunchenWien, Germany: R. Oldenbourg Verlag, pp. 437447. E. J. Boleky, The performance of complementary MOS transistors on insulating substrates, RCA Rev., vol. 80, pp.
372-395, 1970.
SC-9,
NO.
5>
OCTOBER 1974
[51
[61
Kapazitat auf Source und Drain im Ersatzschaltbild eines und Ent wicldwrgsMOS-Transistors, Siemenx Forxchungsberichte 1, no. 3$ pp. X4-286, 1972. [121 J. R. Burns, Switching response of complementary+symmetry MOS transistors logic circuits, RCA Rev., vol. 25, pp. 627481, 1964. [131 R. w. Ahrons and P. D. Gardner, [Introduction of technology and performance in complementary symmetry circuits, IEEE J. Solid-State Circuits (Special Issue on Technology jor Integrated-Circuit Design), vol. SC-5, pp. 2429, Feb. 1970. [141 F. F. Fang and H. Rupprecht, High performance MOS integrated circuits using ion implantation technique, presented at the 1973 ESSDERC, Munich, Germany,
[71
formation in an insulated gate field [81 K. Goser, [Channel effect transistor ( IGFET) and its emrivalent circuit . Sienzen.s Forschungsund Entwiclclungsbekhte, no. 1, pp. 3-9, 1971. Accurate metallization [91 A. E. Ruehli and P, A. Brennan, capacitances for integrated circuits and packages, IEEE J. Solid-State Circwits (Corresp.), vol. SC-8, pp. 289-290, Aug. 1973. (Siemens Netzwerk Analyse Programm Paket), [101 SINAP Siemens AG, Munich, Germany. [Aufteilung der Gate-Kanal[111 K, Goser and K. Steinhubl,
for a photograph
and biography,
please see p.
a photograph
and
biogra~hy,
please
see p.
Design
MOSFETS Dimensions
HWA-NIEN YU,
with
Very
ROBERT H. DENNARD, RIDEOUT,
R. LEBLANC,
AbsfracfThis paper considers the design, fabrication, and characterization of very small MOSI?ET switching devices suitable for digital integrated circuits using dimensions of the order of 1 p. Scaling relationships are presented which show how a conventional MOSFET can be reduced in size. An improved small device structure is presented that uses ion implantation to provide shallow source and drain regions and a nonuniform substrate doping profile. One-dimensional models are used to predict the substrate doping profile and the corresponding threshold voltage versus source voltage characteristic. A two-dimensional current transport model is used to predict the relative degree of short-channel effects for different device parameter combinations. Polysilicon-gate MOSFETS with channel lengths as short as 0.5 ~ were fabricated, and the device characteristics measured and compared with predicted values. The performance improvement expected from using these very small devices in highly miniaturized integrated circuits is projected.
LIST
OF SYMBOLS
Inverse
semilogarithmic
D AW,
threshold characteristic. Width of idealized step function fde for chaDnel implant. Work function difference for between silicon and substrate. Dielectric constants silicon dioxide. Drain current. Boltzmanns Unitless MOSFET Effective Intrinsic constant. constant. scaling
Manuscript received May 20, 1974; revised July 3, 1974. The aubhors are with the IBM T. J. Watson Research Center, Yorktown Heights, N.Y. 10598.
Substrate acceptor concentration. Band bending in silicon at the onset of strong voltage. inversion for zero substrate
DENN.4m
et al. : ION-IMPLANTED
MOSFETS
257
*,
(1
junction
potential.
GATE ~
tox=loooh +
=. _-
GATE
:N+ ___,
&*200A
~N+o .___,
Q.,,
t ox
+ --0 Ll a
1, /l /L\
5P
-OILhp
VaubDrain,
Source Source
source, gate and substrate relative relative voltage. drain depletion width. to source.
N~=25x106/cm
NA=5 x 105/cm3
to substrate. layer
(a)
(b)
Fig. 1. Illustration of device scaling principles with K = 5. (a) Conventional commercially available device structure. (b) Scaled-down device structure.
channel
relatively
INTRODUCTION EW N over monly new widely while ing Full tion new can the HIGH resolution semiconductor a decrease contact the in lithographic integrated linewidth of approach industry pattern device [5] exhibited the benefits and optical techniques circuit five to which today. writing fabrication projection has patterns ten times is comOf the been for
lightly reduces
doped
starting
substrate,
this
implant This
the sensitivity
to changes in the source-to-substrate reduced substrate sensitivity off for a thicker Second, shallow able taining of these switching with ion gate insulator
be traded
masking
tends to be easier to fabricate implantation source and drain respect features an acceptable in device which
semiconductor beam
allows regions
to short-channel an
lithography also of
combination
design with
development structures
if desired, which has well-controlled and which has significantly capacitances capacitances). begins by describing to a conventional structure verification device the scaling MOSFET of the (e.g., drain-to-gate
technologies, very the small integrated is known the small design, MOSFET
optimized paper
dimensions. fabrication, switching using reducing length) device when and drain and chardevices dimensions the of sourcean FET
for order
digital of
circuits that
are applied
1 p. It (i.e.,
capable
of improved
channel in the
formance.
Experimental
characteristics. the extend substrate the in on, been by depletion over under most the a
significant source in
proach is then presented. Next, the fabrication process for an improved scaled-down device structure using ion implantation is described. Design considerations for this all-implanted a simple strate structure are based on two analytical model that predicts that devices, tools: the suband a predicts of chanare to two-dione-dimensional
the
silicon
gate
electrode.
sensitivity
for long channel-length current-transport characteristics results experimental the model from
device It
short-channel vertical
analyses design
junction
horizontal the
sensitivity
dimensions, applied
while
proportionately
decreasing
voltages
and increasing
the substrate
doping
con-
Yarious parameters is shown. Then, detailed attention is design, intendedfor zero substrate givcll to all alternate
bins, which offers some advantages with respect to thresh-
[7], [8]. Applying this scaling approach to a centration properly designed conventional-size MOSFET shows that
Finally,
concludes
with to
a discusFETs.
a 200-A
gate insulator
is required
if the channel
length
performance circuits
improvements
be expected small
is to be reduced to 1 ~. A major consideration the use of ion implantation for very small centration scaled-down atoms of ion implantation of doping of this paper is to show how leads to an improved design MOSFETS. allows First, the ability a low condoping
from integrated
DEVICE The concise in principles manner the switching n-channel of the device general and devices. lllOSFET
show be
in
to accurately
introduce
followed of
the substrate
dccreming
size
performance a state-ofscaled-down
lIOSFET the-art
1 compares with a
258 clevice be 1 (a) vices niques. substrate clesignecl following The typical by using the larger device scaling principles shown in to Fig. detechwith a
IEEE
JOURFJ.iL
OF SOLIC)-STATE
CIRCUITS,
OCTOBER
1974
described
later.
structure
p-type substrates) the work function difference A~Vf is of opposite sign, ancl approximately cancels out *,J. ~. is the band bencling tial) bias. It would and (2) prevent mately increased constant, in the silicon that (i.e., the surface appearing potenin (1) approxi). at the onset of strong appear exact scaling actually inversion the * terms increasing = for zero substrate
a 1OOO-A gate
bias
chosen doping
to
give
gate threshold
to the source potential. cm-3 is low enough strate tant source sensitivity. criterion followers in
eloping since ~0 @ *,
However, the fixed substrate bias supply normally used with n-channel devices can be adjusted so that (~,, + V,,,,; ) = plied (~,, + substrate V,,,,,) /K. Thus, by scaling clown the apbias more than the other applied voltages, across the source region that under or drain junctions, the gate, can be reclevice by
circuits becomes
voltage
increases
drop
two o~er the full range of variation clevice limit illustrated the channel in Fig. length from rameters restriction depletion normally mum tration In
values three all
or across the depletion All of the equations characteristics example, the MOSFET
L to about, 5 p. This
the penetration of the
arises primarily
region surrounding the clrain into the area controlled by the gate electrode. For a maxiof approximately the surface potential voltage. suitable
and by a
[9]
~ ,_
d
P,,,e.. tOx/K-
modify
()(
w/K V,
L/ii
)
of
K,
(Vet/K)
Id/K
(3) set
L, the device is
dimension, are &K, clirnensions t,ox =
of
applied
no
change
mobility.
scaled
by
a transformation
slightly
variables: linear
voltage, reduced
the scaling
faCtOr
K, e.g.,
where
refer
to
the
new
scaled-down
such as gate insulator as the horizontal and width. the substrate using
is maintained
etc., as well
of channel
Second, the voltby the same factor doping concenfactor 1 (b) was
clown device except for a change in scale for the spatial coordinates. Furthermore, the electric field strength at any corresponding Thus, effects point is un?hanged velocity and, hence, in both any fixed because
V/c
=
ve-
V/x.
locity
changed
saturation crystal
clesign shown
be similar
using
5 which
microscopic dimensions, by
K, the
in channel
length to 1 p,.
(3), since the clevice current current of channel with This is consistent
The scaling
relationships
that the depletion layer widths in the scaled-down device arc reducecl in proportion to the device climensions due to the reducecl potentials ancl the increased doping. For example,
w, = {[2cs,(V~ +
is unchanged
by scaling,
sheet density of carriers (i.e., electrons per unit gate area) moving at the same velocity. In the vicinity of the clrain, the carriers extent Thus, will move away of mobile from carriers the surface per unit around to a lesser cliffusions. volume the will drain, in the new device, the clensity due to the shallower region
}2 m w,/Ic.
(1)
The
threshold
voltage
is also decreased
bc higher
in the space-charge
to the reduced device voltages so function properly in a circuit with This is shown by the threshold device.
+ Vs-s,,,/K)]}
complementing the higher density clue to the heavier doped substrate, tionships in Table on circuit
voltage equation
levels.
for power density, delay time, etc., are given I ancl will be discussed in a subsequent section performance. the scaling relationships, two sets of
(tO=/KCOf) { Q,.,,
(AJVf
~.)
V,/K.
In
(2)
in Tt is primarily
tOx/K,
devices were fabricated with gate insulators o of 1000 ancl 200 A (i. e., K = 5). The measured drain voltage characteristics of these devices, normalized to W/L = 1,
are quite of the shown similar smaller the in Fig. when ~ 9. The tin-o with sets voltage by of characteristics and current of 2, five, the are scales which exact plotted
creased
thickness,
while
terms tend to cancel out, In most gates of doping or aluminum type gates on
clevice scaling
reduced predictions.
a factor In Fig.
confirms
MOSFETS
259
1,5~ GATEVOLTAGE [V] 20
tox =100ox
10L *w=5p v~ub=-?v 0.5+= sO.65V
Q 0
5 5 10 15 20
DRAINVOLTAGE[V]
(a)
0.3 GATE VOLTAGE[v]
0.2
3 .
0 IF
+ ~
o 0
Expcrimentzd drain voltage characteristics ventional, and (b) scaled-down structures shown malized to W/L = 1.
2.
3.5
drain
tox ds
voltage
is large exhibit
enough the
and the
charact cristics
3.0 2.5 fid
2,o
expected
relationship,
=10001! =5V
When linear
proj ccted to intcrccpt the gate voltage axis this relationship defines a threshold voltage useful for design purposes.
[#A] 2,,5
Onc area in which the dcvicc characteristics fail to scale is in the subthrcshold pr weak inversion region of the turn-on characteristic. on V, with B@ow threshold, an inverse 1,, is exponentially slopc, by device is given dependent scmilogarithrnic
1.0 0.5
volts
a ()dccadc
= d log,, 1,,
Fig. 3. Experimental turn-on characteristics for conventional and scaled-down devices shown in Fig. 1 normalized to W/L
= 1.
which
larger
clcvice.
The
the
current
scale
is
to
be
fortuitous magni-
lJaramctcr ~ is important to dynamic memory circuits because it determines the gate voltage excursion rcq~lired to go froln tile low Current off state to the high current (on state ( 11]. In an attempt in (4)
increase thereby (3). In
experimental length used L1orc and Icngth reduction the also heavier scales 3, which
to also extend
~/K), effective
the linear
cause
scaling
relationships
temperature
<% significant, [ 12] and of
but this
surface
lT70uld
the design
scaling for
dcviccs
order and
to
operation the,
accept scale
verified versus
in ~T.
Fig. turn-on
as desired. character-
nonscaling is of particular
property concern
subthreshold Ininiaturc
the
scaled-down
Shon-nj
dynamic
menl-
260
IEEE
JOURNAL
OF SOLID-STATE
CIRCUITS,
OCTOBER
1974
t~*
under drain
and
surface
4(b), for the case of a regions under the source into the lighter
material in the
drain
tend cause
extend With
to merge
much
in
further
the lighter
doped regions
which extreme,
substrate.
deeper junctions
of threshold high give drain a more effects chosen
these depletion
doped or, control
a loss at
the field
shalpatdoping
tern
avoids
these
V,,,=
-Iv
concentration
is properly
device
capacitances
due to the
depletion
Figs.
and 4(b) ], and due to the natural by the ion implantation of the polysilicon but regions. gate The thicker
alignment
afforded
)
Fig.
.
-... -.. .-. . . . . . . . . . . . . . . .1( .. . . . . . .
overlap
gate insulator
capacitance,
in this respect is offset by the decreased for the thicker increase, a design objective
To compensate
1
4. Detailed ture, and (b)
,y~~=-lv
cross sections for (a) scaled-down device struccorresponding ion-implanted device structure.
of Fig. 4(a).
FABRICATION OF ION-IMPLANTED MOS~ETs
ory
circuits
which
require
low
source-to-drain
leakage
currents. ION-IMPLANTED The 4 (a). utilizing shown initial scaling In considerations with the The DEVICE DESIGN just presented length lead to the shown in Fig. design is uses an a factor
ion-implanted
n-channel
a test
lengths
contrast, 4 (b).
capability doping
by ion implantation
ion-implanted is lower
exposure, it was more convenient to use contact masking with high quality master masks for process development. which starting oxide For are this purpose in high the resolution uses lines subsequent was is required as small processing. (i.e., only The about for the gate pattern reduced substrate isolation which resistivity The method between suitable as 1.5 p
of four, and an implanted boron surface layer having a coneentration somewhat greater than the concentration used throughout the unimplanted structure of Fig. 4(a). The concentration and the depth of the implanted surface layer will are chosen so that within is turned the this heavier doped region layer be completely surface depletion
2 CI. cm
7.5 x 1015cm-).
as it is not essential cause several ing dry (40 keV), implanted the silicon thermal
to the work
on with
are available.
Thus, when the source is biased above ground potential, the depletion layer will extend deeper into the lighter doped substrate, and the additional exposed bulk charge will be reasonably small and will cause only a modest increase in the gate-to-source voltage required to turn on the device. With this improvement thickness maintain in substrate gate sensitivity the gate insulator and still can be increased a reasonable
growth
B ions were low dose (6.7 x 101 atoms/cmz) into the wafers, raising the boron doping near surface. All implantations in order were performed diffusion thick of to restrict a 3500-A drain
after gate oxide growth the implanted regions. After silicon the channel Next,
implantation, and
poly 2000-A
to as much as 350A
doped n, and the gate regions regions (100 keV), high dose the the
threshold voltage as will be shown later. Another aspect of the design philosophy low implanted n+ regions surface of depth layer. The implanted p-type
n source As
comparable depletion
105 atoms/cm)
implantation this
through
step, however,
DENNARD
ei!
al.:
ION-IMPLANTED
MOSS13Ts
261
V,(fo, vw~=-l) [v] z:~
6-
,,-., _ :\
ORIGINAL IMPLANT
t5 mE ~ ~
l!i !
\, i
Ns ; ~
: ; \\ 1 ;, ; 0 ., ~ \l ... +IOEALIZED STEP FUNCTION AFTER ANNEALING
e -i ~
X3 -/ Z* / 2+ # 4 I-
_ tdb
Fig. 5. Predicted substrate doping profile for basic ion-implanted device design for 40 keV BI1 ions implanted through the 350-A gate insulator.
Fig. 6. Calculated and experimental substrate sensitivity characteristics for non-implanted devices with 200- and 350-A gate insulators, and for corresponding ion-implanted device with 350-A gate insulator.
gate masks
the channel
region
from
the imThe in a
is redistributed
as shown profiles
dose incident
there.
were obtained
a slight
of ASP5
a computer program developed by F. F. Morehead of our laboratories. The program assumes that boron atoms diffusing face and modeling in the silicon thereby purposes reflect from the silicon-oxide the surface concentration. to use a simple, of the doping profile profiles rather shown interFor idealprofile, apand 5 well raise
ions underneath (or source) implantations 1000C, implanted for polysilicon sulating contact fined, Electrical which implantation the source oxide
the edges of the gates, The gate-to-drain is estimated 20 min without to be of the order of 0.2 steps that and follow 11 min the at at 900C, adequate greatly processing
overlap
it is convenient representation
is more than
to anneal spreading
as shown by the solid line in Fig. 5. The step profile final predicted that The three the advantage parameters. it can be described
all have the same active dose. Using the step profile, old voltage of Poissons tions the vertical short-channel Fig. 6 which equation dimension plots a model for determining from piecewise boundary considers has been developed with one-dimensional and cannot
2000-A
using
low-temperature
chemical-vapor
holes to the n+ and polysilicon and the metalization contact directly regions the final to avoid was applied to the
appropriate model
[11 ], The
account voltage
effects. Results
of the model
the threshold
to-substrate bias for the ion-implanted step profile shown in Fig. 5. For comparison, Fig. 6 also shows the substrate sensitivity with doping, and characteristics gate insulator for like for the nonimplanted having and device a 350-A a constructure. a 200-A and a constant device structure background
density.
ANALYSIS 6.7 350-A X 101 gate 3 persilicon time GausiV~.
doping
channel
implant
is shown the
in Fig.
incident
dose,
The nonirnplanted 200-A case exhibits a low substrate sensitivity, but the magnitude of the threshold voltage is also low. On the other hand, the nonimplanted 350-A case shows a higher threshold, but with an undesirably high substrate sensitivity. The ion-implanted case offers both a sufficiently high threshold voltage and a reason-
implantation function
is given
level,
40 keV
and
standard process-
were
as 1300 treatments
i%, respectively
the heat
of the subsequent
> ably low substrate sensitivity, particularly for V~.S,,I~ 1 V. For Vs..,,,,, < 1 V, a steep slope occurs because the
262
IEEE
JOURNAL
OF SOLID-STATE
CIRCUITS,
OCTOBER
1974
surface
not
inversion
~, the
layer
step
is obtained the
heavier doped
while
im~
the depletion
exceed planted region.
region For
in the
V,.,u,b
under
1 V, at inversion
the lighter a fixed doped
-m
~U K K ~ g a K n
H
10-7
Vd K4V v,u~=-IV
T-
,+/ //
+// /rf +//
,+ //
+/
,+f //
tion with
region V..,,,,,
into with
and thethreshold
then increases
relatively substrate
+1 /
i ;1
sensitivity similar
range
to 4 V)
to the slope of the nonthe threshold design worst reduce still under which will voltage which case. the implanted so that, effects
1.2
.1.4
is significantly adequate conditions threshold conduction plications. Experimental measurements 10 ,p) which agree this 35 keV, 6 result, of 40 kel
x
design margin
Fig. 7. Calucu!at.ed
acteristic for basic lengths with V,u, z
be high ap-
enough so that the device can be turned as required results have are
dynamic given in
These
with
curve. design
101 atoms/cm2
implant
the slightly
1011 atoms/cm2.
(SHORT CHANNEL) ANALYSIS
TWO-DIMENSIONAL
For devices with one-dimensional threshold drain by the which uniform field voltage into the for gate. account doping While
short-channel due to
lengths,
the
model
is inadequate region
I_J???5
12345678 SOURCE-DRAIN SPACING, L (MICRONS) 910
Fig. 8. Experimental and calculated dependence voltage on channel length for basic ion-implanted V,.b = 1 v, v. = 4 v.
complicated
structure leads to
by the nonidentified jected acteristics from threshold were Fig. 3 as the V*. in the actual When manner current the of at
an electric
the
prochargave
pattern that is difficult to approximate. For implanted case, the two-dimensional numerical transport utilized. Chang strate The calculate model The doping of Kennedy and Mock was [15], modified the computer profiles program considered behavior
voltage, plotted
4 X 10-8 A at threshold for all device lengths. The band bending, +,, at this threshold condition is approximately
abrupt
device
with at
used
for these devices. model was used to decurof the ion-implanted of the device
substrate
concentrations
current
the value
small error
the turn-on
vice by a point-by-point
Vt,
were measmodel. the chanin the Apin Fig. 7 curves, in
rent Ior increasing values of gate voltage. Calculated results are shown in Fig. 7 for two values of channel longlength in the range of 1 p., as well as for a relatively
channel device with L =
with
various
channel
lengths
of the two-dimensional determining is described are plotted the calculated data different is plotted voltage devices results with
10 ,fl. All
ratio
of unity,
Ike
cases. As the
show
considering
occurs at about
10-7 A where
the turn-on
characteristics
of channel
The
tion from the exponential subthreshold behavior response on this semilogarithmic plot) to the
square-law behavior. This current level can
as L is decreased
reductions
DENNARD et
al.:
ION-IMPLANTED MoSF@i3
9 y_?-.~ BASIC DESIGN (a) ] 7.5E15 L=o.8p N+ laop lo2p 10p vd=+4v CALCULATED THRESHOLD VOLTAGE [ VOLTS ) FOR Ids = 10-7 AMPERES
263
J v,~ ~ Vm
?
DEEPER SOURCE/ORAIN JUNCTIONS ~+ ---N+
Fig.
9.
drain
voltage load
characteristics
for
basic voltage
ion12.2 ~, 4
(b)
P
Vs.~ =
1 V, L = resistance each
1.1 p, and
W =
xj=o.4p
parameters;
.30 Q, drain
V in 8 steps
0.5 V apart.
circuit
applications greater
value
be set
somewhat
1 ,P so that,
an expected is reason0.3 pwould to this shortfor many of different characterchansource is the No exdrain imin Figs.
Fig. 10. Threshold voltage calculated using two-dimensional rent transport model for various parameter conditions. band voltage of 1.1 V is assumed, curA flat-
range of deviation ably well controlled. give V~ = l. O& channel circuit devices ofL istics effect
alone.
for an ion-implanted are shown The general short-channel taken energy from
shape of the characteristics for much effects devices were using observed a B
:TRATE @vm
Su
channel
respectively. were also used to test parameters. values representalength The of for the tabulates of the design to various in Fig. Fig. as a function considering is about more pared calculated that. the boron effects dose implanted would With occur. in the silicon that the comhave However, thresholds bias and still a grounded
The two-dimensional the sensitivity results threshold indicated are given voltage voltages.
20 percent values
of channel
short-channel
10(a) is an idealized
tion for the basic design that The first perturbation in junction preciable device (from reduction would
has been discussed thus far. to give an apfor the shorter by 20 percent comparable of the from to the shallower
the shallower
implanta-
tion it is possible good substrate is completely with show about design. a heavier
such a cascj again for this case In L = fact, the 0.8 p. is due
Another a factor
the same as for an L = 1.0 ,p. device important depletion improvement layer widths around
is apparently
surface
Also,
less
with
for smaller
to be similar
depletion
under
the
gate
case of deeper j unc~ions. The next possible departure from the basic design is the use of a shallower boron implantation in the channel region, to only give half the as deep, wiih long-channel profile, and a heavier threshold concentration [Fig. same
at threshold,
help prevent
10(d) ]. With
the shallower
is controlled.
264
OCTOBER 1974
1.2 ~ Lo -1 0 ~ 0.8
i
k
1
**
,<p:*-- * 2
% a i ~ 0,6.+ f o 1 >
2 0.4 g g 0,2 I 11 0.0 01234
Fig. 11.
Experimental on channel
and length
calculated for
dependence zero
of
threshold bias
voltage design.
ion-implanted
substrate
CHARACTERISTICS OF THE ZERO SUBSTRATE BIAS DESIGN Since the last be better worthwhile mental 20 keV, obtain and tested behaved devices with 6.0 x design shown in Fig. 10(e) appears effects, fully. In this were to it is built case a
rl
in terms
to review
Experi-
a shallower
~,o, o
4 V applied drain
Fig. 12. Substrate sensitivity characteristics for ion-implanted zero substrate bias design with channel length as parameter.
voltage
is also given in this figure, of threshold design version, old offers improved threshold For control such for strong inthe 13 is its agis of threshold voltage
less variation
as expected.
The dependence
subthreshof Fig.
turn-on suitable
of L, The drain-to-source
low value for this measurement. The results the substrate sensitivity is indeed about the
same for this design with zero substrate bias as for the original design with V.u~ = 1 V. Note that the smaller devices values design, show a somewhat with relatively (and drain) characteristics experimental different values flatter lower voltage. for the zero substrate and calculated, of L. The relatively bias in small are shown substrate thresholds sensitivity at high characteristic The turn-on both I?ig. 13 for
the situation
for dynamic
memo~,
1 V presented
of source
CIRCUIT PERFORMANCE WITH SCALED-DOWN DEVICES The very performance small improvement in expected from of using com-
MOSFETS
integrated
circuits
shift in threshold for the short-channel devices is evident; however, the turn-on rate is considerably slower for this case than for the V,ub = 1 V case shown in Fig. 7. This is due to the fact that the depletion region in the silicon under the gate is very shallow for this zero substrate bias case so that a large portion of a given gate voltage change is dropped across the gate insulator capacitance rather than across the silicon depletion layer capacitance. This is discussed in some detail for these devices in another apbias paper [11 ]. The consequence for dynamic memory plications is that, even though the zero substrate
parably small dimensions is discussed in this section. First, the performance changes due to size reduction alone are obtained from the scaling considerations given earlier. The influence on the circuit performance due to the structural changes of the ion-implanted design is then discussed. Table I lists the changes in integrated circuit performance which follow from scaling the circuit dimensions, voltages, and substrate doping in the same manner as the device changes described with respect to Fig. 1. These changes are indicated in terms of the dimensionless scal-
265
TABLE SCALING RESULTS II LINES Factor
K K
20 KeV,6EII
]o-6 . V,=4V
v$,~=o ,0-7
voltage
drop
L=lOp
1
K
.
*) A .
and reduced
layer by the
widths. transition
These times
caa re-
;
o
,[
0.2 0.4
CTvj
0.6 .0.8 vg [v]
unchanged
resist-
decreased
sultant reduction factor of K. The so the power-delay area of a given the power more density circuits
in the delay time of each circuit by a power dissipation of each circuit is reand current by
K8.
1.0
1.2
1.4
duced by K due to the reduced voltage product device remains problem in Table
K2
Since the
Fig.
or circuit
even if many
are placed
II,
a number
FOR CIRCUIT
the cross-sectional while the length along resolution for the thicknesses with
area of conductors of the conductors widths because (e.g.j down on to for solid limit with the is K voltto remain becomes
Parameter L, W Na
Factor
It is assumed here that are necessarily of the more etching, constant very small comparable degenerately volubility ing factor would It K. Justifying
so
reduced stringent
1/. 1/.
l/K
conductivity (until
is reasonable
VC/Z VI
1/. 1 /K2
1
any
assumptions
be tedious, that
is given.
is argued
all nodal
circuits MC)SFET
decreased
This follows
in comparison
of an unterminated
transmission constant
or some intermediate
of two or more devices, and because the resistof each device is unchanged by scaling, threshold scaling. elements which will properly voltage is made that parasitic or unchanged because subsequently. resistance operate
V/I
R~C, which is unchanged by scaling; however, this makes it difficult to take advantage of the higher switching speeds inherent in the scaled-down devices when signaI propagation over long lines is involved, Also, the current density which MOSFET latively widths of in a scaled-down causes minor, micron a reliability but in high buses and they circuits, conductor concern, become The is increased In problems significant problems circuits by
K,
assumption examined
negligible voltages
Vt
scales as shown in (2), and furthermore because the reduced tolerance spreads on Vt should be proportionately as well if each parameter percentage accuracy. the same time ages are reduced internally in (2) is controlled margins signal generated to the same but at voltNoise are reduced, noise coupling voltage swings,
by widening
by the lower
Due to the reduction in dimensions, all circuit elements (i.e., interconnection lines as well as devices) will have their capacitances reduced by a factor of K. This occurs because of the reduction ponents, which is partially by
K
ion-implanted
similar
devices
performance
in this to that
by K in the area of these comcancelled by the decrease insulating in due to thinner films
of the scaled-down device with K = 5 given in Table I. For the implanted dcviccs with the higher operating voltages (4 V instead (0.9 V instead of 3 V) and higher threshold level will voltages be reduced of 0.4 V), the current
the electrode
spacing
266
IEEEJOURNAL OFSOLID-STATE CIRCUrrS, OCTOBER 1974 to (Vg Tt) /tox to about 80 percent of the
12 II 109 8~ 47 /! + +/ / L= Lma~k-AL 20 KeV,6EII Cniz Vg=Vt+ 0.5 VOLTS V,j =0.05 VOLTS v~ub=f) +/ +/: I
I 1 J
device. The power dissipation a factor of two less in the lines will substrate lines overall would imshow dopelebe less
is thus about the same in both cases. All device devices, and n+ interconnection due to the lighter depth. interconnection so that circuit the would junction
the same improvement ing ancl dec~:ased ments such essentially improvement than a factor is proportional micron as metal unchanged in a typical
Some capacitance
capacitance
be somewhat
of two. The delay time per circuit which to VC/I thus appears to be about the and for the directly scaled-down
z x6 z z & 54-
1 / / /1 ,f/,!,!I AL=0.86P
paper These
I-
small circuits
2
Lmosk (})
lithographic that
techniques
A consistent
technique L.
used
to determine
such as very thin an all ion-implanted these difficulties formance. modified particularly A for
It was then shown how can bc used to overcome device area. or permodel proved degree of current transport structures the relative
on the observation
that
WR.w
LP.m
(Al)
sacrificing ion-implanted
where li,(.~~~ 1s the channel resistance, and pCl,~~the sheet resistance of the channel. For a fixed value of V~ Vt >
in predicting
0, and with
region, pendent intercept sion the
turned
short-channel effects rameter combinations. was to design with lower a l-P circuits channel
arising from different The general objective polysilicon-gate for high-density length
sheet
resistance
an n-channel
versus L~,,k will of L. Then, a plot of w&han t,he Lm.,ll axis at AL because AL = L~.~~ L, reduction etching. in the mask dimenAn example of this and values as follows.
such as those used in dynamic combination control, and substrate lMOSFET B channel As an applied bias that
of subthreshold sensitivity
in Fig. 14. of W and l?,,,. First, used in Fig. the sheet resistance
achieved by an experimental keV, 6,0 x 10 atoms/cm keV, 350-A tended from from the using 4 x 105 atoms/cm and presented of view ~gate insulator, for
that used a 35 implant, a 100 implant, bias attractive but suffers Finally from of design a of in-
source/drain
of the icm-implanted n+ region was determined using a relatively large four-point probe structure. Knowing the n+ sheet resistance drain resistance resistance allows us to compute the source and the resist-
1 V. Also
zero substrate
of a long, slender,
turn-on
range. expected
R,,hnn= VC,,J1,
= (V, I,(R,
+ R. + 2Rc + R1omi))/Id,
(A2)
performance
integrated
circuits
comparably
was projected.
resistance
and Rla,,l is the load resistance of the measurement circuit. Id was determined at VO = Vt + 0.5 V with a small applied drain voltage of 50 or 100 mV. The procedure is more simple and accurate if one uses a set of MOSFETS having different values of L~.,~ but all with the same value
L.,..],
technique length
for
determining small
the
effective
channel. perimental
L for very
MOSFETS
of W,nnsl,.Then
plot
Rchan
versus
data is described
in cwder to determine
DENNARD et U1. : ION-IMPL.4NTED MOSFETS ACKNOWLEDGMENT We wish to acknowledge the valuable contributions of the Also J. J. pre. at the
advanced datz Robert Terrell, H. Dennard (M65) was Tex., in 1932. He received
B. L. Crowder and F. F. ?vIorehead who provided ion implantations and related design information. important Chang Walker paration to were the contributions two-dimensional and testing Research device assisted The and V. DiLonardo of l?. Hwang computations. with the mask facility devices
and M .S. degrees in electrical engineering from ,Southern Methodist University, Dallas, Tex., in 1954 and 1956, respectively, and the Ph.D. degree from Carnegie Institute of Technology, Pittsburgh, Pa., in 1958. In 1958 he joined the IBM Research Division where his experience included study of new devices and circuits for logic and memory communication applications, techniques. and devclo~ment of Since 1963 he has been
and W.
activities. center.
were fabri-
technology
RE~EREN-CES [11
F. Fung, M. Hatzakis, and C, H. Tingj Electron-beam cation of ion implanted high-performance FET J. Vat. Sci. Technol., vol. 10, p. 1082, 1973. J. M. Pankrantz, gain, low-noise lithography, in Dec. [31 fabricircuits, highbeam
at the IBM T. J. Watson Research Center, Yorktown Heights, N. Y., where he worked with a group exploring large-scale integration (LSI), while making contributions in cost and yield models, MOSFET device and integrated circuit design, and FET memory cells and organizations. Since 1971 he has been manager of a group which is exploring high density digital integrated circuits using advanced technology concepts such as electron beam ~attcrn exposure.
[2]
H. T. Yuan, and L. T. Creagh, A transistor fabricated with electron Dig. Int. Electron Devices
Meeting,
H, N. Yu, R. H. Dennard, T. H. P. Chang, .kn experimental high-density memory with electron beam, in IAWCC Dig. inn . . !wwl R. C. Henderson, R. F. W. Pease, Tech. A. M,
[41
P. ranlithoDec.
Fritz
H,
Gaensslen
was born
in Tuebingen,
Mallcry, and R. L. Wadsack, [A high speed dom access 1024-bit memory made with graphy, in Tech. Dig. Int. Electron Devices 1973, pp. 138-140. [51
Germany, on October 4, 1931, He received the Dipl. Ing and Dr. Ing. degrees in electrical engineering from the Technical University of Munich, Munich, 1959 and 1966, respectively. Prior to 1966 he served fessor in the Department gineering, Munich,
u.~
Germany,
in Pro. En-
D. L. ~pears and H. I. Smith, X-Ray lithographya new high resolution replication process, Solid State Tech.nol., vol. ~:~ 15, p. 21, 1972. Projection effects, masking, IBM thin photoresist vol. layers 14, p. N. Yu, at the D. C., inn~fereuce J. Res. Develop.,
as Assistant of Electrical
[61 S. Middlehoek,
Technical Germany.
[71
R. H. D&nard, F, H, Gaensslen, Design of micron MOS switching IEEE Int, Dec. 1972. Electron Devices
L. Kuhn, devices,
and H. presented
Meeting,
Washington,
[81
A. N. Broers and R. H. Dennard, Impact technology on silicon device fabrication, (Electrochem. Sot. Publication), gess, eds., pp. 830-841, 1973. D. L. Critchlow, R. H, Dennard, characteristics sisters, IBM H. and
of electron beam Semicond, Silicon and R. R. BurDesign tran. compleIEEE J. 1972. {Device MOS-
was working on the synthesis of linear and the IBM T. J. Watson Re. ~i~ital networks. In 1966 he joined search Center, Yorktown Heights, N,Y., where he is currently a member of a semiconductor device and process design group. His current technical interests involve various aspects of advanced integrated circuits like miniaturization, device simulation, and ion implantation, assignment at the Dr. Gaensslcn Sclmft. From IBM Scptcmbcr Laboratory, of the 1973 he was on a. one Boeblingen, Germany, Nachrichtentechnische year
R. Huff
is a member
Gesell-
[91
S. E. Schuster,
[101
R. M. Swanson and J. D. Meindl, Ion-implanted mentary MOS transistors in low-voltage circuits, Solid-State Circuits, V. L. Rideout, F, design FETs, H. vol. SC-7, pp. 146-153, April Gaensslen, and A. LeBlanc, for ion Develop., implanted n-channel to be published. Ph~s. Rev.
vol.
[111
[121
A. B. Fowler, Transport
Si surfaces,
properties of
elec169, p. 619,
Hwa-Nien
Yu
(M6,5)
was born
in Shanghai,
[131 [141
W. S. Johnson, IBM System Products Division, E. Fishkill, N. Y., private communication. H, S. Lee, An analysis of the threshold voltage for short channel IGFETs, Soiid-State Electron., 1973. D. P. Kennedy and P. C. Murley, Steady cal theory for the insulated gate field IBi14 J. Res. Develop,, vol. 17, p. 1, 1973. vol. state effect 16, p. 1407,
China, on January 17, 1929. He received the B. S., M.S,, and Ph,D, degrees in electrical engineering from the University of Illinois, Urbana, in 1953, 1954, and 1958, respectively. While at the University, he was a Research Assistant in the Digital Computer Laboratory and worked on the design of the Illiac-11 computer, Since joining the IBM Research Laboratory in 1957, he has been engaged in various exploratory solid-state device research activities. After working with the Advanced Systems Development Division from 1959 to 1962, he rejoined the Research Division in 1962 to work on the ultra-high speed germenium dcvicc technology. Since 1967, he has been engaged in advanced silicon LSI device technology research. He is currently the Manager of Semiconductor Technology at the IBM T. J. Watson Research Center, Yorktown Heights, N.Y. Dr. Yu is a member of Sigma Xi.
[151
mathematitransistor,
[161
[171 [181
M. S. Mock, A two-dimensional mathematical model of the insulated-gate field-effect transistor, Solid-State Electron., vol. 16, p, 601, 1973. Division, W. Chang and P. Hwang, IBM System Products Essex Junction, Vt,, private communication. R. R. Troutman, Subthreshold design considerations for insulated gate field-effect transistors, IEEE J. Solid-State Circuits, vol. SC-9, p. 55, April 1974.
268 V. N.J. gree Leo Rideout (S61M65) received the in 1963 from was born in
IEEE
sity of Wisconsin, Madison, the M.S.E.E. degree in 1964 from Stanford University, Stanford, Calif., and the Ph.D. degree in materials science in 1970 from the University of Southern California (U.S.C.), Los Angeles. His thesis work at U.S.C. under Prof. C. R. Crowell concerned thermally assisted cide Schottky barriers, of current transport in platinum silion semiconductors. ently the the Mr. engaged fabrication Bassous American in of
red detectors at the Centre National dEtudes des Telecommunications, Issy-lesMoulineaux, Seine. From 1960 to 1964 he worked at the Thomas Laboratory in West A. Edison Research Orange, N.J., where
his activities included studies in arc discharge phenomena, ultra violet absorption spectroscopy, and organic semiconductors, In 1964 he joined the IBM Research Laboratory, Yorktown Heights, N.Y., to work As a member of the Research staff he is presthe study of of for materials the and processes used in and silicon integrated the circuits. Electrochemical Advancement of Society Science.
From 1963 to 1965 he was a member of the technical staff Bell Telephone Laboratories where he worked on high-frequency germanium transistors on potassium tantalate. and metal-semiconductor Schottky barriers In 1966 he spent a year as a Research
is a member Association
Assistant in the department of Materials Science at the Technological University of Eindhoven, Eindhoven, The Netherlands, studying acoustoelectric effects in cadmium sulphide. In 1970 he joined IBM Research in the device research group of Dr. L. Esaki where he worked on fabrication and contact technology for multiheterojunction superlattice structures using gallium-arsenidephosphide and gallium-aluminum-arsenide. Since 1972 he has been a member of the semiconductor device and circuit design group of Dr. R. Dennard at the IBM T. J. Watson Research Center, Yorktown Heights, N.Y, His present research interests concern high density silicon FET technology. He is the author or co-author Tau of 20 technical Dr. Rideout Beta Pi, Eta papers and 3 U.S. Patents.
Andre
R. LeBlanc
(M74)
received
the
B.S.
degree in electrical engineering, and the M.S. degree in physics from the University of Vermont, Burlington, in 1956 and 1959, respectively, and trical engineering New Prior Mexico, to joining the D. SC. degree in from the University in 1962. Junction, IBM, Essex elecof
Albuquerque,
is a member of the Electrochemical Society, Kappa Nu, Phi Kappa Phi, and Sigma Xi.
Ernest Bassous was born in Alexandria, 1931. He received the B. SC. degree in versity of London, degree in physical Brooklyn, Brooklyn, From British 1954 Boys to School, Londonj chemistry N.Y. in he England from the 1965. Egypt.
Egypt, chemistry
on
Cor~oration in conjunction with the University of New Mexico. in 1959 he took an educational leave of absence to complete his doctorate. He is presently a member of the Exploratory Memory Group at the IBM Laborato~, Essex Junction, where his current technical interest includes a study of short-channel MOSFET devices. He has authored five publics. tions ports. Dr. and twelve papers, is a member as well as several Xi and IBM Tau Technical Beta Pi. Re-
1959
taught
Chemistry
Physics to
at
Alexandria,
France
LeBlanc
of Sigma