scieee Science in your language
[en] (orig)

Impact of intrinsic parameter fluctuations on the performance of In0.75Ga0.25As implant free MOSFETs

Abstract

We investigate the level of statistical variability in implant free (IF) MOSFETs, which are one of the most promising candidates III–V channels implementation. We report results for the threshold voltage (VT) fluctuations in aggressively scaled IF III–V MOSFETs induced by random discrete dopants in the δ-doping plane obtained using 3D drift–diffusion (D–D) device simulations. The D–D simulator is meticulously calibrated against results obtained from ensemble Monte Carlo device simulations. The simulated 30, 20 and 15 nm gate length In0.75Ga0.25As channel IF transistors exhibit threshold voltage standard deviations of 42, 58 and 61 mV, respectively, at a drain voltage of 0.1 V. At a drain voltage of 0.8 V, the threshold voltage standard deviations increase to 55, 71 and 81 mV, respectively. While the standard deviations of VT in the 30 and 20 nm IF MOSFETs are close to those observed in bulk Si MOSFETs with equivalent gate lengths, the threshold voltage standard deviation in the 15 nm gate length IF MOSFET is lower.

Read accessible full text

Impact of intrinsic parameter fluctuations on the performance of In0.75Ga0.25As implant free MOSFETs

Author: Seoane Iglesias, Natalia; García Loureiro, Antonio Jesús; Aldegunde Villar, Manuel Alejo; Kalna, K.; Asenov, A.
Publisher: IOP Publishing
Year: 2009
DOI: 10.1088/0268-1242/24/5/055011
Source: https://minerva.usc.es/bitstreams/5c98baa0-e684-4857-a890-0c769a320924/download
Impac o in insic pa ame e luc ua ions on he
pe o mance o In0.75Ga0.25As implan ee MOSFETs
N. Seoane1, A. Ga cia–Lou ei o1, M. Aldegunde1, K. Kalna2
and A. Aseno 2
1Dep . o Elec onics & Compu e Science, Uni . o San iago de Compos ela,15782
San iago de Compos ela, Spain
2Dep . o Elec onics & Elec ical Enginee ing, Uni . o Glasgow, Glasgow G12
8LT, Sco land, Uni ed Kingdom
E-mail: [email p o ec ed]s
Abs ac . We in es iga e he le el o s a is ical a iabili y in implan ee (IF)
MOSFETs, which a e one o he mos p omising candida es III–V channels
implemen a ion. We epo esul s o he h eshold ol age (VT) luc ua ions in
agg essi ely scaled IF III–V MOSFETs induced by andom disc e e dopan s in he
δ–doping plane ob ained using 3D d i –di usion (D–D) de ice simula ions. The D–
D simula o is me iculously calib a ed agains esul s ob ained om ensemble Mon e
Ca lo de ice simula ions. The simula ed 30, 20 and 15 nm ga e leng h In0.75Ga0.25As
channel IF ansis o s exhibi h eshold ol age s anda d de ia ions o 42, 58 and
61 mV, espec i ely, a a d ain ol age o 0.1 V. A a d ain ol age o 0.8 V, he
h eshold ol age s anda d de ia ions inc ease o 55, 71 and 81 mV, espec i ely.
While he s anda d de ia ions o VTin he 30 and 20 nm IF MOSFETs a e close
o hose obse ed in bulk Si MOSFETs wi h equi alen ga e leng hs, he h eshold
ol age s anda d de ia ion in he 15 nm ga e leng h IF MOSFET is lowe .
PACS numbe s: 81.05.Ea, 85.30.De
Submi ed o: Semicond. Sci. Technol.
Impac o in insic pa ame e luc ua ions on he pe o mance o IF MOSFETs 2
1. In oduc ion
Wi h he scaling o he CMOS echnology owa ds he 22 nm gene a ion, silicon
channel con en ional and no el MOSFET a chi ec u es ace di icul ies o mee he
pe o mance equi emen s [1]. The e o e, in ensi e esea ch is unde way o eplace
he Si channel wi h o he high mobili y semiconduc o which can deli e signi ican
pe o mance imp o emen . MOSFETs wi h Ge channels [2] o p–channel applica ions
o compound semiconduc o s [3] o n–channel applica ions a e now a widely accep ed
op ion o o e coming he pe o mance limi a ions inhe en o he low mobili y o Si
[4]. Recen de elopmen s such as he demons a ion o a high–κga e dielec ic s ack
o GaAs wi h an unpinned in e ace [5], and sui able epi axial laye s uc u es ha e
subs an ially imp o ed he p ospec o he in oduc ion o III–V MOSFETs o high
pe o mance low powe CMOS applica ions. III–V n–channel MOSFETs in combina ion
wi h high pe o mance p–channel Ge ansis o s a e becoming se ious candida es o co–
in eg a ion in dual channel CMOS pla o ms [1], which may ha e unique ad an ages in
ega d o RF, swi ching and powe con ol unc ions.
Wi h he scaling o silicon MOSFETs o nanome e dimensions, he in insic
pa ame e a ia ion in oduced mainly by disc e eness o cha ge and ma e has become
a majo bo leneck o scaling and in eg a ion [6]. One o he majo sou ces o s a is ical
a iabili y a e he andom disc e e dopan s in he ac i e egion o he ansis o s. The IF
MOSFETs may ha e supe io obus ness agains dopan s induced a iabili y as dopan s
a e sepa a ed om he mobile elec ons in he channel. Howe e , he p oximi y o he
δ–doping plane may s ill signi ican ly a ec he elec on popula ion and anspo in
he channel. The e o e, he analysis o he po en ial in insic a ia ions is i al o he
p ope assessmen o he scaling po en ial o III–V IF MOSFETs.
In his wo k, we s udy he h eshold ol age a iabili y due o he p esence o
andom disc e e dopan s in he δ–doping laye o IF In0.75Ga0.25As MOSFETs. The
de ices, scaled o ga e leng hs o 30, 20 and 15 nm, ha e been simula ed using a 3D
pa allel d i –di usion (D–D) de ice simula o desc ibed in de ails elsewhe e [7]. The
pa allelisa ion o he code has been ca ied ou using he s anda d message passing
in e ace (MPI) [8].
In sec ion 2 we in oduce he basic ea u es o ou 3D D–D de ice simula o and
he adop ed simula ion me hodology. The de ice s uc u e and he calib a ion o he
d i –di usion simula o agains Mon e Ca lo simula ion da a a e p esen ed in sec ion 3.
The simula ed a iabili y in he h eshold ol age is p esen ed in sec ion 4. Finally, he
las sec ion summa ises he main conclusions o his wo k.
2. Simula ion me hodology
Figu e 1 shows he low cha o he 3D pa allel ini e elemen simula o o IF
he e os uc u e MOSFETs illus a ing he compu a ional p ocedu e used o sol e he
d i –di usion anspo model equa ions. In his model, he Poisson and he con inui y
Impac o in insic pa ame e luc ua ions on he pe o mance o IF MOSFETs 3
equa ion o elec ons a e sol ed consis en ly wi h mixed Di ichle and Neumann
bounda y condi ions. The disc e isa ion o he Poisson equa ions has been ca ied ou
using he Ri z–Gale kin app oxima ion wi h linea ini e elemen basis unc ions [9].
On he o he hand, he disc e isa ion o he con inui y equa ions equi es a special
app oach. In pa icula , he Scha e e –Gummel disc e isa ion scheme has been used
in his wo k [9].
The non–linea Poisson equa ion and he cu en con inui y equa ion o elec ons
a e decoupled using Gummel i e a ions and linea ised using he New on–Raphson
me hod. The linea sys ems a e hen sol ed in pa allel using domain decomposi ion
me hods [10]. To sol e he local linea sys ems wi hin each subdomain we ha e
used K ylo subspace sol e s, such as he FGMRES and he BiCGSTAB me hods,
p econdi ioned by Incomple e LU ac o isa ions dependen on bo h a nume ical
h eshold and a ce ain le el o ill–in (s anda d ILUT p econdi ione ). Mo e de ails
abou he simula o can be ound in [7].
Quan um co ec ions like he densi y g adien me hod [11] ha e no been included
in o he D–D app oach a his s age, al hough a such simula ed dimensions he use
o quan um co ec ions would be desi able. The quan um co ec ions would a ec he
desc ip ion o he indi idual dopan s in he δ–doping laye and he shape o ca ie
densi y in he channel. Elec os a ically he dopan s induce a deep Coulomb well
esul ing in an a i icially high ca ie densi y in he a ac i e po en ial case. The
quan um co ec ions like densi y g adien me hod would es ic he high ca ie densi y
by app oxima ing he quan um con inemen e ec s. This a i icial high ca ie densi y
c ea ed by he Coulomb po en ial o a poin –like cha ge is educed in ou case by he
use o a cloud–in–cell assigmen wich sp eads he cha ge o e he e ahed al elemen
[12]. Besides, only he ails o he Coulomb po en ial o he disc e e dopan s in he
δ–doping laye a ec he po en ial dis ibu ion in he channel and he a i icial ca ie
apping in his case is less p onounced. The e o e, ou classical simula ions, which a e
compu a ionally less expensi e, s ill p o ide use ul quali a i e esul s.
3. De ice s uc u e and calib a ion
The de ices analysed he e a e In0.75Ga0.25As enhancemen mode MOSFETs wi h
implan ee sou ce/d ain egions [13]. The MOSFET s uc u e is g own on an InP
subs a e and consis s o a 49 nm In0.52Ga0.48As bu e laye and a Si δ–doping laye wi h
3×1012 cm−2. The In0.75Ga0.25As channel laye is sandwiched be ween In0.53Ga0.47As
embed laye s. The high mobili y channel o ms a quan um well wi h ene gy ba ie s
compa able o he supply ol ages a and beyond he 22 nm node. In his way,
he ca ie s a e well con ined o he channel, p o iding ul a– hin body like scaling
pe o mance. Finally, an amo phous Ga2O3/GdGaO dielec ic s ack is used o sepa a e
a me al ga e as shown in igu e 2. The de ices ha e been scaled, in bo h e ical and
ho izon al dimensions in espec o ga e leng hs o 30, 20 and 15 nm. All dimensions
used in he scaling p ocess a e collec ed in able 1.
Impac o in insic pa ame e luc ua ions on he pe o mance o IF MOSFETs 4
The 3D D–D simula o is me iculously calib a ed agains ID–VGcha ac e is ics
ob ained om MC simula ion esul s [14] a d ain biases o 0.1 and 0.8 V as shown in
igu e 3 o ga e leng hs o 30, 20 and 15 nm, espec i ely. The wo k unc ion o he
me al ga e was ixed o 4.0 eV. In he calib a ion, we ha e used a high ield mobili y
model [15]. In he calib a ion p ocess, he low ield mobili y and he sa u a ion eloci y
a e adjus ed o ma ch accu a ely he MC simula ion da a. The calib a ion p ocess
yields a low ield mobili y o 5000 cm2/Vs and a sa u a ion eloci y o 4 ×107cm/s
in he channel o he 30 nm ga e leng h implan ee In0.75Ga0.25As MOSFET, and o
4000 cm2/Vs and 108cm/s o he 20 nm ga e leng h and, inally, o 5000 cm2/Vs and
6.5×107cm/s o he 15 nm ga e leng h.
Figu e 3 also show he da a ob ained om he D–D simula ions on a loga i hmic
scale in he inse s in o de o illus a e he sub h eshold slope. Table 2 collec s he
sub h eshold slope a low and high d ain ol ages o he h ee scaled de ices. The
sub h eshold slope wo sen only sligh ly when he IF ansis o is scaled om 30 o
20 nm ga e leng h, bu i de e io a es by 17% when scaled o he 15 nm ga e leng h.
The sub h eshold slopes a e simila o hose measu ed in Si hin–body ansis o s [16].
Finally, we would like o s ess he limi a ions o ou simula ion model. The D–D
app oach does no p ope ly ep esen he non–equilib ium ca ie dynamics and ballis ic
anspo e ec s in sub–50 nm IF MOSFETs. Howe e , when accu a ely calib a ed
agains he esul s om ensemble Mon e Ca lo de ice simula ions [14] he D–D
app oach gi es a good indica ion o he expec ed in insic ma e ial pa ame e s a ia ions
associa ed wi h de ice elec os a ics like he h eshold ol age luc ua ions [17].
4. Th eshold ol age a iabili y
To in es iga e he h eshold ol age a ia ions in he IF MOSFETs in oduced by he
andom disc e e dopan s in he δ–doped laye , a andom dis ibu ion o dopan s is
gene a ed om he con inuous doping p o ile. The andom placemen o he dopan s in
he δ–doped laye is pe o med on an a omis ic g id de ined by he posi ions o he In,
Al and As a oms, as can be seen in igu e 4. The cha ge associa ed wi h his dis ibu ion
is hen mapped o he de ice simula ion mesh using a cloud–in–cell algo i hm [12]. Fo
e e y dopan , i s cha ge is spli among he ou nodes o he e ahed on enclosing i .
The ac ion o he cha ge assigned o he pa icula node is a ba ycen ic coo dina e
o he cha ge posi ion wi h espec o ha node. In his way, he poin –like cha ge is
smoo hed on he nea es neighbou mesh nodes. Figu e 5 depic s elec on densi y and
conduc ion band p o ile ac oss he he e os uc u e a he posi ion o a andom dopan
in he del a–doped laye . The ionised impu i y a ec s he conduc ion band, p oducing
a po en ial well which esul s in a e y sha p peak o he elec on densi y. The in luence
o his e ec in a h ee–dimensional simula ion can be seen in igu e 6 which shows an
example o he elec on densi y a loga i hmic scale inside he channel and he δ–doping
laye . The posi ion o he dopan s in luences he cha ge dis ibu ion in he channel
al hough he e is a physical sepa a ion be ween hem. The ails o Coulomb po en ial
Impac o in insic pa ame e luc ua ions on he pe o mance o IF MOSFETs 5
associa ed wi h he ionised dono s p oduce a lowe ing o he conduc ion band le el and
an inc ease in he elec on densi y in he channel. The ail o Coulomb po en ial abo e
he disc e e dopan s c ea es islands o nega i e cha ge esul ing in di e en a ou ed
pa hs o he elec ons and consequen ly di e en I–V cha ac e is ics.
We ha e gene a ed s a is ical samples o 100 mic oscopically di e en ansis o s
wi h di e en andom posi ions o dopan s in he δ–doping laye o analyse he impac o
andom disc e e dopan s induced luc ua ions on he h eshold ol age. We ha e used
a cons an cu en c i e ion o ex ac he h eshold ol age. A low d ain bias, he
elec os a ic beha iou below he channel is e y simila o he scaled de ices because
he dimensions o he channel and he laye s below a e kep cons an du ing he scaling
and he same cons an cu en c i e ion has been used (ID= 10 nA·W/L). Howe e ,
wi h an inc easing he d ain bias o 0.8 V he elec os a ic beha iou o each scaled
ansis o becomes a e y di e en . The e o e, a he high d ain bias, we ha e used a
di e en cons an cu en c i e ion o each one o he h ee scaled de ices (ID= 100
nA·W/L, ID= 50 nA·W/L and ID= 300 nA·W/L o he 30, 20 and 15 nm ga e
leng h de ices, espec i ely). The c i e ion has been chosen conside ing he esul s om
simula ions wi h a con inuous doping p o ile whe e ID–VGcha ac e is ics on a log scale
exhibi s a linea beha iou in he sub– h eshold egion. The dis ibu ions o h eshold
ol age a d ain biases o 0.1 V and 0.8 V a e p esen ed in igu es 7 and 8 espec i ely,
o each ga e leng h de ice. The s a is ical mean o he h eshold ol age ob ained
om he sample o de ices wi h di e en dopan dis ibu ions in he δ–doping laye is
always la ge ha he nominal alue om he con inuous simula ions, a bo h low and
high d ain biases. This beha iou is opposi e o he gene al ends o a e age h eshold
ol age lowe ing obse ed in sub–100 nm ga e leng h bulk Si MOSFETs simula ed wi h
andom disc e e dopan s in he channel egion only [18]. This is no o ally su p ising
because he luc ua ions in hese wo de ice a chi ec u es a e coming om wo di e en
sou ces: accep o s in he channel in he bulk Si MOSFET and dono s in he δ–doping in
he IF MOSFET. The disc e e dopan s om he δ–doping laye induce he localisa ion
o elec ons in o Coulomb wells. This localisa ion, on a e age, educes he cu en low
along he channel when compa ed o he con inuous δ–doping simula ions and he ga e
has o lowe , on a e age, he po en ial ba ie much mo e. The a e age h eshold ol age
inc ease a a low d ain ol age is mo e p onounced when he de ices a e scaled down
o 15 nm. Howe e , his beha iou is no seen a high d ain ol age, when he a e age
inc ease is mos impo an o he 20 nm ga e leng h de ice.
Figu e 9 shows he conduc ion band p o ile and elec on densi y in a plane om
sou ce o d ain along he middle o he channel o he 15 nm ga e leng h de ice, a
VD=0.1 V and VG=0.5 V. The e ec o he posi ion and numbe o andom dopan s
on he channel is shown o h ee di e en andom dopan con igu a ions and compa ed
o a de ice wi h con inuous doping (smoo h). The con igu a ion ha p oduces a
highes cu en has a la ge numbe o dopan s, 29, compa ed o he nominal alue
o 21. In his case, he po en ial ba ie is lowe han in he o he cases and he he
elec on densi y is highe , which leads o a highe d ain cu en . The o he wo andom

Impac o in insic pa ame e luc ua ions on he pe o mance o IF MOSFETs 6
dopan con igu a ions ha e 18 and 16 dopan s p o iding cu en s o 20 A/m and 10
A/m espec i ely (since we simula e de ices wi h pa icula wid h and no ansla ion
symme y i is no app op ia e o p o ide cu en densi y bu he exac simula ed
cu en ). The e ec o he posi ion o he dopan s is also no iceable in he igu e.
Fo ins ance, o he lowes cu en de ice, he po en ial ba ie below he ga e is lowe
on one side o he channel, which a ou s he cu en low he e. Howe e , o he de ice
wi h a cu en o 20 A/m he po en ial ba ie below he ga e is highe in he middle o
he de ice, so he cu en low will be h ough he la e al egions o he de ice.
Table 3 shows he s anda d de ia ion o he h eshold ol ages ob ained om his
a iabili y s udy as a unc ion o he ansis o ga e leng h and he applied d ain ol age.
The s anda d de ia ion alues o scaled bulk Si MOSFETs a e p o ided in he same
igu e o compa ison [17]. The impac o luc ua ions becomes mo e p onounced wi h
he applied d ain bias since he s anda d de ia ion inc eases. The s anda d de ia ion o
he h eshold ol age also inc eases wi h educing ga e leng h, ollowing app oxima ely
an ae ial dependence o he andom dopan induced h eshold ol age a ia ion which
magni ude is p opo ional o he i s o de o 1
√W L . Resul s show ha he s anda d
de ia ion o In0.75Ga0.25As IF MOSFET is close o ha o bulk silicon MOSFETs wi h
compa able ga e leng h as can be seen in able 3. The magni ude o he h eshold
ol age luc ua ions indica es ha he andom dopan induced a iabili y om he δ–
doping laye in IF MOSFETs has a simila impac on hei scaling and in eg a ion as
in he same ga e leng h implan ed Si ansis o s [17].
5. Conclusion
In his wo k, a calib a ed pa allel ini e elemen 3D D–D de ice simula o [7] has been
adop ed o he simula ion o IF III–V MOSFETs. The simula o has been me iculously
calib a ed agains ID–VGcha ac e is ics a low and high d ain ol ages ob ained om
comp ehensi e Mon e Ca lo de ice simula ions [14]. The simula o has been used
o in es iga e he in luence o andom disc e e dopan s in he δ–doping laye on he
h eshold ol age in 30 nm, 20 nm and 15 nm ga e leng h IF In0.75Ga0.25As MOSFETs.
The sub h eshold slopes o he IF MOSFETs in he ange o 100 mV/dec, ex ac ed
om he d i –di usion simula ions, a e compa able wi h hose obse ed in no el Si
hin–body ansis o s [16]. The sub h eshold slope sligh ly inc eases when he de ice is
scaled om 30 nm o 20 nm ga e leng h. When he IF ansis o is scaled in espec o
a ga e leng h o 15 nm, he sub h eshold slope de e io a es by 17%.
Resul s show ha he impac o luc ua ions becomes mo e p onounced wi h he
inc ease in he applied d ain bias and wi h he educ ion o he ga e leng h. The
magni ude o he h eshold ol age s anda d de ia ion is simila o ha obse ed in
equi alen bulk Si MOSFETs. The e o e, he in insic pa ame e induced a ia ions in
he h eshold ol age may p esen a simila p oblem o in eg a ion and ci cui design
using IF MOSFETs as o he equi alen ga e leng h Si MOSFETs.
Impac o in insic pa ame e luc ua ions on he pe o mance o IF MOSFETs 7
Acknowledgemen s
This wo k was suppo ed by Spanish Go e nmen (TIN2007–67537–C03–01) and by
Xun a de Galicia (DXIDI07TIC01CT and INCITE08PXIB206094PR). NS and MA
hank Xun a de Galicia and Minis e io de Educaci´on y Ciencia de Espa˜na o hei
awa ded ellowships (A. Al a i˜no and FPU espec i ely). NS also hanks Xun a de
Galicia o p o iding inancial suppo o he posi ion as a isi ing esea che a he
Uni e si y o Glasgow. KK would like o acknowledge he UK EPSRC suppo h ough
g an (EP/D070236/1).
Re e ences
[1] In e na ional Technology Roadmap o Semiconduc o s, 2007 [h p://public.i s.ne ]
[2] Shang H, Lee K –L, Kozlowski P , D’Emic C, Babich I, Siko ski E, Ieong M, Wong H –S P, Gua ini
K and Haensch W 2004 Sel –aligned n–channel ge manium MOSFETs wi h a hin Ge oxyni ide
ga e dielec ic and ungs en ga e Elec on De ice Le . 25 135–137
[3] Chau R, Da a S, Doczy M, Doyle B, Jin B, Ka alie os J, Majumda A, Me z M, and Radosa lje ic
M 2005 Benchma king nano echnology o high–pe o mance and low–powe logic ansis o
applica ions IEEE T ans. Nano echnol. 4153–158
[4] Sko nicki T, Hu chby J A, King T –J, Wong H –S P and Boeu F 2005 The end o CMOS
scaling: owa d he in oduc ion o new ma e ials and s uc u al changes o imp o e MOSFET
pe o mance IEEE Ci cui s De ice 21 1 16–26
[5] Passlack M 2005 De elopmen me hodology o high–κga e dielec ics on III–V semiconduc o s:
GdxGa0.4−xO0.6/Ga2O3dielec ic s acks on GaAs J. Vacuum Sci. Technol. 23 4 1773–1781
[6] Wong H –S P, F ank D J, Solomon P M, Wann H –J, and Welse J 1999 Nanoscale CMOS P oc.
IEEE 87 537–570
[7] Seoane N, Ga cia–Lou ei o A, Kalna K and Aseno A 2007 Impac o in insic pa ame e
luc ua ions on he pe o mance o HEMTs s udied wi h a 3D pa allel d i –di usion simula o
Solid–S a e Elec on. 51 481–488
[8] Message Passing In e ace (MPI) Fo um 2008 h p://www.mpi- o um.o g
[9] Ma kowich P A 1986 The S a iona y Semiconduc o De ice Equa ions (Compu a ional
Mic oelec onics. Sp inge –Ve lag)
[10] Saad Y 1996 I e a i e Me hods o Spa se Linea Sys ems (PWS Publishing Co.)
[11] Aseno A, B own A R, Da ies J H, Kaya S and Sla che a G 2003 Simula ion o In insic Pa ame e
Fluc ua ions in Decananome e and Nanome e scale MOSFET’s IEEE T ans. Elec on De 50
1837–52.
[12] Hockney R W and Eas wood J W 1988 Compu e Simula ion Using Pa icles (IOP Publishing)
[13] Passlack M, Ha in O, Ray M and Medendo p N 2004 US Pa en 6,963,090 2005
[14] Kalna K, D oopad R, Passlack M, and Aseno A 2007 Mon e Ca lo simula ions o InGaAs nano–
MOSFETs Mic oelec on. Eng. 84 2150–2153
[15] Caughey D M and Thomas R E 1967 Ca ie mobili ies in silicon empi ically ela ed o doping
and ields P oc. IEEE. 55 2192–2193
[16] Ka alie os J, Doyle B, Da a S, Dewey G, Doczy M, Jin B, Lionbe ge D, Me z M, Rachmady
W, Radosa lje ic M, Shah U, Zelick N and Chau R 2006 T i–Ga e ansis o a chi ec u e wi h
high–k ga e dielec ics, me al ga es and s ain enginee ing 2006 Symp. VLSI Technol. Dig. Tech.
Pap. 62–63
[17] Roy G, B own A R, Adamu–Lema F, Roy S, and Aseno A 2006 Simula ion s udy o indi idual
and combined sou ces o in insic pa ame e luc ua ions in con en ional nano–MOSFETs IEEE
T ans. Elec on De . 53 12 3063–3070
Impac o in insic pa ame e luc ua ions on he pe o mance o IF MOSFETs 8
[18] Aseno A 1998 Random dopan induced h eshold ol age lowe ing luc ua ions in sub–0.1 µm
MOSFET’s: A 3–D a omis ic simula ion s udy IEEE T ans. Elec on De . 45 12 2505–2513
[19] Roy G, B own A R, Roy S, and Aseno A 2003 Bipola quan um co ec ions in esol ing indi idual
dopan s in ’a omis ic’ de ice simula ions Supe la . Mic os uc . 34 327–334
Impac o in insic pa ame e luc ua ions on he pe o mance o IF MOSFETs 9
Lis o Tables
Table 1. Dimensions conside ed o he scaling o IF In0.75Ga0.25As MOSFETs.
Ga e leng h [nm]
Thickness o [nm] 30 20 15
Sou ce- o-ga e sepa a ion 30 20 15
Ga e- o-d ain sepa a ion 30 20 15
High-κdielec ic 3 2 1.5
In0.52Al0.48As op space 2 1 1
In0.53Ga0.47As op embed 1 0.5 0.5
In0.75Ga0.25As channel 5 5 5
In0.53Ga0.47As bo om embed 1 1 1
In0.52Al0.48As bo om space 2 2 2
δ-doping 3 ×1012cm−2
In0.52Al0.48As bu e 49 49 49
Table 2. The sub h eshold slope a low (VD= 0.1 V) and high (VD= 0.8 V) d ain
ol ages o he IF In0.75Ga0.25As MOSFETs scaled in espec o indica ed ga e leng hs.
Ga e leng h Sub h eshold slope [mV/dec]
[nm] VD= 0.1 V VD= 0.8 V
30 87.5 94.7
20 88.1 106.2
15 105.6 131.7
Table 3. S anda d de ia ion o he h eshold ol age as a unc ion o he ga e leng h
and he d ain ol age o scaled In 0.75Ga0.25As IF MOSFETs compa ed wi h he
s anda d de ia ion obse ed in scaled bulk Si MOSFETs.
In0.75Ga0.25As IF MOSFETs Bulk Si MOSFETs [17]
VDGa e leng h [nm] Ga e leng h [nm]
[V] 30 20 15 35 25 18 13
S anda d de ia ion, σVT[mV]
0.1 42 58 61 33 46 61 100
0.8 55 71 81 – – – –
Impac o in insic pa ame e luc ua ions on he pe o mance o IF MOSFETs 16
(a) (b)
Figu e 9. Conduc ion band p o ile (a) and elec on densi y (b) in a plane om sou ce
o d ain along he middle o he channel o he 15 nm ga e leng h IF MOSFET de ice.
Fou si ua ions a e ep esen ed: a de ice wi h con inuous doping (smoo h) and h ee
de ices wi h di e en andom dopan con igu a ions. The ob ained d ain cu en o
each case is also indica ed. The sou ce and d ain egions a e on he igh and le ,
espec i ely.