The Dependence of the Performance of Strained
NMOSFETs on Channel Width
Lingyen Yeh, Member, IEEE, Ming Han Liao, Student Member, IEEE, Chun Heng Chen, Jun Wu,
Joseph Ya-min Lee, Chee Wee Liu, Senior Member, IEEE, T. L. Lee, and M. S. Liang
Abstract—The dependence of the performance of strained
NMOSFETs on channel width was investigated. When the
chan-nel width was varied, the stress in the chanchan-nel varied
accord-ingly. This changed the electron effective mass and, consequently,
the ON-state current
I
on. By shrinking the channel width of a
strained NMOSFET from 1 to 0.1
µm and by keeping the channel
length at 55 nm, the ON-state drain current per unit channel width
was enhanced by 22%. The gate leakage current was also affected
by the stress in the channel, which can be explained by the increase
in hole barrier height at the Si
/SiO
2interface. Furthermore, when
the film stress was increased by 1 GPa, the gate leakage current
density
J
gof a strained NMOSFET with a channel width of
0.1
µm and a length of 55 nm under a negative bias −3 V was
reduced by 63%.
Index Terms—Contact etch stop layer (CESL), high-stress
silicon nitride, MOSFET, strained silicon.
I. I
NTRODUCTIONT
HE USE OF a contact etch stop layer (CESL) is one of
the key methods to boost the performance of
nanometer-scale MOSFETs [1]–[3]. The stress in the channel of a
CESL-strained MOSFET can be enhanced by increasing the CESL
film stress and the CESL thickness and by optimizing the device
structure [1], [2]. The stress in the channel is also observed
to depend on the layout of a MOS transistor [3]. However,
few studies have been carried out to increase the stress in the
channel by properly choosing the channel width. The optimal
channel width to reach an
I
ongain for a CESL-strained
NMOS-FET with a channel length of 55 nm has already been reported
[4]. The stress in the channel is also found to affect the gate
leakage current density
J
g. However, the dependence of
J
gon
the channel width has not been fully investigated. This brief
examined the dependence of
I
onand
J
gon the channel width,
Manuscript received January 22, 2009; revised July 22, 2009. Current version published October 21, 2009. The review of this brief was arranged by Editor J. Woo.
L. Yeh is with the Institute of Electronics Engineering, National Tsing Hua University, Hsinchu City 300, Taiwan, and also with Taiwan Semicon-ductor Manufacturing Company, Hsinchu City 300, Taiwan (e-mail: lyyeh. jcsu@msa.hinet.net).
M. H. Liao, T. L. Lee, and M. S. Liang are with Taiwan Semiconductor Manufacturing Company, Hsinchu City 300, Taiwan.
C. H. Chen and J. Y. Lee are with the Institute of Electronics Engineering, National Tsing Hua University, Hsinchu City 300, Taiwan.
J. Wu is with Taiwan Semiconductor Manufacturing Company, Hsinchu City 300, Taiwan, and also with the Graduate Institute of Materials Science and Engineering, National Chiao Tung University, Hsinchu City 300, Taiwan.
C. W. Liu is with the Department of Electrical Engineering, National Taiwan University, Taipei 106, Taiwan, and also with the National Nano Device Laboratories, Hsinchu City 300, Taiwan.
Digital Object Identifier 10.1109/TED.2009.2030542
Fig. 1. Calculated inverse of the electron effective mass plotted as a function ofSxxin the channel. VariousSyy/Sxxratios−0.28, 0.5, 1 (biaxial stress), and 1.5 are depicted.
and its scope has been extended to the channel width for
CESL-strained NMOSFETs with channel lengths ranging from 1
µm
to 55 nm.
II. E
XPERIMENTIn this brief, NMOSFETs were processed using shallow
trench isolation (STI), nitrided-SiO
2gate dielectric, n+
poly-crystalline silicon gate, and tensile silicon nitride CESL [5].
The filling material of the STI was silicon oxide that was
deposited by a high-density-plasma chemical-vapor-deposition
(CVD) process. The STI process followed the conventional
STI process reported in [6]. The effective oxide thickness was
about 1.9 nm, which was measured on a large gate square and
operated at 1.1 V. The CESL was deposited by a CVD process.
The 3-D stress distribution with different device geometric
structures was simulated by a 3-D finite element mechanical
stress simulation program, ANSYS. In the simulation, the
elastic analysis was restricted to the effect of the nitride layer
on the channel [7]. The change in the effective mass under the
stress was calculated by the
k · p model [8], [9].
III. R
ESULTS ANDD
ISCUSSIONFig. 1 shows the calculated inverse electron effective mass
m
∗e
as a function of the simulated
S
xx. Various
S
yy/S
xxratios
of
−0.28, 0.5, 1 (biaxial stress), and 1.5 are depicted. S
xxand
S
yyrepresent the stresses in the channel, which are in parallel
and transverse to the direction of the current flow. The inverse
m
∗e
is observed to increase with an increase in
S
xxwhen the
Fig. 2. (a) CalculatedSyy/Sxxplotted as a function ofSxxwith various combinations of channel length and channel width. (b) Stress components plotted as a function of channel width for NMOSFETs withL = 55 nm.
value of
S
xxis smaller than about 1.5 GPa. Furthermore, the
film stress after annealing must be greater than 3 GPa to give
an
S
xxof 1.5 GPa in the simulation. It is unlikely that silicon
nitride would ever reach such a high tensile film stress [10].
Thus,
S
xxthat is below approximately 1.5 GPa has been the
focus of this study. Moreover,
m
∗eis observed to further reduce
with an increasing ratio of
S
yy/S
xxfrom
−0.28 to 1.5.
Fig. 2(a) shows the ways to reach a high
S
yy/S
xx(small
m
∗e)
by optimizing the dimensions of the NMOSFETs. The
calcu-lated
S
yy/S
xxwas plotted as a function of
S
xxwith various
combinations of channel length
L and channel width W . For
example, consider the curve with
L = 55 nm in Fig. 2(a). It
can be observed that, when
W decreases from 10 to 0.6 µm,
S
yy/S
xxincreases; however,
S
xxdecreases with the reduction
in
W . When W decreases further from 0.6 µm to 75 nm,
S
yy/S
xxkeeps increasing with the reduction in
W , but S
xxchanges from a decreasing to an increasing value with the
reduction in
W . Fig. 2(b) (in our previous work, , Fig. 1(b)[4])
provides an explanation for this phenomenon. When
W is
greater than 0.6
µm, S
yyincreases, whereas
S
xxdecreases,
and, therefore,
S
yy/S
xxincreases with a reduction in
W . Both
S
yyand
S
xxincrease with a reduction in
W when W is lesser
than 0.6
µm. The rate of increase in S
yywith decreasing
W is
observed to be about eight times of that for
S
xx. As a result,
S
yyis found to dominate the width dependence of
S
yy/S
xx, and the
S
yy/S
xxratio is observed to increase with a reduction in
W . In
Fig. 3. ExperimentalIon per unit width compared to the theoretical cal-culation is plotted as a function of channel width. (—) Theoretical Ion, Lphysical= 55 nm (quasi-ballistic model). (· · ·) Theoretical Ion,Lphysical= 80 nm (quasi-ballistic model). (-. -) Theoretical Ion, Lphysical= 0.5 µm (drift-diffusion model). (- - -) Theoretical Ion, Lphysical= 1 µm (drift-diffusion model). (◦) Measured Ion,Lphysical= 55 nm. () Measured Ion, Lphysical= 80 nm. () Measured Ion,Lphysical= 0.5 µm. () Measured Ion, Lphysical= 1 µm. The value for the devices with W = 10 µm is
normalized as one. The inset in Fig. 3 shows Vt− Vt (W = 10 µm) as a function of channel width.(◦) Lphysical= 55 nm. () Lphysical= 80 nm. () Lphysical= 0.5 µm. () Lphysical= 1 µm.
addition,
S
xxis found to take up a minimum value when
W is
about 0.6
µm, which gives rise to the turnaround in the curve
with
L = 55 nm [Fig. 2(a)]. A similar analysis can be applied
to other
L values.
As shown in Fig. 2(a), to reduce
m
∗eby a large
S
yy/S
xxratio, the smallest
W should be used. Furthermore, S
xxand
S
yybecame more tensile, and
S
zzbecame more compressive
with a decrease in
W , and this trend was preferred to increase
the electron mobility [2], [4]. Thus, the preferred
S
xx,
S
yy, and
S
zzstresses and
S
yy/S
xxratio could be reached by adopting
the smallest
W value.
To prove the mechanism mentioned earlier, the measured
I
onper unit width of the strained NMOSFETs with various
W and L values is presented in Fig. 3. By reducing the value
of
W from 1 to 0.1 µm, the I
onper unit width was observed
to increase to about 1%, 2%, 7%, and 22% for
L of 1 µm,
0.5
µm, 80 nm, and 55 nm, respectively. Furthermore, the
magnitude of the increase in the
I
onper unit width increased
with decreasing channel length mainly because of the smaller
separation distance between the center of the channel and the
sources of the applied stress with decreasing
L. Thus, the film
stress became more effective to affect the stress in the channel
and reduce the electron effective mass. Furthermore, the film
stress was also more effective in increasing
I
on. The
stress-induced enhancement of the electron mobility was applied to
the theoretical calculation of NMOSFETs with an
L of 1 and
0.5
µm, based on the drift-diffusion model [11]. However,
ballistic transport was observed in nano-MOSFETs [12]. To
resolve this problem, a quasi-ballistic transport model was used
for nano-NMOSFETs with an
L of 80 and 55 nm (Fig. 3). The
details of the calculation using quasi-ballistic transport were
reported in [8] and [12]–[14].
To further validate the stress mechanism discussed earlier,
J
gunder the stress described in Fig. 2(a) was also studied.
Fig. 4. Ig of the highly and lightly strained NMOS devices is plotted as a function of channel width. The CESL stress for the highly strained devices is 1 GPa higher than that for the lightly strained ones.Vsubstrate= 0 V in this measurement.
accumulation. Both the highly and lightly strained NMOSFETs
were measured, and the CESL stress for the highly strained
devices was 1 GPa higher than the lightly strained ones. The
J
gof the highly strained NMOSFET with a
W/L of 0.1 µm/55 nm
was reduced by 63% compared that of the lightly strained one
under a negative bias of
−3 V. Furthermore, the J
gof the lightly
strained devices was reduced by about 17%, 33%, and 39% by
decreasing
W from 1 to 0.1 µm under V
gbiases of
−1, −2,
and
−3 V, respectively, and that of the highly strained devices
by about 40%, 48%, and 56%, respectively. Thus,
J
gcould be
reduced by decreasing
W from 1 to 0.1 µm, and this effect was
more obvious for the highly strained devices. Generally, the
J
gof NMOSFETs with a smaller
W was higher than those with a
higher
W because J
gnear the STI edge was higher [15]–[17].
Hence, the
J
greduction in Fig. 4 may not have been caused by
the STI process. Furthermore, the energy bands of the strained
silicon and carrier distribution were altered by stress [2]. The
influence of stress on the electron barrier height was relatively
insignificant because the work functions of conductors and
heavily doped polycrystalline silicon changed relatively little
by stress [18]. Thus, the effect was mainly on holes. The
calculation of the splitting of the sixfold degenerate valence
band was reported in [19]. It involved linear splitting and
spin-orbit splitting, expressed in terms of deformation potentials.
The spin-orbit splitting for silicon was very small (
∼0.04 eV),
and, hence, this term was neglected in the calculation. The
required parameters to calculate the deformation potential can
be obtained from [19]–[21]. The calculated result is shown in
Fig. 5. A higher
S
yy/S
xxvalue was observed to cause a greater
shift of the valence band edge. Thus, the barrier height for
light holes increased, and the probability of the hole injecting
into the gate was reduced, which explains the reduction in
J
gin Fig. 4. This
J
greduction was found to be more effective
with decreasing channel length. This can be explained by the
fact that the separation distance between the center of the
channel and the sources of the applied stress became closer with
decreasing channel length. As a result, for an NMOSFET with
smaller channel length, the CESL film stress may become more
effective to affect the stress in the channel and to increase the
barrier height for light holes. Thus, the CESL film stress may
be more effective in reducing
J
g.
Fig. 5. Calculated valence band splitting of light holes (V1), heavy holes (V2), and split-off holes (V3) are plotted as a function of stressSxx. Various Syy/Sxxratios from−0.28 to 1.5 are depicted.
IV. C
ONCLUSIONThus, in this study, the dependence of the performance
of silicon-nitride-strained NMOSFETs on channel width was
investigated. When the value of
W was varied, the stress in
the channel varied accordingly. This changed
m
∗eand,
conse-quently,
I
on. By reducing the value of
W , S
xxand
S
yybecame
more tensile,
S
zzbecame more compressive, and the
S
yy/S
xxratio became higher. All these changes aided in the reduction
in
m
∗eand increased the value of
I
on. By decreasing the value
of
W from 1 to 0.1 µm, the I
onper unit width was observed to
increase by about 1%, 2%, 7%, and 22% for an
L of 1 µm,
0.5
µm, 80 nm, and 55 nm, respectively. The stress in the
channel also reduced the value of
J
g. This was explained by
an increase in the barrier height for light holes. When the film
stress was increased by 1 GPa, the
J
gof an NMOSFET with a
W/L of 0.1 µm/55 nm was reduced by 63% under a negative
bias of
−3 V.
R
EFERENCES[1] S. Pidin, T. Mori, R. Nakamura, T. Saiki, R. Tanabe, S. Satoh, M. Kase, K. Hashimoto, and T. Sugii, “MOSFET current drive optimization using silicon nitride capping layer for 65-nm technology node,” in VLSI Symp. Tech. Dig., 2004, pp. 54–55.
[2] S. E. Thompson, M. Armstrong, C. Auth, M. Alavi, M. Buehler, R. Chau, S. Cea, T. Ghani, G. Glass, T. Hoffman, C.-H. Jan, C. Kenyon, J. Klaus, K. Kuhn, Z. Ma, B. Mcintyre, K. Mistry, A. Murthy, B. Obradovic, R. Nagisetty, P. Nguyen, S. Sivakumar, R. Shaheed, L. Shifren, B. Tufts, S. Tyagi, M. Bohr, and Y. El-Mansy, “A 90-nm logic technology featuring strained silicon,” IEEE Trans. Electron Devices, vol. 51, no. 11, pp. 1790– 1797, Nov. 2004.
[3] A. Oishi, O. Fujii, T. Yokoyama, K. Ota, T. Sanuki, H. Inokuma, K. Eda, T. Idaka, H. Miyajima, S. Iwasa, H. Yamasaki, K. Oouchi, K. Matsuo, H. Nagano, T. Komoda, Y. Okayama, T. Matsumoto, K. Fukasaku, T. Shimizu, K. Miyano, T. Suzuki, K. Yahashi, A. Horiuchi, Y. Takegawa, K. Saki, S. Mori, K. Ohno, I. Mizushima, M. Saito, M. Iwai, S. Yamada, N. Nagashima, and F. Matsuoka, “High performance CMOSFET tech-nology for 45 nm generation and scalability of stress-induced mobility enhancement technique,” in IEDM Tech. Dig., 2005, pp. 229–232. [4] M. H. Liao, L. Yeh, T.-L. Lee, C. W. Liu, and M.-S. Liang, “Superior
nMOSFET performance by optimal stress design,” IEEE Electron Device Lett., vol. 29, no. 4, pp. 402–404, Apr. 2008.
[5] C. C. Wu, Y. K. Leung, C. S. Chang, M. H. Tsai, H. T. Huang, D. W. Lin, Y. M. Sheu, C. H. Hsieh, W. J. Liang, L. K. Han, W. M. Chen, S. Z. Chang, S. Y. Wu, S. S. Lin, H. C. Lin, C. H. Wang, P. W. Wang, T. L. Lee, C. Y. Fu, C. W. Chang, S. C. Chen, S. M. Jang, S. L. Shue, H. T. Lin, Y. C. See, Y. J. Mii, C. H. Diaz, B. J. Lin, M. S. Liang, and Y. C. Sun, “A 90-nm CMOS device technology with high-speed, general-purpose, and low-leakage transistors for system on chip applications,” in IEDM Tech. Dig., 2002, pp. 65–68.
[6] C. H. Yu, M. H. Chi, Y. H. Liou, Y. L. Tu, C. S. Wu, Y. S. Chen, C. Y. Pai, and C. S. Tsai, “Plasma charging defect characterization, inspection, and monitors in poly-buffered STI,” IEEE Trans. Semicond. Manuf., vol. 15, no. 4, pp. 478–485, Nov. 2002.
[7] ANSYS Complete User’s Manual, ANSYS Incorporation, Canonsburg, PA, 2004, p. 22, ver,. 9.0.
[8] M. H. Liao, C. W. Liu, L. Yeh, T.-L. Lee, and M.-S. Liang, “Gate width dependence on backscattering characteristics in the nanoscale strained complementary metal–oxide–semiconductor field-effect transistor,” Appl. Phys. Lett., vol. 92, no. 6, p. 063 506, Feb. 2008.
[9] K. Uchida, T. Krishnamohan, K. C. Saraswat, and Y. Nishi, “Physi-cal mechanisms of electron mobility enhancement in uniaxial stressed NMOSFETs and impact of uniaxial stress engineering in ballistic regime,” in IEDM Tech. Dig., 2007, pp. 129–132.
[10] P. R. Chidambaram, C. Bowen, S. Chakravarthi, C. Machala, and R. Wise, “Fundamentals of silicon material properties for successful exploitation of strain engineering in modern CMOS manufacturing,” IEEE Trans. Electron Devices, vol. 53, no. 5, pp. 944–964, May 2006.
[11] Y. Taur and T. H. Ning, Fundamentals of Modern VLSI Devices. Cambridge, U.K.: Cambridge Univ. Press, 2002, p. 423.
[12] M. Lundstrom, “Elementary scattering theory of the Si MOSFET,” IEEE Electron Device Lett., vol. 18, no. 7, pp. 361–363, Jul. 1997.
[13] M. Lundstrom, “On the mobility versus drain current relation for a nanoscale MOSFET,” IEEE Electron Device Lett., vol. 22, no. 6, pp. 293– 295, Jun. 2001.
[14] M.-J. Chen, H.-T. Huang, K.-C. Huang, P.-N. Chen, C.-S. Chang, and C. H. Diaz, “Temperature dependent channel backscattering coefficients in nanoscale MOSFETs,” in IEDM Tech. Dig., 2002, pp. 39–42. [15] T. Ohashi, T. Kubota, and A. Nakajima, “Ar annealing for suppression
of gate oxide thinning at shallow trench isolation edge,” IEEE Electron Device Lett., vol. 28, no. 7, pp. 562–564, Jul. 2007.
[16] A. H. Perera, J.-H. Lin, Y.-C. Ku, M. Azrak, B. Taylor, J. Hayden, M. Thompson, and M. Blackwell, “Trench isolation for 0.45µm active pitch and below,” in IEDM Tech. Dig., 1995, pp. 679–682.
[17] W. Tonti, R. Bolam, and W. Hansch, “Impact of shallow trench isolation on reliability of buried- and surface-channel sub-µm PFET,” in Proc. IRPS, 1995, pp. 24–29.
[18] M. L. Lee and E. A. Fitzgerald, “Hole mobility enhancements in nanometer-scale strained-silicon heterostructures grown on Ge-rich re-laxed Si1−xGex,” J. Appl. Phys., vol. 94, no. 4, pp. 2590–2596, Aug. 2003.
[19] C. G. V. de Walle and R. M. Martin, “Band line ups and deformation potentials in the model-solid theory,” Phys. Rev. B, Condens. Matter, vol. 39, no. 3, pp. 1871–1883, 1989.
[20] C. Ohler, C. Daniels, A. Foster, and H. Luth, “Heterojunction band offsets and Schottky-barrier heights: Tersoff’s theory in the presence of strain,” Phys. Rev. B, Condens. Matter, vol. 58, no. 12, pp. 7864–7871, 1998. [21] R. Hull, Properties of Crystalline Silicon. London, U.K.: INSPEC, 1999,
ser. EMIS Data Reviews Series, p. 404.
Lingyen Yeh (M’04) received the M.S. degree in
materials science and engineering from National Tsing Hua University, Hsinchu City, Taiwan, in 1994. Since 1994, he has been worked in the semi-conductor industry in Taiwan. From 1994 to 1995, he was with the R&D, Holtek Semiconductor In-corporation. From 1995 to 2000, he was with the R&D, Winbond Electronics Corporation, and its sub-sidiary company Worldwide Semiconductor Manu-facturing Company. Since 2000, he has been with the R&D, Taiwan Semiconductor Manufacturing Com-pany, Hsinchu City. He is currently with the Institute of Electronics Engi-neering, National Tsing Hua University, Hsinchu City. He has done research work in CMOS devices, including shallow trench isolation, strained silicon, ultrashallow junction, thin dielectric film, and CMOS process integration. He has authored and coauthored more than ten papers and received more than ten U.S. patents.
Ming Han Liao (S’07) received the B.S. degree in
mechanical engineering and the Ph.D. degree in elec-tric engineering from National Taiwan University, Taipei, Taiwan, in 2003 and 2007, respectively.
Since 2008, he has been with the R&D, Taiwan Semiconductor Manufacturing Company, Hsinchu City, Taiwan, where he is involved in strained Si technology and ultrashallow junction develop-ment. He has authored or coauthored more than 30 publications/conference proceeding papers/ presentations.
Dr. Liao was the recipient of the Zhuo Zhang-Zong Scholarship Award in 2005 for his outstanding research work.
Chun Heng Chen was born in Taipei, Taiwan, in
1982. He received the B.S. degree in electrical en-gineering from National Chung Hsing University, Taichung, Taiwan, in 2005 and the M.S. degree in electronics engineering from National Tsing Hua University, Hsinchu City, Taiwan, in 2007, working on the electrical characterization and reliability of high-k dielectrics, where he is currently working toward the Ph.D. degree in silicon solar cells.
Jun Wu received the B.S. and M.S. degrees in
mate-rials sciences and engineering in National Tsing Hua University, Hsinchu City, Taiwan, in 1993 and 1997, respectively.
He is currently with the Graduate Institute of Materials Science and Engineering, National Chiao Tung University, Hsinchu City. He is currently with the R&D, Taiwan Semiconductor Manufacturing Company, Hsinchu City, where his work involves thin-film processing and strained silicon technol-ogy. His research interests include semiconductor processing.
Joseph Ya-min Lee received the B.S. degree from
the Department of Physics, National Taiwan Uni-versity, Taipei, Taiwan, in 1965 and the Ph.D. de-gree from the Department of Physics, University of Maryland, College Park, in 1972. His Ph.D. thesis work was done in the area of experimental solid-state physics.
From 1972 to 1975, he was an Associate Professor with the Department of Physics and the Department of Materials Science, National Tsing Hua Univer-sity, Hsinchu City, Taiwan. From 1976 to 1987, he worked in the U.S. semiconductor industry. From 1977 to 1981, he was with Carlsbad Research Center, Hughes Aircraft Company, Carlsbad, CA, and, from 1981 to1987, with Hughes Research Laboratories, Malibu, CA. From 1987 to 1989, he was a Full Professor with the Department of Electrical Engineering and Applied Physics, Case Western Reserve University, Cleveland, OH. Since 1989, he has been a Full Professor with the Department of Electrical Engineer-ing, National Tsing Hua University. He has done research work in experimental solid-state physics, semiconductor photodetectors, CMOS and CMOS/SOS devices, GaAs/AlGaAs real space transfer devices, high dielectric constant, and ferroelectric thin films for VLSI memory devices. He has published more than 115 scientific journal papers and authored seven books.
and M.S. degrees in electrical engineering from Na-tional Taiwan University (NTU), Taipei, Taiwan, in 1985 and 1987, respectively, and the Ph.D. degree in electrical engineering from Princeton University, Princeton, NJ, in 1994.
He is currently a Professor with the Department of Electrical Engineering, NTU, where he is also with the Graduate Institute of Electronics Engineering. He is also with the National Nano Device Laboratories, Hsinchu City, Taiwan. His current research interests include strained-Si/Ge FETs, SiGe photonics, Ge/Si on anything, high-k, metal gate, and bulk/thin-film solar cells. He invented the first MOS tunneling LEDs