Ammonia pretreatment for high-k dielectric growth on silicon

Article abstract of DOI:10.1063/1.1807024

Thermal nitridation of H/Si(100) surfaces with NH₃ gas has been studied as a pretreatment for atomic layer deposition of Al₂O ₃. The chemical nature of both the nitride interface and the Al ₂O₃ growth was characterized using in situ transmission infrared spectroscopy and medium energy ion scattering. Nitride layers thicker than 3-4 A provide an effective barrier against interfacial SiO ₂ formation and promote the nucleation of Al₂O₃ growth.

Full text of DOI:10.1063/1.1807024

Ammonia pretreatment for high- κ dielectric growth on silicon
R. T. Brewer, M.-T. Ho, K. Z. Zhang, L. V. Goncharova, D. G. Starodub, T. Gustafsson, Y. J. Chabal, and N.
Moumen
Citation: Applied Physics Letters 85, 3830 (2004); doi: 10.1063/1.1807024
View online: http://dx.doi.org/10.1063/1.1807024
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/85/17?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Effects of film reoxidation on the growth and material properties of ultrathin dielectrics grown by rapid thermal
nitridation in ammonia
J. Vac. Sci. Technol. B 26, 1382 (2008); 10.1116/1.2953730
Atomic scale characterization of Hf O 2 ∕ Al 2 O 3 thin films grown on nitrided and oxidized Si substrates
J. Appl. Phys. 96, 6113 (2004); 10.1063/1.1808245
Microstructure and thermal stability of HfO 2 gate dielectric deposited on Ge ( 100 )
Appl. Phys. Lett. 85, 2334 (2004); 10.1063/1.1794849
Enhanced initial growth of atomic-layer-deposited metal oxides on hydrogen-terminated silicon
Appl. Phys. Lett. 83, 740 (2003); 10.1063/1.1595719
Ultrathin nitrided-nanolaminate ( Al 2 O 3 / ZrO 2 / Al 2 O 3 ) for metal–oxide–semiconductor gate dielectric
applications
J. Vac. Sci. Technol. B 20, 1143 (2002); 10.1116/1.1481864
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.230.73.202 On: Tue, 23 Dec 2014 00:20:38

APPLIED PHYSICS LETTERS
VOLUME 85, NUMBER 17
25 OCTOBER 2004
Ammonia pretreatment for high-␬ dielectric growth on silicon
a)
R. T. Brewer, M.-T. Ho, K. Z. Zhang, L. V. Goncharova, D. G. Starodub,
T. Gustafsson, and Y. J. Chabal
Departments of Chemistry & Chemical Biology and Physics & Astronomy, Laboratory for Surface
Modification, Rutgers University Piscataway, New Jersey 08854
N. Moumen
International SEMATECH, 2706 Montopolis Drive, Austin, Texas 78741
(
Received 24 May 2004; accepted 17 August 2004)
Thermal nitridation of H/Si͑100͒ surfaces with NH gas has been studied as a pretreatment for
3
atomic layer deposition of Al O . The chemical nature of both the nitride interface and the Al O
2
3
2
3
growth was characterized using in situ transmission infrared spectroscopy and medium energy ion
scattering. Nitride layers thicker than 3–4 Å provide an effective barrier against interfacial SiO2
formation and promote the nucleation of Al O growth. © 2004 American Institute of Physics.
2
3
[
DOI: 10.1063/1.1807024]
The implementation of high-␬ dielectrics to replace SiO
silicon nitride surface promotes nucleation and linear growth
similar to Al O growth on hydrophilic SiO surfaces.
2
1
,2
gate oxide is urgently needed. Appropriate high-␬ dielec-
tric materials must meet a series of requirements, such as
thermodynamic stability with Si to avoid the formation of an
2
3
2
Si͑100͒ samples are prepared using standard RCA clean-
1
6
ing techniques and a 20 s 10% –20% HF dip to create a
1
7
interfacial SiO layer, and thus to permit the deposition of
well-defined hydrogen surface termination.
The
2
gate oxides with the thin ͑Ͻ1 nm͒ effective oxide thickness
required by the semiconductor industry road map.
Atomic layer deposition (ALD) is a promising, industri-
ally attractive method for depositing ultrathin gate oxides
H-terminated Si͑100͒ samples are placed in a dry nitrogen
purged ALD cell operating at atmospheric pressure and lo-
cated inside the sample compartment of a Nicolet Magna
infrared spectrometer. Nitridation is performed by intro-
ducing a 4% NH3/N2 gas mixture into the reaction chamber,
raising the sample temperature to the desired temperature at
3
,4
1
8
without the formation of an interfacial SiO layer. However,
2
attempts to use ALD to deposit high-␬ dielectrics directly on
about 100 °C/min, and then annealing for 2 min. Al O is
H-terminated silicon have resulted in the formation of an
2
3
5
deposited by alternating exposures of trimethylaluminum
unwanted SiO interfacial layer. On H-terminated Si͑100͒
2
1
9
(
TMA) and D O (respectively 0.001 and 0.02 mole fraction
for instance, Al O growth cannot be initiated by replacing
2
2
3
in 0.81/min N carrier gas, atmospheric pressure), with at
the hydrogen termination with a hydroxyl group. Previous
experiments have shown that exposure of H-terminated
Si͑100͒ to either O or H O at elevated temperatures results
2
least 30 min purge time between precursor exposures. The
IR windows are protected during the precursor exposure by
shutters and nitrogen purge. The shutters are only opened to
take IR measurements between deposition cycles after the
chamber is fully purged. Al O ALD is performed between
2
2
in oxygen insertion into the silicon back-bond without the
6
removal of the surface hydrogen. Additionally, in situ IR
2
3
absorption studies of Al O ALD have revealed that SiO
2
3
2
3
18 and 380 °C. Temperatures are measured using a thermo-
formation is associated with the presence of metal precursor
at the surface, i.e., after the initial trimethylaluminum depo-
couple clipped to the substrate and calibrated using an infra-
red optical pyrometer. The difference between the thermo-
couple and actual sample temperature (as measured
optically) is significant (see Table I) due to convective cool-
7
sition, which often only partially covers the surface. While
longer exposures of the metal precursor prior to ALD growth
do minimize SiO formation and promote the growth of
2
8
high-␬ dielectrics, complete elimination of the SiO interfa-
2
cial layer has not been demonstrated.
TABLE I. Summary of SixNy growth by annealing H-teminated Si ͑100͒ in
4
% NH /N . The temperatures are measured using a thermocouple (TM)
In this letter, we investigate ammonia thermal pretreat-
ment of H-terminated Si͑100͒ to passivate the surface
against SiO formation during Al O ALD, and we monitor
3 2
clipped to the sample and by infrared optical pyrometry. The nitride thick-
nesses were calculated using the measured MEIS nitrogen areal density and
2
2
3
−3
assuming the Si Ny has the density of bulk Si N ͑3.3 g/cm ͒, then esti-
x
3
4
the surface chemical state with in situ infrared spectroscopy.
mating other film thicknesses grown at other temperatures by comparing the
9
−
1
Si N is a well-known diffusion barrier and has been used to
nitride LO mode absorption intensity ͑ϳ850 cm ͒.
3
4
1
0
11
protect materials from semiconductor dopants metals,
1
2
13,14
water, and oxygen.
Silicon nitride has also been pro-
Temperature ͑°C͒
posed as a barrier between HfO and Si to avoid silicate
2
MEIS
N density ͑atoms/cm ͒
Calculated
formation in high-␬ dielectric gate stacks and thus to mini-
2
TM
Optical
Si N thickness ͑Å͒
x
y
1
5
mize the equivalent oxide thickness (EOT). While silicon
nitride diffusion barriers are often tens of nanometers thick,
we find that 3–4 Å is sufficient to prevent the formation of
interfacial SiO during Al O ALD. We also find that the
350
400
500
600
700
439
496
605
715
826
0.7
1.2
3.3
5.5
7.5
2.0ϫ1015
3.9ϫ1015
2
2
3
a)
Electronic mail: rhett.t.brewer@intel.com
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
0
003-6951/2004/85(17)/3830/3/$22.00 3830 © 2004 American Institute of Physics
1
31.230.73.202 On: Tue, 23 Dec 2014 00:20:38

Appl. Phys. Lett., Vol. 85, No. 17, 25 October 2004
Brewer et al.
3831
FIG. 2. IR absorbance spectra of Al O3 grown by ALD at 330 °C on a
2
FIG. 1. IR transmission absorbance spectra from hydrogen terminated
Si͑100͒, taken in situ at 45° incidence angle, as the surface was exposed to
a 4% NH /N gas mixture (atmospheric pressure) for 2 min at the indicated.
nitrided silicon surface, prepared by NH3 exposure of a H-terminated
Si͑100͒ surface at (a) 495 °C and (b) 605 °C, at atmospheric pressure. The
reference spectrum is the respective nitrided surface in both cases.
3
2
temperature. The reference spectrum for (a) is from the H-terminated
Si͑100͒ surface and for (b) it is from the sample exposed to NH at 825°C.
3
nitrogen and assumed a film density of bulk Si N
3
4
3
͑
3.3 g/cm ͒ to determine the thickness of the Si N layers
x y
ing of the thermocouple in the atmospheric pressure nitrogen
purge ambient.
grown at 605 and 825 °C. The Si N thickness for the films
x y
grown at other temperatures were determined by scaling the
intensity of the Si–N–Si IR absorption peak at 850 cm−1 to
the values measured at these two points (Table I). These
values represent a lower limit on the film thickness because
these thin Si N films may be less dense than bulk Si N .
Exposure of Si to NH at high temperatures is a standard
3
2
0,21
process for growing Si N ,
an effective oxygen diffusion barrier
which has been shown to be
x
y
9
,13,15
thus preventing
interfacial SiO formation. However, its thickness should be
2
x
y
3
4
minimized to attain low EOT because its dielectric constant
Combining Si N density and Si–N–Si IR peak measure-
x y
͑
ϳ7͒, while larger than SiO , is still lower than other high-␬
ment uncertainties, the thickness of the 605 °C NH nitride
2
3
1
materials like HfO and Al O (25 and 9, respectively).
film could be as high as 4.1 Å (3.3 Å in Table I), so we
estimate our thickness measurements to be within 1 Å.
ALD Al O film thicknesses are extracted from MEIS
2
2
3
We first monitor the formation of Si N from the expo-
x
y
sure of H-terminated Si͑100͒ to a 4% NH /N gas mixture
3
2
2
3
as a function of temperature using in situ transmission IR
spectroscopy (see Fig. 1). There is no decrease of the Si–H
Al areal densities. We find that as a result of the very large
exposures used in our experiments, and possible incomplete
purging between growth cycles, some chemical vapor depo-
sition may take place resulting in a higher growth rate (up to
5 Å per cycle) than what was reported (ϳ1.8 Å per cycle)
for pure (and possibly incomplete) ALD growth. Our re-
sults are therefore characteristic of fully saturated layers and
observations of interfacial silicon oxidation processes are in-
trinsic to the materials system, rather than a result of incom-
plete precursor surface coverage. However, for all the studies
presented in the following, great care is taken to maintain
consistent growth conditions to ensure meaningful compari-
sons.
−
1
stretching modes, ϳ2100 cm [Fig. 1(a)] or growth of the
−
1
Si–N–Si phonon modes, ϳ840 cm (Ref. 22) [Fig. 1(b)]
until the temperature reaches 380 °C, at which point we ob-
serve a small reduction in the Si–H stretching modes. At
ϳ440 °C, we observe a drastic reduction in the surface hy-
drogen (only traces of the mono-hydride remain), accompa-
nied by the formation of Si–N–Si bonds. Further increases
in the 4% NH3 exposure temperature result in larger
Si–N–Si peak absorption areas, which we assume to be pro-
portional to the thickness of the Si N layer. Evidently, the
2
4
x
y
H-terminated surface is inert to NH at low temperatures and
3
the surface only begins to react at temperatures where hydro-
gen begins to desorb from the surface. At no time during our
The effectiveness of NH3 grown SixNy in preventing
SiO2 formation is first studied by growing ALD Al2O3 on
SixNy surfaces at various temperatures and measuring the
formation of interfacial SiO2 with transmission IR spectros-
copy. Figure 2 shows the absorption spectra after one and
four cycles of TMA/D2O exposures, at 331°C, on Si͑100͒
H-terminated surfaces that had been pretreated with a 4%
NH /N exposure at ϳ495 and 605°C for Figs. 2(a) and
experiment do we observe Si–NH absorption modes be-
x
cause the temperatures at which NH3 reacts with the
H-terminated surface are above the NH decomposition tem-
x
2
3
peratures observed for NH absorption on porous Si.
3
We use the Si–N–Si mode IR absorption area as a quan-
titative measurement of Si N thickness by calibrating it with
x
y
3
2
medium energy ion scattering (MEIS) for two Si N films
2(b), respectively. We note Al O growth during the first
x
y
2 3
grown by exposing an H-terminated Si͑100͒ surface to 4%
exposure cycle on Si N (evident from the Al O absorption
x y 2 3
−
1
NH /N at 605 and ϳ825 °C. The MEIS spectra are best
peak 850 cm ), which is not observed under identical depo-
sition conditions on H-terminated Si͑100͒. From the TO and
3
2
modeled by a thick SiO N layer on a thin Si N layer, in
x
y
x
y
−
1
contrast to the IR data which do not exhibit a strong SiO N
LO absorption modes of SiO (1020 and 1175 cm , respec-
x
y
2
−
1 22
peak at 950 cm . This apparent discrepancy indicates that
the as grown, predominantly Si N film is heavily oxidized
tively), it is clear that ϳ1 Å (average thickness) of SiO is
2
formed on the 1.2-Å-thick Si N grown at 495 °C [Fig. 2(a)],
x
y
x
y
to form oxynitride during the transfer from the growth to the
while no measurable SiO ͑Ͻ0.1 Å͒ is formed on the 3.3-
2
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
MEIS chamber. We used the MEIS measured areal density of
Å-thick Si N grown at 605 °C [Fig. 2(b)]. We conclude that
x y
1
31.230.73.202 On: Tue, 23 Dec 2014 00:20:38

3832
Appl. Phys. Lett., Vol. 85, No. 17, 25 October 2004
Brewer et al.
growth on hydrophilic SiO , indicating that the Si N surface
2
x
y
is reactive to TMA and does not pose a barrier to Al O3
2
nucleation and growth. In contrast, the first cycle of
TMA/D O on the H-terminated surface, even at these very
2
large precursor exposures, results in no measurable Al O3
2
because of the low reactivity of the H-terminated Si͑100͒
surface with TMA.
In conclusion, we have demonstrated that interfacial
SiO formation is prevented by Si N films only 3–4 Å thick
2
x
y
during ALD growth of Al O . Immediate growth of Al O
2
3
2
3
during the first TMA/D O exposure cycle demonstrates that
2
there is no nucleation barrier to Al O growth on the Si N
2
3
x
y
surface.
R.T.B. and Y.J.C. are grateful to M.M. Frank, S. Rivil-
lon, and E. Garfunkel for stimulating discussions. T.G (L.G.,
D.S) and Y.C. (R.B., M.H.) acknowledge support from the
National Science Foundation (No. DMR 0218406) and (No.
CHE-0415652), respectively.
FIG. 3. IR absorbance spectra showing the oxidation of 3-Å-thick Si N on
x
y
Si͑100͒ as a function of the D O exposure temperature. The reference is the
2
H-terminated surface.
NH thermal Si N films must be between 3 and 4 Å thick
1
3
x
y
G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys. 89, 5243
before they are continuous and that a continuous Si N is
effective in preventing the formation of interfacial SiO2.
(2001).
x
y
2
A. I. Kingon, J. P. Maria, and S. K. Streiffer, Nature (London) 406, 1032
x
y
3(
2000).
The ability of a continuous Si N layer to prevent inter-
D. G. Schlom and J. H. Haeni, MRS Bull. 27, 198 (2002).
K. J. Hubbard and D. G. Schlom, J. Mater. Res. 11, 2757 (1996).
B. W. Busch, O. Pluchery, Y. J. Chabal, D. A. Muller, R. L. Opila, J. R.
Kwo, and E. Garfunkel, MRS Bull. 27, 206 (2002).
facial SiO formation is further investigated by exposing a
4
5
2
3
-Å-thick Si N film to the ALD oxidant, D O, at increasing
x y 2
temperatures (Fig. 3). It is clear that, while Si N is oxidized
x
y
6
7
X. Zhang, E. Garfunkel, Y. J. Chabal, S. B. Christman, and E. E. Chaban,
Appl. Phys. Lett. 79, 4051 (2001).
M. M. Frank, G. D. Wilk, and Y. J. Chabal, Appl. Phys. Lett. 82, 4758
by D O even at 190 °C to form Si O N , there is no SiO
2
x
y
z
2
formation up to 605 °C. In contrast, an H-terminated Si͑100͒
surface first treated with TMA at 330 °C (half of our stan-
dard Al O ALD cycle) and then exposed to D O at 380 °C
8(
2003).
2
3
2
M. M. Frank, Y. J. Chabal, M. L. Green, A. Delabie, B. Brijs, G. D. Wilk,
M. Y. Ho, E. B. O. da Rosa, I. J. R. Baumvol, and F. C. Stedile, Appl.
Phys. Lett. 83, 740 (2003).
results in the formation of over 1 Å SiO2.
Finally, the nucleation of Al O on Si N is examined by
2
3
x
y
9
I. J. R. Baumvol, Surf. Sci. Rep. 36, 5 (1999).
comparing the rate of Al O growth for the first four cycles
10
2
3
M. D. Maeem, C. Parks, K. Wangemann, and H. Glawischnig, Thin Solid
Films 290–291, 485 (1996).
C. Ahrens, C. Depta, F. Schittgelm, and S. Wilhelm, Appl. Surf. Sci. 91,
of TMA/D O exposure to the growth rate of Al O on
2
2
3
11
H-terminated Si͑100͒ surfaces and on SiO surfaces, pre-
2
285 (1995).
pared by dipping an H-terminated Si͑100͒ sample into a so-
1
2
2
5
S. Yan, H. Maeda, J.-I. Hayashi, K. Kusakabe, S. Morooka, and T. Okubo,
J. Mater. Sci. 28, 1829 (1993).
13
lution of 1 HCl:1 H O :4 H O at 80 °C for 10 min. In
2
2
2
−
1
Fig. 4, the Al O absorption phonon area ͑950 cm ͒, which
is proportional to Al O film thickness, is plotted as a func-
2
3
F. Fontaine, A. Deneuville, E. Lucazeau, E. Gheeraert, C. Savall, and J. C.
Bruyers, Diamond Relat. Mater. 4, 596 (1995).
4
2
3
1
R. R. Wang, G. E. Welsch, and O. Monteiro, J. Biomed. Mater. Res. 46,
tion of TMA/D O exposure cycle. On the OH-terminated
2
152
62 (1999).
SiO surface we observe linear growth of Al O from the
2
2
3
H. Takeuchi and T. J. King, Appl. Phys. Lett. 83, 788 (2003).
RCA cleaning procedure: Acetone, methanol, water rinse, followed by
treating the sample with 1 H2O2 :1 NH4OH:4 H2O at 80°C and
1 H2O2 :1 HCl:4 H2O at 80°C for 10 min each with water rinse in be-
tween.
first TMA/D O exposure cycle, confirming that there is no
16
2
2
4
barrier to Al O nucleation. Al O deposition under the
2
3
2
3
same conditions on the Si N surface proceeds as fast as
x
y
1
1
1
7
8
9
Y. J. Chabal, G. S. Higashi, K. Raghavachari, and V. A. Burows, J. Vac.
Sci. Technol. A 7, 2104 (1989).
M. M. Frank, Y. J. Chabal, and G. D. Wilk, Mater. Res. Soc. Symp. Proc.
745, 41 (2003).
Although not important for these experiments, we use D O as an oxidizer
2
to allow for spectroscopic differentiation between water fluctuations in the
spectrometer purge gas and surface reactions.
S. Ishidzuka, Y. Igari, T. Takaoka, and I. Kusunoki, Appl. Surf. Sci. 130–
2
2
0
1
132, 107 (1998).
J. W. Kim, H. W. Yeom, K. J. Kong, B. D. Yu, D. Y. Ahn, Y. D. Chung, C.
N. Whang, H. Yi, Y. H. Ha, and D. W. Moon, Phys. Rev. Lett. 90, 106103
2(
2003).
2
2
2
B. Semmache, M. Lemiti, C. Chaneliere, C. Dubois, A. Sibai, B. Canut,
and A. Laugier, Thin Solid Films 296, 32 (1997).
3
4
A. C. Dillon, P. Gupta, M. B. Robinson, A. S. Bracker, and S. M. George,
J. Vac. Sci. Technol. A 9, 2222 (1991).
L. G. Gosset, J. F. Damlencourt, O. Renault, D. Rouchon, P. Holliger, A.
Ermolieff, I. Trimaille, J. J. Ganem, F. Martin, and M. N. Semeria, J.
Non-Cryst. Solids 303, 17 (2002).
FIG. 4. Peak areas of the Al O longitudinal optical (LO) absorption band
2
3
2
5
(
assumed proportional to film thickness) as a function of the TMA/D O
Y. J. Chabal, M. K. Weldon, A. B. Gurevich, and S. B. Christman, Solid
2
This a rdt iec pl eo s iist i oc no pc yy rc i lge h ft oe rd d ai fs f ei nr edn i tc as ut er fda ci ne s t: h Se i –a Hr ti ,c Sl ei .O R ea un sd eS oi fN A .I P content is subjec St t taot et hP eh et en or mm s. a6 t5 :– h6 t6t p, :2/ /5s 3c i(t 1a 9t i9o 9n ) .. aip.org/termsconditions. Downloaded to IP:
2
x
y
131.230.73.202 On: Tue, 23 Dec 2014 00:20:38