Furnace gas-phase chemistry of silicon oxynitridation in N2O

Article abstract of DOI:10.1063/1.115909

During furnace N2O-based silicon oxynitride growth, the total concentration of NOx species varies strongly with the flow rate of N2O. At low flow rates the N2O decomposes at least partially in the cooler region of the furnace near the gas inlet. This results in lower than expected NOx (x=1,2) concentrations in the oxidizing ambient. At high flow rates, the exothermic decomposition of N2O can heat the inlet region of the furnace, resulting in decomposition at a temperature above the nominal furnace temperature and higher than expected NOx concentrations. These effects can lead to a substantial variation in the concentration of N in the oxynitride as a function of N2O flow.

Full text of DOI:10.1063/1.115909

Furnace gasphase chemistry of silicon oxynitridation in N2O
K. A. Ellis and R. A. Buhrman
Citation: Applied Physics Letters 68, 1696 (1996); doi: 10.1063/1.115909
View online: http://dx.doi.org/10.1063/1.115909
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/68/12?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Enhanced nitrogen incorporation and improved breakdown endurance in nitrided gate oxides prepared by anodic
oxidation followed by rapid thermal nitridation in N2O
Appl. Phys. Lett. 69, 3875 (1996); 10.1063/1.117134
On the formation process of luminescing centers in sparkprocessed silicon
J. Appl. Phys. 80, 5318 (1996); 10.1063/1.363470
Nitrogen transport during rapid thermal growth of silicon oxynitride films in N2O
Appl. Phys. Lett. 69, 2385 (1996); 10.1063/1.117645
Nanometerscale local oxidation of silicon using silicon nitride islands formed in the early stages of nitridation
Appl. Phys. Lett. 69, 2222 (1996); 10.1063/1.117172
Dry oxidation mechanisms of thin dielectric films formed under N2O using isotopic tracing methods
Appl. Phys. Lett. 68, 2366 (1996); 10.1063/1.116135
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:
142.157.183.114 On: Fri, 28 Nov 2014 16:12:00

Furnace gas-phase chemistry of silicon oxynitridation in N2O
K. A. Ellis and R. A. Buhrman
School of Applied and Engineering Physics, Cornell University, Ithaca, New York 14853-2501
͑Received 22 November 1995; accepted for publication 22 January 1996͒
During furnace N2O-based silicon oxynitride growth, the total concentration of NOx species varies
strongly with the flow rate of N2O. At low flow rates the N2O decomposes at least partially in the
cooler region of the furnace near the gas inlet. This results in lower than expected NOx ͑xϭ1,2͒
concentrations in the oxidizing ambient. At high flow rates, the exothermic decomposition of
N2O can heat the inlet region of the furnace, resulting in decomposition at a temperature above the
nominal furnace temperature and higher than expected NOx concentrations. These effects can lead
to a substantial variation in the concentration of N in the oxynitride as a function of N2O
flow. © 1996 American Institute of Physics. ͓S0003-6951͑96͒05012-X͔
The use of N2O in the growth of gate oxides dates to
1990, when it was discovered that oxidation in N2O results in
the enhancement of electrical properties of metal-oxide-
semiconductor field-effect transistors ͑MOSFETs͒.1 This
benefit has been linked to the position and bonding state of
incorporated nitrogen.2 X-ray photoelectron spectroscopy
͑XPS͒ measurements indicate that when N is incorporated
within 20 Å of the interface, it is triply bonded to Si, and
appears to be responsible for suppressing latent interface
states. N further from the interface, where it is doubly
bonded to Si, appears to play no such role. It has also been
shown that the presence of N in the oxide is useful in sup-
pressing the indiffusion of B from pϩ-polysilicon in
p-MOSFET devices.3
higher flow rates, the exothermic decomposition of N2O can
heat the inlet, resulting in greater NO production and an
increase in the amount of N incorporated into the oxide. As a
test of this model, we have studied the temperature profile
and exhaust gas NOx concentration ͑xϭ1,2͒ as a function of
flow rate for a small tube furnace. We have also directly
measured the N content in oxides grown at different flow
rates in a conventional gate oxidation furnace.
The furnace used in this study was a 1 in. ͑ID͒ quartz
furnace tube, resistively heated along 15 in. of its length.
Although the furnace diameter was obviously much smaller
than that typically used for growing gate oxides, the reaction
mechanisms involved in N2O decomposition have been
shown to be independent of surface effects.8 Thus, we antici-
pate that most of our observations will scale with furnace
size. However, we acknowledge that heat dissipation and tur-
bulence effects will not scale with furnace size. We also
identify the gas flow velocity, in cm/s at standard tempera-
ture and pressure, as the most fundamental measure of flow
rate.
One of the concerns associated with N2O furnace growth
of gate oxides is the variation of N concentration with the
N2O flow rate.4 This was recently attributed by Tobin et al.,5
to the influence of reaction ͑1͒,
2NOϩO2↔2NO2,
͑1͒
which results in the gradual conversion of NO, previously
produced during the decomposition of N2O, into NO2.4
However, it is well-established that NO2 dissociates rapidly
and completely into NO and O2 above 600 °C.5,6 Meaningful
amounts of NO2 will therefore never accumulate in the fur-
nace. Additionally, in the model of Tobin et al.,4 flow rate
had a negligable effect on NOx ͑xϭ1,2͒ concentrations.
However, here we present evidence that the total concentra-
tion of NOx species in a furnace oxidation process changes
with flow rate. In light of this discovery, we have developed
a new model for the furnace gas-phase chemistry of N2O
oxidation, one which leads to the possibility of greater con-
trol over nitrogen incorporation.
The basis of this model is an observation by Briner
et al.,7 who studied N2O decomposition in a fundamentally
different type of reaction vessel. They found that the amount
of NO produced during decomposition increases with tem-
perature. The nature of this effect is not precisely known, as
it involves competition between several reaction pathways.
This temperature dependence provides a mechanism by
which the flow rate can have a direct effect on the NO con-
centration. At lower flow rates, the N2O decomposes at least
partially in the cool inlet of the furnace, resulting in lower
NO concentrations and less N incorporated into the oxide. At
In discussing the composition of the oxidation ambient
for such a furnace, we note that in most cases the gas resi-
dence time is significantly greater than the lifetime of N2O at
the oxidation temperature, estimated at 2 s at 900 °C.4 From
this we can conclude that the oxidation takes place entirely
in the decomposition products of N2O. The first step in this
decomposition is the unimolecular dissociation of N2O, rep-
resented in reaction ͑2͒. The atomic oxygen generated in this
process reacts rapidly through one of the reactions ͑3͒–͑7͒,
which with reaction ͑1͒ give rise to O2, NO, and N2 as de-
composition products:9
N2O→N2ϩO,
OϩO→O2,
OϩN2O→2NO,
OϩN2O→N2ϩO2,
OϩNO→NO2ϩh␯,
⌬H0ϭ38.3 kcal/mole,
͑2͒
͑3͒
͑4͒
͑5͒
͑6͒
͑7͒
NO2ϩN2O→N2ϩO2ϩNO.
Note that the NO2 generated by reaction ͑6͒ decomposes
immediately into NO and O2.5 The relative amounts of the
three decomposition products, N2, O2, and NO, depends
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:
1696 Appl. Phys. Lett. 68 (12), 18 March 1996 0003-6951/96/68(12)/1696/3/$10.00 © 1996 American Institute of Physics
142.157.183.114 On: Fri, 28 Nov 2014 16:12:00

upon pressure, the presence of inert species, and most impor-
tantly the temperature at which the process occurs.7,10 Since
NO is the species responsible for N incorporation into the
oxide,11 and the amount of N incorporated into the oxide can
be linked to the growth rate,12 we can expect both the N
content and the growth rate of the oxide to vary directly with
the NO/O2 ratio.
The temperature dependence of the NO concentration,
coupled with the short life time of N2O, raises the possibility
that the temperature profile at the gas inlet plays an impor-
tant role in determining the NO concentration. If the gas
heat-up time, i.e., the time required to bring the gas from
room temperature to the nominal furnace temperature, is
long compared to the lifetime of N2O at these temperatures,
then the NO concentration will be below the expected iso-
thermal value. On the other hand, the exothermic nature of
reaction ͑2͒ raises the possibility that at sufficiently high flow
rates enough heat will be generated in the inlet to raise the
gas temperature beyond the nominal furnace temperature, re-
sulting in higher NO concentrations. This exothermic heating
phenomenon has been noticed previously.13
FIG. 1. Temperature profiles for the 1 in. furnace at 900 and 1000 °C. The
1000 °C data were taken at a flow velocity of 1.5 cm/s. The 900 °C data
were taken at a flow velocity of 0.65 cm/s.
wall temperature, and correspondingly we did not observe
chemiluminescence in the furnace tube. This indicated that
decomposition took place over a region at least several cm in
length. As a consequence of the lack of significant exother-
mic heating the NO concentration varied over a smaller
range, from 2.2% to 4.4%, for a variation in flow velocity of
0.15–2.25 cm/s. This was consistent with a concentration of
2.7% observed by Briner et al.,7 for an isothermal decompo-
sition process.
If we assume that the NO concentrations scale with fur-
nace size, our observations lead to the conclusion that under
standard operating conditions we can vary the NO concen-
tration by up to a factor of two. For a standard 3 in. diffusion
furnace, flow velocities are typically 0.15–0.6 cm/s, or 1–4
standard liters per minute ͑SLM͒. At 1000 °C this corre-
sponds to a NO concentrations of 3.5%–7.5%. At 900 °C, we
predict that the NO concentration should change from 2.2%–
4%.
To directly examine the consequences of such a variation
in NO concentration, we grew a series of gate oxides in a
standard 3 in. diffusion furnace. We grew all of the oxides on
3 in. Si͗100͘ wafers, following a recipe employed by Saks
et al.,15 consisting of a 5 min O2 step, 30 min N2O, 30 min
O2, and 30 min N2O, all at 900 °C. Our motivation for using
this particular protocol was to produce oxides with a bimodal
nitrogen distribution. We grew three of these oxides, at flow
To examine these possible thermal effects, we measured
the NOx concentrations in the exhaust gas using a technique
loosely based on one developed in Ref. 11. We used an
evacuated flask to draw a sample of the furnace exhaust
gases. We introduced H2O into the flask, and allowed the
system to come to equilibrium. The NO2 in the exhaust gas
underwent a slow oxidation-reduction reaction with the
H2O, resulting in the generation of nitric acid.14 Since this
reaction was complete, the net result was one mole of
HNO3 for each mole of NO or NO2 in the furnace gas. With
a standard volumetric titration we determined the amount of
HNO3 in the flask, and therefore the concentration of NOx
species in the furnace gas. Knowing the amount of NOx spe-
cies drawn into the flask, and the total volume of exhaust gas
involved, we calculated the NO concentration inside the fur-
nace, assuming that all NOx species were in the form of NO.
We measured the furnace NO concentrations in this
manner at two temperatures, 900 and 1000 °C. We also mea-
sured the temperature of different furnace gases as a function
of distance into the furnace, by placing a thermocouple in the
gas stream. For these temperature profiles, we made mea-
surements for both N2 and N2O, to isolate the heating effect
of decomposition. As shown in Fig. 1͑a͒, at 1000 °C with
N2O flowing into the furnace, the gas at the inlet was heated
by as much as 200 °C above the temperature observed for
N2. We attributed this heating to the exothermic nature of
N2O decomposition. The hot region in Fig. 1, at about 5 cm
into the furnace, corresponded to the location of a white
flame, visible to the naked eye. We attributed this flame to
the chemiluminescence of reaction ͑6͒.9 In Fig. 2, for the
same temperature we show the measured NO concentration
as a function of flow rate. The concentration varied from 4%
NO at the lowest flow velocity, to a maximum of 19% NO at
flow velocities above 1.5 cm/s. This was consistent with data
from Briner et al.,7 who observed a NO concentration of
5.9% for an isothermal decomposition process at 1000 °C.
We found that the temperature profile at 900 °C differed
qualitatively from the profile for 1000 °C, as shown in Fig. 1.
The gas inlet was heated by at most 10 °C above the furnace
FIG. 2. NO concentration as a function of flow velocity for a 1 in. furnace
at 900 and 1000 °C. The expected values for an isothermal decomposition
process are 2.7% and 5.9%, respectively, from Ref. 7.
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:
Appl. Phys. Lett., Vol. 68, No. 12, 18 March 1996 K. A. Ellis and R. A. Buhrman 1697
142.157.183.114 On: Fri, 28 Nov 2014 16:12:00

cessing. Principally, this variation leads to some difficulty in
comparing oxides grown in different furnaces, as growth
recipes often do not include sufficient information on flow
velocity and gas residence time. This effect has been known
for some time.4 However, our work suggests that at 900 °C
the variation of oxide properties with flow velocity is far
greater than might be anticipated from higher temperature
studies. We also acknowlede that, because of our neglect in
dealing with turbulence and the details of heat transfer, there
may be other factors influencing furnace performance which
prevent the scalability of our NO concentration measure-
ments. Most notably, the shape of the furnace inlet and its
size in comparison with the rest of the furnace may have
secondary effects on the decomposition process. However,
the mechanism by which these factors and the flow velocity
influence the nitrogen incorporation are identical: the NO
concentration at the wafer depends directly upon the tem-
perature profile of the gas near the inlet. These changes in the
temperature profile can arise from finite gas heat-up time, or
can be produced by heat liberated during the decomposition
of N2O.
FIG. 3. N depth profile for two furnace N2O gate oxides, showing the effect
of flow rate on N concentration. The oxides were grown at 1 and 4 SLM,
corresponding to flow velocities of 0.15 and 0.6 cm/s, at a temperature of
900 °C, in O2 ͑5 min͒, N2O ͑30 min͒, O2 ͑30 min͒, and N2O ͑30 min͒.
rates of 1, 2, and 4 slm, corresponding to flow velocities of
0.15, 0.3, and 0.6 cm/s. The oxide thicknesses were 143, 139,
and 117 Å, respectively, as measured ellipsometrically with
nϭ1.462. A comparison of the N profiles of the 1 and 4 slm
oxides is given in Fig. 3. These N profiles were obtained by
an XPS etchback process described elsewhere.2 The inte-
grated N concentration of the 4 slm oxide was twice that of
the 1 slm oxide.
This research was supported by the Semiconductor Re-
search Corporation. Additional support was provided by the
National Science Foundation, through the use of the Cornell
Nanofabrication Facility and the use of the facilities of the
Cornell Materials Science Center.
It is important to note that the wafers grown by Saks
et al.,15 had nitrogen exclusively at the interface, while our
oxides clearly showed a bimodal nitrogen distribution. This
difference appears to be due solely to a change in the flow
rate. Their observation was that nitrogen which is incorpo-
rated during the initial N2O oxidation step is removed during
the final N2O step. This was attributed by Saks et al., to the
removal of N through a reaction with NO. Since NO is also
clearly present during the growth of our oxides, and action
by O2 was ruled out by Saks,15 we suggest that the species
responsible for N removal is atomic oxygen. It has been
shown previously, with low pressure O3 annealing, that
atomic oxygen is capable of scavenging N from an oxide.16
Under sufficiently high flow rates, atomic oxygen from reac-
tion ͑2͒ may survive to reach the wafer. The short residence
time used by Saks, reported as less than 10 s, is longer than
the calculated N2O lifetime of 2 s;4 however, given the un-
known time required to heat the gas to 900 °C, and the pos-
sibility of thermal inhomogeneities, we suspect that some
N2O survived to decompose near the wafer, resulting in ex-
posure of the wafers of Saks et al., to atomic oxygen.
We conclude that the variation of flow rate presents an
important tool for optimizing nitrogen incorporation. It also
brings to the surface several concerns regarding N2O pro-
1 W. Ting, H. Hwang, J. Lee, and D. L. Kwong, Appl. Phys. Lett. 57, 2808
͑1990͒.
2 E. C. Carr and R. A. Buhrman, Appl. Phys. Lett. 63, 54 ͑1994͒.
3 T. Aoyama, K. Suzuki, H. Tashiro, T. Toda, T. Yamazaki, Y. Arimoto, and
T. Ito, J. Electrochem. Soc. 140, 3624 ͑1993͒.
4 P. J. Tobin, Y. Okada, S. A. Ajuria, V. Lakhotia, W. A. Feil, and R. I.
Hedge, J. Appl. Phys. 75, 1811 ͑1994͒.
5 K. Jones, Comprehensive Inorganic Chemistry ͑Pergamon, Oxford, 1973͒,
Vol. 2, p. 350.
6 W. A. Rosser and H. Wise, J. Chem. Phys. 24, 493 ͑1956͒.
7 E. Briner, Ch. Meiner, and A. Rothen, J. Chim. Phys. 23, 609 ͑1926͒.
8 C. N. Hinshelwood and F. J. Lindars, Proc. R. Soc. London Ser. A 231,
162 ͑1955͒.
9 F. Kaufman, N. J. Gerri, and R. E. Bowman, J. Chem. Phys. 25, 106
͑1956͒.
10 F. F. Musgrave and C. N. Hinshelwood, Proc. R. Soc. London Ser. A 168,
23 ͑1923͒.
11 Z.-Q. Yao, H. B. Harrison, S. Dimitrijev, D. Sweatman, and Y. T. Yeow,
Appl. Phys. Lett. 64, 3584 ͑1994͒.
12 M. L. Green, D. Brasen, L. C. Feldman, W. Lennard, and H.-T. Tang,
Appl. Phys. Lett. 67, 1600 ͑1995͒.
13 T. S. Chao, W. H. Chen, S. C. Sun, and H. Y. Chang, J. Electrochem. Soc.
140, L160 ͑1993͒.
14 K. Jones, Comprehensive Inorganic Chemistry ͑Pergamon, Oxford, 1973͒,
Vol. 2, p. 354.
15 N. S. Saks, D. I. Ma, and W. B. Fowler, Appl. Phys. Lett. 67, 374 ͑1995͒.
16 E. C. Carr, K. A. Ellis, and R. A. Buhrman, Appl. Phys. Lett. 66, 1492
͑1995͒.
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:
1698 Appl. Phys. Lett., Vol. 68, No. 12, 18 March 1996 K. A. Ellis and R. A. Buhrman
142.157.183.114 On: Fri, 28 Nov 2014 16:12:00