ECS Solid State Letters, 2 (10) P91-P93 (2013) P91

2162-8742/2013/2(10)/P91/3/$31.00 © The Electrochemical Society

Effects of Tris(tert-pentoxy)silanol Purge Time on SiO2 Thin-Film

Growth Rate in Rapid Atomic Layer Deposition
Joon Rae Kim,a Hajin Lim,a Sanghyun Park,a Yu Jin Choi,a Sungin Suh,a Bong Seob Yang,a
Jaeyeong Heo,b,z and Hyeong Joon Kima,z
a Departmentof Materials Science and Engineering & Inter-University Semiconductor Research Center, Seoul
National University, Seoul 151-742, Korea
b Departmentof Materials Science and Engineering, Chonnam National University, Gwangju 500-757, Korea

Unusual growth per cycle (GPC) behaviors, depending on the tris(tert-pentoxy)silanol purge time, were observed in catalytic SiO2
growth by atomic layer deposition (ALD). Short purge times led to decreases in GPC, and normal saturated GPC was observed
for longer purge times, following ideal ALD behavior. When an extremely long purge time was used, a decreased growth rate was
observed; loss of surface hydroxyl groups as a result of increased cross-linking was considered to be a possible cause. SiO2 films of
comparable quality to those obtained using longer purge times were obtained by increasing the Ar flow rate during purging.
© 2013 The Electrochemical Society. [DOI: 10.1149/2.005310ssl] All rights reserved.

Manuscript submitted June 18, 2013; revised manuscript received July 11, 2013. Published July 24, 2013.

Silicon dioxide (SiO2 ) is one of the most widely used materi-
als in the semiconductor industries. Its role includes use as gate
dielectrics in metal oxide semiconductor field-effect transistors, di-
electric layers in capacitors in dynamic random access memories,
insulating layers between metal interconnects such as Al or Cu, and
moisture barriers in organic light-emitting diodes and organic thin-
film transistors.1 There are several methods for forming SiO2 thin
films, including thermal oxidation of Si, evaporation, chemical va-
por deposition (CVD), plasma-enhanced CVD (PECVD), and atomic
layer deposition (ALD). Among these, ALD has been highlighted as
an ideal method for forming very conformal and ultrathin films in
the continued scaling-down of semiconductor devices. ALD is a self-
limiting technique, characterized by alternating exposure of the metal
precursor and the reactant.2,3 The chemical reaction occurs only at the
film surface, so extremely conformal growth of SiO2 films is enabled
on complex three-dimensional structures. However, the biggest hurdle
for implementing ALD in semiconductor devices is its low growth per
cycle (or per time), adding high costs for mass production.
An ALD process consists of four steps: precursor pulse, purging,
reactant pulse, and purging. These four steps make one growth cycle.
In general, the growth per cycle (GPC) of ALD SiO2 ranges from 0.05
to 0.1 nm/cycle. Given that one reaction cycle of the ALD process
takes 1–30 s, the growth per time corresponds to 0.1–6 nm/min. The
conventional time required to process one wafer in the semiconduc-
tor industry is 5 min.4 This calculation indicates that the formation
of films thicker than 50 nm by ALD will decrease the productiv-
ity of the overall process. Much effort has therefore been made to
increase the GPC of ALD SiO2 , while maintaining good quality of
the deposited films.5 Several plasma-enhanced ALD processes have
been reported to grow high-quality SiO2 films, with increased growth
rates of up to ∼0.2 nm/cycle.6–8 For thermal ALD, pyridine (C5 H5 N)
and ammonia (NH3 ) have been reported to promote the surface re-
action at lower temperatures than those used for conventional reac-
tions of chlorine-based precursors with water, by working as cata-
lysts. The GPC obtained was ∼0.21 nm/cycle.9,10 Lee et al. reported
an increased growth rate of ∼0.32 nm/cycle using ozone as an ox-
idant reacting with an Si2 Cl6 precursor. Hausmann et al. reported
rapid SiO2 ALD growth using tris(tert-butoxy)silanol as the precur-
sor, with the aid of trimethylaluminum (TMA) as a catalyst, achieving
a GPC of ∼12 nm/cycle.11 In this case, it was noted that no oxi-
dant gas was used for the growth, and it was argued that repeated
insertions of silanol molecules into the Al–O bond led to the for-
mation and subsequent cross-linking of siloxane polymers. Tris(tert-
pentoxy)silanol (TPS) has also been used for rapid SiO2 ALD, with a
GPC of ∼14 nm/cycle.12 These catalytic ALD growth processes have
received much attention since self-limiting behavior is still main-
z
E-mail: jheo@jnu.ac.kr; thinfilm@snu.ac.kr
tained, while gaining a greater than 100-fold increase in terms of
GPC. However, because extremely long pulses (10–110 s) and purge
times (10–600 s) are needed for these silanol molecules,11,12 the ac-
tual growth per time do not fully satisfy the requirements of device
fabrication.
In this study, a detailed re-examination of the TPS purge time was
carried out to increase the actual growth per time for catalytic SiO2
ALD, while obtaining high-quality films. In particular, extremely long
purge times (>400 s) were investigated. The characteristics of rapid
SiO2 ALD were analyzed as a function of TPS purge time, focusing on
the growth mechanism and related process optimization. A reaction
model based on the TPS purge time is also proposed.
ALD SiO2 thin films were deposited on Si substrates using TPS and
TMA in a commercial ALD system (MP1100, ASM Genitech Corp.,
volume = ∼150 cm3 ). The growth temperature and pressure were
fixed at 140◦ C and 0.6 Torr (∼80 Pa), respectively. The temperatures
of TPS and TMA were maintained at 80◦ C and room temperature,
respectively. Injection of TPS was performed using Ar as a carrier gas,
and a flow rate of 50 standard cubic centimeters per minute (sccm) was
used. One reaction cycle consisted of TMA pulse (2 s)/TMA purging
(4 s)/TPS pulse (50 s)/TPS purging (10–1300 s) in continuous flow
mode.13 An Ar flow rate of 180 sccm was used for each purge step in
normal operation. Three reaction cycles were normally performed, to
obtain SiO2 films of thickness ∼30–70 nm.
The film thickness was measured using an ellipsometer (Gaertner
L116D). The wet etch rates (WERs) of the SiO2 films were examined
using a 1% HF solution in deionized water. SiO2 films of thickness
100 nm thermally grown in a furnace and by PECVD at 350◦ C were
used for WER comparisons.
In the conventional ALD process, the purge step following the
precursor pulse or reactant pulse step is for the removal of unreacted
precursor/reactant and byproducts. If the purge time is too short, the
GPC tends to increase as a result of additional gas-phase reactions.14
If purging is performed for long enough, true saturation in the GPC
should be observed, given that the surface reaction follows ideal ALD
behavior.
Figure 1 shows plots of film thickness as a function of TPS purge
time for three growth cycles. It is clearly seen that the final thickness
of the film formed in this catalytic SiO2 ALD does not follow the ideal
saturation curve seen for other generally reported ALD processes.14
Changes in the film thickness can be roughly divided into four regimes.
When the TPS purge time is 10–120 s, designated regime 1 in the fig-
ure, the film thickness decreases as the purge time decreases, contrary
to ideal ALD behavior. When a longer purge time of 120–400 s is
used (regime 2), the film thickness appears to saturate at ∼65–70 nm
(calculated GPC = ∼21–23 nm/cycle). If a longer purge time (400–
800 s) is used (regime 3), the film thickness again shows a sharp drop,
and finally saturates at a constant thickness of ∼30 nm, in regime 4,
for a TPS purge time of 800–1300 s. Figure 1 suggests that this rapid

Downloaded on 2018-09-16 to IP 216.185.156.28 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 P92 ECS Solid State Letters, 2 (10) P91-P93 (2013)

80 24
(1) (2) (3) (4)
70 22
Thickness (nm)

60
50
40
30
20
0 500 1000
TPS purge time (s)
Figure 1. Film thickness as a function of TPS purge time for three growth
cycles. The Ar flow rate is 180 sccm. The growth rate trend can be divided into
four different regimes (1–4). Regime 1 (10–120 s): film thickness increases
with purge time; regime 2 (120–400 s): growth saturation; regime 3 (400–
800 s): film thickness decreases with purge time; regime 4 (>800 s): growth
saturation.
SiO2 ALD has its own unique growth mechanism, unlike that of other
conventional ALD processes.
Figure 2a shows a schematic reaction flow diagram of normal rapid
SiO2 ALD, in which surface TMA acts as chemisorption sites for
incoming TPS molecules.11,12 When a short TPS purge time is used, it
is assumed that a large amount of water, which is one of the byproducts,
will exist inside the reactor. When TMA is pulsed in the following
step, some fraction of the TMA molecules will react with water in the
gas phase, and be consumed. This eventually decreases the number
of TMA molecules chemisorbed on the substrate and decreases the
GPC. Even if TMA is chemisorbed on an OH-terminated surface,
the remaining water molecules can oxidize Al. If this occurs, OH-
terminated Al cannot catalyze chemisorption of the incoming silanol
molecules, as depicted in Figure 2b. This is also supported in part by
the following experiment. When only TPS was injected into an Al2 O3
thin film, which was deposited by the reaction of TMA and water,
there was no indication of SiO2 growth on this Al2 O3 surface. This
result suggests that fully hydroxylated Al and/or Al–O–Al surfaces
may not effectively catalyze the insertion of TPS.
(a)
RO Si OR
CH3 CH3
Al Al
OH OH O O O O
Thickness (nm)
20
18
16
14
12
10
0 20 40 60 80 100 120
1500 TPS purge time (s)
Figure 3. Film thicknesses of SiO2 for one deposition cycle while varying
the TPS purge time from 10 to 120 s. Almost identical film thicknesses are
obtained.
To check whether the remaining water molecules in the reactor
affect the next-injected TMA, the following experiments were con-
ducted. Only one cycle of deposition was conducted, and the final
TPS purge time was varied from 10 to 120 s. Figure 3 indicates that
the deposited film thicknesses were almost identical (∼1.5 nm differ-
ence), irrespective of the TPS purge time. From a combination of these
results with those for regime 1 in Figure 1, it is concluded that certain
molecules that are not sufficiently purged out after the TPS pulse step
hinder the next reaction. It is believed that water molecules play a
role in decreasing the growth rate, and water is one of the byprod-
ucts of the cross-linking of siloxane polymers. In this regard, regime
2 in Figure 1 appears to be the optimum condition in terms of the
TPS purge time. It is understood that long purge times of 120–400 s
lead to complete removal of interfering species generated during the
cross-linking reaction, resulting in production of thick SiO2 films.
Purge times longer than 400 s led to a decrease in the obtained film
thickness, as shown in Figure 1 (regimes 3 and 4). A consequence of
extremely long purge times is believed to be a decrease in the num-
ber of surface hydroxyl groups, which act as surface chemisorption
sites for incoming TMA. As discussed above, using only one depo-
sition cycle gave SiO2 films of the same thickness. Extremely long
purge times result in a reduction in the number of surface hydroxyl
groups by increasing cross-linking, and this decreases the amount of
chemisorbed TMA. Less TMA chemisorption reduces the efficient
insertion and cross-linking of silanols.
To effectively purge out byproducts generated during cross-linking
reactions, while decreasing the overall process time, several methods
were considered. Increasing the growth temperature promotes efficient
OR removal of the byproducts, but it also increases the cross-linking
reaction, leading to too-fast termination of SiO2 growth. One simple
way of increasing the purge efficiency is to add more Ar gas. The effect
O of increasing the Ar flow rate during TPS purging on the growth rate
was determined as follows. The purge time was fixed at 10 s, and
Al additional Ar flows of 50, 100, and 150 sccm were used, resulting
O O in total Ar flow rates of 230, 280, and 330 sccm. Figure 4a shows

10
(b) 70 (a) (b)
Wet Etch Rate (Å /s)

8 PECVD oxide
Thickness (nm)

60 6

50 4 Normal purge (180 sccm)

Add 50 sccm (230 sccm)
Normal purge (180 sccm) Add 100 sccm (280 sccm)
Add 50 sccm (230 sccm) 2 Add 150 sccm (330 sccm)
CH3 OH OH 40 Add 100 sccm (280 sccm)
Add 150 sccm (330 sccm)
Thermal oxide

0
Al Al Al 0 30 60 90 120 0 300 600 900 1200
TPS purge time (s) TPS purge time (s)
OH OH O O O O O O
Figure 4. (a) Film thicknesses of SiO2 for different purge conditions. Filled
Figure 2. Schematic reaction flow diagram of rapid SiO2 ALD. (a) Surface black squares indicate the film thickness for normal purge condition (Ar 180
TMA acts as chemisorption sites for incoming TPS molecules. (b) Remaining sccm). Additional Ar flow rate increases the film thickness (triangles, purge
water molecules resulting from short purge time oxidize chemisorbed Al and time 10 s). (b) Calculated WERs of corresponding SiO2 films in 1% HF
insertion of TPS is reduced, resulting in low growth rate. solution.

Downloaded on 2018-09-16 to IP 216.185.156.28 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 ECS Solid State Letters, 2 (10) P91-P93 (2013)
the final thicknesses of the grown SiO2 films. The final thickness of
the SiO2 film with normal Ar purging (180 sccm) is also included
for comparison. Increasing the Ar flow rate did not deteriorate the
thickness uniformity within the film (±2%). As the flow rate increases,
a monotonic increase in the growth rate is obtained. This suggests that
a higher growth rate can be obtained with a shorter purge time by
using more efficient purge conditions. The quality of the grown films
was checked by comparing their WERs in 1% HF solution. SiO2 films
grown by a thermal method and PECVD were included as control
films. Figure 4b shows a plot of the calculated WER with purge
time for each film; the figure shows that the WER decreases with
increasing TPS purge time, indicating that a denser film is formed. As
discussed earlier, a short purge time may decrease the density of TMA
chemisorbed on the OH-terminated surface. This leads to less efficient
cross-linking of siloxane polymers, and the final film has a less dense
structure. In contrast, a longer purge is sufficient for the formation of
a densely cross-linked SiO2 structure, so a lower WER is expected.
The WER of the SiO2 film with the longest purge time, 1300 s,
is lower than that of the SiO2 film grown by PECVD, although it is
still somewhat higher than that of the thermally grown film. Even the
film grown at the shortest purge time, 10 s, has a WER comparable to
that of the SiO2 film grown by PECVD. As more Ar is added during
purging, a gradual decrease in WER is observed. The WER of the
film grown with the aid of an additional 150 sccm Ar flow is 5.4 Å/s,
which is comparable to that of the SiO2 film grown with the normal
TPS purge time of 1000 s. Our results also suggest that optimization
of the gas flow and chamber design should be carried out in parallel
with process parameter optimization. By enhancing the efficiency of
the removal of byproducts generated during cross-linking reactions,
high-quality SiO2 films can be obtained with increased GPCs and
growth per time, making rapid/catalytic SiO2 ALD more compatible
with the mass-production scale.
Downloaded on 2018-09-16 to IP 216.185.156.28 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
P93
The effects of the TPS purge time on the GPC of rapid SiO2 ALD
were investigated. A short TPS purge time led to a decrease in the
GPC of SiO2 ; this may be the result of consumption of TMA by reac-
tion with water molecules remaining in the reaction chamber during
insufficient purging. When a longer purge time is used, a decrease in
GPC was observed, and this was attributed to the loss of surface hy-
droxyl groups, which act as chemisorption sites for TMA. Increasing
the flow rate of Ar during the purge step enabled the formation of
high-quality SiO2 films in considerably shorter purge times.
Acknowledgments
This work was supported by Samsung Electronics. Financial sup-
port by Chonnam National University, 2012 is also acknowledged.
References
1. J. Plummer, M. Deal, and P. Griffin, Silicon VLSI technology 2000, New Jersey:
Prentice Hall.
2. M. Pessa, R. Makela, and T. Suntola, Appl. Phys. Lett, 38, 131 (1981).
3. S. M. George, Chem. Rev., 110, 111 (2010).
4. S.-J. Won, J. R. Kim, S. Suh, N.-I. Lee, C. S. Hwang, and H. J. Kim, ACS Appl.
Mater. Interfaces, 3, 1633 (2011).
5. D. Choi, B.-K. Kim, K.-B. Chung, and J.-S. Park, Mater. Res. Bull., 47, 3004 (2012).
6. J. W. Lim, S. J. Yun, and J. H. Lee, ETRI Journal, 27, 118 (2005).
7. S. W. King, J. Vac. Sci. Technol. A, 29, 041501 (2011).
8. G. Dingemans, C. A. A. van Helvoirt, D. Pierreux, W. Keuning, and
W. M. M. Kessels, J. Electrochem. Soc., 159, H277 (2012).
9. Y. Du, X. Du, and S. George, Thin Solid Films, 491, 43 (2005).
10. J. Klaus, O. Sneh, A. Ott, and S. George, Surf. Rev. Lett., 6, 435 (1999).
11. D. Hausmann, J. Becker, S. Wang, and R. G. Gordon, Science, 298, 402 (2002).
12. B. Burton, M. Boleslawski, A. Desombre, and S. George, Chem. Mater., 20, 7031
(2008).
13. S. K. Karuturi, L. Liu, L. T. Su, Y. Zhao, H. J. Fan, X. Ge, S. He, and A. T. I. Yoong,
J. Phys. Chem. C, 114, 14843 (2010).
14. R. L. Puurunen, J. Appl. Phys., 97, 121301 (2005).