Growth inhibition to enhance conformal coverage in thin film chemical vapor deposition

Full text of DOI:10.1021/ja807802r

Published on Web 12/09/2008
Growth Inhibition to Enhance Conformal Coverage in Thin Film Chemical
Vapor Deposition
†
†
‡
,‡
,†
Navneet Kumar, Angel Yanguas-Gil, Scott R. Daly, Gregory S. Girolami,* and John R. Abelson*
Department of Materials Science and Engineering, Frederick Seitz Materials Research Laboratory, UniVersity of
Illinois at Urbana—Champaign, Urbana, Illinois 61801 and Department of Chemistry, UniVersity of Illinois at
UrbanasChampaign, 600 South Mathews AVenue, Urbana, Illinois 61801
Received October 2, 2008; E-mail: girolami@scs.uiuc.edu; abelson@illinois.edu
Many important technological applications depend crucially on
the ability to deposit conformal (i.e., uniformly thick) films on
substrates having 3-dimensional relief features. For example, in the
microelectronics industry, conformal deposition processes are
needed to line and fill trenches and vias as part of the fabrication
molecule that dissociates from the precursor and desorbs without
decomposition from the surface at growth temperatures of 175-200
°C. Film growth from Ti(BH ) (dme) alone occurs with a relatively
4 3
high ꢀ value of 0.30 and is therefore conformal only for features
with relatively low ARs (∼3:1). The low conformality results
1
,2
of capacitors, interconnects, and other key device components.
4 3
because Ti(BH ) (dme) is less volatile than many other precursors;
Physical vapor deposition (PVD) processes such as sputtering,
which are widely used to grow thin films, are unfortunately unable
to afford good conformality inside features with aspect ratios (ARs)
above ∼5. For this reason, the International Technology Roadmap
for Semiconductors predicts that it will be necessary to replace PVD
with chemical vapor deposition (CVD) and atomic layer deposition
under the nonequilibrium flow conditions in a low-pressure CVD
deposition apparatus, the partial pressure of precursor at the film
-3
surface is well below the ∼10 Torr required for the onset of self-
site-blocking effects.
(
ALD) methods to line and fill features having the increasingly
large ARs that will be characteristic of future generations of
2
,3
integrated circuits. ALD is inherently conformal, but the film
growth rates are very low (typically <0.2 nm/cycle), especially in
4
high AR features. In contrast, CVD can routinely deposit films
much more rapidly, but a key challenge is how to ensure high
5
conformality while maintaining the high growth rates.
The conformality of the growth process in a feature of given
AR depends on the reactive sticking probability (ꢀ) of the gas phase
precursor on the film growth surface (Supporting Information, eq
2 4 3
Figure 1. (a) TiB film growth rate at 175 °C and (b) Ti(BH ) (dme)
precursor pressure distribution as a function of normalized depth (depth/
1
), lower values of ꢀ giving rise to better conformality because the
6,7
trench width) in a macrotrench in the absence of added dme (b) and in the
precursor will diffuse deeper into the feature before reacting.
-4
presence of 6 × 10 torr added dme (2).
Although ꢀ depends strongly on the chemical identity of the growth
species, it also can vary with deposition parameters such as pressure
and temperature. For example, we have recently shown that vias
Remarkably, howeVer, when dme is added to the Ti(BH ) (dme)
growth stream, the conformality of the TiB2 film is markedly
4
3
8
of 100 nm diameter and 30:1 AR can be uniformly coated and
improVed. Thickness profiles for films grown under the same
5
,9
-
6
even completely filled without voids by CVD.
The highly
precursor pressure (3 × 10 Torr) and temperature (175 °C), one
in the absence of dme and another in the presence of 6 × 10
-
4
conformal deposition is made possible by employing precursors
that can grow useful films at low temperatures and that have
relatively high vapor pressures, so that the surface can be saturated
with adsorbed precursor. Saturation leads to self-blocking of surface
Torr of dme, are dramatically different (Figure 1). We define a
“conformality index” as the normalized depth within a trench at
which the film thickness falls to half of the value at the top of the
trench (the normalized depth is the depth divided by the aperture
width of the feature). Larger values for the index correspond to
greater conformality. In a macrotrench with a 25 µm opening, the
films fall to half their thickness at depths of 62 µm in the absence
of dme and 156 µm in its presence (Figure 1a), corresponding to
a nearly 3-fold increase in the conformality index from 2.5 to 6.2.
Note that these depositions were conducted in a very deep trench,
in which the films eventually taper to zero thickness. In a trench
with a smaller aspect ratio, the boundary condition imposed by the
trench bottom reduces the precursor pressure drop and causes the
5
reactive sites, which reduces ꢀ and improves conformality.
In this paper, we present an alternative approach that can greatly
improve the conformality of films from precursors that under normal
circumstances afford poorly conformal deposits. The general idea
is to decrease the value of ꢀ without changing the growth
temperature or precursor partial pressure, and thereby promote
greater diffusion of the precursor deep into features with high ARs.
The decrease in ꢀ, and improvement in conformality, is achieved
by adding a suitable second species to the growth stream that can
serve as a growth inhibitor.
Here we describe the application of this idea to the CVD of
2
coating to be much more conformal. For the present TiB system,
adding dme can more than double the AR of a trench that can be
filled conformally.
titanium diboride (TiB
dme ) 1,2-dimethoxyethane.
2
) from the precursor Ti(BH
4
)
3
(dme), where
1
0,11
The dme ligand is a neutral
The precursor pressure along the axis of the trench, calculated from
5
†
the film thickness profile using the continuity equation, falls rapidly
Department of Materials Science and Engineering.
Department of Chemistry.
‡
in the absence of dme but relatively slowly in its presence (Figure
1
7660 9 J. AM. CHEM. SOC. 2008, 130, 17660–17661
10.1021/ja807802r CCC: $40.75  2008 American Chemical Society

C O M M U N I C A T I O N S
1
b). The smaller depletion of the gas phase concentration of the
pressure of the inhibitor B (see Supporting Information). The
experimental growth rate data fit this functional form extremely
well (Figure 2). The value of ꢀ calculated from the growth rate is
given on the right abscissa in Figure 2; it falls from 0.35 to 0.07 as
the dme pressure increases from 0 to 6 × 10 Torr. This reduction
in ꢀ, which is the key effect of the inhibitor, leads to a dramatic
improvement in the conformality of the CVD growth process.
Experimentally, the film growth rate at constant precursor and
inhibitor pressure increases slightly as a function of temperature,
corresponding to an apparent activation energy of 0.67 eV (65 kJ/
mol). This increase could be due to one or both of two effects: an
precursor as a function of depth is the key factor that improves the
conformal coverage. The dme pressure remains relatively unchanged
inside the trench, because it is not consumed in the growth reaction.
In separate experiments, we have shown that dme does in fact
-4
serve as an inhibitor of TiB
2
deposition. At a dme pressure of 6 ×
film growth rate from Ti(BH (dme) at 175
C is reduced by a factor of 5 (Figure 2) without any significant
-
4
1
°
0
Torr, the TiB
2
4 3
)
change in film quality and purity (see Supporting Information).
increase in the reaction constant k
k′des for desorption of the inhibitor, which implies an increase in the
precursor surface coverage θ . We have shown in our earlier work
that the TiB growth rate from the Ti(BH (dme) precursor is very
weakly activated across a wide temperature range (175-600 °C).
r
or an increase in the rate constant
A
2
4 3
)
11
We believe that the increased growth rate at higher temperatures in
the presence of the inhibitor is due to enhanced desorption of the dme.
The apparent activation energy value compares well with the desorption
activation energy of various ethers (0.4-0.6 eV) bound to metal
14
surfaces. At higher growth temperatures, the partial pressure of
inhibitor must be increased in order to maintain the surface coverage
of Bads necessary to reduce the sticking probability of the precursor
and to obtain conformal coverage.
Figure 2. Growth rate (left axis) and sticking probability ꢀ (right axis) of
2 4 3
a TiB film grown at 175 °C from Ti(BH ) (dme) as a function of dme
pressure: experimental (9) and fit to theory (solid line).
In summary, we have demonstrated that the conformality of CVD
films can be increasedswithout changing the temperature or
precursor fluxssimply by adding a second component to the gas
stream that reduces the precursor sticking probability. This approach
should be able to enhance the conformality of many other low-
temperature CVD processes.
At least two different but related mechanisms can be invoked to
explain the reduction in a CVD growth rate upon addition of an
inhibitor to the growth stream: (1) The inhibitor occupies, and
therefore blocks, reactive surface sites, thus reducing the net rate
of precursor adsorption and therefore the film growth rate, or (2)
the inhibitor promotes the reversible desorption of adsorbed
precursor, thus reducing the net rate of precursor adsorption and
therefore the growth rate. The first mechanism is equivalent to
reducing the rate of the forward reaction leading to film growth,
whereas the second mechanism is analogous to Le Chatelier’s
principle, in which the addition of a reaction product increases the
rate of the back reaction. As far as we are aware, there has been
no previous recognition that these mechanisms are potentially
relevant to CVD film growth kinetics, although it is known that
recombinative desorption pathways relevant to CVD deposition
Acknowledgment. The authors thank the National Science
Foundation for support under grants DMR 03-15428 and CHE 07-
5
0422. A.Y.-G. thanks La Fundaci o´ n la Caixa and Ministerio de
Ciencia e Innovaci o´ n for support. Ex-situ materials characterization
was carried out in the Center for Microanalysis of Materials at the
Frederick Seitz Materials Research Laboratory, University of
Illinois, which is partially supported by the U.S. Department of
Energy under grant DEFG02-07ER46453 and DEFG02-
0
7ER46471.
1
2
processes can take place on surfaces, and that film deposition
from solutes dissolved in supercritical CO can be slowed by
Supporting Information Available: Details of the deposition experi-
ments, derivation of rate law for deposition of film as a function of
precursor and inhibitor pressures, and figure of the temperature dependence
2
1
3
addition of a reaction product.
The inhibition effect can be modeled with the following steps:
i) the precursor AB reversibly chemisorbs on the surface to form
a reactive species Aads and a byproduct Bads; (ii) the adsorbed
reactive species Aads further reacts to afford film M ; and (iii) the
2 4 3
of the TiB growth rate deposited from Ti(BH ) (dme) in the presence of
(
g
-4
3
× 10 Torr of added dme. This material is available free of charge via
the Internet at http://pubs.acs.org.
s
addition of an inhibitor, in this case identical to the reaction
byproduct B, raises the steady state population of Bads on the surface
References
(1) Granneman, E. H. A. Thin Solid Films 1993, 228, 1–11.
(
2) International Technology Roadmap for Semiconductors. www.itrs.net, 2007.
(Supporting Information, eq 2-4). We further assume that Aads and
(
(
(
3) Leskela, M.; Ritala, M. Thin Solid Films 2002, 409, 138–146.
4) Lim, B. S.; Rahtu, A.; Gordon, R. G. Nat. Mater. 2003, 2, 749–754.
5) Yang, Y.; Jayaraman, S.; Kim, D. Y.; Girolami, G. S.; Abelson, J. R. Chem.
Mater. 2006, 18, 5088–5096.
Bads compete for the same surface reactive sites. Note that
mechanism 1 will require a large degree of site blocking whereas
mechanism 2 can take place even if the population of Bads is not
sufficient to block most of the adsorption sites. It is possible that
other intermediate steps can occur, for example, the formation of
a weakly bound (physisorbed) state preceding chemisorption, but
provided that these additional steps are not rate limiting, the
conclusions of the model will be unchanged.
Solving the appropriate rate equations for steady state coverage
of species Aads and Bads under the conditions of our experiment
leads to a solution for the surface coverage of Aads (and therefore
(
6) Gates, S. M. Chem. ReV. 1996, 96, 1519–1532.
(7) Hsieh, J. J. J. Vac. Sci. Technol., A 1993, 11, 78–86.
(
8) Gil, A.-Y.; Yang, Y.; Kumar, N.; Abelson, J., to be submitted.
9) Jayaraman, S.; Klein, E. J.; Yang, Y.; Kim, D. Y.; Girolami, G. S.; Abelson,
J. R. J. Vac. Sci. Technol., A 2005, 23, 631–633.
(
(
10) Jensen, J. A.; Gozum, J. E.; Pollina, D. M.; Girolami, G. S. J. Am. Chem.
Soc. 1988, 110, 1643–1644.
(11) Kumar, N.; Yang, Y.; Noh, W.; Girolami, G. S.; Abelson, J. R. Chem.
Mater. 2007, 19, 3802–3807.
(
12) Chung, Y. S.; Lee, H. S.; Lee, Y. S.; Kim, S. Surf. Sci. 2001, 482, 312–317.
(13) Zong, Y. F.; Watkins, J. J. Chem. Mater. 2005, 17, 560–565.
14) Sexton, B. A.; Hughes, A. E. Surf. Sci. 1984, 140, 227–248.
(
the growth rate) of the form C
1
/(1 + C
2
p
B
), where p
B
is the partial
JA807802R
J. AM. CHEM. SOC. 9 VOL. 130, NO. 52, 2008 17661