Infrared and near-infrared spectroscopic probing of atomic layer deposition processes

Article abstract of DOI:10.1016/j.molstruc.2010.03.087

The focus of this paper is the application of Fourier transform infrared spectroscopy (FTIR) and near infrared-tuneable diode laser spectroscopy (NIR-TDLAS) as in situ tools to monitor the formation of gas phase products during atomic layer deposition (ALD). Two studies are chosen to highlight the importance of monitoring the gas phase, in this surface driven process, for both mechanistic and reactor dynamics considerations. The first study is the FTIR monitoring of the ALD of HfO₂, from Hf[N(CH₃)₂]₄ and H₂O where it was found that, in addition to the expected HN(CH₃)₂, gas phase species were generated including CO, CH₄ and HCN. The second ALD system investigated was the growth of Al₂O₃ from Al(CH₃)₃ and H₂O employing NIR-TDLAS as a real time in situ probe to monitor the evolution of CH₄.

Full text of DOI:10.1016/j.molstruc.2010.03.087

Journal of Molecular Structure 976 (2010) 324–327
Contents lists available at ScienceDirect
Journal of Molecular Structure
journal homepage: www.elsevier.com/locate/molstruc
Infrared and near-infrared spectroscopic probing of atomic layer
deposition processes
*
A. O’Mahony, M.E. Pemble, I.M. Povey
Tyndall National Institute, University College Cork, Lee Maltings, Cork, Ireland
a r t i c l e i n f o
a b s t r a c t
Article history:
The focus of this paper is the application of Fourier transform infrared spectroscopy (FTIR) and near infra-
red-tuneable diode laser spectroscopy (NIR-TDLAS) as in situ tools to monitor the formation of gas phase
products during atomic layer deposition (ALD). Two studies are chosen to highlight the importance of
monitoring the gas phase, in this surface driven process, for both mechanistic and reactor dynamics con-
siderations. The first study is the FTIR monitoring of the ALD of HfO2, from Hf[N(CH3)2]4 and H2O where it
was found that, in addition to the expected HN(CH3)2, gas phase species were generated including CO,
CH4 and HCN. The second ALD system investigated was the growth of Al2O3 from Al(CH3)3 and H2O
employing NIR-TDLAS as a real time in situ probe to monitor the evolution of CH4.
Received 30 January 2010
Received in revised form 31 March 2010
Accepted 31 March 2010
Available online 8 April 2010
Keywords:
Atomic layer deposition
Infrared spectroscopy
Reaction mechanisms
Ó 2010 Elsevier B.V. All rights reserved.
1. Introduction
structures or structures with high-aspect ratios which may other-
wise prove difficult to process with other deposition techniques.
In situ monitoring of chemical vapour deposition (CVD) and
more recently atomic layer deposition (ALD) processes has been
widely used for gas phase and surface studies. In situ studies of
ALD deposition include the use of ellipsometry [1], infrared spec-
troscopy [2–6], quartz crystal microbalance and mass spectrome-
try [7–10], X-ray photoelectron spectroscopy (XPS) [11–13],
attenuated total internal reflection Fourier transform infrared
spectroscopy (ATR-FTIR) [14,15]. In situ characterisation of ALD
processes can provide information regarding both the reaction
pathways and dynamics of the growth process, revealing details
on precursor nucleation and dissociation, the formation of inter-
mediates and products and limitations in the process and reactor
design.
The ALD method of deposition relies on alternating pulsing of
gaseous precursors, interspersed with system purges [16]. The pre-
cursors chemisorb on the substrate surface and excess gaseous
material is removed during the purges. The pulse/purge sequence
is repeated numerous times to build up a layer of material with a
well defined thickness. An idealised ALD process adheres to a
self-limiting growth mechanism with monolayer control, with
adhesion of the precursor to the substrate defined by the number
of available surface sites and the size of the precursor ligand. In
the absence of complex nucleation effects the growth is linear with
respect to the number of pulse cycles. Furthermore, the high de-
gree of conformality of material growth exhibited by ALD makes
it ideal for coating and infiltrating materials with complicated
ALD has applications in many areas but has performed particularly
well in aspects of semiconductor processing, such as deposition of
high-k materials for MOSFETS and DRAM capacitors [17] and metal
deposition for interconnects [18].
The cyclic nature of ALD and the self-limiting half reactions
lends itself very well to surface spectroscopic probing methods.
As a consequence the focus of mechanistic studies is often directed
to surface processes. However, although surface studies may well
identify the key components of a mechanism the gas phase should
not be neglected as minor reaction channels and secondary reac-
tions of products may be of significance, particularly in applica-
tions where very low levels of impurities may perturb the
pertinent properties of the material grown, for example, the ALD
of high-k materials for device applications [18].
To date, in situ FTIR monitoring of ALD processes has been used to
study a widerange of processesincluding the deposition of La2O3 [2],
Al2O3 [3], HfO2 [4–6,19,20], SiO2 [21], SnO2 [22], TiO2 [15] and TiN
[23]. In situ FTIR studies of HfO2 provide valuable information about
the nucleation on different silicon surfaces with substrate termina-
tion as well as reaction by-products. A study of HfO2 ALD by Wang
et al. identified organic species with carboxylic and nitro radicals
as contaminants of the ozone and tetrakisethylmethylamidohafni-
um (TEMAH) HfO2 ALD process [5]. Studies confirm the formation
of methylethylamine (HN(CH3)(C2H5)) as a by-product of the reac-
tion of TEMAH and water, as expected [19,20].
In this article we highlight that the importance of performing
gas phase studies in addition to surface probing. To this end we
have probed the gas phase of the ALD deposition of hafnia (HfO2)
using FTIR. It should be recognised that although the detection of
* Corresponding author.
E-mail address: ian.povey@tyndall.ie (I.M. Povey).
0022-2860/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.molstruc.2010.03.087

A. O’Mahony et al. / Journal of Molecular Structure 976 (2010) 324–327
325
intermediates in CVD reactors has been performed by infrared
spectroscopy for many years the pulsed nature of the ALD process
does not lend itself to real time FTIR studies. Products and interme-
diates formed are rapidly pumped to exhaust and hence as the
pulse time of the reactant is often very short (sub-second) the
average concentration of the products can be very low over the re-
quired spectral integration time. For this reason the ALD process is
often monitored using a stop flow approach where the precursor
and its products are held within the chamber prior to purging to
allow the spectroscopic probe to record data. In situ tuneable diode
laser absorption spectroscopy is commonly used in gas sensing
applications as diode lasers have the capacity for sub-second, accu-
rate and sensitive monitoring of gas phase species such as NH3
[15,24], CO [25], O2 and water [26]. Near infrared-tuneable diode
laser spectroscopy has been previously employed to monitor CVD
based processes [29] and has recently been adopted in the moni-
toring of ALD precursor pulse trains [30].
The ALD reactor utilized in this study was based on a home built
cross flow system that has been previously used to grow excellent
quality HfO2 [27,28]. The reactor was cold wall in design, sub-
strates being heated by a resistive cartridge heated graphite sus-
ceptor. The two optical probes used in these studies were a
Bruker IFS66 near/mid infrared Fourier transform spectrometer fit-
ted with an MCT detector and secondly, a diode laser that was cen-
sor lines to equilibrate the precursor vapour pressure, the lines
were re-evacuated to remove any pre-reacted precursor products
and finally re-equilibrated before growth. A pulse sequence was
adopted that ensured the precursors were delivered in excess
and purged sufficiently between cycles to prevent gas phase cross
over of material. The FTIR spectra were referenced to a background
and integrated over 200 scans (ꢀ3 min) at a spectral resolution of
1 cmÀ1, over the spectral range of 4000–400 cmÀ1. Three pulse re-
gimes were adopted to compendsate for the low residence times of
products and reactants compared to that of the spectral integration
period necessary for adequate signal to noise.
(a) CVD regime: A CVD comparable regime was monitored using
in situ FTIR studies, to compare directly to the ALD half-cycle re-
gimes and to eliminate or confirm any CVD contributions. A
50 ms pulse of Hf[N(CH3)2]4 was pulsed into the sealed reaction
chamber, followed by a 50 ms pulse of water. The formation of
products was monitored in the sealed reaction chamber using
in situ FTIR over 200 scans.
(b) ALD half-cycle; TDMAH pulse in water conditioned reaction
chamber: The reactor was conditioned with a 50 ms pulse of water
and after the duration of the ALD cycle (approximately 2 min), the
reactor was evacuated. 50 ms of TDMAH was pulsed into the reac-
tion chamber and the resulting formation of products monitored in
the sealed reaction chamber, over 200 scans.
tred at 1.6
l
m in conjunction with an InGaAs detector. Both probe
(c) ALD half-cycle; water pulse in TDMAH conditioned reaction
chamber: The reactor was conditioned with a 50 ms pulse of
TDMAH and after the duration of the ALD cycle-approximately
2 min, the reactor was evacuated. 50 ms of water was pulsed into
the reaction chamber and the formation of the resulting products
monitored over 200 scans using in situ FTIR.
Figs. 2 and 3 illustrates the gas phase spectra obtained as a func-
tion of susceptor temperature from TDMAH in our ALD reactor un-
der three sets of experimental conditions. Fig. 2 corresponds to
conditions closely resembling CVD, where both the TDMAH and
water were introduced into the reactor at the same time. Fig. 3a
and b correspond to more ALD-like conditions where the two pre-
cursors (TDMAH and water) are isolated after previously exposing
the reactor to the other precursor followed by a 30 s evacuation. In
all experiments the precursor pulses used were akin to those used
real growth conditions [28].
beams were collimated through the reactor via optical viewing
ports and passed 5 mm above the graphite susceptor as illustrated
in Fig. 1.
1.1. Gas phase in situ monitoring of HfO2 deposition using FTIR
spectroscopy
ALD and CVD of HfO2 using TDMAH or TEMAH and water has
been shown to grow good quality material [4,28]. In addition, the
surface chemistry of these precursors has been studied by infrared
techniques [4,14] and density functional theory modelling [14]. In
these studies evidence to support a mechanism based on H
abstraction by an amine ligand of the TDMAH was presented. Here
we examine the gas phase products of both the CVD and ALD reac-
tions and describe how they may influence the growth process.
In the study presented in this work HfO2 deposition was per-
formed with tetrakis(dimethylamido)hafnium (TDMAH) [Hf[N
Examination of Figs. 2 and 3a shows that the spectra are
remarkably similar. From these spectra it can be seen that the ma-
jor gas phase product in both scenarios is as expected dimethyl-
amine (HN(CH3)2), with characteristic spectral features, N–H
bend at 735 cmÀ1, C–N stretch at 928 and 1022 cmÀ1, and the
CH3 stretch region 2791–2982 cmÀ1 [33]. When the case of water
(CH3)2]4-electronic grade (SAFC Hitech)] and water (18 M
X cm
pure, Elga Purelab UHQ) maintained at 70 °C (vapour pressure
2.57 torr) and room temperature ꢀ24 °C (vapour pressure
13.60 torr) respectively. The transfer lines were heated to 110 °C
to prevent condensation of the metal precursor and the reaction
chamber heated to 100 °C. Nitrogen flow was set to 30 sccm to
transport the precursor vapours to the reactor and to purge the
reactor following a pulse sequence. Prior to each growth run the
water and TDMAH lines were evacuated to remove any residual
products, the precursor bubblers were opened to allow the precur-
being pulsed onto
a TDMAH coated substrate is examined,
(Fig. 3b), it is immediately apparent that a different product distri-
bution is produced with spectral features that were not obvious in
the CVD and TDMAH ALD half-cycle spectra, notably at 1730 cmÀ1
.
It is important to note that these features must be a product of a
purely surface chemistry as there are no gas phase reactants that
could be fragmented to produce such spectral features. At present
we have no unambiguous assignment of these spectral features but
can confirm that they are not due to gaseous HN(CH3)2.
Optical viewports
As the temperature is raised it is apparent that both under the
CVD and ALD conditions trace amounts of smaller organic and
inorganic species are evident (Figs. 2 and 3a and b). The concen-
tration of these species increases with increasing reactor temper-
ature but significantly, although low in concentration, these
fragments are present at typical ALD growth temperatures. The
complete identification of all of these species is not yet fully real-
ized, however we can confirm that they include CH4, CO and
HCN, characteristic spectral features and their temperatures of
observation are tabulated in Table 1 for the CVD and ALD half-
cycle studies.
Gas
pulse
train
Heated
susceptor
Fig. 1. Schematic of ALD system set-up for in situ real time detection of CH4 as a
product of the Al2O3 ALD process.

326
A. O’Mahony et al. / Journal of Molecular Structure 976 (2010) 324–327
400 oC
CO
HCN
300 oC
250 oC
200 oC
*
*
150 oC
*
* = dimethylamine
*
*
3500
3000
2500
2000
1500
1000
500
Wavelength / cm-1
Fig. 2. FTIR spectra of the TDMAH and water reaction under CVD conditions.
a
400 oC
CO
HCN
350 oC
300 oC
250 oC
200 oC
150 oC
* = dimethylamine
*
*
*
*
*
3500
3000
2500
2000
Wavelength / cm -1
1500
1000
500
b
400 oC
CO
350 oC
300 oC
250 oC
1730
200 oC
150 o
C
+
+
+
+ = unknown
+
+
+
3500
3000
2500
2000
Wavelength / cm-1
1500
1000
500
Fig. 3. FTIR spectra of the TDMAH and water reaction under (a) isolated TDMAH pulse, (b) isolated water pulse conditions.
Table 1
imply that at growth temperatures 6200 °C the formation of CH4,
CO and HCN is less favourable and as such, it may be that they arise
not as primary reaction products, but as the result of additional
reactions.
Product fragment vibrational frequencies (cmÀ1) and observation temperature (°C)
for CVD and ALD half-cycle experiments.
CVD
Hf pulse cycle
H2O pulse cycle
In summary, we have demonstrated that the decomposition of
TDMAH in the presence of water produces HN(CH3)2 under both
CVD and the ALD TDMAH pulse measurement conditions. Further-
more, in the ALD water pulse cycle HN(CH3)2 is not the dominant
gas phase product observed. Additionally, increasing amounts of
CH4, CO and HCN, that could give rise to impurity incorporation,
are produced as the growth temperature is increased.
HCN t2 712 t3 3280
CH4 3018, 1307
CO 2144
300
200
250
300
200
250
250
x
250
The production of these small molecules at elevated tempera-
tures is not too surprising. However, whether their appearance
plays a distinct role in determining the nature of the deposited
oxide remains to be established. For example, Kukli et al. observed
a marked increase in impurity levels in films grown with this pre-
cursor at higher temperatures which could be attributed to sec-
ondary surface reactions involving such species [32]. The
mechanisms by which such species form are currently the subject
of further investigation. The spectral data obtained in this study
1.2. Real time in situ probing of Al2O3 deposition using NIR-TDLS
The long integration times required for FTIR monitoring of gas
phase processes in CVD processes are not viewed as a significnt
hinderance with respect to the determination of mechanistic de-

A. O’Mahony et al. / Journal of Molecular Structure 976 (2010) 324–327
327
B
C
A
0
50
100
150
Time / second
Fig. 4. Real time diode laser signal centred at 1.6 lm over a series of ALD cycles of TMA and water. The production of CH4 is clearly observed at the pulsing of both precursors.
tails. This is due to the fact that although species maybe individu-
ally short lived their average concentration approximates a steady
state concentration through the spectral measurement. Unfortu-
nately, this is not the case in ALD products and intermediates that
are swept from the probe region are not replaced until the cycle is
repeated. Thus it is essentially that a real time probe is developed
that can accumulate data on a millisecond time scale. To this end
we have employed a NIR-TDLAS system to study the ALD of
Al2O3 [31]. The process is monitored via the evolution of methane
(CH4) employing an Inphenix distributed feedback laser diode as
the light source and an Agilent 8164B InGaAsP light wave measure-
ment system as the detector. The fibre coupled laser source was
demonstrate that gas phase IR studies of ALD processes can reveal
a great deal regarding the efficacy of an established ALD process
(Al2O3 growth) with respect to the dynamics within a reactor
and efficiency of precursor pulsing. These studies should be consid-
ered along with IR data regarding the structure of surface interme-
diates in order to select the most appropriate precursors and to
design the optimum ALD growth process.
References
[1] E. Langereis, S.B.S. Heil, H.C.M. Knoops, W. Keuning, M.C.M. van de Sanden,
W.M.M. Kessels, J. Phys. D: Appl. Phys. 42 (2009) 073001.
[2] J. Kwon, M. Dai, M.D. Halls, E. Langereis, Y.J. Chabal, R.G. Gordon, J. Phys. Chem.
C. 113 (2009) 654–660.
[3] D.N. Goldstein, J.A. McCormick, S.M. George, J. Phys. Chem. C. 112 (2008)
19530–19539.
[4] M.-T. Ho, Y. Wang, R.T. Brewer, Y.J. Chabal, N. Moumen, M. Boleslawski, Appl.
Phys. Lett. 87 (2005) 133103.
[5] Y. Wang, M. Dai, M.-T. Ho, L.S. Wielunski, Y.J. Chabal, Appl. Phys. Lett. 90 (2007)
022906.
[6] Y. Wang, M.-T. Ho, L.V. Goncharova, L.S. Wielunski, S. Rivillon-Amy, Y.J. Chabal,
T. Gustafsson, N. Moumen, M. Boleslawski, Chem. Mater. 19 (2007) 3127–
3138.
[7] M. Juppo, A. Rahtu, M. Ritala, M. Leskela, Langmuir 16 (2000) 4034–4039.
[8] A. Rahtu, T. Alaranta, M. Ritala, Langmuir 17 (2000) 6506–6509.
[9] R. Matero, A. Rahtu, M. Ritala, Langmuir 21 (2005) 3498–3502.
[10] S.B.S. Heil, P. Kudlacek, E. Langereis, R. Engeln, M.C.M. van der Sanden, W.M.M.
Kessels, Appl. Phys. Lett. 89 (2006) 131505.
[11] F.S. Aguirre-Tostado, M. Milojevic, B. Lee, J. Kim, R.M. Wallace, Appl. Phys. Lett.
93 (2008) 172907.
[12] M. Milojevic, F.S. Aguirre-Tostado, C.L. Hinkle, H.C. Kim, E.M. Vogel, J. Kim, R.M.
Wallace, Appl. Phys. Lett. 93 (2008) 202902.
[13] M. Milojevic, C.L. Hinkle, F.S. Aguirre-Tostado, H.C. Kim, E.M. Vogel, J. Kim, R.M.
Wallace, Appl. Phys. Lett. 93 (2008) 252905.
[14] M.J. Kelly, J.H. Han, C.B. Musgrave, G.N. Parsons, Chem. Mater. 17 (2005) 5305–
5314.
[15] V.R. Rai, S. Agarwal, J. Phys. Chem. C. 113 (2009) 12962–19625.
[16] T. Suntola, J. Antson, US Patent Number 4,058,340, 1977.
[17] O. Sneh, R.B. Clark-Phelps, A.R. Londergan, J. Winkler, T.E. Seidel, Thin Solid
Films 402 (2002) 248.
[18] International Technology Roadmap for Semiconductors, 2007. <http://
www.itrs.net/n>.
[19] J.E. Maslar, W.S. Hurst, D.R. Burgess, W.A. Kimes, N.V. Nguyen, ECS Trans. 12
(7) (2007) 133–143.
[20] J.E. Maslar, W.S. Hurst, D.R. Burgess, W.A. Kimes, N.V. Nguyen, E.F. Moore, J.T.
Hodges, ECS Trans. 13 (2) (2008) 139–149.
[21] B.B. Burton, S.W. Kang, S.W. Rhee, S.M. George, J. Phys. Chem. C. 113 (2009)
8249–8257.
[22] X. Du, Y. Du, S.M. George, J. Phys. Chem. A 112 (2008) 9211–9219.
[23] J.W. Elam, M. Schuisky, J.D. Ferguson, S.M. George, Thin Solid Films 436 (2003)
145–156.
centred at 1.6
lm corresponding to the 2m3 transition of CH4 and
set to deliver an 80 mA current, 0.3
lW power (Arroyo Instruments
4210 laser source). The laser source was collimated prior to enter-
ing the reactor and focussed to the detector collection fibre after
exiting the reactor. Twenty millisecond Al(CH3)3 [TMA-electronic
grade (SAFC Hitech)] and 400 ms water (18 M
X cm pure, Elga Pur-
elab UHQ) were pulsed, interspersed with 30 s nitrogen purges.
The production of CH4 was monitored at 100 ms intervals using
the diode laser. The signal intensity transmitted through the reac-
tor is illustrated in Fig. 4 from which it is clearly evident that CH4 is
produced from both gas pulses. As the TMA is pulsed into the reac-
tor (A) a sharp decrease in detected diode intensity is observed
associated with the absorption of the IR radiation from the diode
by CH4. As can be seen, the signal then recovers rapidly. However,
after the purging of the initial CH4 burst a temporally broader de-
crease in diode intensity begins to develop (B), which eventually
fades to the baseline intensity within the 30 s purge cycle. The
pulsing of water into the TMA ‘coated’ reactor again produces a ra-
pid response in signal intensity (C) this however does not recover
as rapidly and merges into the second broad response as seen in
the TMA half-cycle. This pattern of CH4 evolution, i.e. initial rapid
CH4 production followed by CH4 evolution on a much lower time-
scale is very reproducible and indeed not surprising considering
the cold wall nature of the ALD reactor used in this study. The
broad features between the pulses are indicative of slower wall
reactions or simply desorption of physisorbed precursors from
the colder reactor surfaces, highlighting the difficulties in rapidly
purging cold wall reactors. Clearly the existence of these slower
steps must be considered in order to optimize the design of the
ALD process. For example, if the water pulse was introduced imme-
diately after point A, the fact that CH4 is still being produced from
the TMA pulse, and the possibility that unreacted TMA still exists
in the reactor, could result in the appearance of CVD type processes
wherein the TMA and the water molecules meet in the reactor.
Such processes could severely impact on the material quality and
uniformity.
[24] M.E. Webber, D.S. Baer, R.K. Hanson, Appl. Opt. 40 (12) (2001) 2031–2042.
[25] H. Xia, W.Q. Liu, Y.J. Zhang, Spectrosc. Spec. Anal. 28 (11) (2008) 2478–2481.
[26] L. Persson, M. Lewander, M. Andersson, K. Svanberg, S. Svanberg, Appl. Opt. 47
(12) (2008) 2028–2034.
[27] É. O’Connor, R.D. Long, K. Cherkaoui, K.K. Thomas, F. Chalvet, I.M. Povey, M.E.
Pemble, P.K. Hurley, B. Brennan, G. Hughes, S.B. Newcomb, Appl. Phys. Lett. 92
(2008) 022902.
[28] É. O’Connor, S. Monaghan, R.D. Long, A. O’Mahony, I.M. Povey, K. Cherkaoui,
M.E. Pemble, P.K. Hurley, G. Brammertz, M. Heyns, S.B. Newcomb, V.V.
Afanas’ev, Appl. Phys. Lett. 94 (2009) 102902.
[29] R. Holdsworth, P.A.N. Martin, D. Raisbeck, H. Rivero, H.E. Sanders, D.W. Sheel,
M.E. Pemble, Chem. Vap. Depos. 7 (2001) 39–43.
2. Conclusions
[30] M.E. Pemble, I.M. Povey, F. Chalvet, ECS Trans. 11 (7) (2007) 155–166.
[31] R.L. Puurunen, J. Appl. Phys. 97 (2005) 121301.
[32] K. Kukli, T. Pilvi, M. Ritala, T. Sajavaara, J. Lu, M. Leskela, Thin Solid Films 491
(1–2) (2005) 328–338.
We have demonstrated that although ALD is a surface driven
process the study of gas phase chemistry is essential as it may im-
pinge directly on the quality of the material grown. We further
[33] D.-H. Kang, M. Trenary, Surf. Sci. 519 (2002) 40–56.