Chem. Mater.

2010, 22, 4669–4678 4669

DOI:10.1021/cm100874f

Atomic Layer Deposition of CdS Films

Jonathan R. Bakke,† Hee Joon Jung,‡ Jukka T. Tanskanen,†,§ Robert Sinclair,‡ and
Stacey F. Bent*,†
†
Department of Chemical Engineering, and ‡Department of Materials Science and Engineering,
Stanford University, Stanford, California 94305, and §Department of Chemistry, University of
Eastern Finland, Yliopistokatu 7, 80130 Joensuu, Finland
Received March 26, 2010. Revised Manuscript Received June 23, 2010

Pure, polycrystalline CdS deposited by atomic layer deposition (ALD) on Si(100) or glass using
dimethyl cadmium and in situ generated H2S is investigated in detail. This ALD system follows
saturation behavior typical of ALD systems, and the growth rate monotonically decreases with
temperature from 100
C-300
C; by 400
C linear growth rate behavior is no longer seen. The crystal
structure as determined by X-ray diffraction and transmission electron microscopy gradually
transitions from zincblende to wurtzite with increasing temperature until the film is primarily
wurtzite by 400
C. Further, the average grain size increases with temperature. Transmission electron
microscopy images and selected area diffraction patterns confirm the presence of zincblende and
wurtzite crystals because of stacking faults and demonstrate that {111} crystal planes are more
oriented parallel to the substrate at lower temperatures. Ultraviolet-visible spectroscopy shows that
the bandgap is 2.3-2.42 eV in the 100
C-400
C range with a slight increase occurring with
temperature. The roughness of the films is found to increase both with temperature and cycle number
as observed with atomic force microscopy and scanning electron microscopy. Density functional
theory calculations were used to understand observations concerning the growth rate and the
bandgap of the films deposited at different temperatures.

Introduction interesting because of that crystal’s long-term stability.3

Recently, CdS quantum dots have gained attention for use
Atomic layer deposition (ALD) is a thin film growth
in light emitting diodes and in photovoltaics as sensitizers.4,5
technique with a myriad of applications because of its
CdS has been deposited by many methods6 including
unique capabilities. Since the precursors are supplied
chemical bath deposition,7 atomic layer epitaxy (ALE),8,9
sequentially into a reactor and the half-reactions are
metal organic chemical vapor deposition (MOCVD),1,10,11
surface-reaction rate limited, ALD has the capability to
pulsed laser deposition,12 and spray pyrolysis.13 Various
produce uniform thin films over large areas at a maximum
deposition methods using the dimethyl cadmium (DMCd)
rate of one monolayer per cycle. Nevertheless, steric and/or
precursor have been used to deposit CdS and CdxZn1-xS
electronic effects almost always lead to submonolayer
films.1,14
growth (i.e., a rate lower than the interplanar distance
The deposition of CdS via ALD has not yet been
per cycle). The ALD process is characterized by a growth
investigated under the normal operating conditions for
rate that is linear as a function of the number of cycles even
most ALD systems; however, two earlier studies have
when an excess of precursor is dosed into the reactor.
described ALE of CdS in an ultrahigh vacuum (UHV)
CdS is a wide bandgap II-VI semiconductor with an
environment. Tadokoro et al. deposited CdS on GaAs-
energy gap of 2.42 eV,1 and it is utilized for a number of
(100) under UHV with elemental Cd and S precursors by
applications. A primary use of CdS is as a buffer layer for
thin film photovoltaic devices, and it is often deposited in
(4) Chang, C. H.; Lee, Y. L. Appl. Phys. Lett. 2007, 91, 3.
zincblende form for these applications. In fact, the world (5) Zhao, J. L.; Bardecker, J. A.; Munro, A. M.; Liu, M. S.; Niu, Y. H.;
record efficiencies for CdTe and CuInxGa1-x(S,Se)2 Ding, I. K.; Luo, J. D.; Chen, B. Q.; Jen, A. K. Y.; Ginger, D. S.
photovoltaics have been obtained with cells using CdS Nano Lett. 2006, 6, 463–467.
(6) Mane, R. S.; Lokhande, C. D. Mater. Chem. Phys. 2000, 65, 1–31.
as the n-type buffer layer.2 Applications of wurtzite CdS (7) Liu, Q.; Mao, G. B. Surf. Rev. Lett. 2009, 16, 469–474.
as a window for HgCdTe infrared detectors are especially (8) Tadokoro, T.; Ohta, S.; Ishiguro, T.; Ichinose, Y.; Kobayashi, S.;
Yamamoto, N. J. Cryst. Growth 1993, 130, 29–36.
(9) Luo, Y.; Han, M.; Slater, D. A.; Osgood, R. M. J. Vac. Sci.
*To whom correspondence should be addressed. E-mail: sbent@stanford. Technol. 2000, 18, 438–449.
edu. Phone: (650) 723-0385. Fax: (650) 723-9780. (10) Obrien, P.; Walsh, J. R.; Watson, I. M.; Hart, L.; Silva, S. R. P.
(1) Cockayne, B.; Wright, P. J. J. Cryst. Growth 1984, 68, 223–230. J. Cryst. Growth 1996, 167, 133–142.
(2) Repins, I.; Contreras, M. A.; Egaas, B.; DeHart, C.; Scharf, J.; (11) Yim, W. M.; Stofko, E. J. J. Electrochem. Soc. 1972, 119, 381.
Perkins, C. L.; To, B.; Noufi, R. Prog. Photovoltaics 2008, 16, 235– (12) Ullrich, B.; Sakai, H.; Segawa, Y. Thin Solid Films 2001, 385, 220–
239. 224.
(3) Boieriu, P.; Sporken, R.; Xin, Y.; Browning, N. D.; Sivananthan, S. (13) Ma, Y. Y.; Bube, R. H. J. Electrochem. Soc. 1977, 124, 1430–1435.
J. Electron. Mater. 2000, 29, 718–722. (14) Bakke, J. R.; Bent, S. F. ECS Trans. 2009, 25, 9–14.

r 2010 American Chemical Society Published on Web 07/27/2010 pubs.acs.org/cm

4670 Chem. Mater., Vol. 22, No. 16, 2010 Bakke et al.

ALE using a molecular beam epitaxy (MBE) setup at spectroscopy (XPS), and transmission electron microscopy (TEM)
340
C. This system produced epitaxial c(200) CdS with measurements was a Czochralski grown n-type Si(100) wafer with
stacking faults and dislocations.8 Luo et al. deposited CdS by a resistivity of 1-5 Ω-cm and a native silicon oxide layer which is
ALE at room temperature on ZnSe(100) using a UHV approximately 2 nm thick. Films were also deposited on glass for
system as well.9 The precursors dimethyl cadmium and H2S comparison of crystal orientation and bandgap using X-ray diffrac-
tion (XRD) and ultraviolet-visible spectroscopy (UV-vis), res-
were used to deposit up to 15 bilayers of stoichiometric CdS.
pectively. The substrates were cleaned via 5 min sonication steps
Annealing at 250
C produced zincblende CdS, which
each in acetone and ethanol with deionized water rinses between
matched the substrate crystal structure. The reaction mecha- each solvent bath. Residual organics were then removed with a
nism was also studied in detail, and the results showed that piranha clean (70% sulfuric acid and 30% hydrogen peroxide) for
the DMCd precursor etched the surface to release DMZn. 15 min followed by a deionized water rinse and drying with
Another similar II-VI material, CdSe has been grown via pressurized air. Fresh piranha is hot and corrosive, and extreme
ALE using cadmium and selenium elemental sources.15 caution must be used when handling this solution.
In this study, we detail the deposition of CdS via ALD After ALD, resulting film thicknesses were measured by a
on hydroxyl terminated surfaces of Si(100) or glass using Gaertner L116C single-wavelength ellipsometer using 632.8 nm
DMCd and an in situ H2S source in an ALD flow reactor. light at a 70
angle of incidence and with the polarizer set to 45
.
In depth characterization of the growth rates of the films At least two 1 cm
1 cm samples from each run were used, and
measurements were performed on three spots for each sample to
is performed, and DFT calculations are used to support
account for any non-uniformity. Elemental composition of the
observed growth rates. The material properties of the
films was determined by XPS with a Surface Science Instru-
films including the crystal structure, bandgap, roughness, ments S-Probe monochromatized spectrometer using Al KR
and surface morphology are carefully documented. In 1486 eV radiation at a pressure of 6.7
10-11 kPa (5
10-10
addition, we make direct comparisons between CdS ALD Torr). The Argon sputter was performed at a pressure of 1.3

and the related ZnS ALD processes. ZnS is another II-VI 10-8 kPa (1
10-7 Torr) using 5 keV Arþ at 2 mm
2 mm raster
semiconductor which has been studied more extensively at 45
incidence to the sample. Survey scans were performed
when deposited by ALD. Analogous precursors dimethyl with a step size of 1 eV. Surface morphology and roughness were
zinc (DMZn)16 and diethyl zinc (DEZn)17-19 have been characterized with a Veeco Multimode AFM in tapping mode
used with the counter reactant H2S to yield ZnS films, and with MikroMasch NSC16 tips. Crystal structure was deter-
density functional theory calculations have also been mined with a PANalytical X’Pert PRO XRD system in parallel
beam mode using Cu KR radiation. Surfaces were imaged by
carried out to explore the growth mechanisms of ZnS.20
SEM using an FEI XL30 Sirion SEM at a 5 kV operating
Hence, it serves as an ideal basis for comparison.
voltage. Bandgaps were determined by UV-vis with Varian’s
Cary 6000i UV-vis-NIR spectrophotometer. TEM samples
Experimental and Computational Details with a thickness of ∼80 nm were prepared using a focused ion
ALD growth of CdS was performed in a custom built warm beam (FIB, FEI Strata 235DB dual-beam FIB/SEM) lift-out
wall reactor which has been described in a previous publication.17 Omniprobe technique with a Gaþ ion beam at 30 keV. Cross-
The precursors for this reaction were dimethyl cadmium (DMCd) sectional bright-field and high resolution transmission electron
(Strem)14 and in situ generated H2S which is produced by heating microscopy images and selected area diffraction (SAD) patterns
thioacetamide to 150
C under an inert atmosphere; this method were taken by an FEI Tecnai G2 F20 X-TWIN operated at an
has been previously described in detail.17 DMCd is a liquid at accelerating voltage of 200 kV. Fast Fourier Transformation
room temperature with a melting point of -4.5
C and a boiling (FFT) images were obtained by DigitalMicrograph from the
point of 105.5
C at atmospheric pressure. DMCd is a highly high resolution TEM (HRTEM)
reactive chemical and should only be handled in a glove box free Periodic hybrid density functional theory (DFT) calcula-
of water and oxygen. The H2S is mostly pure with small amounts tions were performed to investigate the electronic character-
of acetonitrile. Nitrogen was used as the carrier and purge gas at a istics of the deposited films and the temperature dependence
constant flow rate of 60 sccm. The substrate temperature was of the growth rate. The calculations were carried out by
varied for the deposition of CdS with the temperature affecting CRYSTAL200621 quantum chemistry software using the PBE0
parameters such as growth rate, crystal structure, root-mean- density functional,22 which is one of the better performing hybrid
square (rms) roughness, and bandgap. For all studies except pulse functionals for solid state and for late-transition-metal reac-
and purge length dependencies, the optimized cycle of ALD tions.23,24 Default optimization convergence thresholds and
consisted of 0.4 s DMCd, 10 s purge, 0.4 s H2S, and a 10 s purge. an extra large integration grid were utilized in the calcula-
All precursors were maintained at room temperature (∼22
C). tions. Karlsruhe split-valence basis set with polarization func-
Needle valves controlled the rate of dosing of each precursor. tions (def2-SVP)25 was used for C and H, while a standard
The substrate for ellipsometry, scanning electron microscopy 6-31G* basis was adopted for S. For Cd a triple valence quality
(SEM), atomic force microscopy (AFM), X-ray photoelectron
(21) Dovesi, R. S.; Saunders, V. R.; Roetti, C.; Orlando, R.; Zicovich-
(15) Mikhaevich, D. P.; Ezhovskii, Y. K. Russ. J. Appl. Chem. 2003, 76, Wilson, C. M.; Pascale, F.; Civalleri, B.; Doll, K.; Harrison, N. M.;
1197–1200. Bush, I. J.; D’Arco, Ph.; Llunell, M. CRYSTAL2006 User’s
(16) Hunter, A.; Kitai, A. H. J. Cryst. Growth 1988, 91, 111–118. Manual; University of Torino: Torino, Italy, 2006.
(17) Bakke, J. R.; King, J. S.; Jung, H. J.; Sinclair, R.; Bent, S. F. Thin (22) Adamo, C.; Barone, V. J. Chem. Phys. 1999, 110, 6158–6170.
Solid Films 2010, 518, 5400–5408. (23) Paier, J.; Marsman, M.; Kresse, G. J. Chem. Phys. 2007, 127,
(18) Kim, Y. S.; Yun, S. J. Appl. Surf. Sci. 2004, 229, 105–111. 24103.
(19) Stuyven, G.; De Visschere, P.; Hikavyy, A.; Neyts, K. J. Cryst. (24) Quintal, M. M.; Karton, A.; Iron, M. A.; Boese, A. D.; Martin,
Growth 2002, 234, 690–698. J. M. L. J. Phys. Chem. A 2006, 110, 709–716.
(20) Tanskanen, J. T.; Bakke, J. R.; Bent, S. F.; Pakkanen, T. A. (25) Weigend, F.; Ahlrichs, R. Phys. Chem. Chem. Phys. 2005, 7, 3297–
Langmuir 2010, 26, 11899-11906. 3305.
Article Chem. Mater., Vol. 22, No. 16, 2010 4671

Figure 1. CdS growth rate on Si(100) as a function of DMCd pulse time

and N2 purge time. 100 cycle films were deposited at 150
C. Purge time
was 10 s and H2S pulse time was 0.4 s for the DMCd pulse study, and pulse
times for H2S and DMCd were 0.4 s for the purge time study.

9-7-6-311-d631G basis26,27 with the outermost exponent vari- Figure 2. Film thickness as a function of temperature for CdS ALD.
ationally optimized for bulk CdS (0.1739 -> 0.1797) was used.
A modified basis set was necessary for the periodic calculations, Using the optimized operating parameters at 150
C
and the basis reproduced the characteristics of bulk CdS well. described above, the behavior of the CdS deposition
As an example, the optimized lattice parameter of zincblende process was tested over a temperature range of 100
C-
CdS was within 2% of the experimental value. The active 400
C. Figure 2 demonstrates that the growth is linear
growth surface was modeled by a four atomic layer thick for the 100
C-300
C range. The growth rates obtained
hydrogenated S-terminated (111) slab of zincblende CdS, and by linear regression decrease with temperature, and
the outermost atomic layer of the slab was allowed to relax in DFT calculations provide an explanation for this effect
the optimizations. Frequency calculations were performed to
(vide infra). The growth rate monotonically decreases
investigate the effect of thermodynamics on the film growth
characteristics. The electronic characteristics of the deposited
from approximately 1.96 Å/cycle at 100
C to 0.72 Å/
films were analyzed on the basis of PBE0-calculated density of cycle at 300
C. These growth rates are equivalent to 0.58
states-plots (DOS) and by using 2
2
2 and 3
3
3 CdS monolayers/cycle and 0.21 monolayers/cycle, respec-
supercells with a stoichiometry of Cd8S8 and Cd27S27, respec- tively, when referenced to the c(111) interplanar distance
tively. The expansion of the DOS plots was performed using 18 of 3.36 Å. At 300
C a significant portion of the film has
Legendre polynomials. the h(103) phase, and the growth rate is 0.38 monolayers/
cycle using its interplanar distance of 1.9 Å; consequently,
Results and Discussion the actual growth rate at 300
C is in the range of 0.21-
0.38 monolayers/cycle. No incubation period is evident
1. ALD Growth Characterization. CdS growth rates as
from the growth rate studies indicating that nuclea-
a function of precursor pulse time and nitrogen purge
tion occurs fairly readily on the SiO2/Si(100) substrate;
time were studied, keeping all other process parameters
nevertheless, further experiments specifically focusing on
constant. The resulting growth rate curves are shown in
nucleation are of interest for future studies.
Figure 1 as a function of DMCd pulse time and N2 purge
The boiling point of DMCd is 105.5
C, and an
time, respectively. The data in Figure 1 show that satura-
increased residence time in the reactor at lower tempera-
tion behavior, which is a characteristic of ALD, with
tures may account for the significantly increased growth
respect to DMCd is reached by 0.4 s, and the growth is
rate at 100
C because the remaining reactant could react
self-limiting. Also shown, after 10 s of purging with N2,
with H2S during the following pulse, leading to a small
excess reactants have been removed. The growth rate
CVD component. However, as shown by Groner et al.
dependence on H2S pulse time was not measured since
for deposition of Al2O3 using trimethylaluminum at low
the gas was delivered in excess to account for a decrease in
temperatures, the growth rate can be held constant
the pressure of H2S gas during deposition. Unless other-
by increasing the purge time at lower temperatures.28 At
wise stated, experiments in this paper were performed at
400
C, the growth rate does not display the linear behavior
the optimum conditions for 150
C, which is a pulse
expected from ALD: the growth rate increases with cycle
sequence of 0.4 s DMCd, 10 s N2 purge, 0.4 s H2S, and
number and is not uniform across a 2 in. sample range.
10 s N2 purge with a total nitrogen flow rate of 60 sccm. A
This observation is likely due to instability or decomposition
purge time of around 7 s is sufficient in this reactor for
of the ligands. Thus, true ALD does not occur at 400
C, but
ALD growth; however, a 10 s purge time was used to
pure CdS is still deposited without contamination from
ensure that no CVD component is incorporated during
ligands. The cross-section TEM images in Figure 3a confirm
the detailed study of ALD growth rate and properties.
the growth rate for the 150
C film. The thickest portion of
(26) CRYSTAL basis set library at http://www.crystal.unito.it/Basis_ the 400
C film was analyzed by TEM, and it verifies that the
Sets/Ptable.html. film is very rough.
(27) Dou, Y.; Egdell, R. G.; Law, D. S. L.; Harrison, N. M.; Searle, B.
G. An experimental and theoretical investigation of the electronic
structure of CdO. J. Phys.: Condens. Matter 1998, 10(38), 8447– (28) Groner, M. D.; Fabreguette, F. H.; Elam, J. W.; George, S. M.
8458. Chem. Mater. 2004, 16, 639–645.
4672 Chem. Mater., Vol. 22, No. 16, 2010 Bakke et al.

Figure 3. TEM images of 500 cycles of CdS at (a) 150
C and (b) 400
C. Figure 4. PBE0-calculated dissociative DMCd chemisorption energies
per precursor (ΔG, 1 atm) on hydrogenated S-terminated (111) surface of
zincblende CdS for a precursor fractional surface coverage of 0.25, 0.50,
The growth rate curves obtained for CdS display many and 1.00 as a function of temperature.
similarities to the diethylzinc and H2S ALD process.17
The ZnS system displays a growth rate that monotoni- monolayer with respect to the lower fractional coverages,
cally decreases with temperature from a rate of 0.44 and thus entropy increases at an increasing rate for the lower
monolayers/cycle at 100
C to 0.24 monolayers/cycle coverages with temperature. (2) In addition, the effective size
at 300
C (compared to 0.58 monolayers/cycle and 0.21 of the chemisorbed species increases with temperature be-
monolayers/cycle for CdS). Moreover, density functional cause of the increased atomic motion, which leads to more
theory calculations by Tanskanen et al.20 suggest that a significant ligand repulsion at higher coverages and to the
higher growth rate occurs with DMZn compared to observed preference for the less densely packed surfaces at
DEZn. Consequently, the increased growth rate in terms higher temperatures.
of Å/cycle is not unexpected compared to the DEZn 2. Film Characterization. XPS was used to analyze the
system because of the increased lattice constant and the composition of the ALD films and shows that the films
difference in ligands. Also, neither system shows a “tradi- are stoichiometric to within the error of the XPS method;
tional” ALD window in which the growth rate is flat.17,19 the chemical composition as calculated from XPS data
To understand the origin of the decreasing growth rate as is bounded by a few atomic percent because of the many
a function of increasing temperature, the film growth char- sources of error inherent in quantitative analysis by
acteristics were investigated by periodic PBE0 calculations. XPS;30-32 moreover, the cross-section (sensitivity) for
We focused on studying the reaction energies for the dis- Cd in XPS is much larger than for S.33 Thus, films
sociative chemisorption of DMCd according to the reaction deposited above 150
C are likely stoichiometric as the
mechanism SH*(s) þ Cd(CH3)2(g) f S-Cd-CH3*(s) þ bandgap and crystal lattice parameters match that of pure
CH4(g), which has been suggested to be the major pathway CdS (vide infra). Films deposited at 100
C display
of DMCd dissociation with a hydrogenated sulfur surface behaviors which may be due to sulfur vacancies or
where “*” refers to the surface species.29 The SH*(s) growth hydrogen impurities (vide infra) and are not unexpected
surface was represented by a hydrogenated (111) slab of from literature.28,34 Figure 5 shows an XPS spectrum of a
cubic CdS, and precursor surface coverages of 0.25, 0.50, and film deposited at 200
C: only Cd and S are visible in the
1.00 monolayers were considered. The calculated Gibbs- film. Similar to the ZnS deposition in a previous study,17
corrected reaction energies at temperatures between 50 and no noticeable carbon or nitrogen from the acetonitrile
400
C are illustrated in Figure 4. The energies are given per byproduct of the in situ H2S generation is incorporated
mole of precursor to allow direct comparison of different into the film. No contaminants were seen in any analyzed
fractional coverages. At low temperatures, the DMCd dis- samples in the temperature range of 100
C-400
C, and
sociation energies for the different coverages are within 5 the composition did not change with temperature. Final-
kcal/mol of each other. On the other hand, low-coverage ly, the composition was analyzed at various sputter times,
systems become clearly favored over a monolayer coverage and the composition was constant, confirming that the
as a function of increasing temperature so that by 400
C the sputtering did not have a noticeable effect on the mea-
surface coverages of 0.25 and 0.50 are favored over the full sured composition.
coverage system by approximately 10 kcal/mol per chemi- CdS is well-known to exist in either the zincblende
sorbed DMCd. As a consequence, the slow growth rates (cubic) or the wurtzite (hexagonal) crystal struc-
at high temperatures are attributed to partial surface cover- tures, and the phase strongly depends upon factors
age by the metal precursor because of steric hindrance at
(30) Analytical Instrumentation Handbook; Marcel Denker: Las Vegas,
highly covered surfaces. The calculated energetic trends NM, 1990.
originate primarily from (1) entropy and (2) steric effects: (31) Ohring, M., The Materials Science of Thin Films; Academic Press
(1) Namely, atomic motion, which increases as a function Limited: New York, 1992.
(32) Wagner, C. D. Anal. Chem. 1977, 49, 1282–1290.
of temperature, is more limited for the tightly packed (33) Wagner, C. D.; Davis, L. E.; Zeller, M. V.; Taylor, J. A.; Raymond,
R. H.; Gale, L. H. Surf. Interface Anal. 1981, 3, 211–225.
(34) Weber, M.; Krauser, J.; Weidinger, A.; Bruns, J.; Fischer, C. H.;
(29) Han, M.; Luo, Y.; Moryl, J. E.; Osgood, R. M. Surf. Sci. 1999, 425, Bohne, W.; Rohrich, J.; Scheer, R. J. Electrochem. Soc. 1999, 146,
259–275. 2131–2138.
Article Chem. Mater., Vol. 22, No. 16, 2010 4673

At lower temperatures, the films are primarily zincble-

nde, and the wurtzite content increases with temperature.
The phase change is gradual from 100 to 300
C with the
films over this temperature range primarily containing
zincblende CdS. A large shift from zincblende to mostly
wurtzite occurs by 400
C as signified by the strong
appearance of the h(101) peak at 28.2
and an increase
of intensity in the h(103) peak. The wurtzite XRD pattern
is similar to patterns obtained from wurtzite films or
nanoparticles deposited by other methods.36 The dif-
Figure 5. XPS survey spectrum after 10 s Arþ ion sputter of CdS film ference in peak intensities of the main peak at 26.5

deposited at 200
C. at different deposition temperatures is due to several
reasons. First, there are slight differences in the thickness
(see Figure 6 caption) that largely account for the change
in intensities. Next, as both phases appear in the thin film,
the average intensity of this peak decreases. Finally,
as discussed in detail for a ZnS system, the tendency
for{111} planes of zincblende to preferably align parallel
to the sample surface decreases with increasing temper-
ature.17 The crystal structure presented herein is strongly
dependent on the deposition temperature. The 100 and
400
C samples were annealed at 500
C in a pure H2S
atmosphere and no change occurred in the XRD patterns
(data not shown).
Using the Scherrer equation and the full width half max
(fwhm) of the peak intensity for the main peak at 26.5
,
the average grain size for the main crystal grains can be
estimated; however, it should be noted that the Scherrer
equation provides a minimum average crystal size be-
cause of line broadening that may occur from the equip-
ment. The results of this analysis, shown in Figure 7,
indicate that the grain size linearly increases with tem-
perature. We can also analyze the average grain size of the
wurtzite crystal structure obtained from the 47.8
peak
which is present in all the films (Figure 9). The grain sizes
at lower temperatures are small at ∼12 nm; however, the
size quickly increases at 225
C and increases with tem-
perature thereafter. This examination shows that at low
temperatures the wurtzite grains are small, and this
phenomenon is attributed to their formation from stack-
Figure 6. Temperature dependent XRD patterns of CdS on glass. The ing faults in the cubic structure (vide infra). At higher
patterns correspond to (a) 100
C (49 nm) (b) 150
C (64 nm), (c) 225
C temperatures, the main growth mechanism forms wurt-
(29 nm), (d) 300
C (28 nm), and (e) 400
C (37 nm). Asterisks designate
a peak. The peak at 26.5
is not labeled with a star. zite so the grains are larger. The film thickness (given in
the Figure 6 caption) does not appear to significantly
such as deposition method, temperature, time, and pre- affect the grain size. To confirm the above trends in
cursors.3,7,9,10,35 The crystal structure in this work was crystal size, the crystal sizes were analyzed from cross-
determined by XRD for CdS films grown at tempera- section TEM images (vide infra), and the crystals were
tures ranging from 100-400
C on glass. The data is determined to be 11 ( 3 nm at 150
C and 29 ( 16 nm
summarized in Figure 6, and peaks are labeled with at 400
C. As mentioned before, XRD patterns of the
a “*”. All CdS films share two peaks, one at 26.5
150 and 400
C films after a 10 h anneal in H2S at 500
C
corresponding to the h(002)/c(111) peak and one at 47.8
were analyzed, and the phases present did not change;
corresponding to h(103). The large background slope at however, the average grain size as determined by the
low intensities is due to the amorphous glass substrate. fwhm for the 150
C film increased to be equivalent to
Table 1 provides a summary of the relative intensities of that obtained for the as deposited 400
C film. The grain
each peak and compares this ALD system to that of the sizes did not change for the 400
C sample after annealing.
similar ZnS system. A comparison to the ZnS ALD system as shown
in Table 1 demonstrates that for CdS films, wurtzite
(35) Froment, M.; Bernard, M. C.; Cortes, R.; Mokili, B.; Lincot, D.
J. Electrochem. Soc. 1995, 142, 2642–2649. (36) Singh, V.; Chauhan, P. Chalcogenide Lett. 2009, 6, 421–426.
4674 Chem. Mater., Vol. 22, No. 16, 2010 Bakke et al.

Table 1. Summary of CdS XRD Intensities Relative to the h(002)/c(111) Peak and a Comparison to an Analogous ZnS ALD Processa
400
C 300
C 225
C 150
C 100
C
cubic hexagonal CdS ZnS CdS ZnS CdS ZnS CdS ZnS CdS ZnS

100 0.14
111 002 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
101 1.20 0.14
200 0.02
102 0.30 0.13
220 110 0.06 0.05 0.01 0.003 0.02 0.03
103 0.41 0.06 0.44 0.07 0.13 0.003 0.06 0.002 0.08
311 112 0.23 0.10 0.14 0.02 0.003
201 0.07
222 004 0.08 0.01 0.03 0.01 0.03 0.01 0.05
202 0.06
%h(002)/c(111) 29% 79% 52% 90% 86% 98% 90% 99% 86% 100%
a
The %h(002)/c(111) row is the percentage of total intensity belonging to that peak.

Figure 7. Grain sizes of the main grains attributed to the principal cubic
and wurtzite peaks. The solid markers are measured after a 500
C anneal
in H2S for 10 h.

characteristics are displayed at lower temperatures and

also increase at a much faster rate with increasing tem-
perature. For example the h(002)/c(111) peak accounts
for 100% and 79% of the intensity for ZnS films at 100 Figure 8. Selected area diffraction (SAD) patterns of 500 cycle films of
and 400
C, respectively. However, in CdS films the CdS with 150 nm diameter aperture at 150
C (a and b) and at 400
C (c
values are 87% and 29%. If one accounts for the peaks and d): a and c SAD patterns are of Si and CdS, b and d SAD patterns are
of CdS only. The SAD pattern in b for 150
C shows mainly zincblende
that are solely attributable to wurtzite phases, the CdS spots and the CdS (111) planes are strongly oriented parallel to the film
and ZnS films at 400
C are at least 63% and 5% wurtzite, surface (underlying substrate is Si (001). The SAD in d for 400
C shows
respectively. In fact, the h(101) peak dominates for CdS at that zincblende (111) planes and wurtzite (002) planes have same direc-
tional stacking which is evidence for stacking fault formation in the
400
C. The increased stability of the wurtzite phase in wurtzite phase. The preferential orientation of crystals is lost at 400
C
CdS compared to ZnS has been shown in a previous study compared to 150
C.
where the phase of the CdxZn1-xS at a constant tempera-
ture is strongly affected by the composition.37 (111) spots are aligned to the Si (002) spots (present
The crystal grain growth is confirmed to be distinct at because of dynamical diffraction) with bright intensity.
different temperatures by studying the TEM images in This alignment may also be inferred from Figure 6b
Figure 3 and the SAD patterns in Figure 8, and they are because of the presence of only one high intensity X-ray
supported by the XRD patterns in Figure 6. Large peak in the pattern corresponding to c(111). Very weak
columnar zincblende grains for the 150
C film extending intensity wurtzite 110 (h) type spots in the SAD pattern in
from the surface are visible in the TEM image in Figure 8b confirm the XRD analysis that there is a small
Figure 3a, and the SAD pattern over Si and CdS using portion of wurtzite phase present in films deposited at
a 150 nm diameter aperture shown in Figure 8a corro- 150
C. The SAD pattern contains spots which are
borates that the zincblende (111) planes are highly ori- primarily a contribution from the c(111) crystals strongly
ented parallel to the surface since the CdS zincblende oriented to the Si (002) direction. An HRTEM image and
its FFT image (not shown) further confirm that both
(37) Chen, W. W.; Zhang, J. M.; Ardell, A. J.; Dunn, B. Mater. Res. wurtzite and zincblende crystals are present in the 150
C
Bull. 1988, 23, 1667–1673. film. In Figure 8 various diffraction spots of zincblende
Article Chem. Mater., Vol. 22, No. 16, 2010 4675

Figure 9. High resolution TEM image of 150
C CdS film: (a) overview of a grain containing stacking faults and (b) magnified view of the region in the
white box in (a). Intrinsic and extrinsic stacking faults (SFs) form by a/6 Æ211æ type displacement of plane position on zincblende (111) planes to form local
wurtzite grains. Closely spaced SFs may construct a larger wurtzite structure.

and wurtzite were assigned on the basis of the previously spots in Figure 8d as opposed to Figure 8b presumably
reported lattice parameters36,38 by the relationship of indicates a higher portion of wurtzite in that local area.
Rd = Lλ = fixed constant, which is the camera length An HRTEM image of the 400
C film and its FFT image
equation of a TEM diffraction pattern, where R is the (not shown) elucidate that both zincblende and wurtzite
distance from the center spot to a specific spot corre- CdS are present. The SAD pattern matches the FFT
sponding to a specific plane, d is the interplanar spacing image in that the c(111) plane is coincident with the
between the specific planes, L is the camera length, and h(002) plane.
λ is 2.51 pm, which is the wavelength of the electron beam A high resolution TEM image of 150
C CdS film
with a 200 kV acceleration. The Si spots in the SAD (Figure 9a) confirms that wurtzite grains are present in
patterns of Figure 8a and Figure 8c were used as a zincblende grains because of stacking faults. A magnified
reference because Si has a well-known lattice parameter view of the HRTEM image in Figure 9b shows mixed
(aSi = 5.4308), and direct comparison of the spots bet- local zincblende and wurtzite form from a combination of
ween Si and CdS gives better verification of CdS spots. intrinsic and extrinsic stacking faults, and the stacking
Comparatively, the 400
C film does not show a strong sequence is clearly defined and labeled. c(111) and h(002)
alignment of the h(002)/c(111) planes and is primarily planes are stacked parallel and a/6 Æ211æ type displacement
wurtzite in character. The XRD patterns and Table 1 of zincblende (111) planes yields intrinsic and extrinsic
show that many crystal phases are visible because of the stacking faults which provide transformation from zinc-
random orientation. The SAD patterns such as those in blende (thermodynamically stable phase at 150
C) to
Figure 8c,d show that the crystals are randomly oriented wurtzite (metastable phase at 150
C). This indicates that
compared to the silicon substrate. Thus, the reduction in wurtzite AB stacking forms because of stacking faults from
peak intensities of the main peak as seen in XRD is due to zincblende ABC stacking. Finally, closely spaced stacking
both the increasing wurtzite content and the decreasing faults can construct a larger local wurtzite structure as seen
texture. The clear wurtzite diffraction array with higher in Figure 9b.
index planes is seen in the SAD pattern of the CdS region Optical characterization of the CdS films on glass was
in Figure 8d. c(111) and h(002) have the same diffraction performed with UV-vis to determine the bandgap. CdS
spot position because the ABC stacking direction of the is a direct bandgap semiconductor, and literature values
c(111) planes is the same as the AB stacking direction of of the bandgap range from 2.25 eV-2.45 eV depending
h(002) planes. This coincidence of c(111) and h(002) on deposition method, crystal structure, grain size, and
suggests the presence of stacking faults which local form quantum effects;1,7,12,36,39 however, the accepted band-
hexagonal stacking in the cubic structure. Random ori- gap for pure CdS is 2.42 eV.1 CdS quantum dots display
entation of the spots without coincidence would signify much larger bandgaps; however, they approach bulk
that wurtzite forms freely without orientation relative to values at >10 nm, which means significant quantum
zincblende; thus, the coincidence of those spots is con- confinement effects are not expected for these films based
sistent with wurtzite AB stacking resulting from stacking on the Scherrer equation calculations.40,41 In this work,
faults in the zincblende ABC stacking. Further, the
clearer and stronger intensity of the wurtzite diffraction (39) Zelayaangel, O.; Hernandez, L.; Demelo, O.; Alvaradogil, J. J.;
Lozadamorales, R.; Falcony, C.; Vargas, H.; Ramirezbon, R.
Vacuum 1995, 46, 1083–1085.
(38) Benkabou, F.; Aourag, H.; Certier, M. Mater. Chem. Phys. 2000, (40) Wang, Y.; Herron, N. Phys. Rev. B 1990, 42, 7253.
66, 10–16. (41) Weller, H. Angew. Chem., Int. Ed. Engl. 1993, 32, 41–53.
4676 Chem. Mater., Vol. 22, No. 16, 2010 Bakke et al.

Figure 11. PBE0-calculated projected and total DOS plots using a 2

2
2 supercell for cubic CdS. (a) Defect-free CdS, (b) S vacancy density of
1.7 nm-3, (c) S substitution density of 1.7 nm-3. Fermi levels are indicated
with dashed lines.

account for this, the “bulk” bandgap shifts closer to the

Figure 10. Tauc plots (a) of as deposited CdS at 150
C and (b) of the bandgap for the annealed and 400
C films.42 This
same film after a 500
C anneal in an H2S atmosphere and (c) of as gradual change in the bandgap may be related to the less
deposited CdS at 400
C.
crystalline nature of the 150
C film as discussed above
with the XRD patterns and Scherrer equation calcula-
Tauc plots of (Rhν)2 versus hν are used to estimate the tions, or it may be related to the presence of some sulfur
bandgap for several thicknesses and ALD temperatures. vacancies in the film (vide infra). These vacancies may
The Tauc plots for 500 cycle films deposited at 150 and also serve to “wash out” the absorption due to the
400
C are shown in Figure 10. For films deposited at 100 wurtzite crystals since they are present in Figure 10b after
and 150
C, the bandgap is ∼2.3 eV, and for films the H2S anneal.
deposited in the range 225
C-400
C, the bandgap DOS plots for zincblende CdS were calculated at the
exactly matches the literature value of 2.42 eV. The PBE0 level of theory to determine if the described band-
thickness did not appear to have an effect on the bandgap. gap behavior of the films deposited at 150
C could be
An interesting observation is that an inflection point explained by a model in which defects affect the bandgap
appears in the Tauc plot over the 2.5-2.6 eV range for at 150
C but are removed upon annealing. As previously
films deposited between 225-400
C because of a new noted by XPS (see Figure 5), the films may be slightly
absorption edge that occurs at ∼2.5 eV. The high absorp- rich in cadmium. Two possible routes for excess cad-
tion energy is probably due to the increased wurtzite mium were modeled: one due purely to sulfur vaca-
content in the films deposited at higher temperatures, ncies and another due to cadmium substitutions. The
which would be similar to observations for the ZnS calculated defect densities were 1.7 defects 3 nm-3 and 0.5
system using DEZn/H2S.17 The 150 and 400
C films defects 3 nm-3. The calculated total DOS plots with con-
were also annealed at 500
C in an H2S atmosphere for tributions from S and Cd are shown for the 1.7 nm-3
10 h to study the difference compared to the as depo- defect densities in Figure 11. The S vacancies have an
sited films. After the anneal, UV-vis spectroscopy and effect on the electronic structure of CdS by introducing
XRD were performed again on the samples. According mainly Cd states to the valence region and decreasing the
to XRD, no significant alteration in crystal structure optical gap with respect to zincblende CdS. The bandgap
occurred for either film, but the grain sizes appeared is decreased by 0.4 eV when the S vacancy defect density is
to increase for the 150
C film (vide supra). On the 0.5 nm-3 compared to a pure CdS film. Substitution of S
other hand, Zelaya-Angel et al. noted that zincblende by Cd has a strong impact on the electronic structure of
CdS deposited by CBD transitioned to wurtzite upon CdS by shifting the Fermi level as illustrated in Figure 11c
a 500
C anneal;39 that effect was not observed here. The and resulting in metallic films. Consequently, the calcu-
bandgap of the 400
C film remained unchanged; how- lations suggest that S vacancies in the bulk of the CdS
ever, Figure 10b illustrates that the optical characteristics film deposited at low temperature may contribute to the
of the 150
C film significantly changed in three ways. observed red-shift of the film band gap down to 2.3 eV
First, the bandgap blue-shifts to 2.42 eV matching that of and may cause the slowly increasing slope in the Tauc
the films deposited at higher temperatures. Second, the plot before the bandgap absorption energy if the vacan-
Tauc plot also displays the inflection points at ∼2.5 eV cies introduce a spectrum of states within the bandgap.
because of the presence of a new absorption edge. Finally, We propose that annealing the 150
C film in an H2S
the slope at energies below the CdS absorption is flat like atmosphere results in the refilling of the vacancies,
that for the as deposited 400
C film and the annealed
films. The as deposited 150
C film, however, shows (42) Chen, Z.; Jaramillo, T.; Deutsch, T.; Kleiman-Shwarsctein, A.;
Forman, A.; Gaillard, N.; Garland, R.; Takanabe, K.; Heske, C.;
a gradual change in slope at energies below the CdS Sunkara, M.; McFarland, E.; Domen, K.; Miller, E.; Turner, J.;
bandgap. If the Tauc plot for Figure 10a is baselined to Dinh, H. J. Mater. Res. 2010, 25, 3–16.
Article Chem. Mater., Vol. 22, No. 16, 2010 4677

Figure 12. rms roughness versus cycle number of CdS deposited by ALD
as determined by AFM. Films were deposited at 150
C on Si(100).

which increases the bandgap to match the literature

value of 2.42 eV and flattens the leading edge of the
Tauc plot.
Another possible factor affecting the band gap is the
incorporation of hydrogen into the films, which is com-
mon for CdS films deposited by CBD; however, in the Figure 13. SEM plan view of (a) 100 cycles and (b) 500 cycles CdS
deposited at 150
C on Si(100).
CBD case the majority of the H atoms were present as
interstitials which were removed upon annealing at
150-200
C. The remaining H atoms were present as
CHx or Cd(OH)2.34,43 As our films were mostly deposited
at or above 150
C and did not contain O or C impurities,
we expect the hydrogen concentration to be negligible
at these deposition temperatures, and this is similar to
the findings of ZnS deposited with DEZn and H2S
which showed hydrogen impurities of <0.1 at %.18
Nevertheless, a similar system using trimethylaluminum
and water as precursors for Al2O3 ALD films did show
hydrogen impurity concentrations that decreased with
temperature.28,44 Thus, removal of the light impurities
from the 150
C film upon annealing may contribute to Figure 14. rms roughness versus temperature of CdS deposited by ALD
the observed band gap behavior. on Si(100) as determined by AFM.
AFM and SEM were used to examine the surface
morphology of CdS films deposited on Si(100) surfaces of the rms roughness as a function of temperature. The
as a function of the cycle number at 150
C. The graph graph shows that the rms roughness is constant at low
in Figure 12 shows that the rms roughness increases temperatures and then rises to a higher value by 225
C.
approximately linearly with the number of cycles. This This trend of an increase around 150
C is similar to that
behavior has previously been observed for the DEZn/ noted for the wurtzite grain size measured by XRD
H2S ALD system but the roughness is noticeably greater (Figure 7), leading us to speculate that the increasing
for the CdS system reported herein.17,18 SEM plan view presence of wurtzite crystals which are also randomly
images of the 100 and 500 cycle samples at 150
C as oriented causes an increase in roughness on the film.
shown in Figure 13 delineate two key features of the On the other hand, DFT simulations provide another
films due to their thicknesses. First, the 500 cycle film is possible explanation: if the growth process transitions
much rougher, consistent with the AFM results. Sec- from a monolayer coverage process to a partial mono-
ond, the surface features appear to be much smaller for layer process at higher temperatures as suggested by
the 100 cycle film compared to the 500 cycle film. the change in dissociation energy of the DMCd pre-
Further, the 500 cycle film does not appear to be cursor (Figure 6), this may also lead to increase in
uniformly rough. roughness with temperature. A similar growth method
The surface morphology was also studied for films has been postulated as the cause of roughness for the
deposited at different temperatures. Figure 14 is a plot related DEZn/H2S system.17,18 The SEM plan view
images in Figure 15 show 250 cycle films deposited at
(43) Krauser, J.; Riedle, T.; Klenk, R.; Klaer, J.; Lux-Steiner, M. C.; 100 and 300
C. Both films show uniform roughness
Weidinger, A. Appl. Phys. A: Mater. Sci. Process. 2000, 70, 617– across the surface; however, the 300
C film is both
623. rougher and contains larger features, consistent with
(44) Groner, M. D.; Elam, J. W.; Fabreguette, F. H.; George, S. M. Thin
Solid Films 2002, 413, 186–197. Figure 14.
4678 Chem. Mater., Vol. 22, No. 16, 2010 Bakke et al.

and then increase more suddenly in size at higher tem-

peratures. This phenomenon may be explained by TEM
images showing that smaller wurtzite grains are present
because of stacking faults in the cubic phase. Tauc plots
obtained from UV-vis measurements were used to con-
firm that the band gap matches literature values of 2.42
eV. DFT simulations indicate that absorption at lower
energies in CdS films may be due to sulfur vacancies,
and experiments suggest that sulfur atoms may replace
vacancies by annealing in an H2S atmosphere.
SEM and AFM were used to study the surface mor-
phology as a function of temperature and cycle number.
The roughness increases linearly with cycle number
(thickness) in a similar fashion to ZnS although CdS is
consistently rougher for a given thickness. Surface rough-
ness was confirmed to increase with temperature as well.
These surface properties may be due to the growing
crystal size, to the increasing mixture of different phases,
or to the preference for depositing in a submonolayer
manner as suggested by DFT simulations.
We have demonstrated a successful method for ALD of
Figure 15. SEM plan view of 250 cycle films deposited at (a) 100
C and CdS thin films with a good degree of control over crystal
(b) 300
C on Si(100). phase and morphology with changing growth para-
Conclusion meters. Moreover, our results show that CdS deposited
by ALD can form the wurtzite phase at temperatures
Polycrystalline, near stoichiometric CdS has been de- significantly lower than the thermodynamic zincblende-
posited by ALD using DMCd and in situ generated H2S. to-wurtzite phase transition temperature. Considering
The film thickness versus cycle number is linear for the the array of uses for CdS, we believe this study will be
temperature range 100-300
C, and the growth rate is applicable to many fields which require deposition of CdS
monotonically decreasing with temperature. At 400
C films with tunable material properties.
the deposition is not uniform, and the growth rate varies
with the number of cycles and the position on the wafer. Acknowledgment. J.R.B. acknowledges funding from
Typical saturation behaviors with DMCd pulse time and the Department of Defense (DoD) through the National
purge time are observed for this ALD process. DFT Defense Science and Engineering Graduate Fellowship
calculations suggest that the decreasing growth rate with (NDSEG) and from the National Science Foundation
temperatures occurs because the deposition of material in (NSF) Graduate Fellowship. J.T.T. acknowledges financial
a submonolayer manner is thermodynamically favored at support from the University of Eastern Finland. We also
high temperatures. recognize use of the Stanford Nanocharacterization Labora-
XRD and TEM were used to confirm that at tempera- tory (SNL) and of the Center for Polymer Interfaces and
tures as low as 100
C, the CdS films were primarily Macromolecular Assemblies (CPIMA). The authors gratefully
appreciate support from Varian Semiconductor Equipment
zincblende with wurtzite content; however, the wurtzite
Associates. The TEM and SAD studies were supported as part
character of the films rapidly increased to become the of the Center on Nanostructuring for Efficient Energy Conver-
dominant crystal structure by 400
C. Compared to a sion, an Energy Frontier Research Center funded by the U.S.
similar II-VI metal chalcogenide, the wurtzite nature of Department of Energy, Office of Science, Office of Basic Energy
the CdS films is much stronger than for ZnS at a given Sciences under Award No. DE-SC0001060.
temperature. The TEM images and SAD patterns also
establish that crystals are more highly oriented at lower Supporting Information Available: HRTEM and the corre-
temperatures. Analysis of the fwhm of the XRD patterns sponding FFT images are available for the 150 and 400
C films.
using the Scherrer equation shows that the grain size of AFM micrographs for multiple temperatures and thicknesses
the zincblende portion of the film increases linearly with are also available. This material is available free of charge via the
temperature while the wurtzite grains are a constant size Internet at http://pubs.acs.org.