APPLIED PHYSICS LETTERS VOLUME 76, NUMBER 2 10 JANUARY 2000

Raman spectroscopy of electrochemically self-assembled CdS

quantum dots
A. Balandina)
Department of Electrical Engineering, University of California–Riverside, Riverside, California 92521
K. L. Wang
Device Research Laboratory, Electrical Engineering Department, University of California–Los Angeles,
Los Angeles, California 90095
N. Kouklin and S. Bandyopadhyay
Department of Electrical Engineering, University of Nebraska–Lincoln, Lincoln, Nebraska 68588
共Received 20 September 1999; accepted for publication 10 November 1999兲
We report a Raman spectroscopy investigation of electrochemically self-assembled quasiperiodic
arrays of CdS quantum dots with characteristic feature size of 10 nm. The dots were synthesized
using electrochemical deposition of CdS into a porous anodized alumina film.
Polarization-dependent Raman scattering study over an extended frequency range reveals the
quantization of electronic states in the conduction band and intersubband transitions. Raman peaks
observed at 2919 and 3050 cm⫺1 are attributed to transitions between the lowest two subbands. The
results suggest that quantum dot arrays, produced by inexpensive robust electrochemical means,
may be suitable for infrared detector applications. © 2000 American Institute of Physics.
关S0003-6951共00兲02202-6兴

Quantum dots represent the ultimate test bed for size copy has been chosen as an experimental tool since it allows
quantization of both electrons and phonons. Strong spatial for probing of active optical phonon modes, as well as con-
confinement of these entities create an exciting opportunity fined electronic structure of quantum dots.12 The latter is
for reengineering of the optical, electronic, and thermal prop- particularly important since the sample structure of self-
erties of many technologically important semiconductors assembled quantum dot arrays does not allow for Fourier
through modification of their electronic states and phonon transform infrared 共FTIR兲 spectroscopy in the regular wave-
dispersion.1,2 Intersubband optical transitions in quantum guide geometry. Owing to the small amount of CdS depos-
dots are of particular interest because of their potential use in ited inside the nanopores and the large band gap of bulk CdS
infrared detectors.3,4 Relative to two-dimensional quantum (E G ⬃2.5 eV), it is also difficult to study confined electron
well structures, quantum dots have many advantages in op- levels in this system using photoluminescence excitation
tical applications due to their delta function-like sharp den- 共PLE兲.
sity of electronic states, and relaxed optical selection rules The samples for this study have been prepared by fol-
which allow for detection of normally incident light.5,6 lowing a number of steps. First, a thin foil of aluminum is
To date, optical investigation of quantum dot arrays has anodized for a few hours in 15% H2SO4 at a dc current
been almost exclusively restricted to systems self-assembled density of 40 mA/cm2 to produce a thick nanoporous alu-
in the Stranski–Krastanow 共SK兲 mode of molecular beam mina film on the surface of the foil. This film is stripped off
epitaxy 共MBE兲 growth.1,4,7 The study of SK quantum dots and the anodization is repeated for a few minutes. The po-
has significantly added to the core knowledge of the physical rous alumina film that forms in the second iteration contains
processes in quasi-zero-dimensional 共0D兲 structures. Alter- an ordered hexagonal array of pores with an average diam-
nate candidate, which can potentially lead to low cost manu- eter of 10 nm and a pore density of 4⫻1010 cm⫺2. The pores
facturing, is electrochemical self-assembly.8–10 This tech- are then ‘‘rounded’’ by soaking in phosphoric acid for a few
nique is capable of fabricating dense 共density ␳ ⬃1012 cm⫺2兲 seconds and the porous film is subsequently washed in dis-
highly ordered quasiperiodic two-dimensional 共2D兲 arrays of tilled water to remove debris. Finally, the pores are filled up
semiconductor quantum dots embedded in an insulating po- with CdS using ac electrodeposition. This is accomplished
rous alumina matrix. It is a simple and inexpensive approach by ac electrolyzing the film in H2SO4 for a few seconds
to self-assembling quantum dots and has an extremely fast followed by immersion in a boiling aqueous solution of
throughput because it is a parallel process whereby several CdSO4. The Cd⫹⫹ ion in the solution reacts with the S⫺⫺ ion
arrays can be self-assembled simultaneously. left behind in the walls of the pores from the ac electrolysis
In this letter we report the Raman scattering polarization step to form CdS.
study of arrays of cadmium sulfide 共CdS兲 quantum dots self- The amount of CdS that goes into the pores increases
assembled by the electrochemical technique. CdS is a direct- superlinearly with the duration of immersion in CdSO4. Dur-
band gap, II–VI compound semiconductor that can be used ing the fabrication process, we have carefully avoided over-
for the fabrication of optoelectronic devices such as solar filling the pores by adjusting the immersion time. For this
cells, laser diodes, and photoconductors.11 Raman spectros- study, we have chosen CdS since it requires very simple
deposition steps, and has a linear expansion coefficient ( ␬
a兲
Electronic mail: [email protected] ⫽5.0⫻10⫺6 K⫺1) matched to that of the alumina substrate

0003-6951/2000/76(2)/137/3/$17.00 137 © 2000 American Institute of Physics

138 Appl. Phys. Lett., Vol. 76, No. 2, 10 January 2000 Balandin et al.

of material electrodeposited inside the nanopores. We ob-

served a systematic downshift of the 1LO peak of the quan-
tum dot arrays by one to two wave numbers from its bulk
position. This result is consistent with the data reported in
Ref. 11 for CdS nanoclusters fabricated by the solution
growth technique 共SGT兲 on slide glass substrates, and for
CdS dots (D⬃4 – 10 nm) embedded in silicon dioxide glass
films.12 The fact that the redshift of the phonon peaks due to
spatial confinement of phonon modes13 is not all that pro-
nounced may be explained by its compensation due to strain-
induced blueshift. This situation is similar to the one recently
reported by some of us for MBE grown Ge quantum dots on
the Si substrate.3
All spectra in our study have been excited by the 514 nm
line of an Ar ion laser corresponding to the energy E
⬃2.4 eV, which is lower than the band-gap energy for CdS.
In the past, absorption and photoluminescence studies indi-
cated that the effective band gap 共blueshifted due to quantum
confinement and strain兲 is about 3 eV.8 Under the conditions
of extreme resonance, when the laser frequency is close to
the effective band gap, the background due to the hot lumi-
nescence dominates the spectra.14 In the extended polarized
spectrum 关 x(z,z)x兴 of all examined samples, we observed a
pronounced broad peak around 3000 cm⫺1. It occurred at
2919 cm⫺1 (E⫽361 meV) in sample A and at 3050 cm⫺1
(E⫽377 meV) in sample B. We attribute it to intersubband
transition within the conduction band of CdS dots 共see Fig.
2兲. The large linewidths of these peaks accrue mostly from
inhomogeneous broadening of the intersubband levels due to
variations in the dot size across the irradiated portion of the
array 共the laser spot will typically excite about 10 000 dots
FIG. 1. 共a兲 Atomic force image of the nanoporous alumina (Al2O3) template
simultaneously兲. The dots are actually somewhat cylindrical
used for electrochemical fabrication of the quantum dot arrays. The interdot in shape 共like the pores into which they are ‘‘cast’’兲. Al-
separation for this sample is about 30 nm. Inset shows a schematic drawing though the diameters of these cylinders exhibit less than 5%
of the CdS dot array; 共b兲 Raman spectra of the electrochemically self- variation, the heights of these cylinders vary by much more
assembled CdS quantum dots and corresponding bulk material. The inten-
sity of the signal from the dot array is much weaker. The longitudinal and this variation could easily account for the large line-
optical 共LO兲 phonon peaks of CdS are seen at about 300 and 600 cm⫺1 width. Other than size nonuniformity, asymmetry of the con-
共second order兲. A small redshift of the peaks with respect to the bulk posi- fining potential and interfacial strain could also contribute to
tion may be due to spatial confinement of phonon modes. the linewidth broadening. Finally, it is even possible that the
high frequency shoulders in Fig. 2 are due to overlap of
( ␬ ⫽5.8⫻10⫺6 K⫺1). An atomic force micrograph 共AFM兲 of additional unresolved peaks from other intersubband transi-
the regimented nanoporous alumina template, which was tions. Due to this reason, it is not possible to accurately
subsequently filled with CdS, is shown in Fig. 1共a兲. An inset determine the full width at half maximum.
shows a schematic drawing of a cross section of a typical We have excluded free-electron absorption in the alumi-
electrochemically self-assembled quantum dots array after num foil as an origin of the peaks in Fig. 2 by measuring the
the deposition step. Raman spectra of the bare alumina film without the CdS in
Raman spectra were measured using a Renishaw Raman the pores. The bare film shows no peak. The peak structure
Imaging 2000 microscope at room temperature. All spectra observed when the alumina pores are filled with CdS, and the
were excited by the 514 nm line of an Ar ion laser in the line shape, are analogous to those associated with previously
backscattering configuration and recorded by a Si charge- reported intersubband transitions in Ge quantum dots ob-
coupled device 共CCD兲 camera. The spectral resolution of the served both by Raman spectroscopy3 and by Fourier tran-
instrument was about 0.1 cm⫺1. The polarization dependent sorm infrared 共FTIR兲 spectroscopy.4
Raman spectra were taken over an extended frequency range It is reasonable to assume that the peaks are related to
from 250 to 4200 cm⫺1. the transitions between the two lowest confined electron
The low-frequency part of the spectra for CdS dots and states (E1↔E2). If we proceed from this assumption, then,
bulk material is shown in Fig. 1共b兲. One can clearly see two using the simple infinite barrier model for these dots and
characteristic CdS longitudinal optical 共LO兲 phonon peaks at neglecting exciton effects, we can estimate the dot diameter
about 300 cm⫺1 共1LO兲 and 600 cm⫺1 共2LO兲. The intensity of to be D⫽3.9 nm for sample A and D⫽3.8 nm for sample B
peaks from the quantum dot array is much weaker than that 共dots are assumed to have a cylindrical shape with the height
from bulk CdS. This is expected owing to the small amount equal to the diameter兲. In this calculation, we assumed that
Appl. Phys. Lett., Vol. 76, No. 2, 10 January 2000 Balandin et al. 139

and provided the first evidence of the discretization of the

conduction band states in these dots. The discretization is
evidenced by intersubband transition within the conduction
band. It is natural to ask why we do not observe intersubband
transitions within the valence band. We believe that this is
because the four times heavier effective mass of heavy holes
decreases the hole subband spacing commensurately and
makes the transitions unresolvable at room temperature. Pre-
vious experiments established only the existence of the low-
est electron and highest heavy hole subband through a blue-
shift of the photoluminescence peak8 thus providing
evidence of quantum confinement, but no direct evidence of
the discretization of conduction- or valence-band states.
Those experiments involved interband transitions whereas
the present involves intraband transitions. We have also ob-
served a slight but systematic redshift of LO phonon peaks
characteristic of spatially confined phonon modes. Finally, in
the context of engineering applications, the observation of
intersubband transitions in electrochemically self-assembled
quantum dots holds out the hope of developing midinfrared
detectors capable of detecting normally incident light. The
fabrication technique that was used is robust, inexpensive,
and has the high-throughput required for commercial appli-
cations.
The authors thank Dr. J. L. Liu 共UCLA兲 for discussions
on Raman study of electronic excitations and Dr. Yi Liu
共Nebraska兲 for TEM studies. The work at the University of
Nebraska was supported by the U.S. Army Research Office
under Grant No. DAAG55-98-1-0015. The work at UCLA
was in part supported by the DOD MURI Program on Low
Power/Low Noise Electronics managed by ARO.
FIG. 2. High frequency polarized Raman spectrum of CdS dots 共solid lines兲
of the sample A 共a兲 and sample B 共b兲 and the aluminum substrates with the 1
layer of alumina 共dashed lines兲 used for their fabrication. The effective S. Fafard, Z. R. Wasilewski, C. Ni Allen, D. Picard, P. G. Piva, and J. P.
optical size of the dots of sample B is smaller than that of sample A. The McCaffrey, Superlattices Microstruct. 25, 87 共1999兲.
2
latter is expected from different anodizing processes applied for fabricating K. L. Wang and A. Balandin, in Optics of Nanostructed Materials, edited
these two samples. by V. A. Markel and T. F. George 共Wiley, New York, 1999兲.
3
J. L. Liu, Y. S. Tang, K. L. Wang, T. Radetic, and R. Gronsky, Appl.
Phys. Lett. 74, 1863 共1999兲.
4
J. L. Liu, W. G. Wu, A. Balandin, G. Jin, and K. L. Wang, Appl. Phys.
the electron effective mass retains its bulk value of m e*
Lett. 74, 185 共1999兲.
⫽0.21m 0 . The calculated dot sizes are less than what we 5
D. Pan, Y. P. Zeng, M. Y. Kong, J. Wu, Y. Q. Zhu, C. H. Zheng, J. M. Li,
find in bright field TEM images which show D⬃10 nm. and C. Y. Wang, Electron. Lett. 32, 1726 共1996兲.
6
However, this discrepancy can be explained by invoking the K. L. Wang and R. P. G. Karunasiri, in Semiconductor Quantum Wells
and Superlattices for Long-Wavelength Infrared Detectors, edited by M.
side depletion caused by the Fermi level pinning at the CdS- O. Manasreh 共Artech House, Norwood, MA, 1993兲, p. 139; W. G. Wu, J.
alumina interface, which makes the effective dot diameter L. Liu, Y. S. Tang, G. L. Jin, and K. L. Wang, Proc. SPIE 3630, 98
smaller. What is important is that Ref. 8 reported an effective 共1999兲.
7
dot diameter of 3.5 nm in such samples from an independent N. Kirstaedter, N. N. Ledentsov, M. Grundman, D. Bimberg, V. M. Usti-
nov, S. S. Ruvimov, M. V. Maximov, P. S. Kopev, Zh. I. Alferov, U.
measurement 共blueshift of the band gap measured from in- Richter, P. Werner, U. Gosele, and J. Heydenreich, Electron. Lett. 30,
terband absorption兲. The close agreement between these two 1416 共1994兲; V. Ya. Aleshkin, N. A. Bekin, N. G. Kalugin, Z. F.
independent estimates lends credence to our claim of having Krasilnik, A. V. Novikov, V. V. Postnikov, and H. Seyringer, JETP Lett.
observed intersubband transition. The high frequency shoul- 67, 48 共1998兲.
8
S. Bandyopadhyay, A. E. Miller, H. C. Chang, G. Banerjee, D.-F. Yue, R.
ders in Fig. 2 may be attributed to additional unresolved E. Ricker, S. Jones, J. A. Eastman, and M. Chandrasekhar, Nanotechnol-
peaks from other intersubband transitions with different peak ogy 7, 360 共1996兲.
9
positions due to quantum dot size variation and possible R. E. Ricker, A. E. Miller, D.-F. Yue, G. Banerjee, and S. Bandyo-
asymmetricity. The small features in the interval 1000–1500 padhyay, J. Electron. Mater. 25, 1585 共1996兲.
10
A. Balandin, S. Bandyopadhyay, P. G. Snyder, S. Stefanovich, G. Baner-
cm⫺1, which have been seen in some of the spectra, may jee, and A. E. Miller, Phys. Low-Dimens. Semicond. Struct. 11/12, 155
have several different origins. A peak around 1200 cm⫺1 can 共1997兲.
be attributed to the higher-order optical phonon or water
11
J. Lee and T. Tsakalakos, Nanostruct. Mater. 8, 381 共1997兲.
12
A. G. Rolo, L. G. Vieira, M. J. M. Gomes, and J. L. Ribeiro, Thin Solid
bands.
Films 312, 348 共1998兲.
In conclusion, we have reported a Raman scattering 13
J. H. Campbell and P. M. Fauchet, Solid State Commun. 58, 739 共1986兲.
study of electrochemically self-assembled CdS quantum dots 14
C. Schuller, Physica E 共Amsterdam兲 3, 121 共1998兲.