Size tunable visible and near-infrared photoluminescence from vertically
etched silicon quantum dots
Sameer S. Walavalkar,
1,2,
a
Andrew P. Homyk,
1,2
Carrie E. Hofmann,
2
M. David Henry,
1,2
Claudia Shin,
2
Harry A. Atwater,
2
and Axel Scherer
2
1
Kavli Nanoscience Institute, Caltech, Pasadena California 91125, USA
2
Department of Applied Physics, Caltech, Pasadena California 91125, USA
Received 31 January 2011; accepted 31 March 2011; published online 14 April 2011
Corrugated etching techniques were used to fabricate size-tunable silicon quantum dots that
luminesce under photoexcitation, tunable over the visible and near infrared. By using the fidelity
of lithographic patterning and strain limited, self-terminating oxidation, uniform arrays of pillar
containing stacked quantum dots as small as 2 nm were patterned. Furthermore, an array of
pillars, with multiple similar sized quantum dots on each pillar, was fabricated and tested. The
photoluminescence displayed a multiple, closely peaked emission spectra corresponding to quantum
dots with a narrow size distribution. Similar structures can provide quantum confinement effects for
future nanophotonic and nanoelectronic devices. ©
2011 American Institute of Physics
.
doi:
10.1063/1.3580768
Over the past 60 years silicon devices have evolved to
form the backbone of the micro-electronics industry. Unfor-
tunately, the indirect nature of the silicon band-gap has ham-
pered the material with highly inefficient light emission
properties, thus impeding its spread into the optoelectronic
field. Recent work, however, has focused on utilizing the
physical changes seen in low dimensional nanostructures to
breathe life into the use of silicon as an active optical
material.
Silicon quantum wells,
1
grown nanowires,
2
etched
nanopillars,
3
and quantum dots
4
,
5
have shown promising
electroluminescent and photoluminescent properties. Specifi-
cally silicon nanocrystals embedded in a dielectric matrix
such as silicon dioxide or silicon nitride have been used as a
convenient method to generate complimentary metal on
semiconductor
CMOS
compatible layers of light emitting
material. The most popular technique to fabricate these nano-
crystals has been to deposit a layer of silicon rich oxide or
nitride and anneal at above 1100 °C, allowing the excess
silicon to precipitate into clusters within the dielectric.
6
–
8
Depending on the annealing temperature, a statistical distri-
bution of quantum dot sizes is found. It has been shown that
these dots can be made to luminesce by optical pumping as
well as through electrical excitation by tunneling electrons
and holes through the dielectric and allowing them to recom-
bine in the dots.
Although this precipitation technique has been improved
and well characterized over the years, it comes with some
key limitations. For example, the spatial distribution of these
dots cannot be controlled accurately which makes addressing
individual quantum dots difficult. Furthermore, the inherent
size distribution of the precipitated nanoclusters leads to in-
homogeneous broadening and artificially widened emission
spectra.
7
Recent work
9
has shown a remarkable effort into the
fabrication and characterization of single etched quantum
dots. These top-down devices have demonstrated narrow
linewidths and ‘blinking’ behavior associated with emission
from single nanocrystals.
9
–
11
In this letter we demonstrate the
ability to fabricate uniform arrays of stacked pillars of silicon
quantum dots whose size
and thus peak emission wave-
length
can be precisely tuned via etching parameters and
oxidation conditions. These vertical quantum dots exhibit
bright photoluminescence
PL
from the visible to the near
infrared and can be predictably and repeatably placed
through lithographic techniques.
Aluminum oxide disks to mask vertical quantum dots
were patterned following previous techniques.
3
The samples
were etched using a “pseudo-Bosch” recipe using SF
6
and
C
4
F
8
to simultaneously etch and passivate, respectively. Un-
der a fixed gas ratio and forward power the etch produces
pillars with vertical sidewalls at an etch rate of roughly 250
nm/min;
3
an example of this etch can been seen in frame
b
of Fig.
1
. By tuning the gas ratio as the etch progresses into
the silicon we controllably undercut the etch mask to gener-
ate structures with arbitrarily corrugated features as small as
30 nm; an example of this etch technique can been seen in
a
Electronic mail: walavalk@caltech.edu.
FIG. 1.
a
SEM image of an array of corrugated silicon nanopillars imme-
diately after etching. These pillars were fabricated by oscillating the etching
conditions to controllably undercut and overpassivate the silicon.
b
An
array of nanopillars with vertical sidewalls etched using the same pseudo-
Bosch recipe but without varying the ratio of etch to passivation gas.
APPLIED PHYSICS LETTERS
98
, 153114
2011
0003-6951/2011/98
15
/153114/3/$30.00
© 2011 American Institute of Physics
98
, 153114-1
Downloaded 29 Apr 2011 to 131.215.220.185. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions
Fig.
1
a
. In order to prevent additional undercut as the etch
continues we increased the forward power to better collimate
the ions.
The second step in the fabrication is the self-terminating
oxidation. It has been shown
2
,
3
,
12
that when convex silicon
structures are oxidized at temperatures below 950 °C, the
oxidation self-terminates predictably leaving a silicon core
encapsulated by silicon dioxide. The oxidation process ends
when the compressive strain at the Si–SiO
2
interface pre-
vents oxygen diffusion to the oxidation front. By selecting
the oxidation temperature as well as the initial dimensions of
the convex silicon structure the size of the remaining silicon
core can be tuned.
3
,
11
Samples were fabricated with original mask diameters of
80 and 100 nm and were undercut to produce three uniform
corrugations per pillar. The corrugation size was varied be-
tween samples. The samples were simultaneously oxidized at
915 °C for seven hours and were cooled to room tempera-
ture in a forming gas of 20:1 nitrogen and hydrogen. The
presence of an elliptical silicon quantum dot was confirmed
nondestructively by reflection mode transmission electron
microscopy
TEM
Ref.
3
as well as destructively after PL
testing by transferring the quantum dots onto a TEM grid and
viewing the structures in transmission mode.
Figure
2
shows TEM images of vertical quantum dots
after oxidation. The first image shows the morphology of the
surface after oxidation as well as a bright-field image of the
silicon core in the head and first quantum dot of the pillar.
The inset shows a similar view in dark-field with diffraction
contrast used to highlight the single crystal nature of the
remnant quantum dot. Frames
b
and
c
of Fig.
2
show the
presence of crystalline silicon within the oxidized cores of
the legs of the corrugated nanopillars.
The three samples tested are shown before oxidation as
colored frames in Fig.
3
, the diameters are roughly
a
30 nm
black
,
b
37 nm
blue
, and
c
45 nm
green
and the
corrugations have a period of approximately 60 nm for each
sample. Micro-PL was performed in an inverted optical mi-
croscope setup with a 457 nm free-space argon ion laser used
to pump the sample; the full experimental setup is described
in previous work.
3
The data collected from three samples of
different preoxidation size is shown in Fig.
3
. The color of
the frame around the scanning electron microscope
SEM
images corresponds to the curve plotted in the figure. It can
be seen that there is a correlation between the original size of
the etched corrugations and the peak emission wavelength of
the oxidized quantum dot. Peak emission was found to be at
roughly 600 nm
2.06 eV
, 640 nm
1.94 eV
, and 810 nm
1.53 eV
, for the samples with initial diameters of 30 nm
black
,37nm
blue
, and 45 nm
green
, respectively, with
a full width half maximum of 150 meV or less for each
sample.
Although careful effort was made to produce quantum
dots with diameters that had as narrow a size distribution as
possible, the peak emission wavelength is a strong function
of dot size
8
,
13
and even a change in diameter of 0.25 nm
about one monolayer of Si
can shift the peak emission
energy up to 100 meV
30 nm
.
8
Several causes, including
noncircular mask patterning, debris on the wafer, and local
etch variation could have such an impact on the peak emis-
sion wavelength. Furthermore, it has been shown
2
,
3
that the
strain incorporated into thermally oxidized silicon nanostruc-
tures can have a significant impact on the bandgap, shifting
the peak emission energy by 200 meV between 1% compres-
sive and tensile strain. The assumption made in this work is
that the three stacked quantum dots have the same size and
strain conditions; however this is not necessarily true, espe-
cially when it is considered that the top and bottom quantum
dots are capped with a head and tail, while the central quan-
tum dot is bracketed by two other dots. These two causes can
possibly account for the multiply-peaked structure of the
three measured emission spectra. This explanation is consis-
tent with the observations that the individual peaks, which
make up the total curves, have widths that fall within or
close to the previously measured
10
75–100 meV
20–30 nm
linewidths of room temperature, etched quantum dots.
The sharp peaks seen in the blue curve in Fig.
3
can be
attributed to such a cause or to a similar effect combined
with a limitation of the experiment. Converting the number
of fabricated pillars per pad into an areal density yields a
value of
5
10
9
cm
−2
; roughly three orders of magnitude
less dense than coalesced nanoparticles.
4
In our case, the
50
objective we used allowed sampling a 5
5
m
2
area
corresponding to between 300 and 500 pillars, leading to a
FIG. 2.
a
Bright-field image of the head and first quantum dot of a corru-
gated pillar after oxidation. Inset shows a similar picture with diffraction
contrast to highlight the crystalline nature of the remaining silicon nanocrys-
tals.
b
and
c
These frames utilize diffraction contrast to highlight the
remaining quantum dots in the legs of the corrugated pillars after self-
terminating oxidation. Scale bars are 50 nm.
FIG. 3.
Color online
PL spectra of three samples of etched and oxidized
quantum dots with different initial corrugation diameters. The leftmost curve
centered at 600 nm
corresponds to pillars in frame
a
, the middle curve
centered at 640 nm
corresponds to the pillars in frame
b
, and the right-
most curve
centered at 810 nm
corresponds to the pillars in frame
c
.The
preoxidation size is 30 nm, 37 nm, and 45 nm for the
a
black,
b
blue, and
c
green samples, respectively. Note that the larger the preoxidation size of
the corrugated pillars the longer the peak emission wavelength. Scale bars
are 200 nm in each frame.
153114-2
Walavalkar
etal.
Appl. Phys. Lett.
98
, 153114
2011
Downloaded 29 Apr 2011 to 131.215.220.185. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions
relatively discrete spectrum where each pillar could signifi-
cantly contribute to the total signal. The relatively low col-
lected signal from the few pillars in combination with the
preferential scattering of the light from certain quantum dot
pillars could give those pillars an unequal contribution to the
signal; resulting in such peaks. Future work is needed to
quantitatively explain these spectral details; performing mea-
surements at cryogenic temperatures could narrow the line-
widths and allow the emission from individual quantum dots
to be assessed. Similar effects, resulting in multiply and
sharply-peaked spectra, have been seen in previous work.
6
,
9
From the dark and light-field TEM images it is possible
to estimate the size of the quantum dots in the oxidized pil-
lar; however the distortion found in each image due to the
image being taken through about 50 to 75 nm of silicon
dioxide makes the bounds of error too large for a meaningful
size measurement. Instead, dot sizes were determined by
comparing the peak emission energy with the band-gaps ob-
tained by previous theoretical and experimental work.
8
,
14
It is
also critical to note that these dots are embedded in an oxide
matrix which has been found
13
,
14
to redshift the peak emis-
sion by almost 1 eV compared to bare quantum dots or those
with a hydrogen terminated surface. Based on data presented
in Refs.
8
,
10
,
13
, and
14
we estimate that the measured
nanocrystal sizes are centered around 2 nm
black
, 2.4 nm
blue
,and5nm
green
in diameter.
The ability to fabricate silicon quantum dots with pre-
dictable sizes could prove to be useful when attempting to
incorporate them with existing silicon photonic structures
such as waveguides or photonic crystals. Instead of using a
stochastic distribution of nanocrystals, these can be placed
lithographically to coincide with the peak of the optical
mode. The ability to tailor the peak emission of the quantum
dots allows one to select the emission wavelength to suit a
task or create a vertical stack of different dot sizes to allow
for broad spectral emission. Furthermore, there has been in-
terest in producing transistors with quantum dots that dem-
onstrate quantum effects and Coulomb blockade at noncryo-
genic temperature. To retain quantum behavior at room
temperature the quantum dot must be aligned between two
electrodes and smaller than 5 nm in diameter.
15
By turning
the fabrication vertical, structures similar to those described
in this paper could overcome the challenge of patterning
gates with difficult, lateral electron beam lithography.
In conclusion, we have presented a CMOS compatible
method to fabricate narrow band, luminescent silicon quan-
tum dots. By controlling the size during lithography and
etching, it is possible to utilize the self-terminating nature of
convex silicon oxidation to predictably tune the peak emis-
sion wavelength. We have also demonstrated the ability to
stack quantum dots of various predictable sizes in order to
tailor the spectral behavior of these corrugated pillars. These
devices and behaviors may have important applications in
both future nanophotonic and nanoelectronics devices.
S. Walavalkar would like to thank Erika Garcia as well
as Professor Tom Tombrello and Ryan Briggs for useful dis-
cussion. We would also like to gratefully acknowledge the
Boeing corporation under the CT-BA-GTA-1 grant, the Ad-
vanced Energy Consortium under the BEG10-07 grant, and
DARPA for generous support under the NACHOS
Grant
No. W911NF-07-1-0277
program. A. Homyk would like to
thank the ARCS foundation for their support. M. D. Henry
would like to thank the John and Fannie Hertz Foundation
for their funding.
1
K. D. Hirschman, Proceedings of the UGIM Symposium Microelectronics
Education for the Future Twelfth Biennial University/Government/
Industry Microelectronics Symposium, 1997, Vol. 177.
2
A. R. Guichard, D. N. Barsic, S. Sharma, T. I. Kamins, and M. L. Brong-
ersma,
Nano Lett.
6
, 2140
2006
.
3
S. S. Walavalkar, C. E. Hofmann, A. P. Homyk, M. D. Henry, H. A.
Atwater, and A. Scherer, Nano Lett.
10
, 4423
2010
.
4
X. Wen, L. V. Dao, and P. Hannaford,
J. Phys. D: Appl. Phys.
40
, 3573
2007
.
5
R. J. Walters, G. I. Bourianoff, and H. A. Atwater,
Nature
London
4
,
143
2005
.
6
T.-Y. Kim, N.-M. Park, K.-H. Kim, G. Y. Sung, Y.-W. Ok, T.-Y. Seong,
and C.-J. Choi,
Appl. Phys. Lett.
85
, 5355
2004
.
7
L. Pavesi, L. Dal Negro, C. Mazzoleni, G. Franzò, and F. Priolo,
Nature
London
408
, 440
2000
.
8
M. Wolkin, J. Jorne, P. Fauchet, G. Allan, and C. Delerue,
Phys. Rev. Lett.
82
, 197
1999
.
9
J. Valenta, R. Juhasz, and J. Linnros,
Appl. Phys. Lett.
80
, 1070
2002
.
10
I. Sychugov, R. Juhasz, J. Valenta, and J. Linnros,
Phys. Rev. Lett.
94
,
087405
2005
.
11
B. Bruhn, F. Sangghaleh, and J. Linnros, Phys. Status Solidi A
208
,631
2010
.
12
H. I. Liu, D. K. Biegelsen, F. A. Ponce, N. M. Johnson, and R. F. W.
Pease,
Appl. Phys. Lett.
64
, 1383
1994
.
13
J. Wilcoxon, G. Samara, and P. Provencio,
Phys. Rev. B
60
, 2704
1999
.
14
C. Garoufalis, A. Zdetsis, and S. Grimme,
Phys. Rev. Lett.
87
, 276402
2001
.
15
E. Leobandung, L. Guo, Y. Wang, and S. Y. Chou,
Appl. Phys. Lett.
67
,
938
1995
.
153114-3
Walavalkar
etal.
Appl. Phys. Lett.
98
, 153114
2011
Downloaded 29 Apr 2011 to 131.215.220.185. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions