RARE-EARTH DOPED SILICON-RICH SILICA: EVIDENCE FOR

ENERGY TRANSFER BETWEEN SILICON MICROCLUSTERS AND

RARE-EARTH IONS

A.J. KENYON, P.F. TRWOGA, M. FEDERIGHI and C.W. PITT

Department of Electronic and Electrical Engineering,

University College London, Torrington Place, London, WC IE 7JE

Abstract
We report the fabrication of rare-earth doped silicon-rich silica thin films by PECVD. The films exhibit
absorption edges in the visible region of the optical spectrum consistent with the presence of silicon microclusters.
Weak visible photoluminescence due to silicon microclusters is observed. In addition, strong luminescence from the
rare-earth ion is obtained even when excited away from characteristic absorption bands; indeed, the luminescence
intensity is largely independent of excitation wavelength below 514 nm. We ascribe this to excitation of silicon
microclusters followed by an efficient transfer of energy to the rare-earth ions.
The very broad absorption of this material opens up the possibility for flashlamp-pumped optoelectronic devices.
In addition, we report the fabrication of silicon-rich silica films by PECVD. We show that the optical properties of
these films are consistent with the presence of silicon microclusters and show absorption spectra similar to those of
the rare-earth doped silicon-rich silica samples. This supports the hypothesis that the principal absorbing species in
the rare-earth doped films is microclustered silicon.

Introduction
There is currently a great deal of interest in the optical properties of microcrystalline, amorphous
and porous silicon [1-5], principally due to the technological implications for fabrication of visible
light emitting devices. Although much of the work in this area concentrates on the generation of
light by quantum size structures on silicon substrates (quantum 'dots' or 'wires'), there has
recently been interest in the optical properties of silicon clusters dispersed in a silica matrix [ 1,2,61.
A number of groups have reported light emission from silicon-rich silica (SiOx, x<2). [1,2,61 A
variety of different approaches have been taken to producing this material which consists of
aggregates of silicon held in a silica matrix. These can be produced by implanting silica with excess
silicon (or conversely silicon with excess oxygen) [61, or by employing a chemical vapour
deposition (CVD) process to grow thin films of silicon rich silica [7,8]. As is the case for porous
silicon, there is strong evidence that the luminescence properties of this material (primarily the peak
emission wavelength) are largely determined by the size of the silicon aggregates or crystallites [5].
Both implantation and CVD allow good control over stoichiometry, and it is possible to tailor both
the absorption band-edge and emission wavelength of filmsý. Recent results have demonstrated
broad-band light emission from the blue to near infra-red regions of the spectrum from silicon
aggregates of varying sizes [6,8,111. The silica host has the advantage over porous silicon of
greater stability, strength and superior optical properties. There is some controversy over the
mechanism of light emission in microclustered silicon, but there is now an appreciable body of
evidence to suggest that quantum confinement of electrons within nanometre-size silicon
aggregates produces a modification of the band-gap of the material with a consequent shift in the
absorption band-edge and emission wavelength [6,7].
To date there have been only a limited number of studies of Si rich SiO 2, but this does appear to
be an area of growing interest. Previous work has employed ion implantation and RF sputtering to
incorporate the silicon clusters. Plasma enhanced chemical vapour deposition (PECVD) is an
alternative low temperature technique which is capable of accurately controlling the stoichiometry
and growth conditions of the film. This allows high quality silica to be grown easily with a well

117
Mat. Res. Soc. Symp. Proc. Vol. 358 0 1995 Materials Research Society
defined silicon content.
In this paper we report the production of microclusters of silicon within silica thin films by
PECVD. The effects on the optical properties of the samples of varying the ratio of silicon to
oxygen are investigated. We demonstrate a link between this ratio and both the absorption and
emission spectra. Other films have been grown which contain a rare-earth dopant along with
silicon aggregates. These films contain around 1 at % erbium, the purpose being to investigate the
effect of introducing a technologically useful optically active species into the strongly blue-
absorbing Si/Si0 2 matrix. We observe room temperature fluorescence due to the 41 13 /2 -- 41 15 /2
transition of the Er3 + ion even when the films are pumped at wavelengths away from characteristic
erbium absorption bands. We also observe weak room temperature visible photoluminescence
which we ascribe to the presence of silicon microclusters.

Experimental

The films studied were produced by PECVD of silicon-rich silica on both silica and silicon
substrates. Silane (Sill4 ) was employed as the silicon precursor and nitrous oxide (N2 0) as the
oxygen source. The silane was diluted to 5 % in argon. The N 20 mass-flow rate was fixed at
10 sccm while that of the SiH 4 /Ar mixture was varied from run to run in the range 240 sccm to
140 sccm. In all cases the RF power used to crack the precursors was 10 W at 13.56 MHz, and
the chamber pressure and substrate temperature were 100 mTorr and 125 °C respectively. The
silicon/oxygen ratio in the silicon-rich silica samples was measured by electron-probe
microanalysis (EPMA). Subsequent annealing of the samples was carried out at 900 TC in a
nitrogen atmosphere for 30 minutes.
The erbium doped samples were produced under similar conditions, with the addition of
Er(tetramethyl heptane dione) 3 into the reaction chamber as the rare-earth precursor. The technique
has been more fully described previously [9]. For all the samples studied, film thickness was
approximately 1-3 [tm, and in the case of the erbium-doped samples, the erbium concentration
was 1 at % as measured by secondary ion mass spectrometry (SIMS). Three rare-earth doped
samples were studied: one as-grown, one annealed at 900 'C for 90 minutes and the third
annealed at 1200 °C for 90 minutes. Again, annealing was carried out in a nitrogen atmosphere.
The experimental arrangement for measuring fluorescence spectra was as follows. A Coherent
Innova 100 argon-ion laser provided pump radiation at any one of four discrete wavelengths in the
blue/green region of the optical spectrum: 457 nm, 476 nm, 488 nm and 514.5 nm. The
maximum power available varied from 450 mW at 457 nm to just over 3 W at 514.5 nm, and the
beam diameter at the sample was around 4 mm. A prism placed just after the laser source served to
filter residual plasma lines. The sample was positioned such that the thin film faced the laser at
Brewster's angle to the laser beam. A collection lens focused the fluorescence through the slits of
the monochromator (Bentham M300). Detection was by red-sensitive photomultiplier tube,
InGaAs photodiode or silicon photodiode connected to a current pre-amplifier (Bentham 286) and
lock-in amplifier (Brookdeal 9503-SC). Both the detection system and monochromator were
computer controlled.
For measurement of absorption spectra, the thin film was removed from the silicon substrate by
very rapid cooling which cracked the film and placed onto a glass blank. Spectra were collected in
transmission, perpendicular to the substrate. A white light source (Bentham ILIA) was focused
onto the sample, and collection of transmitted light remained as before. Two spectra were taken for
each sample: one of a plain glass blank as a reference, and one of the sample itself. These were
combined to produce the absorption spectrum.

Results
Figure 1 illustrates the silicon/oxygen ratio in the as-grown silicon-rich silica films as a function

118
of the gasannealing,
After flow ratiothe
SiH4/N20 as measured
silicon-rich silica filmsusing EPMIA.
exhibited colours, varying from yellow to very deep
red. Figure 2 illustrates the variation in absorption spectra for films with different SiH4/N20 gas
flow ratios. Note that thin-film interference bands are superimposed on the spectra.
Photoluminescence spectra of as-grown silicon-rich silica films are shown in Figure 3. Again,
note the thin-film interference bands superimposed on the spectra. In all cases, excitation was
using 457 nm light.

1.2-
- 1.0 I
.- 0.8-
.E!. ().6-
S0 .4r
0.2 • i • , • , • , ' , '
1214 16 18 20 22 24 26
silane/nitrous oxide gas flow ratio
Figure 1: film stoichiometry as a
function of gas flow ratio
during deposition

1.0 SiH4/N20 flow ratio

0.8 ,- ¢1"
-, - -- - 16:1
'•0.6-••. " S•"".............
""..1
",,.•. 1 :
:
"•0 .4 ....
.,..' . . . . 2 4 :1
2 0.2- 14:
- "' ... ". .. ,.
..... 14:1
0.0 iL.. ........ "•

475 575 675

wavelength
Figure 2: absorption spectra of annealed
SiOx films

80- A SiH4/N20O ratio

•"60-" 22:1
......... 20:1
E 2.•20 - •24:1

550 600 650 700 750 800

wavelength nm
Figure 3: Photoluminescence spectra
of as-grown SiOx films
Figure 4 shows an absorption spectrum of the erbium-doped silicon-rich silica sample annealed

119
at 900 TC. Note the strong band edge in the blue/green and the lack of observable Er3 ÷ absorption
features. These are presumably hidden by the strong silicon cluster absorption band.
Figure 5 shows a visible photoluminescence spectrum of the same sample pumped at 457 nm.
There is a weak broad band centred around 550 nm. This is not a wavelength at which Er3 +
fluorescence would be expected in a silica matrix.
Figure 6 shows near-infra-red photoluminescence characteristic of the 4113/2--41 15/2 transition
of the Er 3÷ ion. This spectrum was obtained from the Er-doped sample under excitation with
476 nm light.
Figure 7 shows a plot of integrated fluorescence intensity of the 1535 nm band of the same
sample as a function of pump wavelength normalised to the intensity of 514.5 nm excitation (solid
line). At each wavelength the pump laser power was held at 200 mW in order to provide a true

6 1.2
1.0.

0
.0
4-
21

L---,-
0- t--- I
2
a
C

0.
0.8.
0.6.
0.4-
0.2-
n 1)

4f5O 650 850 1050 1250 1450 475 525 575 625 675
wavelength (nm) wavelength (nm)
Figure 4: absorption spectrum of Figure 5: visible photoluminescence
Er-doped SiOx film from Er-doped SiOx film

150 120-
100.
.• 100
80-
C S

60. r
50 40 4
20 -
0
o
z
440 450 460 470 480 490 500 510 520 530
1400 1450 1500 1550 1600 1650 1700
Wavelength (nm) Excitation wavelength (nm)
Figure 7: PL vs excitation wavelength
Figure 6: erbium fluorescence from solid line: annealed Er-doped SiOx film
Er-doped SiOx film
dotted line: Er-doped fibre

comparison. Also shown is the corresponding plot for a reference erbium-doped silica fibre (dotted
line). The positions of erbium absorption bands are indicated by the vertical lines. The lengths of
these lines correspond to the approximate absorption strengths given in [ 121.
SIMS analysis of the unannealed erbium-doped film shows the composition to be (in atom %)
oxygen 56 %, oxygenated silicon 22 %, elemental silicon 14 %, nitrogen 5 %, carbon 2 %,
erbium 1 %
XRD analysis of the films showed, in all cases, a uniform background. This may be taken as

120
implying that the silicon aggregates are amorphous. However, it is possible that the instrument
used is not sufficiently sensitive to measure films of this thickness. Further measurements are
currently being undertaken.

Discussion
This study has demonstrated that altering the SiH 4/N20 gas flow ratio changes the stoichiometry
of the SiO, films and that this leads to significant changes in the optical properties of the material.
In particular, it appears that the absorption band edge of the films can be shifted by around 100nm
through careful selection of the gas flow ratio. In agreement with other studies on both amorphous
silicon films and SiOx, visible photoluminescence has been observed. The peak position of this can
once again be changed by varying the gas flow ratio during deposition.
XRD measurements of the structural properties of the as-grown SiOx films suggest that the
excess silicon is present in an amorphous phase, but this remains to be confirmed by more accurate
measurements. Further measurements are under way on the annealed films.
The absorption spectra of the erbium-doped SiOx films exhibit strong absorption band edges in
the visible region of the spectrum similar to those of the non-Er doped samples. Moreover, it
appears that the characteristic absorption bands of the rare earth ion at 450 nm, 490 nm, 520 nm,
635 nm and 980 nm are very much weaker than the absorption due to the silicon aggregates. In
addition, the photoluminescence spectrum of the erbium-doped sample exhibits weak emission
around 550 nm. It can reasonably be assumed that this is not be due to the presence of erbium.
Although visible luminescence in this region has been noted due to transitions from the Er 4S3/2
level in low phonon glasses when pumped at 1480 nm via an upconversion process[10], to our
knowledge it has never been observed in silica glasses. However, broad-band emission in this
region has been seen from amorphous silicon and SiOx [11].
PECVD, being a low-temperature process, is a likely candidate for production of Si
microclusters when used to grow Si0 2 films in an oxygen-deficient environment. The thermal
mobility of the ions on the substrate will be sufficiently low so as to inhibit diffusion of species
within the film, and hence Si clusters formed in the deposition process will persist. Preliminary
results from another group working on PECVD growth of Si microclusters in Si0 2 support our
conclusions [11].
Fluorescence spectra of the rare-earth doped samples exhibit a strong band around 1535 nm
characteristic of the 4113/2 -- 4115/2
I Er3+ transition. The erbium present in the films is thus optically
active, although we have not been able to observe any appreciable absorption due to this species.
Moreover, it appears that it is possible to obtain fluorescence from Er3+ even when the pump
wavelength is away from its characteristic absorption bands. Even more striking is the observation
that the intensity of fluorescence remains largely unchanged with decreasing pump wavelength.
Figure 7 illustrates the normalised integrated fluorescence intensity as a function of excitation
wavelength for the erbium-doped SiOx sample and a reference erbium doped silica fibre. In the
case of the fibre, the fluorescence intensity shows a good correlation with the absorption spectrum
of erbium in silica [121. This is in marked contrast to the silicon-rich sample which exhibits
fluorescence at 1535 nm that is largely independent of excitation wavelength.
The presence of Si microclusters in all samples can be inferred from the luminescence data and
from the band-edge effect in the absorption spectra. Further studies are under way to further
characterise these aggregates by XRD and TEM lattice imaging. The marked difference between the
fluorescence in our erbium-doped samples and that from the fibre cannot be explained by a change
in the erbium absorption or emission cross-sections, for which there is no supporting evidence in
the literature. However, if bands due to silicon microclusters and excited levels of the erbium ion
overlap, it is possible that photons with energies to the blue of the absorption edge of the films are
absorbed by the silicon aggregates and the energy transferred non-radiatively to the excited states
of the erbium. The precise transfer mechanism is difficult to identify and is probably a combination
of the mechanisms proposed by [13] as an explanation for ytterbium electroluminescence in indium
phosphide. Supporting evidence comes from a recent paper by Kimura et al who investigated
electrochemical doping of porous silicon with erbium [141. They reported a strong correlation

121
between the intensity of erbium photoluminescence at 1.541im and the peak wavelength of the
photoluminescence of the porous silicon host. This suggests a mechanism involving energy
transfer from the host porous silicon to the rare-earth ion.
In conclusion, we have produced by PECVD silicon-rich silica thin films that exhibit optical
properties strikingly similar to those observed in Si÷ implanted or sputtered silicon-rich silica films.
Both absorption and photoluminescence spectra strongly suggest the presence of silicon
aggregates. In addition, doping the films with erbium produces samples which exhibit
characteristic 1535 nm fluorescence. This is true even when the sample is excited at wavelengths
away from erbium absorption bands. Indeed, the fluorescence yield appears to be largely
independent of the pump wavelength over the range studied. This contrasts with data presented for
an erbium doped silica fibre and suggests some form of excitation transfer from the silicon
aggregates to the rare earth ion. We are currently investigating further the properties and structure
of Si-rich Si0 2 , the nature of the transfer process and the possibility of substituting other ions into
the SiOx matrix in order to obtain luminescence at other technologically important wavelengths.
Acknowledgements
We gratefully acknowledge support received from the Optoelectronics Interdisciplinary Research
Centre during this work. Thanks go to Martin Vickers of Birkbeck College for the XRD
measurements
References
1. T. Ito, T. Ohta and A. Hiraki, Japan. J. Appl. Phys. 3 1, LI, (1992)
2. S. Hayashi, T. Nagareda, Y. Kanzawa and K. Yamamoto, Japan. J. Appl. Phys. 3 2, 3840,
(1993)
3. Y. Kancmitsu, K. Suzuki, H. Uto, Y. Masumoto, K. Higuchi, S. Kyushin and H.
Matsumoto, Japan. J. Appl. Phys. 32, 408, (1993)
4. 1. Sagnes, A. Halimaoui, G. Vincent and P.A. Badoz, Appl. Phys. Lett. 6 2, 1155, (1993)
5. M. Koos and 1.Pocsik, Appl. Phys. Lett. 62, 1797, (1993)
6. T. Shimizu-Iwayama, M. Ohshima, T. Niimi, S. Nakao, T. Fujita and N. Itoh, J. Phys.:
Condens. Matter, 5, 375, (1993)
7. X. Liu, X. Wu, X. Bao, Y. He, Appl. Phys. Lett., 64, 220, (1994)
8. A.J. Kenyon, P.F. Trwoga, M. Federighi, C.W. Pitt, J. Phys.: Condens. Matter, 6, L319,
(1994)
9. 1. Massarek, P.F. Trwoga, A.J. Kenyon, M. Federighi and C.W. Pitt, IEEE Phot. Tech.
Lett. in press
10. C.A Millar and P.W. France, Electron. Lett. 2 6, 634, (1990)
11. F. Timofeev, Private communication.
12. W.J. Miniscalco, J. Lightwave Tech. 9, 234, (1991)
13. K. Takahei, H.Tagaguchi, K. Nakagome, I. Uwa and P.S. Whitney, J. Appl. Phys. 6 6,
4941, (1989)
14. T. Kimura, A.Yokoi, H. Horiguchi, R. Saito, T. Ikoma, A. Sato, Appl. Phys. Lett. 65,
983, (1994)

122