SlideShare a Scribd company logo
1 of 22
Download to read offline
Home Search Collections Journals About Contact us My IOPscience 
Porous silicon 
This content has been downloaded from IOPscience. Please scroll down to see the full text. 
View the table of contents for this issue, or go to the journal homepage for more 
Download details: 
IP Address: 148.228.88.57 
This content was downloaded on 23/09/2014 at 15:46 
Please note that terms and conditions apply.
Semicond. Sci. Technoi. 10 (1995) 1187-1207. Printed in the UK 
~ ~ ~ 
TOPICAL REVIEW 
Porous silicon 
B Hamilton 
Department of 
M60 IQD, UK 
Physics, UMIST, PO Box 88, Sackville Street, Manchester 
Received 5 January 1995, accepted for publication 17 March 1995 
Abstract. This paper attempts to review the field of research into light emission 
from porous silicon. The driving force behind such research is the tantalizing goal 
of adding optoelectronic functions to the already impressive array of electronic 
functions provided by silicon-based devices. A silicon technology with included light 
emission would move even closer to complete dominance of the electronics 
market. After several years of research effort. the fundamental mechanisms of light 
emission are still not completely resolved. This is not surprising: porous silicon has 
many attributes of a new and complex material, and its study requires a truly 
interdisciplinary effort involving electrochemistry, surface science, structural and 
chemical microscopy on the atomic scale and detailed optical spectroscopy. This 
paper tries to connect these various threads; inevitably what emerges will only 
serve as a rather selective 'snapshot' of a still developing and often perplexing field. 
1. Introduction 
The dominance of silicon in the electronics industry 
is almost complete, at least in terms of volume: the 
worldwide market for silicon-based devices and systems 
depending on them is huge. The comparatively small 
but important markets which silicon does not^ fulfil are 
those of ultra high-speed devices and optoelectronics, in 
particular optical communications. In fact virtually all 
optoelectronic functions requiring high-speed modulation 
rely on compound semiconductor devices; fibre-optic-based 
optical communications systems rest firmly on InP-based 
lasers and modulators, whereas GaAs-based devices supply 
the near-infrared and visible emission required for short-range 
communication and disc redwrite functions. 
It is a curious fact that although silicon is the 
material which essentially fed the information technology 
revolution, much of the highly successful international 
research effort into semiconductor physics during the past 
15 years has been devoted to II-V semiconductors. The 
search for novel physical phenomena based on reduced 
dimensionality-superlattice, quantum well and latterly 
quantum wire and dot structures-has been a major driving 
force for condensed matter physics. Improved device 
functionality has emerged both in the fields of high-speed 
transport and optics, which have strengthened the III-V 
industry in these areas. 
A picture emerges of a silicon industry dealing with 
a rather mature technology, able to fulfil many of the 
growing demands of an information-dependent culture. 
Materials-based research which will underpin a future 
silicon indushy currently centres around silicon-germanium 
heterojunction devices, novel configurations for reduced 
power consumption in portable systems, cheap thin film 
0268-1242/95/091187+21519.50 6 1995 IOP Publishing Ltd 
devices and nanoscale fabrication. Of course the latter 
topic holds out the possibility of novel functionality which 
exploits the quantum regime of electron behaviour, and so 
connects with some of the work reviewed here. Porous 
silicon burst into this arena several years ago, offering at 
least a possibility that silicon technology might eventually 
yield light-emitting devices. The fact is that under optical 
excitation. porous silicon does produce light with high 
efficiency, and furthermore with an emission spectrum 
which can be 'tuned' from the near-infrared to the green 
by varying porosity. Further processing by rapid oxidation 
extends the emission into the bludviolet region of the 
spectrum. 
Clearly then, one driving force behind research into 
light emission from porous silicon is the hope that having 
finally understood the basic mechanisms, it might be 
possible to make an electrically excited LED or laser which 
could, ultimately, be integrated into a complex chip. This 
notion should not be seen as one of simply enlarging the 
functionality of silicon; discoveries of new functionality 
traditionally end up finding applications in semiconductor 
technology. In their turn, both the rapidly oxidizing 
silicon surface and the semiconductor laser respectively 
were dismissed either as a nuisance which would prevent 
development or as lacking in applications! It would 
be wrong therefore to rule out, say, optical interconnect 
applications for porous silicon, provided that it could 
be developed into a stable electroluminescent system, 
compatible with integrated circuit processing. 
In order to even contemplate real applications for 
porous silicon devices we must understand the basic 
radiative processes and must have a clear view of how 
to optimize the porous skeleton and how to control and 
perhaps to take advantage of its enormous surface area. 
1187
B Hamilton 
Finally we must learn how to electrically excite the 
luminescence. In the final analysis it may prove impossible 
to achieve these goals, or some other less complex form of 
optically functional silicon may emerge. Many issues are 
currently being pursued, including modification of porous 
silicon and new ways to process material and contacts, and 
these will be touched upon below. However, the central 
issue remains the origin of the light, and its relationship 
to the atomic-scale structure and the associated electronic 
structure of the porous layer. The main aim of this review, 
then, is to try to draw together the threads of evidence which 
are guiding workers in the field towards understanding the 
physics of the material. The story to date, although largely 
qualitative, is complicated and has generated lively debate. 
This being so, it is politic to simply state at the outset the 
four most commonly held views on the origin of the visible 
luminescence: 
(i) The visible and near-infrared light is the result of of 
quantum confinement shifts of the silicon energy gap due 
to particle localization in nanometre scale structures (wire 
or dot) which make up the porous skeleton. 
(ii) The luminescence originates from surface molecular 
species which coat the porous skeleton, and which result 
from the electrochemical processing. 
(iii) The light originates from radiative decay at 
surfacdinterface states, the character of which are partly 
determined by nanocrystaltine particles within the porous 
layer. 
(iv) Hydrogenated amorphous silicon is a product of 
the invasive electrochemistry and is responsible for the 
emission. 
In attempting to revicw the field, it is necessary to 
subdivide the information; first, in section 2, an overview 
of the main optical phenomena is presented. Section 3 
thcn deals, in a.simple way, with the electrochemical 
process involved in pore formation, leading on to a brief 
review of pore morphology and microstructure. In section 
4 some of the issues involved with the surface of porous 
silicon are discussed in order to provide a firmer basis 
for the review of the debate surrounding luminescence 
mechanisms in section 5. Section 6 deals with some issues 
concerning oxidized porous silicon which shed light on 
some fundamental aspects of the material. Finally section 
I outlines some attempts to make simple device stmctures 
and also the considerable problems involved. 
2. An overview of the optical phenomena 
associated with porous silicon 
Highly porous Si, processed using electrochemical etching 
methods, exhibits strong photoluminescence; efficiencies 
of several per cent have been routinely reported. Spectra 
are broad, but peak wavelengths can be 'tuned' over a 
wide range in the near-infrared and visible, by varying 
the porosity. These facts, which were first noted by 
Canham in 1990 [l], remain the key points underpinning 
a wide-ranging and interdisciplinary research effort. It is 
interesting to note though that the observation that ultra 
small silicon crystallites, passivated by hydrogen, and with 
1188 
Wavevector 
Figure 1. The energy band structure of crystalline silicon. 
The indirect energy gap leads to slow band to band 
radiative decay transitions which require the participation of 
momentum-conserving phonons. 
emission wavelengths which depend on size, pre-dates the 
first report of porous silicon luminescence [2]. 
A natural starting point for a review of porous silicon 
is a comparison with the optical emission properties 
associated with crystalline silicon. The energy band 
structure of any semiconductor dictates many of the 
observed luminescence properties. Silicon has an indirect 
enera gap, shown in figure 1. A well known consequence 
of this that the radiative efficiency of Si is low at room 
temperature. The indirect gap dictates that electron-hole 
recombination across the gap requires the involvement of 
momentum-conserving phonons; the matrix element for the 
transition is thus small. Note that this is not a fundamental 
limit to the radiative efficiency, it simply results in a 
long radiative lifetime; the calculated radiative lifetime of 
moderately doped silicon at room temperature is in the 
millisecond regime. With such a slow radiative decay 
process, injected carriers inevitably recombine through non-radiative 
shunt paths, and the net recombination lifetime 
(though very sample dependent) is orders of magnitude 
shorter than the radiative lifetime. However, if all 
competing shunt paths like deep electron states or surfaces 
did not exist, then silicon would be a perfect emitter at 
close to 1 pm. Unfortunately this remains a hypothetical 
case, though data exist which demonstrate clearly that 
effective removal of the surface shunt path by hydride 
passivation results in long minority carrier decay times and 
large increases in radiative efficiency [3]. 
At low temperatures, Si becomes more optically active. 
This is principally because certain optical decay channels 
become thermally stabilized. For example, the free exciton 
population under injection conditions grows, and more 
importantly shallow impurities or defects can stably bind 
excitons. The initial capture event for an exciton into an 
impurity state may be fast; if the exciton can remain trapped 
for a sufficient time, i.e. is not thermally ionized from the
Porous silicon 
1090 1110 113 1150 
2 
Photon energy h"el 
Figure 2. An example of a low-temperature photoluminescence spectrum 
of high-quality p-type bulk clystalline silicon. The sharp structure is due 
to the decay of excitons which are stably bound to the boron impurities at 
low temperatures 
impurity potential, decay may occur with a matrix element 
determined partially by the impurity. Radiative decay times 
for such transitions vary considerably, but even though 
non-radiative 'branching' usually occurs for such impurity-localized 
excitons it is a relatively simple matter to measure 
the associated luminescence spectra. Such spectroscopy is 
an active field of semiconductor physics, and the reader 
is referred to a comprehensive review by Davies for more 
detail [4]. 
The luminescence spectra associated with bulk 
crystalline Si are typically highly structured and well 
defined, whereas porous Si luminescence is strikingly 
different. Figure 2 shows the photoluminescence spectrum 
from a lightly boron-doped sample [4], this is a high-quality 
version of the sort of wafer which might be used as a 
starting point for processing into porous material. Whilst 
the emission from the p-type wafer shows the characteristic 
sharp features due to the decay of excitons trapped at boron 
acceptors, that from porous material is both broad and 
significantly shifted above the three dimensional gap; this 
is clear from figure 3. which shows some typical examples 
of spectra measured on freshly prepared porous material. 
Although the porous Si emission is blue shifted, it is clear 
from figure 1 that the emission energies lie far below the 
lowest direct gap at the r point in the Brouillon zone. In 
fact it is clear that none of the light emission observed 
from porous Si, even the blue emission discussed later, is 
associated with the three-dimensional direct gap. 
The spectral emission from porous silicon is not 
confined to a single band. The band shown in figure 3 [5] 
is often called the visible or slow band. and is the one first 
reported. This band can actually be shifted systematically 
between the near-infrared and the yellowlgreen region of 
the visible spectrum. Other radiating systems exist and are 
crucial to the emerging story of porous Si, but we shall use 
the properties of the visible band to obtain an overview of 
the luminescence properties. Table 1 gives a brief resume 
of the various bands observed to date with comments on 
their origin. 
g1.1 1 
0.9 oi 0.2 o.i 0.6 o:8 1 
(1-Porosity) 
I 
Figure 3. (a) An illustration of the way in which the 'visible 
band' vanes with porosity. All samples in this data set 
came from the same p+ substrate and were processed in 
the same electrochemical cell. (b) The variation of peak 
energy with porosity for the same band. Although porosity 
is only an indirect assessment of size, it is clear that the 
blue shill becomes faster at the highest porosities. 
Although all conductivity types of Si wafer have now 
been demonstrated to yield the visible band, most work 
has been canied out on pt material. This is largely to do 
with the fact that the anodic electrochemical dissolution, 
which requires a large supply of holes, is most readily 
and precisely established in p-type material (see section 3). 
1189
B Hamilton 
Table 1. Luminescence bands obselved for porous silicon. 
Energy range (ev) Key properties 
1.2-2.2 
('visible band') 
Sensitive to porosity in most reported cases: 
sensitive to surface passivation, 
especially but not exclusively hydride passivation; 
strong temperature dependence of decay time; 
convolution of at least two components; 
dominant band in freshly anodized material. 
Usually weak compared with visible band; 
sensitive to surface oxidation condition. 
strong in high temperature oxidized material 
fast decay time, 
0.8-1.3 
('IR band') sensitive to porosity; 
2.5-2.8 Weak dependence on porosity 
('blue band') 
The visible band displays a quite remarkable variation of 
peak wavelength with porosity. Figure 3 shows this for 
small pieces of the same p substrate, processed in the same 
electrochemical cell to different degrees of porosity and 
measured under identical excitation conditions [SI, It is 
common practice to measure the visible band in freshly 
etched material. This behaviour has been reproduced by 
many groups, and though there are small variations the key 
features are now established. The onset of luminescence 
requires a threshold porosity of around 45%; below this 
value only the luminescence attributes of crystalline, non-porous 
silicon are observed. The visible band moves 
smoothly from its threshold peak energy of 1.3 eV to 2.0 eV 
at a porosity of 90%. At and beyond this porosity, the 
film is mechanically fragile and porosities are difficult to 
reproduce and to measure. The sensitivity of wavelength 
to porosity increases dramatically as the porosity increases. 
Since in a general sense the characteristic size of the porous 
skeleton reduces with increasing porosity, this wavelength 
sensitivity was one of first attributes to underpin the search 
for quantum confinement effects. 
It was rapidly discovered [7] that the visible band is not 
completely stable following anodization; both wavelength 
shifts and efficiency changes occur when porous Si layers 
are stored in ambient conditions. In general blue shifting 
of the emission occurs, with a peak shift of 0.5 eV being 
recorded for a three year storage period, though all effects 
seem to saturate after around one year. Interestingly 
the quantum efficiency often changes little with ambient 
storage and may increase [7]. Another early discovery 
was the optical fatigue of the visible band which is 
especially pronounced for short-wavelength, high-power 
Laser excitation [SI. These instability problems relate to 
surface chemical and electronic structure, issues of vital 
importance for the understanding and control of porous Si. 
In addition to the issues of stability, the visible band 
turns out to have rather complex spectral and temporal 
properties. The emission contained within the spectral 
envelope consists of more than one emission band [91. By 
using time-gated detection it has been demonstrated that 
both fast and slow components are typically present with 
distinctive spectral shapes. Figure 4 [9] demonstrates this, 
and shows that the fast component peaks at significantly 
higher photon energies. The data of figure 4 were obtained 
1190 
I I r 
Photon energy [eVI 
Figure 4. The 'visible' emission band is not a single 
system. This figure shows that time domain measurements 
reveal that at least two spectral bands are present. 
from a p- layer of around 80% porosity which had a fully 
stabilized native oxide (i.e. is fully aged). More recently, 
fast high-energy components have been associated with 
specific types of oxidation of porbus Si. However, for 
porous silicon which has received no additional surface 
treatment, aside from ambient aging, the slow component 
is by far the most important and accounts for almost all of 
the measured quantum efficiency. 
The detailed temporal behaviour of the slow component 
of the visible band varies according to which spectral 
bundle of the rather broad band is measured, and also 
on the measurement temperature. Also when measured 
over several decades, the decay profile is never completely 
exponential. There is broad agreement about the general 
form of the temperature variation of the decay time for this 
band. At low temperatures ( 4 0 K) the decay can be very 
long, of the order of milliseconds; as the temperature is 
increased the lifetime quenches to the microsecond regime. 
There are some differences in detail from sample to sample, 
which makes it difficult to fit the data to detailed kinetic 
models; however, figure 5 demonstrates the trend which 
has been established by many workers.
Porous silicon 
10 
1 10 100 
Temperature (K1 
Figure 5. The temperature dependence of the decay time 
of the main component of the visible band measured at 
1.8 eV. The data are taken from t'Hoofl G W et a/ 1992 
Appl. Phys. Lett 61 2344. 
0.16- 
c 0 , 0 8 i I 
- j L 0 -0 
0 50 100 150 ZOO 250 3 0 
Temperature IKI 
Figure 6. The spectral peak of the visible band moves in a 
very uneven way with temperature. The detailed movement 
is sample dependent. The effect is illustrated here for p+ 
material of around 85% porosity. This trend is typical of a 
disordered system, but can be very marked for porous 
silicon. 
This shortening of the total lifetime at high temperatures 
is always accompanied by a quenching of the luminescence 
efficiency. This shows in a rather unambiguous way that 
non-radiative channels are opening up to the excited carrier 
populations as the temperature is raised. This is rather a 
familiar picture in semiconductors and is often associated 
with disorder, or more specifically particle localization 
within the potential energy minima resulting from disorder. 
It suggests that at low temperatures excitons (for example) 
created by the optical pumping rapidly localize into sites 
which have good radiative efficiency. that is to say the local 
non-radiative channel is slow compared with the measured 
lifetime of typically 1 ms. As the temperature is raised. the 
thermalization time out of these 'radiative sites' becomes 
shorter and the exciton is free to explore larger volumes 
of the porous skeleton, and to find much more efficient 
non-radiative paths. 
The notion that disorder plays a role in the 
luminescence mechanisms of porous Si is compelling 
given the enormous complexity of the porous skeleton, 
and indeed other simple observations support this view. 
One such observation is the spectral shift of the visible 
band as the temperature of the sample is raised and an 
example is shown in figure 6. This occurs because at 
the lowest temperatures particles bind efficiently into the 
deepest potential fluctuations and the luminescence signal 
i o 4 
1.0 20 3.0 4.0 5.0 6.0 70 
Excitation Energy (eV) 
Energy (eV) 
Figure 7. The photoluminescence excitation spectrum of a 
porous silicon layer. There is some similarity to that of bulk 
silicon especially in the higher energy regions towards the 
direct gap 
is weighted in favour of the lower energies characteristic 
of the deep fluctuations. At higher temperatures shallower 
potential fluctuations become statistically more significant, 
shifting the mean of the spectral distribution to higher 
energies. This is exactly what is observed in disordered 
alloy quantum wells [IO], but the effect is much more 
dramatic in porous silicon [ I l l , 
Other spectral features may reveal the presence of 
disorder phenomena, such as the Stokes shift between 
emission and absorption bands, and the relationship 
between the Stokes shift and luminescence linewidth. For 
any situation in which luminescence is dependent on the 
details of localizing potentials, the absorption process itself 
is more representative of the band structure of the solid; 
it is not unevenly weighted by the defect phenomena. A 
practical difficulty in obtaining absorption data from porous 
silicon is that one of the spectral regions of interest is 
well above the three-dimensional gap of the underlying 
substrate. The strong substrate absorption inevitably 
masks the processes in the porous layer. For this reason 
many measurements rely on photoluminescence excitation 
spectroscopy (PLE) which is well suited to the measurement 
of thin surface layers. One of the first reported PLE 
measurements is shown in figure 7 [12]. 
The first impression of such data is that it is rather 
reminiscent of the absorption of bulk silicon, with strong 
absorption above 3.4 eV corresponding to the three- 
1191
B Hamilton 
dimensional direct gap. Once again, though, we run into 
the complexity of the material in the interpretation of the 
data. 'we do not know in detail the macroscopic optical 
constants of the porous silicon layer, though the refractive 
index has been shown to decrease with increasing porosity. 
Clearly some caution must be exercised in assigning an 
optical thickness w. The role of internal light scattering 
is likely to complicate the estimate of optical thickness. 
Calculation of the dielectric functions of the altered layer 
is difficult because of the complexity of the layer, though 
some attempts have been made [13]. Even in the limit of an 
optically very thin surface layer, for which the PLE method 
truly measures the absorption processes, the experiment will 
average the whole ensemble of size distributions; this will 
inevitably lead to a smearing of the data. Such smearing 
would be particularly enhanced if size effects were present 
in the optical density of states functions. 
In order to get a better feeling for the material structure 
which gives rise to these very distinctive optical properties 
of porous silicon, we now turn to a review of the fabrication 
process: this includes some insight into the crucial role that 
high-resolution microscopy has played in the interpretation 
of the material properties. 
3. Porous Silicon formation and microstructure 
For many years porous silicon formation has been used 
as one of the mahy processing techniques for device 
isolation. The FIPOS process, (full isolation by porous 
oxidized silicon) makes use of hydrofluoric acid (HF) 
as an electrolyte in an anodic electrochemical reaction; 
HF, it seems, is the only known electrolyte which can 
anodically dissolve silicon in an efficient manner. The basic 
electrochemical phenomena involved in the FIPOS process 
and optically active porous silicon are essentially the 
same, except that the latter usually has significantly higher 
porosity and in the limit of such high porosity additional 
electrochemical reactions may occur. The electrochemistry 
of nanostructured silicon is still a developing field. and only 
the simplest of views can be presented here. 
In principle the production of a porous silicon layer 
is not demanding; a carefully constructed electrochemical 
cell along the lines of that illustrated in figure 8 [14] 
is all that is required. The cell and the electrolyte 
system must be formed from high-purity material, and 
good control of the operating characteristics is required. 
Nevertheless. processing of centimetre-size samples with 
good macroscopic uniformity is not difficult. Ironically, the 
ease of fabrication is in stark contrast to the complex range 
of characterization methods which have been employed in 
an attempt to understand the porous material. 
In the cell, the SifHF interface forms an elec-trode/ 
electrolyte barrier system. The potential harriers and 
electric field distributions across even the equilibrium sys-tem 
are rather involved, depending on the doping character-istics 
of the semiconductor and the chemical composition 
of the electrolyte. However, the gross feature of the barrier 
is its 'double layer' attribute. There exists finite regions of 
space over which the interfacial electric fields are spread 
and the potential barrier evolves; these are the Hehnoltz 
1192 
Ammeter I - 
Cathode- h -Anode 
-Magnetic 
stirrer 
. , , . . . . . . , 
U ca 
e: 
0 
* 
- -U Silicon ca 
Potential Distribution 
.s ................................ U 
Figure 8. A simple schematic diagram of a basic 
electrochemical cell used for anodization. The potential 
distribution across the electrolytelsilicon system is shown 
below. 
layer in the electrolyte, and the depletion layer in the semi-conductor. 
A schematic diagram [ZO] of the barrier system 
is shown in figure 8. 
The dissolution of silicon occurs only under anodic 
conditions, and the primary process leading to massive 
removal of Si atoms is considered to be the formation 
of silicon fluoride molecules, SiF,. Various routes are 
possible in principle 115-171. Perhaps the simplest example 
proposed are for Si dissolution involves the divalent state 
[I81 
Si + 2HF + ne' + ,352 + 2H' + (2 - n)e-. 
Here it is assumed that holes take part, i.e. holes are 
freely available in the silicon to feed the reaction. This 
requirement is easily fulfilled by p' material, but low or 
even n-type conductivity is not a fundamental barrier to the 
process because optical excitation can always be used to 
generate an excited hole population. The SiFz formed in 
the above reaction may then be removed by other chemical 
reactions [18, 191. It has emerged, however [19], that the 
number of electrons consumed in the initial electrochemical 
reaction, that is the number n in the above equation, is
Figure 9. The electrochemical regimes available for silicon 
processing as a function of the I-V characteristic of the 
electrochemical cell. Region A: pore formation, region B: 
transition, region C: electropolishing. 
greater than 2. It seems therefore that both the divalent 
and tetravalent Si dissolution occur simultaneously 
Si + 4HF + (4 - n)e+ + SiF4 + 4H+ +ne-. 
Again, several routes are possible for the removal of the 
SiF4. 
What is achieved in practice depends on the precise 
anodizing conditions, for example the anodic potential, 
and .ranges from a layer of uniform porosity to 
complete removal or electropolishing. The current-voltage 
relationship of the sample-cell system reveals the various 
regimes of electrochemistry. This is shown in figure 9 [ZO]. 
In the region of low applied potential (A) the current is 
generally exponential with voltage (the Tafel region), with 
a slope of typically 60 mV per decade; this value is clearly 
an indication of the physics of the potential banier and the 
in this region, the silicon removal being driven primarily 
by the above reactions. At significantly higher potentials, 
the electropolishing regime is entered, resulting in complete 
removal of the porous layer, Electropolishing results from 
the formation of an anodic oxide which is dissolved by 
the HF, any irregularities in the silicon topography being 
removed due to the divergence of th e electric field lines at 
regions of dielectric with a consequent enhancement at any 
Si features [221. One proposal for the electrochemistry of 
this oxidation process [23] is the following reaction 
carrier transport [20, 211. Pore formation occurs ................ 
Si + 40H- +ne+ -+ Si(OH)4 + (4 - n)e-associated 
~i~~~~1 0, The simpleS. model for pore formationb, ased 
essentially on impedance to current flow, leads to columnar 
pores (a). A more complex model based a 
diffusion-controlled mechanism of pore formation leads to 
the sort of multiply interconnected or spongy porous layers 
ofien obselved in TEM measurements (b), 
the processing of porous silicon sensitive to the HF 
concentration in the electrolyte, low HF concentrations and 
hence low oxide removal rates favouring electropolishing; 
these trends have been established. experimentally [20]. 
Whilst the simplified electrochemistry discussed so 
far can explain Si removal, it does not account for the 
spatial selectivity which results in pore formation. In 
fact Pore morphologY does depend on conductivity type 
and several models have been proposed to explain this 
crucial feature of the processing, most of them resting 
Si02 + 6HF + H2SiF6 + 2H20. on built-in inhomogeneities in the original Si wafer as 
the trigger for pore formation. The wafer conductivity 
At low potentials, in the Tafel region, the oxide formation and electrochemical details then dictate the detailed pore 
rate is too low to compete with Si removal and porous evolution. 
silicon results. At high potentials oxide formation is One of the earliest attempts to explain pore formation 
enhanced and surpasses the oxide dissolution rate, resulting is due to Beale and co-workers [24, 251, and is based on 
in electropolishing. It is to be expected that this interplay the barrier properties coupled with the spatial variation 
between oxide formation and removal rates should make of impedance to current flow. The essence of the 
1193 
4. 
Si02 + 2Hz0. 
me oxide formation rate is in competition with its 
dissolution rate governed by
B Hamilton 
model is that small inhomogeneities on the wafer surface 
cause enhanced current flow and locally rapid removal 
of Si. The original depression is enlarged, leading to 
pore formation. The nature of the inhomogeneity is not 
specified; it could be some macroscopic perturbation of 
surface morphology, or even a defect at the atomistic level. 
In its simplest form this model led to the expectation that 
silicon between the pores will ultimately become depleted, 
simply because the dimensions of the remaining silicon 
'columns' is insufficient to support the space charge width. 
The impedance offered to the current path into the silicon 
column is then held to grow rapidly and current flows 
preferentially down the electrolyte and into the wafer at the 
bottom of the pore, as illustrated in figure 10(a) [271. This 
provides a possible mechanism for producing columnar 
structures. 
This way of describing the pore evolution does seem 
to go some way towards explaining the gross morphology 
of the porous skeleton in p+ silicon. It is suggested 
that the heavily doped wafer leads to a narrow space 
charge layer in the semiconductor, tunnelling phenomena 
are enhanced Gust as in Schottky baniers to degenerate 
semiconductors) and the impedance to current at the base 
of a pore is significantly lowered. However, the model does 
not explain the pore morphology observed in p- material; 
this typically consists of massively interconnected network 
which is uniformly distributed across the film. 
Smith et al [26] have shown that a more complex pore 
morphology may be explained if pore evolution is limited, 
at least partially by the rate of diffusive transport of the 
hole to the reaction point at the electrolytic interface. The 
diffusion-limited case arises because the impedance offered 
by the barrier system is much higher than in the pf case, 
tunnelling being much weaker in the wide, low-field barrier 
system of the lightly doped semiconductor. This analysis 
still accounts for a faster than average reaction rate at a 
pore tip, and hence elongation of the pore. It also makes 
the interconnected network a more reasonable expectation, 
as shown in figure IO@). A pore E, initiated on the sidewall 
of an existing pore A, will be in better communication with 
the diffusing hole flux in that region of silicon between 
pores A and B until the tip of pore E approaches the tip 
of B within two hole diffusion lengths. This is rather a 
complex, at simplest. two-dimensional diffusion problem, 
but the significant sidebranching is a fundamental feature 
of the observed morphology. 
Other models of pore formation have been discussed. 
The possible effect of quantum confinement in residual 
silicon structures has been proposed [27] as way of 
enhancing carrier depletion effects and hence limiting pore 
growth in the limit of very small structures. Alternative 
electrochemical schemes have also been proposed for the 
Si removal process [19, 281, which draw closer analogies 
with the formation of porous aluminium. The theoretical 
simulation of pore structure for a variety of possible 
electrochemical conditions is given by Parkhutik etal 1291. 
No complete understanding of pore morphology exists, 
and it is likely that improvements in our understanding 
will come about through the application of high-resolution 
microscopy. Transmission electron microscopy has already 
1194 
proved essential in probing the porous structure. Porous 
silicon layers are inevitably fragile and this makes their 
evaluation more difficult, and necessitates the development 
of some novel approaches to specimen preparation. 
Measured pore sizes can vary from -100 nm (macroporous) 
down to c2 nm (mesoporous). As a very approximate 
guide to published data it appears that lightly doped p-type 
silicon produces a fine network of pores whereas 
heavily doped p-type material produces more of a columnar 
structure [24, 25, 30, 311. For lightly doped n-type silicon, 
the pores in general take up a more crystallographic form 
with typical dimensions of several tens of nanometres 
propagating in the (100) direction. This attribute has 
even played a role in VLSI device isolation by trench 
formation. Un l i e the case of p-type silicon, as n-type 
doping level increases, the pore dimension increases and 
hence the interpore spacing decreases. 
These general comments on pore morphology must 
be taken as a rough guide only. The detail form of 
the layer depends on the precise anodization conditions 
used, and very high resolution imaging can often real 
more complex geometry, leading to a fractal view of the 
altered layer. For example it has been known since the 
early work of Beale et al [32] that the columnar pore 
arrangement in pt silicon is heavily branched. It is also 
possible to produce mesoporous n+ silicon with -5 nm 
pore dimensions (33). The key question which high-resolution 
electron microscopy has attempted to address 
concerns the detailed relationship between porosity and 
luminescence. This has been reviewed by Cullis [34], who 
highlighted the need to avoid ion beam milling or other 
invasive specimen preparation methods for the preparation 
of electron transparent samples of porous material, which 
is easily amorphized and chemically modified. Figure 
11 is taken from that review; it demonstrates well 
the key issues concerning the microstructure of p-type 
porous silicon in the transition from relatively low-porosity 
weakly luminescent material to high-porosity strongly 
luminescent material. The pictures represent bright field 
(001) projections. For the weakly emitting material, the 
Si skeleton comprises mainly rod-like structures with a 
range of diameters, the smallest being around 5 nm. 
The corresponding electron diffraction patterns indicate 
completely crystalline material. Figure 1 l(b) illustrates 
material of higher porosity than (a) which gave stronger 
luminescence. The microstructure is now finer with silicon 
structures down to 3 nm clearly visible. Arcing of the 
electron diffraction spots indicates misalignment of the Si 
columns. The electron diffraction pattern now shows more 
severe misalignment of the Si skeleton, but still indicates 
crystalline material. As porosity grows, these trends are 
continued. These TEh4 data, then, point to a correlation 
between a reduced characteristic size distribution of the 
silicon porous skeleton and the switching on of strong 
luminescence. In particular, column or particle sizes of 
around 3 nm or smaller are present in highly luminescent 
material. 
Ultra high-resolution microscopy will continue to play a 
key role in porous silicon research and it can be anticipated 
that advances in microscopy will add more vital information
Porous silicon 
regarding the relationship between microstructure and 
luminescence. Whatever new insight is gained from 
microscopy, though, it remains true that increasing the 
porosity of the material inevitably increases the surface 
area. Surface chemical interactions and the influence of 
the surface on electronic-properties area key areas of 
investigation. 
4. Surface effects on porous silicon 
The 'internal' surface area of porous silicon is very large; 
several hundred square metres per cubic centimetre of 
porous material is typical. It is reasonable therefore to 
expect that the surface itself might play a direct role in some 
of the observed luminescence behaviour, or that the surface 
would exert important effects on the 'bulk' behaviour of 
the material. A good deal of effort has been expended on 
investigating these issues, which still remain at the heart of 
the debate on the origin of the light emission. One of the 
earliest [2] and most graphic attributes of surface chemistry 
is the role of hydrogen coverage. After anodization in the 
HF-based electrolyte, a surface rich in Si-H bonds can be 
routinely observed using infrared local mode absorption 
spectroscopy. Bonds involving one (Si-H), two (Si-H2) 
and three (Si-H3) hydrogen atoms are normally present 
and both stretching and scissor vibrational modes can be 
seen [35-381. Figure 12(n) [5] shows a typical absorption 
spectrum measured for a freshly prepared porous layer 
(curve (a)). Compared to that of, say, an unprocessed 
Si wafer, the H bond-related absorption is dramatically 
increased. Other Vibrational features can be seen which 
are common to both porous layers and bare wafers; these 
correspond to SiSi stretching modes and to (probably bulk) 
Si-0-Si, interstitial 0 asymmetric stretching modes. These 
and other 0-related modes assume much more significance 
for oxidized material. The figure shows the evolution 
of the local mode structure as the sample is annealed 
in vacuum (curve (b)) and also in nitrogen at 300 "C 
for 5 (curve (c)) and 10 min (curve (d)). The vacuum 
anneal completely removes the H-related features, whilst 
the nitrogen anneals promote 0-related modes, probably 
due to weak 0 contamination. 
It was noted above that the process of atmospheric 
aging has an indeterminate effect on the luminescence, 
and may cause it to increase. Such aging causes a 
broadening of the H-related absorption modes and also 
a growth of 0-related modes. However, by far the 
most dramatic phenomenon associated with H coverage 
is observed following desorption on a large scale during 
vacuum anneal. This causes a complete quenching of the 
luminescence. Figure 12(b) illustrates the luminescence 
spectrum for a freshly prepared sample. After vacuum 
anneal no luminescence can be seen, though some weak 
recovery is observed if the vacuum anneal is followed 
by a nitrogen anneal. This recovery is significant, even 
though no H-related absorption can be measured. By 
immersing the annealed sample in HF for a few seconds, 
both the luminescence and surface H bonds measured by 
absorption are dramatically restored. Very small shifts in 
peak wavelength are seen due to this cycle, and these are 
1195 
Figure 11. High-resolution data obtained from TEM 
measurements of porous p-type silicon. The trend in 
characteristic sizes of the remaining silicon skeleton as a 
function of porosity is clear. The associated optical 
characteristics are described in the text.
B Hamilton 
(4 si.0-si 
related mods 
J-defamation 
I 
IIIP 1 1 
I I I I I 
500' lMXl 1500 2m 2500 
Wavenumbers cni' 
Energy (eV) 
Figure 12. (a) The infrared absorption spectrum of 
prepared 4550% porosity silicon is rich in Si-H 
bond-related transitions curve (a), vacuum anneal for 2 min 
at 400 'C removes these modes completely. Further 
annealing in nitrogen at 300 'C for 5 (curve c) or 10 min 
serves only to weakly promote 0-related bonds. (b). The 
luminescence spectrum of the same sample: as-prepared 
(full curve), after the vacuum plus the 10 minute nitrogen 
anneals (dotted line) and finally after immersion in HF 
(broken curve) 
not surprising since one expects small changes in the silicon 
skeleton to Lake place; we don't have precisely the same 
sample at the end of the sequence. There is no doubt. 
however. that surface hydrogen coverage plays a key role in 
the luminescence behaviour of freshly prepared porous Si, 
and that removal and replacement of the H coverage leads 
to reversible quenching and restoration of the luminescence. 
The electronic role of H is not fully understood, but 
removal of H has been shown to increase the Si dangling 
1196 
bond density measured by electron spin resonance [39]. 
Since the dangling bond is known to be a powerful non-radiative 
recombination centre [40], a straightforward role 
of H as a passivating centre is suggested. This notion is 
much connected with the debate surrounding the radiative 
mechanisms which operate in porous silicoa 
There is no doubt then that the surface plays an 
important role and that H bonding is necessw to sustain 
the luminescence. Furthermore the vibrational assignments 
suggest that simple Si-H, bonds account for some, possibly 
most, of the surface hydrogen. It is unrealistic, however, to 
expect that this bonding arrangement accounts for all of the 
surface chemistry and some considerable effort has been 
devoted to probing for other surface constituents which 
might bear on electronic processes. There are several 
important candidates for surface bonding, based simply 
on the processing environment of the wafer; to date most 
reported work has been aimed at probing the involvement of 
oxygen, fluorine, or organic radicals of varying complexity. 
Although surface Si-F bonds play an important role in 
the dynamics of pore evolution, it seems that they are not 
stable on the free surface after processing. Probably they 
are replaced via a hydrolysis reaction, by Si-OH bonds 
which themselves can dissociate into Si-H or Si-0-Si 
bonds by reaction with the atmosphere [41]. 
The role of oxygen is important. Simple exposure to 
air causes surface oxidation of all silicon, and the effect is 
enhanced for porous silicon. After all, this fact led directly 
to the development of the FIPOS process mentioned above. 
The luminescence aging of porous silicon is connected the 
incorporation of 0 into the surface bonding arrangement. 
The detailed form of 0-modified surface bonding has been 
analysed by Kat0 et al [36]. Low-temperature ( d o 0 "C, 
for times of less than 50 min) oxidation was used; this 
might be regarded as a sort of accelerated aging process. 
Using IR local mode absorption, the Si-H transitions are 
seen to broaden and shift somewhat. These spectral changes 
were attributed to the incorporation of 0 into the Si back 
bond(s) associated with the S-H, atomic arrangement. For 
example if 0 is incorporated into one of the three back 
bonds of S-H, the Si-H stretching mode transition was 
calculated to shift from 2090 cm-l to 2127 cm-', and by 
considering all possible sites for 0 incorporation into back 
bonds the general changes which occur in 'lightly' oxidized 
material are accounted for. This picture of 0 incorporation 
leaves all surfaces terminated with an Si-H bond; only the 
back bonds are broken. The spectral deconvolution leading 
which led to this picture rests on the assumption that the 
Si-H stretching mode transition is located at 2090 cm-', 
but it should be noted that some debate exists regarding thc 
precise vibrational nature of this transition. 
Aging, and non-aggressive oxidation also cause 
increased absorption in all peaks that relate to Si-0 
vibrational modes. The effect of further increasing the 
0 content of porous Si causes yet more changes, and is 
currently being used as a modification process to stabilize 
the material. Oxidation is also potentially useful in helping 
to evaluate the luminescence mechanisms and the role of 
the dangling bond in quenching luminescence; these issues 
are discussed further in section 7.
Porous silicon 
5.1. The quantum confmement mechanism 
The idea that the surviving silicon skeleton contains within 
it structures small enough to exhibit quantum confinement 
effects such as opening up of the bandgap was the first 
proposal for a mechanism for porous silicon luminescence. 
This suggestion represents a simple explanation based on a 
well established attribute of the material, i.e. the existence 
of nanoparticles in the layer. Figure 13 is an example 
of the way in which the quantum confinement mechanism 
is often viewed. As we have seen, nanometre sizes for 
crystalline Si particles are amply proven from TEM data 
and are also confirmed by an analysis of the optic phonon 
Raman lineshape [45]. Energy shifts due to confinement on 
a scale comparable to the particle size are a universal feature 
of quantum mechanics, but proof that such a mechanism is 
correct must rest on a direct observation of luminescence 
from the nanoparticles and supporting evidence on the 
interconnection between the geometly of the nanoparticle 
and the emission wavelength. 
Such direct evidence does not exist for porous 
silicon. However, a direct observation of red luminescence 
from oxidized isolated Si nanoparticles with characteristic 
sizes of below 5 nm has been reported [46, 471. 
Such observations demonstrate that isolated particles can 
luminesce, but the detailed chemical arrangement of the 
oxidized nanoparticles complicates the interpretation of the 
luminescence process 
It remains the case that the blue shift of the visible 
luminescence with increasing porosity in p+ material is 
one of the key observations linking the light output 
to nanoparticle size. However, porosity in itself is a 
quantitative measurement of the fractional mass removed, 
but is not a quantitative measure of nanopaaicle size. 
So. for example one can imagine crudely that a film of 
a particular porosity might consist of large Si particles 
and extremely large voids, or of very smal1.nanopartick.s 
and moderate size voids. In p+ material, in which the 
morphology is known to exhibit size reduction of the 
nanoparticles as porosity increases, the increased sensitivity 
of the luminescence blue shift to porosity is precisely what 
would be expected from quantum size effects provided 
that the light originated from recombination within the Si 
nanoparticles. 
The slow decay rate of the visible band is really what 
one might expect from an indirect semiconductor; and 
if the interior of the residual silicon nanoparticles were 
perfectly crystalhe and therefore presented a shunt-free 
environment with no non-radiative recombination centres 
and with completely passivated surfaces, they might offer 
the perfect environment for light emission. However, 
the non-exponential decay indicates that such an idealized 
notion is unlikely. In fact this behaviour is not unlike that 
observed in amorphous silicon, in which carrier trapping in 
the tail states plays a dominant role. 
The photoluminescence attributes are rather different 
for n-type material. In general, for such material there is 
no systematic variation of blue shift of the visible band with 
porosity. For example, marginally porous layers, e.g. less 
than 40%, have been shown to exhibit strong visible 
luminescence with the same general character as that seen 
1197 
Although IR absorption has emerged as a powerful tool 
in the surface analysis of porous silicon, other techniques, 
principally x-ray photoelectron spectroscopy (XPS) and 
SIMS, have been used. SIMS analysis has confirmed the 
presence of hydrogen and Ruorine as the major surface 
species of freshly anodized material, whilst oxygen, carbon 
and nitrogen were detected at lower concentrations [42]. 
The SIMS data also c o n h that F is not stable but 
reduces with atmospheric exposure, presumably due to 
the hydrolysis reaction mentioned above, and indeed an 
increase in surface hydrogen to be expected from this 
process is also observed using SIMS. The other important 
changes revealed by SIMS measurements are a build-up 
of carbon and oxygen with prolonged exposure to the 
atmosphere. 
The XPS technique has revealed fluorine, carbon and 
oxygen on porous silicon surfaces [43], in broad agreement 
with the SIhlS data. Evidence in support of a fluorine-admixed 
Si02 surface phase has also been claimed, based 
on XPS analysis [44]. This possibility, of surface layers 
with rather complex chemistry, for example an Si-0-F-H 
system, though more difficult to analyse experimentally 
than simple Si-H bonding arrangements, is the basis for one 
of the models suggested for the luminescence, to which we 
now turn. 
5. The light emission process 
Having reviewed the basic features of the material it is 
possible. by looking in a little more detail at particular 
pieces of experimental evidence, to try to glean what 
is cumently understood about the basic light emission 
mechanism. This of course is the pivotal question 
surrounding porous silicon. Until it is answered progress 
towards any technological goals will be limited. 
It must be readily acknowledged that the complexity 
of the material provides fertile ground for the proposal of 
differing models for the light emission process. Porous 
silicon is a richly interconnected system of small particles 
and intricate surface topology. This fact alone leads 
naturally on to the expectation of electronic disorder 
with its associated defect and interface states; furthermore 
the large surface area, generated in a chemically varied 
environment adds the possibility of partial surface coverage 
with complex molecular films. All of these attributes 
have formed the basis for hypothesis regarding the light 
emission, and at the time of writing all of these generic 
schemes receive support, often zealous. In keeping with 
the spirit of this lively debate, this section is presented 
by analysing some of the data which either support or 
undermine current models. In the current literature, by far 
the largest attention has been paid to investigations which 
have been designed around the hypothesis that quantum 
confinement plays a key role; by far the largest number 
of reports discuss this issue. Accordingly, this aspect is 
given more emphasis here, though the correctness of the 
hypothesis is not proven.
B Hamilton 
Figure 14. The optical transmission spectrum of 
free-standing porous silicon films. 
to the PLE measurements reported above, there have 
been successful attempts to measure absorption in free-standing 
films, free of any complications associated with 
the substrate. The absorption measurements often appear 
to be rather more sensitive to the low-energy tail, near and 
even below the three-dimensional gap of silicon. Several 
such measurements seem to indicate that the threshold for 
absorption in porous silicon is higher in energy than in 
bulk silicon and that the threshold moves to higher energies 
with increasing porosity. Figure 14 demonstrates this for 
films originally processed from both p and p+ substrates 
[27]. The up-shift in energy was more marked for the p+ 
material which was found to have smaller nanoparticle sizes 
than the p material. These observations are consistent with 
the opening up of the gap due to confinement and would 
strongly support it if it were to be confirmed that silicon 
nanoparticles and not some other phase of material were 
dominating the absorption spectrum. 
Looked at in more detail, the absorption edge of 
porous silicon does not wholly support a simple quantum 
confinement model. Photothermal deflection spectroscopy 
has shown [49] that the absorption strength increases 
roughly exponentially above the luminescence peak. Whilst 
it might be argued that a size distribution of nanoparticles 
might partly explain such data, the same experiments show 
that absorption occurs significantly below the gap of three-from 
Figure 13. Schematic diagrams illustrating the some of the 
structures envisaged for the optically active material, 
according to the quantum confinement hypothesis. 
(a) The transition from quantum wire through oxidized 
nanoparticles to porous glass [7].(b )T he aligned 
nanocrystalline or wire structures consistent with EPR data. 
(From Harvey J F et a/ 1993 NATO AS/ Series voI244, 
p 179.) (c) An electronic view of how an exciton localized 
in a nanoparticle might suffer three possible fates for 
radiative decay giving rise to three luminescence bands. 
(From Koch F 1993 Mat. Res. Soc. Symp. Proc. 298 319.) 
P-tyPe material [481. This is an imPortant Point. and dimensional silicon. The existence of significant Urhach 
is further highlighted by the fact that the pore morphology tails in the density of states is of course a qualitative 
is much more macroscopic in nature with tYPicallY large measure of departure from crystallinity and is reminiscent 
widely spaced and crystallographically oriented pores. This of the behaviour of amorphous silicon. Whatever the 
picture is at variance with the quantum confinement model. origin of the density of states low-energy tails, its existence 
However, the fact that the macropores have much smaller implies strongly some significant degree of electronic 
structure on the sidewalls is a further complication which localization. 
means that we cannot rule out a nanoparticle explanation Falling back on the evidence relating to the 
for the luminescence. luminescence spectrum, the modification of the emission 
A key test of low dimensionality in any electronic characteristics by post-anodization processing has been 
system is a measure of the density of states functions for used to variously support or oppose confinement models, 
electrons and holes, and it was noted above that, in the The blue shift of the luminescence peak as a result of 
case of porous silicon, optical ahsorption is in principle oxidation followed by HF dipping was first suggested as 
a fundamentally better measurement of these (or more supporting evidence, since the consumption (by oxidation) 
precisely of the joint optical density of states). In addition and subsequent removal of silicon is expected to reduce the 
1198
Porous silicon 
..;- i 10000 
c 
3 
E 8000 
.. P 
- 
5 - 
6000 
- 
5 h 
4000 
- c 
c 2000 
850 800 750 700 650 
Wavelength lnml 
Figure 15. An example of one effect of immersing porous 
silicon in ethanol: (1) is the ‘as-prepared spectrum’, (2) is 
for 1 min of immersion, (3) for 3 min, (4) for 10 min and (5) 
for 60 min. The luminescence was measured in situ 
overall size of all silicon components in the porous skeleton. 
To counterbalance this, reports of red shifts with hydrogen 
loss [SO] would not be expected to affect the particle size, 
though very small effects due to strain might be expected 
to produce small wavelength shifts. The influence of low-energy 
‘processing’, essentially immersion in a variety of 
organic fluids, has been shown to have large and nearly 
reversible effects on the emission spectrum. Effects due 
to acetic acid, propanol and ethanol have been reported 
[SI, 521. Figure IS [SI] shows the rather dramatic effect 
of immersion in ethanol for times of up to 1 h. The 
detailed chemical interaction with the porous skeleton has 
not been analysed for these organic treatments, but it seems 
reasonable to assume that they involve surface or near-surface 
effects, and such effects are expected to impinge 
only weakly on optical transitions with energies determined 
mainly by size quantization. 
The debate on quantum confinement has been 
underpinned by attempts to calculate the electronic structure 
of silicon nanoparticles, and hence to predict optical 
properties, in particular the transition energies and matrix 
elements. Effective mass theory (EMT), which has been 
so successful in predicting the properties of epitaxially 
grown low-dimensional smctures, has been applied to 
small silicon structures typical of those know to exist 
in porous silicon. One such calculation [53] was based 
on the notion that for cubic structures with sides greater 
than IO atoms long, bounded by (100) planes, EMT 
represents a plausible approximation for the description of 
wavefunctions. Simple envelope functions and confinement 
energy shifts result. The infinite barrier approximation 
at the cube boundary leads to an optical matrix element 
which is an oscillatory function of the cube size. This is 
difficult to test in a real system because the ensemble of 
sizes present in a given film inevitably smears the effect. 
However, the overall trend for radiative lifetime variation 
with confinement energy shift for the optical transition 
shift, which of course relates to cube size, is predicted 
by EMT to vary rapidly: approximately as the inverse 
cube of confinement shift. Some workers have noted that 
the measured lifetime of the visible band can vary with 
1 + + 
0.5 1.5 2.5 3.5 4.5 
Olnml 
Figure 16. The optical gap predicted by LCAO theory for 
nanometre-size silicon crystallites, as a function of size. 
porosity and therefore with peak photon energy [54], and 
there is rough agreement with the EMT prediction and the 
experimental data. 
Of course, the very small sizes of some crystallites 
observed in porous siiicon must eventually limit the 
applicability of EMT, and point to the need for first 
principles calculations. One such calculation [5S] has been 
performed for wire structures, spanning wire thicknesses 
which vary from the thii, molecular, limit of polysilane to 
structures which are essentially bulk-like. The calculation 
was performed for wires with axes in the [OOI] direction, 
bounded by (110) surfaces which were assumed to be fully 
terminated with hydrogen atoms. A supercell approach 
was used with the basic unit cell of the wire repeating 
in space in order to retain three-dimensional periodicity. 
Such a calculation is far removed from the effective mass 
approach, using a first-principles pseudopotential for the 
Si ions and a bare Coulomb potential for the H ions; the 
exchange-correlation energy and potential were included 
using a local density approximation. It is interesting to 
compare the results of such a calculation with EMT. For 
the confinement up-shift, agreement was good for wire 
diameters of greater than 23 A; above this value the EMT 
prediction appears to be an overestimate. The calculation 
also yielded a radiative lifetime of around 380 ps for a 
wire with 72 atoms in the unit cell, i.e. in broad agreement 
with experiment based on the notion that small crystallites 
yield the photon output of the visible band. The authors 
noted, however. that such a long radiative lifetime implies 
that the high quantum efficiency of this band is largely a 
consequence of the small non-radiative competition rather 
than the lifting of the momentum selection rule. 
A recent calculation using the linear combination of 
atomic orbitals (LCAO) technique has been used 1561 to 
calculate the optical properties of wires and crystallites 
(cylindrical and spherical shapes). This form calculation 
should yield information on both conduction and valence 
band properties of the structures. A key result, the 
calculated optical gap as a function of diameter, is shown 
in figure 16. The crystallites show the greatest sensitivity 
to size; this is the intuitive result based on the fact that 
confinement is in three dimensions. The authors also 
illustrate the Coulomb electron-hole interaction energy 
1199
B Hamilton 
which makes only a small difference to the total calculated 
energy gap. An interesting comparison is made with 
the experimentally measured photoluminescence energy 
measured not from porous silicon but from hydrogen-passivated 
silicon crystallites produced by nucleation from 
the gas phase [57]. 
The result for the wires is a little more complex, 
showing anisotropy between different wire directions. The 
authors note that the visible band, which is tunable between 
1.4 and 2.2 V, would be consistent with characteristic 
structure sizes of between 2.5 and 4.5 nm. The exponent 
relationship between gap and diameter, in the visible band 
energy window. was found to follow D-',39 rather than D-2 
predicted by EMT. However, the calculation predicted an 
inverse square law at larger D values where EMT is valid. 
The LCAO calculation also dealt with recombination 
and optical absorption. It was concluded that the strong 
confinement in silicon (43 A) induces band mixing and 
dipole allowed transitions. The optical matrix elements, 
though, remain small and the radiative decay rates as 
function of transition energy show strong scatter. Partly 
this results from the oscillatory behaviour induced by the 
dependence of the matrix element on the overlap in k space 
of the electron and hole wavefunctions; this was a point 
which emerged also from the EMT formalism. For the case 
of the crystallites, the LCAO calculation also demonstrated 
that the radiative rate was also sensitive to the symmetry 
representation of the Td point group which varies greatly as 
the size of the crystallite is varied. This effect was shown 
to be more sensitive at lower temperatures. 
The optical absorption coefficient based on the above 
calculation, for a crystallite of 3.86 nm diameter is shown 
in figure 17(a) [%I. This shows that the major absorption 
strength is in the ultraviolet, with an absorption 'edge' near 
to 3.5 eV, i.e. close to the direct edge of bulk silicon. 
The spectral shape is also very structured and bears a 
superficial resemblance to what might be expected from 
a molecular system. When viewed on a more sensitive 
scale. figure 17(b), the calculated absorption coefficient 
for this crystallite does show that the transition is allowed 
down to the calculated gap energy, but with small oscillator 
strength. The absorption coefficient shows a quadratic 
dependence on photon energy above threshold, unlike that 
of bulk silicon which shows a linear dependence. The 
blue shift of absorption edge with porosity (assuming an 
attendant reduction of crystallite size) reported above is 
then generally predicted by the LCAO calculation, and 
the predicted non linear shape has also been recorded 
experimentally [58]. 
These three illustrations serve only to review the trend 
in calculations of small structures, and are by no means 
exhaustive. They underline the point that in general there 
is no fundamental disagreement between theory and the 
quantum confinement model for the main emission band 
observed from porous silicon. They also highlight the 
fact that the regime of solid at the heart of the debate 
is tantalizingly poised between one which is comfortably 
crystalline and populated with electrons in Bloch states, and 
one which is better described by a molecular framework 
This notion is very much the theme of the surface film and 
defect models reviewed next. 
1200 
E lev1 
Figure 17. LCAO prediction for the optical absorption 
coefficient of a 3.86 nm crystallite. 
5.2. Molecular films, interfaces and defects 
Since it is clear that surface hydrogen coverage is 
an important criterion for light emission, at least in 
unprocessed porous silicon, several groups have explored 
the possibility that the hydrogen does not simply play 
a passivating role (i.e. dangling bond saturation), but is 
somehow involved directly in the radiative process. Two 
main candidates have emerged; surface hydride species 
and a class of compounds known generically as siloxenes. 
Although other variations have been suggested, these two 
examples are illustrative of the key ideas. 
The idea that surface hydride species of the form 
S a I are directly involved in the luminescence process 
stems largely from the fact that particle size distribution, 
and in particular size reduction attempts, are not 
universally consistent with the quantum confinement model. 
Luminescence from only moderately porous p+ silicon 
(20%) has been reported (591 which did not show a 
blue shift with repeated HF dipping, and with increased 
porosity. On the other hand, this material did show all 
of the well known attributes of surface hydride coverage, 
i.e. luminescence could be quenched and restored by 
hydride removal and replacement. A further report of 
luminescence peaking at 1.7 eV measured for n-type 
samples sample of less than 10% porosity bas been made 
[60]. Particle sizes of around 200 nm were found in 
this material and it was claimed that side pores did.not 
exist. Such a size distribution is not appropriate for
Porous silicon 
and clearly show that .the optical gap shrinkage of the 
amorphous silicon matches well to the red shift of the 
porous silicon sample. 
A somewhat more complex model for the involvement 
surface molecular species in the form of siloxene related 
compounds has been suggested 1641. Siloxene in its 
simplest form has the chemical composition SiaO& and 
can be prepared from Cash via the reaction 
3CaSiz + 6HCI + 3Hz0 = Si6O& + 3CaC12 + 3H2. 
The existence of such compounds has been known for 
some time, and their fluorescence in the green region of 
the visible spectrum is also well known in the chemical 
literature [64]. The initial suggestion was that siloxene 
or closely related compounds are a by-product of the 
electrochemical processing of silicon, which is rich in Si. 
H and 0 atoms. The tuning of the luminescence was 
suggested tentatively to result from chemical variations to 
the basic structure, for example by substituting other ligands 
for the H-terminated Si bonds in the sixfold Si ring of 
the isolated molecule. Many of the features of porous 
silicon luminescence were also seen in the fluorescence 
of siloxene: tunability of wavelength, electroluminescence 
during anodic oxidation (see section 7), luminescence 
fatigue and non-exponential decay. 
The tunability of the chemical structure of the siloxenes 
and its link to emission wavelength were the main question 
marks which militated against the siloxene explanation soon 
after it was suggested. In part this shortcoming was due to 
a lack of understanding of the physical chemistry of these 
materials. More recently [65] it has been demonstrated how 
crystalline films on silicon substrates can in principle be 
produced by evaporation of calcium followed by reaction 
with HCI, i.e. a potential planar technology. Perhaps more 
importantly for the present debate, recent quantum chemical 
simulations of siloxene [66] crystals have led to a better 
understanding of the stability of the system and the way 
in which modification by oxygen incorporation can change 
the electronic properties. 
The idealized crystalline siloxene structure is shown in 
figure 19(n) [65], and consists of a silicon plane, terminated 
by OH and H radicals on opposite side of the plane. 
Calculations suggest that this form is metastable; insertion 
of 0 into the bonds of the Si plane gains 1 eV per Si- 
@Si bond. Therefore annealing the structure is likely to 
transform it into that shown in figure 19(b) [651. The 
stoichiometry remains the same, but now that all the 0 
atoms have been incorporated into what was the Si plane 
notice that isolated Si6 rings begin to appear. 
The optical properties of the metastable and annealed 
structures are quite different. The metastable structure 
fluoresces near 2.6 eV, and has a relatively sharp absorption 
edge at only a slightly higher energy, i.e. a fairly small 
Stokes shift, The quantum chemical calculations predict 
that the Si plane present in the metastable form is a 
direct-gap semiconductor with a gap of 2.7 eV at the 
point, broadly consistent with the experimental data. The 
annealed structure fluoresces in the red, near 2 eV, and 
shows a much broader absorption edge and a very much 
larger Stokes shift. This is much less like the properties 
1201 
"18.705 7 
1.70t v = 
T 
1,501 1 ohm-cm porous Si 
1.45 T a6:H (Yamasakl et al.) 
1.40 
100 200 300 400 500 600 
Temperature ["Cl 
Figure 18. A comparison of the measured shift of the 
luminescence spectra of amorphous and porous silicon as 
a result of annealing. In both cases the loss of hydrogen 
from the system is implicated. 
producing quantum size effects. The same report also 
detailed measurements of p-type material which was subject 
to cyclic (atmospheric) oxidation and HF dipping. This is 
without doubt an obvious way to thin the microstructure, 
and pore enlargement with an associated reduction in 
average particle size is to be expected. However, what 
was in fact observed was a cyclic shift of peak wavelength 
from 720 nm to 680 nm, the shorter wavelength reappearing 
after the HF dip. 
A futther observation [60] relating to size distribution 
is that high-temperature (up to 1200 "C) annealing of 
porous silicon under UHV conditions causes a collapse 
of the microstructure, to .the point where the material 
resembles a collection roughly spherical particles having a 
dimension of a few hundred nanometres. The luminescence 
is also quenched. When such material is dipped in HF, 
luminescence is restored, even though the particle size 
distribution is unchanged. 
These inconsistencies' with the quantum confinement 
model appear to leave the presence of the surface 
hydride species as the only completely consistent factor 
in determining whether or not luminescence is present. 
The plausibility of this idea gains support from work on 
hydrogenated amorphous silicon [61], deposited from the 
vapour phase, with high H content. Luminescence from 
such material is in the range 1.3 to 2.08 eV, and blue shifts 
with increasing H content. This was explained in terms 
of polysilane complexes: (SiHz)" or hydride complexes. 
Wavelength variation may be a feature of the luminescence 
of both entities because the 'gap' of the polysilanes depends 
on the chain length, and the SiH, species, according to 
tight binding calculations, produce bonding states deep in 
the silicon at energies which depend on the H content of 
the molecule [62]. 
On a more practical note, an interesting comparison has 
been made [60] between the red shift of the luminescence 
of porous silicon induced by annealing, (in an argon 
atmosphere), compared with the red shift induced in the 
luminescence of hydrogenated amorphous silicon by similar 
processing [63]. The data are shown in figure 18 [60],
B Hamilton 
expected of a direct gap semiconductor, and is similar 
to the measured properties of porous silicon. Another 
similarity between annealed siloxene and porous silicon 
lies in the involvement of the triplet exciton in the low-temperature 
luminescence, probed by optically detected 
magnetic resonance [66]. These experiments also support 
the idea that the exciton is strongly localized, on a scale 
compatible with the size of the sixfold Si ring expected to 
be present in annealed siloxene. 
It has been proposed that the sixfold silicon ring is the 
basic luminescence 'centre' for both annealed siloxene and 
porous silicon [65]. The spectral properties of properties of 
both are similar and the EPR measurements have identified 
the Si dangling bond as the key non-radiative shunt for 
both. It has been argued that coalescing pores will produce 
fragmentation of monolayer silicon that could lead to the 
formation of the ring structure. Furthermore it was noted 
by the authors of [65] that such a process might be a 
simple explanation for the fact that strong luminescence has 
been reported from porous amorphous silicon [671, which 
appears to relax crystallinity as an absolute prerequisite for 
the luminescence. 
The siloxene model, like the quantum confinement 
model, has many appealing features. However, at the 
present time it seems not to explain in a simple way 
the smooth shift of wavelength with porosity which is 
probably the key result. It must he remembered that this 
band can be reliably tuned down to less than 1.4 eV. A 
more complete statement seems to be required about the 
way in which the sixfold silicon ring, or perhaps some 
perturbation to it, might allow such gross tunability. In 
contrast, wavelength tunability with size is a natural feature 
of quantum confinement. 
6. Oxidized porous silicon 
Oxidation of porous silicon has received much attention 
recently mainly because it produces stable material with 
additional emission at short wavelengths, often referred to 
as the blue or fast band. Not only is this luminescence 
relatively immune to thermal degradation but it exhibits 
decay times in the nanosecond regime, a potentially useful 
attribute for devices. It must be stressed, though, that 
the visible or slow band remains (with some spectral 
modification) in oxidized material, and rapid oxidation for 
typically 30 s at 900 "C can give stable material which emits 
strongly at these longer wavelengths. Higher-temperature 
processing than this tends to remove the visible band [5, 
681, leaving only the fast band, and of course a much more 
fully oxidized material structure. 
Figure 20 shows representative spectra of oxidized 
porous silicon. It has been noted by several groups that the 
intensity of the visible band drops rapidly with oxidation 
for oxidizing temperatures up to 600 "C, and then rises 
with processing temperature until the melting point of 
silicon is reached. Figure 20(a) [69] shows that the visible 
band remains, in a broadened blue shifted form, but also 
that a higher energy band emerges which extends into the 
blue. Of course the latter band requires pumping with an 
appropriately short wavelength source. Figure 20(b) shows 
-,- 
Figure 19. (a) The idealized structure of the Si planes in 
as prepared siloxene: the planes are terminated by H or 
OH radicals on opposite sides. (b) Siloxene after the 
ordered insertion of 0. isolated SiB rings now appear, and 
these may be the luminescent centres responsible for 
emission from modified siloxenes. 
1202
Energy (eV) 
Figure 20. Photoluminescence spectra typical of rapidly 
oxidized porous silicon. (a) The broadening and blue 
shifting of the visible band and the appearance of the high 
energy band (from 1691). (b) Besides the visible band (a) 
measured at 10 K, a lower energy band, the infrared band, 
is produced. There is strong competition between the 
infrared and visible bands which depends on temperature; 
spectra (b), (c) and (d) were measured at 10, 70 and 300 K 
respectively. 
another key result; the oxidation also produces a low energy 
band known as the infrared band. This band which is below 
the energy of the bulk gap competes with the visible band, 
but at low temperawes becomes as efficient as, or more 
efficient than, the visible emission. 
Electron paramagnetic resonance has been applied to 
most forms of porous silicon. By far the most important 
defect to be observed is the dangling bond or Pbo centre. 
This centre has been known for some time to be present at 
the SiSi02 interface [70]; it is a [ill] axially symmetric 
system with the unbonded orbital directed along one of the 
four equivalent [ 11 I] directions. The density of this centre 
increases when the visible luminescence is quenched by 
annealing and it accordingly correlates with hydrogen loss 
from the surface [711. The centre has also been detected 
via its influence on the intensity of the visible band as the 
magnetic field is swept through resonance [72]. The general 
conclusion is that the dangling bond is a key non-radiative 
shunt path for the visible band. 
The optically detected magnetic resonance (ODMR) 
experiment shows a large effect for the infrared band; the 
intensity of the hand increases by around 15% at resonance. 
This compares with values of typically 0.01% variation for 
the visible band. The data, as shown in figure 21 [73], 
indicate something slightly more complex than a simple P ~ o 
Porous silicon 
1.22 1.24 1.26 
Magnetic field IT1 
Figure 21. The strong ODMR signal measured for the 
low-energy or IR band [73]. These data prove that the PbO 
centre (together with another centre) is directly involved in 
the luminescence process responsible for the IR band 
centre. The lowest-field peak with an isotropic g value of 
2.013 does not belong to the dangling bond, but may belong 
to a localized hole. The two higher peaks are signatures of 
the Pbo; their g values are 811 = 2.0017 and gl = 2.0085, 
where the parallel direction of !he magnetic field is along 
the [ 11 11 direction. This anisotropy is exactly that found in 
the EPR spectrum of the dangling bond. The strong effect 
in ODMR may imply that the dangling bond is directly 
involved in the infrared band. 
The origin of the blue band is of course of great 
interest. One obvious question to he asked is that since 
it is much faster than the visible band [74] could it be 
the true signature of a direct energy gap in a quantum 
confined system? This idea has been supported by several 
groups who point to the similarities which exist between the 
blue band and emission from direct-gap semiconductors. 
Unfortunately the blue band is most easily seen in 
highly porous (oxidized) material, and so its wavelength 
dependence on porosity cannot be easily explored; i.e. even 
this rather uncertain size control has not been explored for 
particles which remain in the system. Besides which, the 
rather aggressive oxidations used modify the skeleton of 
the porous layer. Other workers have supported the view 
that the blue band originates from SO?. For example, it 
has been reported [69] that the luminescence does not shift 
spectrally with increasing oxidation time (up to 50 min). 
It has also been noted that oxidized planar silicon wafers 
give similar luminescence [75]. In both these references, 
the suggestion that defects in the Si02 glass might be 
responsible for the emission was made. 
7. Electroluminescence and devices 
The goal of research into porous silicon is without doubt the 
realization of an efficient LED. The wavelength range of 
1203
B Hamilton 
Figure 22. An indium tin oxidelporous silicon 
heterojunction LED, showing the device structure and 
electrical characteristics. 
any potential device is actually a secondary consideration 
since any wavelength band has applications. It has become 
painfully apparent that to fabricate such a device is hugely 
difficult, and quantum efficiencies of 10-6 are typical; even 
then large drive voltages, in excess of tens of volts for 
example are often required. 
A variety of structures with solid state contacts have 
been tried, ranging from simple indium tin oxide, to 
attempts at p n junction technology. Figures 22 and 
23 show two devices which are representative of the 
sort of structures appearing in the literature. Figure 22 
demonstrates a simple heterojunction technology using 
conducting (and transparent) indium tin oxide [761. 
Rectification is observed for such contacts with light 
emission occurring only for one polarity of voltage. 
This particular diode structure produced a rather narrow 
emission band, about 20 nm wide, centred on 580 nm, 
significantly narrower than a typical room-temperature 
photoluminescence spectrum. Although the reasons for 
this are not clear, it may he that the electroluminescence 
excitation excites only a subset of the porous layer. 
A device based on a porous p-n technology is shown 
in figure 23 [77]. This device is based on an n-type 
wafer with a surface-implanted pf layer. Anodization was 
carried out using illumination so that the n-type material 
as well as the implanted p layer became porous. The 
vertical porosity profile depends very much on the doping 
profile, but the active region was thought to be a nanoporous 
region which straddled the metallurgical p n junction. 
The quantum efficiency of the device was measured to 
be lO-4. An interesting attribute of this structure is 
that the emission hand seen in electroluminescence can 
he tuned by the wavelength used to excite the layer 
during the anodization process; the shorter the excitation 
wavelength, the shorter the emission wavelength. Although 
the reasons for this are not completely established, the 
authors point out that in order to supply holes in porous 
1204 
Figure 23. A porous Si p n ju nction device fabricated by 
anodization under illumination of a boron implanted n-type 
wafer. (a) Device structure and I-V characteristics. 
(b) The electroluminescence output for devices anodized 
with different wavelengths of light. 
n-type material, and hence for anodization to proceed, 
the exciting photon must be strongly absorbed within 
the small silicon particles characteristic of nanoporous 
material. Following the qualitative expectations of quantum 
confinement, very small particles would require short 
wavelengths for absorption by the opened up energy 
gaps. Whatever the detail, it was demonstrated that 
electroluminescence ranging from the infrared to blue could 
be produced by varying the anodization excitation from 
infrared to ultraviolet. The electroluminescence data are 
also shown in figure 23. 
A variety of device structures have been tried, but the 
key result, the external quantum efficiency, remains low. It 
seems an obvious conclusion that the contact technology to 
date fails because it does not provide volume excitation of 
the porous film. What is required is a contact technology 
which transports energy into the whole of the film and 
then facilitates local minority carrier injection. Such a 
scheme would also demand a continuous current path to the 
wafer. Of course optical excitation fulfils all of the required 
conditions without the need for current continuity; it is the 
perfect excitation source. It has become increasingly clear
Wavelength (nm) 
Figure 24. (a) The transient (integrated) light emission 
from porous silicon during anodization. (b) The spectral 
detail of the light emitted during the transient period. 
that some innovation in solid state contact philosophy is 
required in order to achieve better success. 
Liquid contacts represent a state of suitability which is 
intermediate between photon beams and solid state contacts, 
and the study of such systems has yielded some important 
results for our understanding of luminescence processes in 
porous silicon. This work has utilized electrolytic solutions 
in order to transport charge carriers to the interior of the 
porous film. In 'fact the anodic dissolution process used 
to create the porous layer is accompanied by some rich 
luminescence detail 178, 791. The key observations for 
the light emitted during the anodic oxidation process are 
shown in figure 24. Firstly there is a time delay before 
any emission is observed and the authors ascribe this to 
the fact that the porosity is simply too low in the early 
stages of anodization for any light to be seen. After this 
delay, the luminescence builds up to some peak value and 
then decays to zero. The quenching is always associated 
with a sharp rise in the required anodic potential (to 
sustain a constant anodic current). Although it is estimated 
from the total anodic charge transferred (Qo in figure 24) 
that the film porosity is only around 50% at the point 
of quenching, the authors conclude that at this point the 
electrical connectivity between the small crystallites in the 
film, which are assumed to be the source of light, is broken. 
The electrochemical activity then switches to oxidation of 
the base of the pores; a process which is associated with 
the increased anodic potential. Actually, light emission is 
Porous silicon 
also observed in this high anodic potential regime, but it 
is of short wavelength and is thought to be associated with 
processes occurring in the oxide layers 1801. 
The spectral distribution of the light transiently emitted 
during the initial stages of anodization, though, shows 
remarkable similarity to the visible or slow band, and 
it would be surprising if its origin were not the same 
as the optically pumped emission. This is shown in 
figure 24, which also shows that as anodization proceeds 
the luminescence peak blue shifts. This shift has been 
analysed by the authors in terms of a tunnelling-limited 
escape mechanism for excited carriers which leads to non-radiative 
recombination. A model was described for a 
tunnelling process occurring through small regions which 
connect optically active crystallites to non-radiative bulk 
silicon sinks. The yodel demonstrates that because these 
regions, which may be even smaller than the crystallites, 
are particularly sensitive to shape (and hence bandgap) 
changes, the predicted variation of tunnelling flux with 
time would produce the observed blue shift. They 
also commented that the oxidation-induced thinning of 
nanopdcles mentioned above cannot account for the blue 
shift: it would be too small. These results then provide 
some additional insight into the non-radiative processes, 
and are supported by the qualitative observation that the 
radiative efficiency of the fluorescent species, perhaps the 
nanoparticles, seems to increase as the porosity and hence 
the barriers to tunnelling transport increases. 
Strictly speaking, many would wish to label these 
observations of light emission during anodization as 
chemiluminescence, rather than electroluminescence, and 
there do remain some problems in understanding the 
electron injection process in the overall ,scheme of the 
luminescence. The hole of course is provided directly 
by the anodic reaction. One possibility for the electron 
supply channel may lie in the existence of intermediate 
species being formed during the oxidation process of the 
silicon wafer; such intermediate species may have energy 
levels high enough to allow electron injection [SI]. The 
key point here is that the structure is undergoing chemical 
transformation during the light emission process. However. 
luminescence using liquid contacts has also been achieved 
under cathodic conditions [82], and cathodic injection does 
not modify the chemistry of the material. The principle 
of liquid contact electroluminescence is therefore firmly 
established. 
8. Conclusions 
Research into the physics of light emission from porous 
silicon is motivated by a very practical desire to extend 
the functionality of silicon. The fact that enthusiasm 
remains strong for research into this complex material is a 
testimony to this huge technological prize and the challenge 
of bringing together the impressively wide ranging 
interdisciplinary tools needed to deal with structures of 
almost fractal complexity. It is the sheer scale of the 
complexity which has limited the interpretation of data and 
the success of attempts to make devices. 
1205
B Hamilton 
As regards the basic luminescence mechanism. it 
would be wrong to give the impression that views do 
not remain divided; quantum structures, surface hydride 
species, amorphous silicon and more complex molecular 
arrangements like siloxene derivatives all have their 
protagonists. It may be that a combination of some or 
even all these are present in porous silicon, but it is 
more satisfying to think that there is a single mechanism 
operating. Such satisfaction is rooted in the traditional 
approach of physics, which finds elegance in obtaining the 
most complete solution possible for model systems. This 
luxury, though, is not available to, say, biologists who deal 
with systems in which complexity is the intrinsic dominant 
feature. 
On balance at the present level of knowledge, the 
quantum confinement model has most support. It is easy to 
understand why this should be the case: size confinement 
and the associated energy shifts are absolutely fundamental 
features of quantum mechanics. So, whatever additional 
mechanisms might be invoked in porous silicon, quantum 
size effects should occur if very small singlecrystal entities 
are formed. The quantum particle or wire hypothesis for 
the optical activity, then, is viewed as a development of 
a straightforward and basic rule of physics. Of course, in 
practice this intrinsic effect may be overwhelmed by the 
optical activity of other chemical species, or there may 
be reasons why quantum sized crystallites simply do not 
fluoresce with adequate efficiency to explain the obsewed 
light emission. Such objections, though, are details, 
especially since the alternative hypotheses are on the whole 
more complex than quantum confinement involving exotic 
molecular species for which detailed knowledge of the 
electronic structure may be missing. 
The current debate therefore tends support the view 
that whilst the jury is still out on all of the possible 
mechanisms, the quantum confinement model remains the 
simplest explanation: and until better or more innovative 
experimental techniques prove otherwise, simplicity holds 
sway. 
Looking back over the past several years of research 
into porous silicon, however, it seems that this effort 
is very much part of a paradigm shift in condensed 
matter physics, and particularly in materials research. 
The movement towards understanding materials systems 
of great complexity is now evidenr One can think of 
numerous examples, including semicofiductor superlattices, 
high-T, superconducting materials, hierarchical biological 
structures characteristic of living matter, and many more. 
Probably this research will in the end prove to be valuable 
even if devices do not result, because at the very least it has 
led to one of the most successful periods interdisciplinary 
work and experimental development for many years. 
Acknowledgments 
The author wishes acknowledge the financial support of the 
EPSRC and from the ESPRIT programme for maintaining 
his involvement in porous silicon research. Special thanks 
are due to his collaborators Ursel Bangert, Phil Dawson, 
Spyros Gardelis and Robert Pettifer for their unstinting 
efforts and critical approach. 
1206 
References 
111 Canham L T 1990 A ~ o lP.h vs. Left. 57 1046 
[Zi Takagi H, Ogawa H;Yazaki Y, Ishizai A and Nakagiri T 
1990 Awl. Phys. Luff. 56 2379 
[3] Yablonovilch E ahd Gmitler T 1986 AppL Phys Left. 49 
587 
[4] Davies G 1989 Phys. Rep. 176 83 
[5] Gardelis S and Hamilton B 1994 J. AppL Phys. 76 5328 
[6] Gardelis S, Rimmer I S, Dawson P. Hamilton 6, Kubiak R 
A, Wall T E and Parker E H C 1991 Appl. Phys. Lett. 
59 2118 
171 Canham L T 1993 Optical Properties of Lmy Dimensional 
Silicon Structures (NATO AS1 Series, vol 244) 
(Dordrecht: Kluwer Academic) p 81 
Appl. Phys Lett. 60 639 
Broomhead D 1993 3. Phys.: Condens. Uutfer 5 L91 
[SI llschler M A, Collins R T, Stathis J H and Tsang J C 1992 
[9] Calcott P D I, Nash K J, Canham L T, Kane W I and 
[IO] Skolnick M S , Tapsler P R, Bass S I. Piu A D, Apsley N 
and Aldred S P 1986 Semicond. Sci. TechnoL 1 1455 
[ll] Gardelis S and Hamilton B 1992 Mater. Res. Soc. Symp. 
Proc. 256 149 
[I21 Wang L, Wilson M T, Goorsky M S and Haegel N M 
Mater. Res. Soc. Symp. Proc. 256 13 
1131 Tsu R and Babic C 1993 Oprical Properties of Low 
DimensionaI Silicon Structures (NATO AS1 Sene& vol 
244) (Dordrecht: Kluwer Academic) o 179 
[I41 Tumer D R 1958 3. Electrochem. Soc. CO5 402 
1151 Gardelis S 1993 PhD Thesis UMIST 
(161 Memming Rand Schwandt G 1966 Surf: Sci. 4 104 
[17] Tumer D R 1960 Suiface Chemistry ofMetuls and 
Semiconductors ed H C Gatos (New York Wiley) p 82 
[181 Tumer D R 1961 Elenrochemistry of Semiconductors 
ed P J Holmes (New York: Academic) p 161 
[19] Unagami T 1980 3. Elecfrochpm. Soc. 127 476 
[ZO] Smith R Land Collins S D 1992 J. Appl. Phys. 71 R1 
17.11 Dewald I F 1960 The Surface Chemistry ofMefals and 
Semiconducfors ed H C Gams (New York: Wiley) p 78 
[22] Foll H 1991 Appl. Phys. Left. A 53 8 
[23] Bang X G and Collins S D 1989 J. Elecfmchem. Soc. 136 
1561 
1241 Beale MI, Chew N G, Uren M 1, Cu!Jis A G and 
Benjamin J D 1985 AppL Phys. Lett. 46 86 
[25] Beale M 1, Benjamin I D, Uren MI, Chew N G and Cullis 
A G 1985 J. Crystal Growrh 73 622 
[261 Smith R L, Chuang S F and Collins S D 1988 J. Electron. 
Muter. 17 533 
[27l Lehmann V and Gosele U 1991 Appl. Phys. Left. 58 865 
[281 Parkhutik V P, Clinenko L K and Labunov V A 1983 Su$ 
Technol. 20 265 
12.91 Parkhutik V P, Martinez-Duart J M and Albella I M 1993 
Optical Properties of Low Dimensional Silicon Structures 
(NATO AS1 Series, ~01244p) ordrecht: Kluwer 
Academic) p 55 
[30] Bomchil G, HaIimaoui A and Herino R 1989 Appl. Su$ 
Sci. 41/42 604 
[31] Hcrino R, Bomchil G. Barla K and Benrand C 1987 
J. Electrochem. Soc. 134 1994 
(321 Beale M I J, Benjamin I D, Uren M J, Chew N G and 
Cullis A G 1985 J. Crystal Growrh 76 622 
[33] L'ECuyer J D. Lorreto M H. Far J P G, Keen 1 M, 
Castledine J G and L'Esperance G 1988 Murer. Res. 
Soc. Symp. Proc. 107 441 
[34] Cullis A G 1993 Optical Properfies ofLow Dimensional 
Silicon Stmctures (NATO AS1 Series, vol 244) 
(Dordrecht: Kluwer Academic) p 147 
[35] Gupla P, Colvin V L and George S M 1988 Phys. Rev. B 
37 8234 
[36] Kat0 Y, Toshimichi I and Hiraki A 1988 Japan. 3. Appl. 
Phys. 27 L1046
1.371. C habal Y. Hieashi G S. Raehavachari K and Burrows V A 
1989 3. VaE Sci. Technor A 7 2104 
1.38.1 V enkateswara R A, Ozanam F and Chazalviel J N 1988 
3. Eiecrrochem. Soc. 138 153 
2569 
1391 Brandt M S and Stutzmann M 1992 Appl. Phys. Lett. 61 
[40] Poindexter I H and Caplan P 1 1983 Prog. Sud Sci. 14 201 
[41] Konishi T N, Yao T, Tajima M, Ohshima H, It0 H and 
Hattori T 1991 Japan J. Appi. Phys. 31 L1216 
[42] Canham L T, Houlton M R, Leong W Y, Pickering C and 
Keen I M 1990 J. April. Phvs. 68 2187 
I431 VSquez R P, Fathauer'R W, George T, Ksendzov A and 
[44] Roy A, Chainani, A Sarma D D and Sood A 1990 Appl. 
Lin T L 1992 Appl. Phys. Lett. 60 1004 
Phvs. Lett. 61 2187 
[45] Fauchet P M and Campbell I H 1988 Crit. Rev. Solidstate 
Mater. Sci. 14 7 
[46] Littau K A et a1 1993 J. Chem. Phys. 97 1224 
[47] BNS L 1991 Appi. Phys. Lett. A53 465 
[48] Prokes S M, Glembocki 0 1, Bermudez V M, Kaplan R. 
Friedersdorf L E and Searson P C 1992 Phys. Rev. B 45 
13788 
Gavrilenko V 1992 Mater. Res. Soc. Symp. Proc. 283 
[49] Koch F, Petrova-Koch T, Muschik T, Nikolov A and 
107 .,I 
(501 Petrova-Koch V, Muschik T. Kux A, Meyer B K and 
Koch F 1992 Appl. Phys. Lett. 61 943 
[SI] Fauchet P M. Ettedgui E;Raisanen A, Brillson L J, Seifirth 
F. Kurinec S K, Gao Y, Pene C and Tsvbeskov L 1993 
Mater. Res. Soc. Symp. Proc.298 271 . 
1.521_ Li K H. Choachieh T. Camobell J C. Kovar M and White 1 
M 1492 Mater. Res. Soc.'Symp. Pmc. 298 173 
[53] Hyberslein M S 1992 Mater. Res. Soc. Symp. Pmc. 256 
179 
[54] Xie Y H, Wilson W L. Ross F M, Miucha J A, Fitzgerald 
E A, Macaulay J M and Hams T D 1992 3. Appl. Phys. 
B 71 2403 
[55] Read A 1, Needs R 1, Nash K 1, Canham L T, Calcotl P D 
J and Qteish A 1994 Phys. Rev. Lett. 69 1232 
[56] Delerue C, Allen G and Lannoo M 1993 Phys. Rev. B 48 
48 
I571 Furakawa S and Miyasato T 1988 Phys. Rev. B 38 5726 
[58] Chelikowsky J and Phillips J C 1989 Phys. Rev. Lett. 63 
[59] Prokes S M. Glembocki 0 1, Bermudez V M, Kaplan R, 
I653 
Freidersdorf L E and Searson P C 1992 Mater. Res. Soc. 
Symp. Proc. 256 107 
[60] Prokes S M 1993 J. Appl. Phys. 73 407 
1611 Wolford D J. Scott B A, Reimer J A and Bradley I A 1983 
Physica B 117 + 118 920 
Faoaconstantoooulos D A and Economou E N 1981 Phw. 
Rev. B U j223 
Unaeami T and Seki M 1980 J. Electrochem Soc. 125 1339 
Brandt M S. Fuchs H D. Stutzman M, Weber J and 
Cardona M 1992 Solid State Commw. 81 307 
Deak P. Rosenbauer M, Stutzman M, Weber J and Brandt 
M S 1992 Phvs. Rev. Lett. 69 2531 
Brandt M S, Rosenbauer M and Stutzman M 1993 Mater. 
Buslarret E, Ligeon M and Orlega L 1992 Solid State 
Cullis A G, Canham L T, Williams G M, Smith P W and 
Prokes S M 1993 Appl. Phys. Lett. 62 3224 
Brower K L and Headley T J 1986 Phys. Rev. B 34 3610 
Meyer B K, Hofman D M, Stadler W, Petrova-Koch V, 
Koch F, Omelina P and Emanuelsson P 1993 AD^ 
Phys. Lett. 63 2120 
Brandt M S and Stutzman M 1992A.m. i. Phvs. Lett. 61 
2569 
Kux A and Hofmann D V 1993 Opticai Properties of Lou, 
Dimensional Silicon Structures (NATO AS1 Series, vol 
244) (Dordrecht: Kluwer Academic) p 197 
Kovalev D I, Yaroshetzkii I D, Muschik T, Petrova-Koch 
V and Koch F 1994 Appi. Phys. Lett, 64 214 
Kontkiewicz A I, Konlkiewicz A M, Siejka 1, Sen S, 
Nowak G, Hoff A M. Sakthivel P, Ahmed K, Mukhejee 
P, Wtanachchi and Lagowski J 1994 AppL Phys. Left. 
65 1436 
Kakhoran N M, Namvar F and Maruska H P 1992 Mater. 
Kozlowski F, Steiner P and Lang W 1993 Optical 
Res. Soc. Symp. Pmc. 256 89 
Properties of Low Dimensional Silicon Structures 
(NATO AS1 Series. vol 2.44) (Dordrecht: Kluwer 
Academic) p 123 
Halimaoui A. Bomchil G, Odes C, Bsiesy A, Gaspard F, 
Herino R. Ligeon M and Muller F 1991 Appl Phys. 
I A t s9 3fl4 
I 
1 
Res. Soc. Symp. Proc. 298 301 
Commun. 83 461 
Dosser 0 D 1993 Mater. Res. Soc. Symp. Proc. 283 182 
-. ... . . . . . 
Porous silicon 
.. 
1791 Muller F, Herino R, Ligeon M, Billat S, Gaspard F, 
Romestain R, Vial J C and Bsiesy A 1993 Optical 
Properties of Low Dimensional Silicon Structum 
(NATO AS1 Series, vol 244) (Dordrecht: Kluwer 
Academic) p 101 
[801 Billat S. Bsiesy A. Gaspard F, Herino R, Ligeon M, Muller 
F. Romestain R and Vial J C 1992 Mater. Res. Soc. 
Symp. Proc. 256 215 
J. Electroanal. Chem. 290 229 
Taylor L 1992 Appl. Phys. Lett. 61 2563 
[811 Peter L M, Borazio A, Levemez H J and Stumpcr J 1990 
[82] Canham L T, Leong W Y, Beale M I 1, Cox T I and 
1207

More Related Content

What's hot

Linear effects in optical fibers
Linear effects in optical fibersLinear effects in optical fibers
Linear effects in optical fibersCKSunith1
 
Epitaxial growth
Epitaxial growthEpitaxial growth
Epitaxial growthIYPUMANI
 
High speed semiconductor devices
High speed semiconductor devices High speed semiconductor devices
High speed semiconductor devices AshishJoshi289524
 
DYE SENSITIZED SOLAR CELLS
DYE SENSITIZED SOLAR CELLSDYE SENSITIZED SOLAR CELLS
DYE SENSITIZED SOLAR CELLSThiru Ram
 
Presentation on oled technology
Presentation on oled technologyPresentation on oled technology
Presentation on oled technologyEGHANATHANS
 
Surface Energy Influence On Ion Sputtering Process
Surface Energy Influence On Ion Sputtering ProcessSurface Energy Influence On Ion Sputtering Process
Surface Energy Influence On Ion Sputtering ProcessDaniel Reilly
 
Ic tech unit 5- VLSI Process Integration
Ic tech unit 5- VLSI Process IntegrationIc tech unit 5- VLSI Process Integration
Ic tech unit 5- VLSI Process Integrationkriticka sharma
 
V mn-mcm-41 catalyst for the vapor phase oxidation of o-xylene
V mn-mcm-41 catalyst for the vapor phase oxidation of o-xyleneV mn-mcm-41 catalyst for the vapor phase oxidation of o-xylene
V mn-mcm-41 catalyst for the vapor phase oxidation of o-xylenesunitha81
 
Energia r.p.h.chang
Energia r.p.h.changEnergia r.p.h.chang
Energia r.p.h.changCesar Diaz
 
Vaneet Sharma Carbon Nanotubes
Vaneet Sharma  Carbon NanotubesVaneet Sharma  Carbon Nanotubes
Vaneet Sharma Carbon Nanotubesvsharma78
 
Fabrication of microelectronic devices
Fabrication of microelectronic devicesFabrication of microelectronic devices
Fabrication of microelectronic devicesThulasikanth Vaddi
 
Llobet et al APL 107 223501 (2015)
Llobet et al APL 107 223501 (2015)Llobet et al APL 107 223501 (2015)
Llobet et al APL 107 223501 (2015)Chen Wang
 
Lect2 up020 (100324)
Lect2 up020 (100324)Lect2 up020 (100324)
Lect2 up020 (100324)aicdesign
 

What's hot (20)

SiOx Nanoparticals.PDF
SiOx Nanoparticals.PDFSiOx Nanoparticals.PDF
SiOx Nanoparticals.PDF
 
Mj 3 Dvlsi
Mj 3 DvlsiMj 3 Dvlsi
Mj 3 Dvlsi
 
Linear effects in optical fibers
Linear effects in optical fibersLinear effects in optical fibers
Linear effects in optical fibers
 
Epitaxial growth
Epitaxial growthEpitaxial growth
Epitaxial growth
 
Ssntd ion track technology to nanotechnology
Ssntd ion track technology to nanotechnologySsntd ion track technology to nanotechnology
Ssntd ion track technology to nanotechnology
 
High speed semiconductor devices
High speed semiconductor devices High speed semiconductor devices
High speed semiconductor devices
 
DYE SENSITIZED SOLAR CELLS
DYE SENSITIZED SOLAR CELLSDYE SENSITIZED SOLAR CELLS
DYE SENSITIZED SOLAR CELLS
 
Presentation on oled technology
Presentation on oled technologyPresentation on oled technology
Presentation on oled technology
 
Surface Energy Influence On Ion Sputtering Process
Surface Energy Influence On Ion Sputtering ProcessSurface Energy Influence On Ion Sputtering Process
Surface Energy Influence On Ion Sputtering Process
 
Ic tech unit 5- VLSI Process Integration
Ic tech unit 5- VLSI Process IntegrationIc tech unit 5- VLSI Process Integration
Ic tech unit 5- VLSI Process Integration
 
V mn-mcm-41 catalyst for the vapor phase oxidation of o-xylene
V mn-mcm-41 catalyst for the vapor phase oxidation of o-xyleneV mn-mcm-41 catalyst for the vapor phase oxidation of o-xylene
V mn-mcm-41 catalyst for the vapor phase oxidation of o-xylene
 
New Material:Perovskites presentation
New Material:Perovskites presentationNew Material:Perovskites presentation
New Material:Perovskites presentation
 
Energia r.p.h.chang
Energia r.p.h.changEnergia r.p.h.chang
Energia r.p.h.chang
 
Vaneet Sharma Carbon Nanotubes
Vaneet Sharma  Carbon NanotubesVaneet Sharma  Carbon Nanotubes
Vaneet Sharma Carbon Nanotubes
 
Niquelado
NiqueladoNiquelado
Niquelado
 
bio sensoring
bio sensoringbio sensoring
bio sensoring
 
IC PROCESSING
IC PROCESSING IC PROCESSING
IC PROCESSING
 
Fabrication of microelectronic devices
Fabrication of microelectronic devicesFabrication of microelectronic devices
Fabrication of microelectronic devices
 
Llobet et al APL 107 223501 (2015)
Llobet et al APL 107 223501 (2015)Llobet et al APL 107 223501 (2015)
Llobet et al APL 107 223501 (2015)
 
Lect2 up020 (100324)
Lect2 up020 (100324)Lect2 up020 (100324)
Lect2 up020 (100324)
 

Similar to P si

Presentation1 seminar defense
Presentation1 seminar defensePresentation1 seminar defense
Presentation1 seminar defensedemewez amtate
 
Opal and inverse opal structures for optical device applications
Opal and inverse opal structures for optical device applicationsOpal and inverse opal structures for optical device applications
Opal and inverse opal structures for optical device applicationsM. Faisal Halim
 
Nano tubes technology in solar
Nano tubes technology in solarNano tubes technology in solar
Nano tubes technology in solarsriviswanadh gubba
 
24 ijaprr vol1-3-32-37nasir
24 ijaprr vol1-3-32-37nasir24 ijaprr vol1-3-32-37nasir
24 ijaprr vol1-3-32-37nasirijaprr_editor
 
Application of Photoluminescence
Application of PhotoluminescenceApplication of Photoluminescence
Application of PhotoluminescenceMonica Chen
 
Moletronics
MoletronicsMoletronics
Moletronicsmaddyz03
 
The Future Of Solar Technology
The Future Of Solar TechnologyThe Future Of Solar Technology
The Future Of Solar TechnologyHeidi Owens
 
Silicon Falls Into Line
Silicon Falls Into LineSilicon Falls Into Line
Silicon Falls Into Lineioneec
 
Making of a silicon chip
Making of a silicon chipMaking of a silicon chip
Making of a silicon chipsurabhi8
 
CIVE685_SolarPotovoltaics_07Oct20-Updated.pdf
CIVE685_SolarPotovoltaics_07Oct20-Updated.pdfCIVE685_SolarPotovoltaics_07Oct20-Updated.pdf
CIVE685_SolarPotovoltaics_07Oct20-Updated.pdfSara972447
 
International Journal of Engineering and Science Invention (IJESI)
International Journal of Engineering and Science Invention (IJESI)International Journal of Engineering and Science Invention (IJESI)
International Journal of Engineering and Science Invention (IJESI)inventionjournals
 
Design and Analysis of Thin Film Silicon Solar cells Using FDTD Method
Design and Analysis of Thin Film Silicon Solar cells Using FDTD MethodDesign and Analysis of Thin Film Silicon Solar cells Using FDTD Method
Design and Analysis of Thin Film Silicon Solar cells Using FDTD MethodDr. S. Saravanan
 
Silicon Photonics: A Solution for Ultra High Speed Data Transfer
Silicon Photonics: A Solution for Ultra High Speed Data TransferSilicon Photonics: A Solution for Ultra High Speed Data Transfer
Silicon Photonics: A Solution for Ultra High Speed Data TransferIDES Editor
 
solar cell by jerox
solar cell by jeroxsolar cell by jerox
solar cell by jeroxjaygo91
 
Prospect of nano electronics material for marine energy
Prospect of nano electronics material for marine energyProspect of nano electronics material for marine energy
Prospect of nano electronics material for marine energyOlanrewaju O Sulaiman
 
New organic infiltrants for 2-D and 3-D photonic crystals
New organic infiltrants for 2-D and 3-D photonic crystalsNew organic infiltrants for 2-D and 3-D photonic crystals
New organic infiltrants for 2-D and 3-D photonic crystalsKonstantin Yamnitskiy
 

Similar to P si (20)

Presentation1 seminar defense
Presentation1 seminar defensePresentation1 seminar defense
Presentation1 seminar defense
 
Opal and inverse opal structures for optical device applications
Opal and inverse opal structures for optical device applicationsOpal and inverse opal structures for optical device applications
Opal and inverse opal structures for optical device applications
 
Nano tubes technology in solar
Nano tubes technology in solarNano tubes technology in solar
Nano tubes technology in solar
 
24 ijaprr vol1-3-32-37nasir
24 ijaprr vol1-3-32-37nasir24 ijaprr vol1-3-32-37nasir
24 ijaprr vol1-3-32-37nasir
 
Pdoc
PdocPdoc
Pdoc
 
Application of Photoluminescence
Application of PhotoluminescenceApplication of Photoluminescence
Application of Photoluminescence
 
Moletronics
MoletronicsMoletronics
Moletronics
 
D0321624
D0321624D0321624
D0321624
 
The Future Of Solar Technology
The Future Of Solar TechnologyThe Future Of Solar Technology
The Future Of Solar Technology
 
Silicon Falls Into Line
Silicon Falls Into LineSilicon Falls Into Line
Silicon Falls Into Line
 
Making of a silicon chip
Making of a silicon chipMaking of a silicon chip
Making of a silicon chip
 
CIVE685_SolarPotovoltaics_07Oct20-Updated.pdf
CIVE685_SolarPotovoltaics_07Oct20-Updated.pdfCIVE685_SolarPotovoltaics_07Oct20-Updated.pdf
CIVE685_SolarPotovoltaics_07Oct20-Updated.pdf
 
International Journal of Engineering and Science Invention (IJESI)
International Journal of Engineering and Science Invention (IJESI)International Journal of Engineering and Science Invention (IJESI)
International Journal of Engineering and Science Invention (IJESI)
 
Design and Analysis of Thin Film Silicon Solar cells Using FDTD Method
Design and Analysis of Thin Film Silicon Solar cells Using FDTD MethodDesign and Analysis of Thin Film Silicon Solar cells Using FDTD Method
Design and Analysis of Thin Film Silicon Solar cells Using FDTD Method
 
Silicon Photonics: A Solution for Ultra High Speed Data Transfer
Silicon Photonics: A Solution for Ultra High Speed Data TransferSilicon Photonics: A Solution for Ultra High Speed Data Transfer
Silicon Photonics: A Solution for Ultra High Speed Data Transfer
 
solar cell by jerox
solar cell by jeroxsolar cell by jerox
solar cell by jerox
 
Prospect of nano electronics material for marine energy
Prospect of nano electronics material for marine energyProspect of nano electronics material for marine energy
Prospect of nano electronics material for marine energy
 
Silicon technology
Silicon technologySilicon technology
Silicon technology
 
Physical chem
Physical chemPhysical chem
Physical chem
 
New organic infiltrants for 2-D and 3-D photonic crystals
New organic infiltrants for 2-D and 3-D photonic crystalsNew organic infiltrants for 2-D and 3-D photonic crystals
New organic infiltrants for 2-D and 3-D photonic crystals
 

Recently uploaded

Get Premium Pimple Saudagar Call Girls (8005736733) 24x7 Rate 15999 with A/c ...
Get Premium Pimple Saudagar Call Girls (8005736733) 24x7 Rate 15999 with A/c ...Get Premium Pimple Saudagar Call Girls (8005736733) 24x7 Rate 15999 with A/c ...
Get Premium Pimple Saudagar Call Girls (8005736733) 24x7 Rate 15999 with A/c ...MOHANI PANDEY
 
Top Rated Pune Call Girls Shirwal ⟟ 6297143586 ⟟ Call Me For Genuine Sex Ser...
Top Rated  Pune Call Girls Shirwal ⟟ 6297143586 ⟟ Call Me For Genuine Sex Ser...Top Rated  Pune Call Girls Shirwal ⟟ 6297143586 ⟟ Call Me For Genuine Sex Ser...
Top Rated Pune Call Girls Shirwal ⟟ 6297143586 ⟟ Call Me For Genuine Sex Ser...Call Girls in Nagpur High Profile
 
Low Rate Call Girls Nashik Vedika 7001305949 Independent Escort Service Nashik
Low Rate Call Girls Nashik Vedika 7001305949 Independent Escort Service NashikLow Rate Call Girls Nashik Vedika 7001305949 Independent Escort Service Nashik
Low Rate Call Girls Nashik Vedika 7001305949 Independent Escort Service NashikCall Girls in Nagpur High Profile
 
High Profile Call Girls In Andheri 7738631006 Call girls in mumbai Mumbai ...
High Profile Call Girls In Andheri 7738631006 Call girls in mumbai  Mumbai ...High Profile Call Girls In Andheri 7738631006 Call girls in mumbai  Mumbai ...
High Profile Call Girls In Andheri 7738631006 Call girls in mumbai Mumbai ...Pooja Nehwal
 
VIP Call Girls Dharwad 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Dharwad 7001035870 Whatsapp Number, 24/07 BookingVIP Call Girls Dharwad 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Dharwad 7001035870 Whatsapp Number, 24/07 Bookingdharasingh5698
 
Lucknow 💋 Call Girls Adil Nagar | ₹,9500 Pay Cash 8923113531 Free Home Delive...
Lucknow 💋 Call Girls Adil Nagar | ₹,9500 Pay Cash 8923113531 Free Home Delive...Lucknow 💋 Call Girls Adil Nagar | ₹,9500 Pay Cash 8923113531 Free Home Delive...
Lucknow 💋 Call Girls Adil Nagar | ₹,9500 Pay Cash 8923113531 Free Home Delive...anilsa9823
 
VVIP Pune Call Girls Warje (7001035870) Pune Escorts Nearby with Complete Sat...
VVIP Pune Call Girls Warje (7001035870) Pune Escorts Nearby with Complete Sat...VVIP Pune Call Girls Warje (7001035870) Pune Escorts Nearby with Complete Sat...
VVIP Pune Call Girls Warje (7001035870) Pune Escorts Nearby with Complete Sat...Call Girls in Nagpur High Profile
 
Call Girls in Thane 9892124323, Vashi cAll girls Serivces Juhu Escorts, powai...
Call Girls in Thane 9892124323, Vashi cAll girls Serivces Juhu Escorts, powai...Call Girls in Thane 9892124323, Vashi cAll girls Serivces Juhu Escorts, powai...
Call Girls in Thane 9892124323, Vashi cAll girls Serivces Juhu Escorts, powai...Pooja Nehwal
 
(PARI) Alandi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
(PARI) Alandi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts(PARI) Alandi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
(PARI) Alandi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escortsranjana rawat
 
Call Girls Dubai Slut Wife O525547819 Call Girls Dubai Gaped
Call Girls Dubai Slut Wife O525547819 Call Girls Dubai GapedCall Girls Dubai Slut Wife O525547819 Call Girls Dubai Gaped
Call Girls Dubai Slut Wife O525547819 Call Girls Dubai Gapedkojalkojal131
 
Makarba ( Call Girls ) Ahmedabad ✔ 6297143586 ✔ Hot Model With Sexy Bhabi Rea...
Makarba ( Call Girls ) Ahmedabad ✔ 6297143586 ✔ Hot Model With Sexy Bhabi Rea...Makarba ( Call Girls ) Ahmedabad ✔ 6297143586 ✔ Hot Model With Sexy Bhabi Rea...
Makarba ( Call Girls ) Ahmedabad ✔ 6297143586 ✔ Hot Model With Sexy Bhabi Rea...Naicy mandal
 
Top Rated Pune Call Girls Chakan ⟟ 6297143586 ⟟ Call Me For Genuine Sex Serv...
Top Rated  Pune Call Girls Chakan ⟟ 6297143586 ⟟ Call Me For Genuine Sex Serv...Top Rated  Pune Call Girls Chakan ⟟ 6297143586 ⟟ Call Me For Genuine Sex Serv...
Top Rated Pune Call Girls Chakan ⟟ 6297143586 ⟟ Call Me For Genuine Sex Serv...Call Girls in Nagpur High Profile
 
Book Sex Workers Available Pune Call Girls Yerwada 6297143586 Call Hot India...
Book Sex Workers Available Pune Call Girls Yerwada  6297143586 Call Hot India...Book Sex Workers Available Pune Call Girls Yerwada  6297143586 Call Hot India...
Book Sex Workers Available Pune Call Girls Yerwada 6297143586 Call Hot India...Call Girls in Nagpur High Profile
 
Call Girls Chikhali Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Chikhali Call Me 7737669865 Budget Friendly No Advance BookingCall Girls Chikhali Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Chikhali Call Me 7737669865 Budget Friendly No Advance Bookingroncy bisnoi
 
Kothanur Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Bang...
Kothanur Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Bang...Kothanur Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Bang...
Kothanur Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Bang...amitlee9823
 
VVIP Pune Call Girls Karve Nagar (7001035870) Pune Escorts Nearby with Comple...
VVIP Pune Call Girls Karve Nagar (7001035870) Pune Escorts Nearby with Comple...VVIP Pune Call Girls Karve Nagar (7001035870) Pune Escorts Nearby with Comple...
VVIP Pune Call Girls Karve Nagar (7001035870) Pune Escorts Nearby with Comple...Call Girls in Nagpur High Profile
 

Recently uploaded (20)

Get Premium Pimple Saudagar Call Girls (8005736733) 24x7 Rate 15999 with A/c ...
Get Premium Pimple Saudagar Call Girls (8005736733) 24x7 Rate 15999 with A/c ...Get Premium Pimple Saudagar Call Girls (8005736733) 24x7 Rate 15999 with A/c ...
Get Premium Pimple Saudagar Call Girls (8005736733) 24x7 Rate 15999 with A/c ...
 
Top Rated Pune Call Girls Shirwal ⟟ 6297143586 ⟟ Call Me For Genuine Sex Ser...
Top Rated  Pune Call Girls Shirwal ⟟ 6297143586 ⟟ Call Me For Genuine Sex Ser...Top Rated  Pune Call Girls Shirwal ⟟ 6297143586 ⟟ Call Me For Genuine Sex Ser...
Top Rated Pune Call Girls Shirwal ⟟ 6297143586 ⟟ Call Me For Genuine Sex Ser...
 
Low Rate Call Girls Nashik Vedika 7001305949 Independent Escort Service Nashik
Low Rate Call Girls Nashik Vedika 7001305949 Independent Escort Service NashikLow Rate Call Girls Nashik Vedika 7001305949 Independent Escort Service Nashik
Low Rate Call Girls Nashik Vedika 7001305949 Independent Escort Service Nashik
 
High Profile Call Girls In Andheri 7738631006 Call girls in mumbai Mumbai ...
High Profile Call Girls In Andheri 7738631006 Call girls in mumbai  Mumbai ...High Profile Call Girls In Andheri 7738631006 Call girls in mumbai  Mumbai ...
High Profile Call Girls In Andheri 7738631006 Call girls in mumbai Mumbai ...
 
CHEAP Call Girls in Hauz Quazi (-DELHI )🔝 9953056974🔝(=)/CALL GIRLS SERVICE
CHEAP Call Girls in Hauz Quazi  (-DELHI )🔝 9953056974🔝(=)/CALL GIRLS SERVICECHEAP Call Girls in Hauz Quazi  (-DELHI )🔝 9953056974🔝(=)/CALL GIRLS SERVICE
CHEAP Call Girls in Hauz Quazi (-DELHI )🔝 9953056974🔝(=)/CALL GIRLS SERVICE
 
VIP Call Girls Dharwad 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Dharwad 7001035870 Whatsapp Number, 24/07 BookingVIP Call Girls Dharwad 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Dharwad 7001035870 Whatsapp Number, 24/07 Booking
 
Vip Call Girls Noida ➡️ Delhi ➡️ 9999965857 No Advance 24HRS Live
Vip Call Girls Noida ➡️ Delhi ➡️ 9999965857 No Advance 24HRS LiveVip Call Girls Noida ➡️ Delhi ➡️ 9999965857 No Advance 24HRS Live
Vip Call Girls Noida ➡️ Delhi ➡️ 9999965857 No Advance 24HRS Live
 
Lucknow 💋 Call Girls Adil Nagar | ₹,9500 Pay Cash 8923113531 Free Home Delive...
Lucknow 💋 Call Girls Adil Nagar | ₹,9500 Pay Cash 8923113531 Free Home Delive...Lucknow 💋 Call Girls Adil Nagar | ₹,9500 Pay Cash 8923113531 Free Home Delive...
Lucknow 💋 Call Girls Adil Nagar | ₹,9500 Pay Cash 8923113531 Free Home Delive...
 
🔝 9953056974🔝 Delhi Call Girls in Ajmeri Gate
🔝 9953056974🔝 Delhi Call Girls in Ajmeri Gate🔝 9953056974🔝 Delhi Call Girls in Ajmeri Gate
🔝 9953056974🔝 Delhi Call Girls in Ajmeri Gate
 
CHEAP Call Girls in Mayapuri (-DELHI )🔝 9953056974🔝(=)/CALL GIRLS SERVICE
CHEAP Call Girls in Mayapuri  (-DELHI )🔝 9953056974🔝(=)/CALL GIRLS SERVICECHEAP Call Girls in Mayapuri  (-DELHI )🔝 9953056974🔝(=)/CALL GIRLS SERVICE
CHEAP Call Girls in Mayapuri (-DELHI )🔝 9953056974🔝(=)/CALL GIRLS SERVICE
 
VVIP Pune Call Girls Warje (7001035870) Pune Escorts Nearby with Complete Sat...
VVIP Pune Call Girls Warje (7001035870) Pune Escorts Nearby with Complete Sat...VVIP Pune Call Girls Warje (7001035870) Pune Escorts Nearby with Complete Sat...
VVIP Pune Call Girls Warje (7001035870) Pune Escorts Nearby with Complete Sat...
 
Call Girls in Thane 9892124323, Vashi cAll girls Serivces Juhu Escorts, powai...
Call Girls in Thane 9892124323, Vashi cAll girls Serivces Juhu Escorts, powai...Call Girls in Thane 9892124323, Vashi cAll girls Serivces Juhu Escorts, powai...
Call Girls in Thane 9892124323, Vashi cAll girls Serivces Juhu Escorts, powai...
 
(PARI) Alandi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
(PARI) Alandi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts(PARI) Alandi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
(PARI) Alandi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
 
Call Girls Dubai Slut Wife O525547819 Call Girls Dubai Gaped
Call Girls Dubai Slut Wife O525547819 Call Girls Dubai GapedCall Girls Dubai Slut Wife O525547819 Call Girls Dubai Gaped
Call Girls Dubai Slut Wife O525547819 Call Girls Dubai Gaped
 
Makarba ( Call Girls ) Ahmedabad ✔ 6297143586 ✔ Hot Model With Sexy Bhabi Rea...
Makarba ( Call Girls ) Ahmedabad ✔ 6297143586 ✔ Hot Model With Sexy Bhabi Rea...Makarba ( Call Girls ) Ahmedabad ✔ 6297143586 ✔ Hot Model With Sexy Bhabi Rea...
Makarba ( Call Girls ) Ahmedabad ✔ 6297143586 ✔ Hot Model With Sexy Bhabi Rea...
 
Top Rated Pune Call Girls Chakan ⟟ 6297143586 ⟟ Call Me For Genuine Sex Serv...
Top Rated  Pune Call Girls Chakan ⟟ 6297143586 ⟟ Call Me For Genuine Sex Serv...Top Rated  Pune Call Girls Chakan ⟟ 6297143586 ⟟ Call Me For Genuine Sex Serv...
Top Rated Pune Call Girls Chakan ⟟ 6297143586 ⟟ Call Me For Genuine Sex Serv...
 
Book Sex Workers Available Pune Call Girls Yerwada 6297143586 Call Hot India...
Book Sex Workers Available Pune Call Girls Yerwada  6297143586 Call Hot India...Book Sex Workers Available Pune Call Girls Yerwada  6297143586 Call Hot India...
Book Sex Workers Available Pune Call Girls Yerwada 6297143586 Call Hot India...
 
Call Girls Chikhali Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Chikhali Call Me 7737669865 Budget Friendly No Advance BookingCall Girls Chikhali Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Chikhali Call Me 7737669865 Budget Friendly No Advance Booking
 
Kothanur Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Bang...
Kothanur Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Bang...Kothanur Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Bang...
Kothanur Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Bang...
 
VVIP Pune Call Girls Karve Nagar (7001035870) Pune Escorts Nearby with Comple...
VVIP Pune Call Girls Karve Nagar (7001035870) Pune Escorts Nearby with Comple...VVIP Pune Call Girls Karve Nagar (7001035870) Pune Escorts Nearby with Comple...
VVIP Pune Call Girls Karve Nagar (7001035870) Pune Escorts Nearby with Comple...
 

P si

  • 1. Home Search Collections Journals About Contact us My IOPscience Porous silicon This content has been downloaded from IOPscience. Please scroll down to see the full text. View the table of contents for this issue, or go to the journal homepage for more Download details: IP Address: 148.228.88.57 This content was downloaded on 23/09/2014 at 15:46 Please note that terms and conditions apply.
  • 2. Semicond. Sci. Technoi. 10 (1995) 1187-1207. Printed in the UK ~ ~ ~ TOPICAL REVIEW Porous silicon B Hamilton Department of M60 IQD, UK Physics, UMIST, PO Box 88, Sackville Street, Manchester Received 5 January 1995, accepted for publication 17 March 1995 Abstract. This paper attempts to review the field of research into light emission from porous silicon. The driving force behind such research is the tantalizing goal of adding optoelectronic functions to the already impressive array of electronic functions provided by silicon-based devices. A silicon technology with included light emission would move even closer to complete dominance of the electronics market. After several years of research effort. the fundamental mechanisms of light emission are still not completely resolved. This is not surprising: porous silicon has many attributes of a new and complex material, and its study requires a truly interdisciplinary effort involving electrochemistry, surface science, structural and chemical microscopy on the atomic scale and detailed optical spectroscopy. This paper tries to connect these various threads; inevitably what emerges will only serve as a rather selective 'snapshot' of a still developing and often perplexing field. 1. Introduction The dominance of silicon in the electronics industry is almost complete, at least in terms of volume: the worldwide market for silicon-based devices and systems depending on them is huge. The comparatively small but important markets which silicon does not^ fulfil are those of ultra high-speed devices and optoelectronics, in particular optical communications. In fact virtually all optoelectronic functions requiring high-speed modulation rely on compound semiconductor devices; fibre-optic-based optical communications systems rest firmly on InP-based lasers and modulators, whereas GaAs-based devices supply the near-infrared and visible emission required for short-range communication and disc redwrite functions. It is a curious fact that although silicon is the material which essentially fed the information technology revolution, much of the highly successful international research effort into semiconductor physics during the past 15 years has been devoted to II-V semiconductors. The search for novel physical phenomena based on reduced dimensionality-superlattice, quantum well and latterly quantum wire and dot structures-has been a major driving force for condensed matter physics. Improved device functionality has emerged both in the fields of high-speed transport and optics, which have strengthened the III-V industry in these areas. A picture emerges of a silicon industry dealing with a rather mature technology, able to fulfil many of the growing demands of an information-dependent culture. Materials-based research which will underpin a future silicon indushy currently centres around silicon-germanium heterojunction devices, novel configurations for reduced power consumption in portable systems, cheap thin film 0268-1242/95/091187+21519.50 6 1995 IOP Publishing Ltd devices and nanoscale fabrication. Of course the latter topic holds out the possibility of novel functionality which exploits the quantum regime of electron behaviour, and so connects with some of the work reviewed here. Porous silicon burst into this arena several years ago, offering at least a possibility that silicon technology might eventually yield light-emitting devices. The fact is that under optical excitation. porous silicon does produce light with high efficiency, and furthermore with an emission spectrum which can be 'tuned' from the near-infrared to the green by varying porosity. Further processing by rapid oxidation extends the emission into the bludviolet region of the spectrum. Clearly then, one driving force behind research into light emission from porous silicon is the hope that having finally understood the basic mechanisms, it might be possible to make an electrically excited LED or laser which could, ultimately, be integrated into a complex chip. This notion should not be seen as one of simply enlarging the functionality of silicon; discoveries of new functionality traditionally end up finding applications in semiconductor technology. In their turn, both the rapidly oxidizing silicon surface and the semiconductor laser respectively were dismissed either as a nuisance which would prevent development or as lacking in applications! It would be wrong therefore to rule out, say, optical interconnect applications for porous silicon, provided that it could be developed into a stable electroluminescent system, compatible with integrated circuit processing. In order to even contemplate real applications for porous silicon devices we must understand the basic radiative processes and must have a clear view of how to optimize the porous skeleton and how to control and perhaps to take advantage of its enormous surface area. 1187
  • 3. B Hamilton Finally we must learn how to electrically excite the luminescence. In the final analysis it may prove impossible to achieve these goals, or some other less complex form of optically functional silicon may emerge. Many issues are currently being pursued, including modification of porous silicon and new ways to process material and contacts, and these will be touched upon below. However, the central issue remains the origin of the light, and its relationship to the atomic-scale structure and the associated electronic structure of the porous layer. The main aim of this review, then, is to try to draw together the threads of evidence which are guiding workers in the field towards understanding the physics of the material. The story to date, although largely qualitative, is complicated and has generated lively debate. This being so, it is politic to simply state at the outset the four most commonly held views on the origin of the visible luminescence: (i) The visible and near-infrared light is the result of of quantum confinement shifts of the silicon energy gap due to particle localization in nanometre scale structures (wire or dot) which make up the porous skeleton. (ii) The luminescence originates from surface molecular species which coat the porous skeleton, and which result from the electrochemical processing. (iii) The light originates from radiative decay at surfacdinterface states, the character of which are partly determined by nanocrystaltine particles within the porous layer. (iv) Hydrogenated amorphous silicon is a product of the invasive electrochemistry and is responsible for the emission. In attempting to revicw the field, it is necessary to subdivide the information; first, in section 2, an overview of the main optical phenomena is presented. Section 3 thcn deals, in a.simple way, with the electrochemical process involved in pore formation, leading on to a brief review of pore morphology and microstructure. In section 4 some of the issues involved with the surface of porous silicon are discussed in order to provide a firmer basis for the review of the debate surrounding luminescence mechanisms in section 5. Section 6 deals with some issues concerning oxidized porous silicon which shed light on some fundamental aspects of the material. Finally section I outlines some attempts to make simple device stmctures and also the considerable problems involved. 2. An overview of the optical phenomena associated with porous silicon Highly porous Si, processed using electrochemical etching methods, exhibits strong photoluminescence; efficiencies of several per cent have been routinely reported. Spectra are broad, but peak wavelengths can be 'tuned' over a wide range in the near-infrared and visible, by varying the porosity. These facts, which were first noted by Canham in 1990 [l], remain the key points underpinning a wide-ranging and interdisciplinary research effort. It is interesting to note though that the observation that ultra small silicon crystallites, passivated by hydrogen, and with 1188 Wavevector Figure 1. The energy band structure of crystalline silicon. The indirect energy gap leads to slow band to band radiative decay transitions which require the participation of momentum-conserving phonons. emission wavelengths which depend on size, pre-dates the first report of porous silicon luminescence [2]. A natural starting point for a review of porous silicon is a comparison with the optical emission properties associated with crystalline silicon. The energy band structure of any semiconductor dictates many of the observed luminescence properties. Silicon has an indirect enera gap, shown in figure 1. A well known consequence of this that the radiative efficiency of Si is low at room temperature. The indirect gap dictates that electron-hole recombination across the gap requires the involvement of momentum-conserving phonons; the matrix element for the transition is thus small. Note that this is not a fundamental limit to the radiative efficiency, it simply results in a long radiative lifetime; the calculated radiative lifetime of moderately doped silicon at room temperature is in the millisecond regime. With such a slow radiative decay process, injected carriers inevitably recombine through non-radiative shunt paths, and the net recombination lifetime (though very sample dependent) is orders of magnitude shorter than the radiative lifetime. However, if all competing shunt paths like deep electron states or surfaces did not exist, then silicon would be a perfect emitter at close to 1 pm. Unfortunately this remains a hypothetical case, though data exist which demonstrate clearly that effective removal of the surface shunt path by hydride passivation results in long minority carrier decay times and large increases in radiative efficiency [3]. At low temperatures, Si becomes more optically active. This is principally because certain optical decay channels become thermally stabilized. For example, the free exciton population under injection conditions grows, and more importantly shallow impurities or defects can stably bind excitons. The initial capture event for an exciton into an impurity state may be fast; if the exciton can remain trapped for a sufficient time, i.e. is not thermally ionized from the
  • 4. Porous silicon 1090 1110 113 1150 2 Photon energy h"el Figure 2. An example of a low-temperature photoluminescence spectrum of high-quality p-type bulk clystalline silicon. The sharp structure is due to the decay of excitons which are stably bound to the boron impurities at low temperatures impurity potential, decay may occur with a matrix element determined partially by the impurity. Radiative decay times for such transitions vary considerably, but even though non-radiative 'branching' usually occurs for such impurity-localized excitons it is a relatively simple matter to measure the associated luminescence spectra. Such spectroscopy is an active field of semiconductor physics, and the reader is referred to a comprehensive review by Davies for more detail [4]. The luminescence spectra associated with bulk crystalline Si are typically highly structured and well defined, whereas porous Si luminescence is strikingly different. Figure 2 shows the photoluminescence spectrum from a lightly boron-doped sample [4], this is a high-quality version of the sort of wafer which might be used as a starting point for processing into porous material. Whilst the emission from the p-type wafer shows the characteristic sharp features due to the decay of excitons trapped at boron acceptors, that from porous material is both broad and significantly shifted above the three dimensional gap; this is clear from figure 3. which shows some typical examples of spectra measured on freshly prepared porous material. Although the porous Si emission is blue shifted, it is clear from figure 1 that the emission energies lie far below the lowest direct gap at the r point in the Brouillon zone. In fact it is clear that none of the light emission observed from porous Si, even the blue emission discussed later, is associated with the three-dimensional direct gap. The spectral emission from porous silicon is not confined to a single band. The band shown in figure 3 [5] is often called the visible or slow band. and is the one first reported. This band can actually be shifted systematically between the near-infrared and the yellowlgreen region of the visible spectrum. Other radiating systems exist and are crucial to the emerging story of porous Si, but we shall use the properties of the visible band to obtain an overview of the luminescence properties. Table 1 gives a brief resume of the various bands observed to date with comments on their origin. g1.1 1 0.9 oi 0.2 o.i 0.6 o:8 1 (1-Porosity) I Figure 3. (a) An illustration of the way in which the 'visible band' vanes with porosity. All samples in this data set came from the same p+ substrate and were processed in the same electrochemical cell. (b) The variation of peak energy with porosity for the same band. Although porosity is only an indirect assessment of size, it is clear that the blue shill becomes faster at the highest porosities. Although all conductivity types of Si wafer have now been demonstrated to yield the visible band, most work has been canied out on pt material. This is largely to do with the fact that the anodic electrochemical dissolution, which requires a large supply of holes, is most readily and precisely established in p-type material (see section 3). 1189
  • 5. B Hamilton Table 1. Luminescence bands obselved for porous silicon. Energy range (ev) Key properties 1.2-2.2 ('visible band') Sensitive to porosity in most reported cases: sensitive to surface passivation, especially but not exclusively hydride passivation; strong temperature dependence of decay time; convolution of at least two components; dominant band in freshly anodized material. Usually weak compared with visible band; sensitive to surface oxidation condition. strong in high temperature oxidized material fast decay time, 0.8-1.3 ('IR band') sensitive to porosity; 2.5-2.8 Weak dependence on porosity ('blue band') The visible band displays a quite remarkable variation of peak wavelength with porosity. Figure 3 shows this for small pieces of the same p substrate, processed in the same electrochemical cell to different degrees of porosity and measured under identical excitation conditions [SI, It is common practice to measure the visible band in freshly etched material. This behaviour has been reproduced by many groups, and though there are small variations the key features are now established. The onset of luminescence requires a threshold porosity of around 45%; below this value only the luminescence attributes of crystalline, non-porous silicon are observed. The visible band moves smoothly from its threshold peak energy of 1.3 eV to 2.0 eV at a porosity of 90%. At and beyond this porosity, the film is mechanically fragile and porosities are difficult to reproduce and to measure. The sensitivity of wavelength to porosity increases dramatically as the porosity increases. Since in a general sense the characteristic size of the porous skeleton reduces with increasing porosity, this wavelength sensitivity was one of first attributes to underpin the search for quantum confinement effects. It was rapidly discovered [7] that the visible band is not completely stable following anodization; both wavelength shifts and efficiency changes occur when porous Si layers are stored in ambient conditions. In general blue shifting of the emission occurs, with a peak shift of 0.5 eV being recorded for a three year storage period, though all effects seem to saturate after around one year. Interestingly the quantum efficiency often changes little with ambient storage and may increase [7]. Another early discovery was the optical fatigue of the visible band which is especially pronounced for short-wavelength, high-power Laser excitation [SI. These instability problems relate to surface chemical and electronic structure, issues of vital importance for the understanding and control of porous Si. In addition to the issues of stability, the visible band turns out to have rather complex spectral and temporal properties. The emission contained within the spectral envelope consists of more than one emission band [91. By using time-gated detection it has been demonstrated that both fast and slow components are typically present with distinctive spectral shapes. Figure 4 [9] demonstrates this, and shows that the fast component peaks at significantly higher photon energies. The data of figure 4 were obtained 1190 I I r Photon energy [eVI Figure 4. The 'visible' emission band is not a single system. This figure shows that time domain measurements reveal that at least two spectral bands are present. from a p- layer of around 80% porosity which had a fully stabilized native oxide (i.e. is fully aged). More recently, fast high-energy components have been associated with specific types of oxidation of porbus Si. However, for porous silicon which has received no additional surface treatment, aside from ambient aging, the slow component is by far the most important and accounts for almost all of the measured quantum efficiency. The detailed temporal behaviour of the slow component of the visible band varies according to which spectral bundle of the rather broad band is measured, and also on the measurement temperature. Also when measured over several decades, the decay profile is never completely exponential. There is broad agreement about the general form of the temperature variation of the decay time for this band. At low temperatures ( 4 0 K) the decay can be very long, of the order of milliseconds; as the temperature is increased the lifetime quenches to the microsecond regime. There are some differences in detail from sample to sample, which makes it difficult to fit the data to detailed kinetic models; however, figure 5 demonstrates the trend which has been established by many workers.
  • 6. Porous silicon 10 1 10 100 Temperature (K1 Figure 5. The temperature dependence of the decay time of the main component of the visible band measured at 1.8 eV. The data are taken from t'Hoofl G W et a/ 1992 Appl. Phys. Lett 61 2344. 0.16- c 0 , 0 8 i I - j L 0 -0 0 50 100 150 ZOO 250 3 0 Temperature IKI Figure 6. The spectral peak of the visible band moves in a very uneven way with temperature. The detailed movement is sample dependent. The effect is illustrated here for p+ material of around 85% porosity. This trend is typical of a disordered system, but can be very marked for porous silicon. This shortening of the total lifetime at high temperatures is always accompanied by a quenching of the luminescence efficiency. This shows in a rather unambiguous way that non-radiative channels are opening up to the excited carrier populations as the temperature is raised. This is rather a familiar picture in semiconductors and is often associated with disorder, or more specifically particle localization within the potential energy minima resulting from disorder. It suggests that at low temperatures excitons (for example) created by the optical pumping rapidly localize into sites which have good radiative efficiency. that is to say the local non-radiative channel is slow compared with the measured lifetime of typically 1 ms. As the temperature is raised. the thermalization time out of these 'radiative sites' becomes shorter and the exciton is free to explore larger volumes of the porous skeleton, and to find much more efficient non-radiative paths. The notion that disorder plays a role in the luminescence mechanisms of porous Si is compelling given the enormous complexity of the porous skeleton, and indeed other simple observations support this view. One such observation is the spectral shift of the visible band as the temperature of the sample is raised and an example is shown in figure 6. This occurs because at the lowest temperatures particles bind efficiently into the deepest potential fluctuations and the luminescence signal i o 4 1.0 20 3.0 4.0 5.0 6.0 70 Excitation Energy (eV) Energy (eV) Figure 7. The photoluminescence excitation spectrum of a porous silicon layer. There is some similarity to that of bulk silicon especially in the higher energy regions towards the direct gap is weighted in favour of the lower energies characteristic of the deep fluctuations. At higher temperatures shallower potential fluctuations become statistically more significant, shifting the mean of the spectral distribution to higher energies. This is exactly what is observed in disordered alloy quantum wells [IO], but the effect is much more dramatic in porous silicon [ I l l , Other spectral features may reveal the presence of disorder phenomena, such as the Stokes shift between emission and absorption bands, and the relationship between the Stokes shift and luminescence linewidth. For any situation in which luminescence is dependent on the details of localizing potentials, the absorption process itself is more representative of the band structure of the solid; it is not unevenly weighted by the defect phenomena. A practical difficulty in obtaining absorption data from porous silicon is that one of the spectral regions of interest is well above the three-dimensional gap of the underlying substrate. The strong substrate absorption inevitably masks the processes in the porous layer. For this reason many measurements rely on photoluminescence excitation spectroscopy (PLE) which is well suited to the measurement of thin surface layers. One of the first reported PLE measurements is shown in figure 7 [12]. The first impression of such data is that it is rather reminiscent of the absorption of bulk silicon, with strong absorption above 3.4 eV corresponding to the three- 1191
  • 7. B Hamilton dimensional direct gap. Once again, though, we run into the complexity of the material in the interpretation of the data. 'we do not know in detail the macroscopic optical constants of the porous silicon layer, though the refractive index has been shown to decrease with increasing porosity. Clearly some caution must be exercised in assigning an optical thickness w. The role of internal light scattering is likely to complicate the estimate of optical thickness. Calculation of the dielectric functions of the altered layer is difficult because of the complexity of the layer, though some attempts have been made [13]. Even in the limit of an optically very thin surface layer, for which the PLE method truly measures the absorption processes, the experiment will average the whole ensemble of size distributions; this will inevitably lead to a smearing of the data. Such smearing would be particularly enhanced if size effects were present in the optical density of states functions. In order to get a better feeling for the material structure which gives rise to these very distinctive optical properties of porous silicon, we now turn to a review of the fabrication process: this includes some insight into the crucial role that high-resolution microscopy has played in the interpretation of the material properties. 3. Porous Silicon formation and microstructure For many years porous silicon formation has been used as one of the mahy processing techniques for device isolation. The FIPOS process, (full isolation by porous oxidized silicon) makes use of hydrofluoric acid (HF) as an electrolyte in an anodic electrochemical reaction; HF, it seems, is the only known electrolyte which can anodically dissolve silicon in an efficient manner. The basic electrochemical phenomena involved in the FIPOS process and optically active porous silicon are essentially the same, except that the latter usually has significantly higher porosity and in the limit of such high porosity additional electrochemical reactions may occur. The electrochemistry of nanostructured silicon is still a developing field. and only the simplest of views can be presented here. In principle the production of a porous silicon layer is not demanding; a carefully constructed electrochemical cell along the lines of that illustrated in figure 8 [14] is all that is required. The cell and the electrolyte system must be formed from high-purity material, and good control of the operating characteristics is required. Nevertheless. processing of centimetre-size samples with good macroscopic uniformity is not difficult. Ironically, the ease of fabrication is in stark contrast to the complex range of characterization methods which have been employed in an attempt to understand the porous material. In the cell, the SifHF interface forms an elec-trode/ electrolyte barrier system. The potential harriers and electric field distributions across even the equilibrium sys-tem are rather involved, depending on the doping character-istics of the semiconductor and the chemical composition of the electrolyte. However, the gross feature of the barrier is its 'double layer' attribute. There exists finite regions of space over which the interfacial electric fields are spread and the potential barrier evolves; these are the Hehnoltz 1192 Ammeter I - Cathode- h -Anode -Magnetic stirrer . , , . . . . . . , U ca e: 0 * - -U Silicon ca Potential Distribution .s ................................ U Figure 8. A simple schematic diagram of a basic electrochemical cell used for anodization. The potential distribution across the electrolytelsilicon system is shown below. layer in the electrolyte, and the depletion layer in the semi-conductor. A schematic diagram [ZO] of the barrier system is shown in figure 8. The dissolution of silicon occurs only under anodic conditions, and the primary process leading to massive removal of Si atoms is considered to be the formation of silicon fluoride molecules, SiF,. Various routes are possible in principle 115-171. Perhaps the simplest example proposed are for Si dissolution involves the divalent state [I81 Si + 2HF + ne' + ,352 + 2H' + (2 - n)e-. Here it is assumed that holes take part, i.e. holes are freely available in the silicon to feed the reaction. This requirement is easily fulfilled by p' material, but low or even n-type conductivity is not a fundamental barrier to the process because optical excitation can always be used to generate an excited hole population. The SiFz formed in the above reaction may then be removed by other chemical reactions [18, 191. It has emerged, however [19], that the number of electrons consumed in the initial electrochemical reaction, that is the number n in the above equation, is
  • 8. Figure 9. The electrochemical regimes available for silicon processing as a function of the I-V characteristic of the electrochemical cell. Region A: pore formation, region B: transition, region C: electropolishing. greater than 2. It seems therefore that both the divalent and tetravalent Si dissolution occur simultaneously Si + 4HF + (4 - n)e+ + SiF4 + 4H+ +ne-. Again, several routes are possible for the removal of the SiF4. What is achieved in practice depends on the precise anodizing conditions, for example the anodic potential, and .ranges from a layer of uniform porosity to complete removal or electropolishing. The current-voltage relationship of the sample-cell system reveals the various regimes of electrochemistry. This is shown in figure 9 [ZO]. In the region of low applied potential (A) the current is generally exponential with voltage (the Tafel region), with a slope of typically 60 mV per decade; this value is clearly an indication of the physics of the potential banier and the in this region, the silicon removal being driven primarily by the above reactions. At significantly higher potentials, the electropolishing regime is entered, resulting in complete removal of the porous layer, Electropolishing results from the formation of an anodic oxide which is dissolved by the HF, any irregularities in the silicon topography being removed due to the divergence of th e electric field lines at regions of dielectric with a consequent enhancement at any Si features [221. One proposal for the electrochemistry of this oxidation process [23] is the following reaction carrier transport [20, 211. Pore formation occurs ................ Si + 40H- +ne+ -+ Si(OH)4 + (4 - n)e-associated ~i~~~~1 0, The simpleS. model for pore formationb, ased essentially on impedance to current flow, leads to columnar pores (a). A more complex model based a diffusion-controlled mechanism of pore formation leads to the sort of multiply interconnected or spongy porous layers ofien obselved in TEM measurements (b), the processing of porous silicon sensitive to the HF concentration in the electrolyte, low HF concentrations and hence low oxide removal rates favouring electropolishing; these trends have been established. experimentally [20]. Whilst the simplified electrochemistry discussed so far can explain Si removal, it does not account for the spatial selectivity which results in pore formation. In fact Pore morphologY does depend on conductivity type and several models have been proposed to explain this crucial feature of the processing, most of them resting Si02 + 6HF + H2SiF6 + 2H20. on built-in inhomogeneities in the original Si wafer as the trigger for pore formation. The wafer conductivity At low potentials, in the Tafel region, the oxide formation and electrochemical details then dictate the detailed pore rate is too low to compete with Si removal and porous evolution. silicon results. At high potentials oxide formation is One of the earliest attempts to explain pore formation enhanced and surpasses the oxide dissolution rate, resulting is due to Beale and co-workers [24, 251, and is based on in electropolishing. It is to be expected that this interplay the barrier properties coupled with the spatial variation between oxide formation and removal rates should make of impedance to current flow. The essence of the 1193 4. Si02 + 2Hz0. me oxide formation rate is in competition with its dissolution rate governed by
  • 9. B Hamilton model is that small inhomogeneities on the wafer surface cause enhanced current flow and locally rapid removal of Si. The original depression is enlarged, leading to pore formation. The nature of the inhomogeneity is not specified; it could be some macroscopic perturbation of surface morphology, or even a defect at the atomistic level. In its simplest form this model led to the expectation that silicon between the pores will ultimately become depleted, simply because the dimensions of the remaining silicon 'columns' is insufficient to support the space charge width. The impedance offered to the current path into the silicon column is then held to grow rapidly and current flows preferentially down the electrolyte and into the wafer at the bottom of the pore, as illustrated in figure 10(a) [271. This provides a possible mechanism for producing columnar structures. This way of describing the pore evolution does seem to go some way towards explaining the gross morphology of the porous skeleton in p+ silicon. It is suggested that the heavily doped wafer leads to a narrow space charge layer in the semiconductor, tunnelling phenomena are enhanced Gust as in Schottky baniers to degenerate semiconductors) and the impedance to current at the base of a pore is significantly lowered. However, the model does not explain the pore morphology observed in p- material; this typically consists of massively interconnected network which is uniformly distributed across the film. Smith et al [26] have shown that a more complex pore morphology may be explained if pore evolution is limited, at least partially by the rate of diffusive transport of the hole to the reaction point at the electrolytic interface. The diffusion-limited case arises because the impedance offered by the barrier system is much higher than in the pf case, tunnelling being much weaker in the wide, low-field barrier system of the lightly doped semiconductor. This analysis still accounts for a faster than average reaction rate at a pore tip, and hence elongation of the pore. It also makes the interconnected network a more reasonable expectation, as shown in figure IO@). A pore E, initiated on the sidewall of an existing pore A, will be in better communication with the diffusing hole flux in that region of silicon between pores A and B until the tip of pore E approaches the tip of B within two hole diffusion lengths. This is rather a complex, at simplest. two-dimensional diffusion problem, but the significant sidebranching is a fundamental feature of the observed morphology. Other models of pore formation have been discussed. The possible effect of quantum confinement in residual silicon structures has been proposed [27] as way of enhancing carrier depletion effects and hence limiting pore growth in the limit of very small structures. Alternative electrochemical schemes have also been proposed for the Si removal process [19, 281, which draw closer analogies with the formation of porous aluminium. The theoretical simulation of pore structure for a variety of possible electrochemical conditions is given by Parkhutik etal 1291. No complete understanding of pore morphology exists, and it is likely that improvements in our understanding will come about through the application of high-resolution microscopy. Transmission electron microscopy has already 1194 proved essential in probing the porous structure. Porous silicon layers are inevitably fragile and this makes their evaluation more difficult, and necessitates the development of some novel approaches to specimen preparation. Measured pore sizes can vary from -100 nm (macroporous) down to c2 nm (mesoporous). As a very approximate guide to published data it appears that lightly doped p-type silicon produces a fine network of pores whereas heavily doped p-type material produces more of a columnar structure [24, 25, 30, 311. For lightly doped n-type silicon, the pores in general take up a more crystallographic form with typical dimensions of several tens of nanometres propagating in the (100) direction. This attribute has even played a role in VLSI device isolation by trench formation. Un l i e the case of p-type silicon, as n-type doping level increases, the pore dimension increases and hence the interpore spacing decreases. These general comments on pore morphology must be taken as a rough guide only. The detail form of the layer depends on the precise anodization conditions used, and very high resolution imaging can often real more complex geometry, leading to a fractal view of the altered layer. For example it has been known since the early work of Beale et al [32] that the columnar pore arrangement in pt silicon is heavily branched. It is also possible to produce mesoporous n+ silicon with -5 nm pore dimensions (33). The key question which high-resolution electron microscopy has attempted to address concerns the detailed relationship between porosity and luminescence. This has been reviewed by Cullis [34], who highlighted the need to avoid ion beam milling or other invasive specimen preparation methods for the preparation of electron transparent samples of porous material, which is easily amorphized and chemically modified. Figure 11 is taken from that review; it demonstrates well the key issues concerning the microstructure of p-type porous silicon in the transition from relatively low-porosity weakly luminescent material to high-porosity strongly luminescent material. The pictures represent bright field (001) projections. For the weakly emitting material, the Si skeleton comprises mainly rod-like structures with a range of diameters, the smallest being around 5 nm. The corresponding electron diffraction patterns indicate completely crystalline material. Figure 1 l(b) illustrates material of higher porosity than (a) which gave stronger luminescence. The microstructure is now finer with silicon structures down to 3 nm clearly visible. Arcing of the electron diffraction spots indicates misalignment of the Si columns. The electron diffraction pattern now shows more severe misalignment of the Si skeleton, but still indicates crystalline material. As porosity grows, these trends are continued. These TEh4 data, then, point to a correlation between a reduced characteristic size distribution of the silicon porous skeleton and the switching on of strong luminescence. In particular, column or particle sizes of around 3 nm or smaller are present in highly luminescent material. Ultra high-resolution microscopy will continue to play a key role in porous silicon research and it can be anticipated that advances in microscopy will add more vital information
  • 10. Porous silicon regarding the relationship between microstructure and luminescence. Whatever new insight is gained from microscopy, though, it remains true that increasing the porosity of the material inevitably increases the surface area. Surface chemical interactions and the influence of the surface on electronic-properties area key areas of investigation. 4. Surface effects on porous silicon The 'internal' surface area of porous silicon is very large; several hundred square metres per cubic centimetre of porous material is typical. It is reasonable therefore to expect that the surface itself might play a direct role in some of the observed luminescence behaviour, or that the surface would exert important effects on the 'bulk' behaviour of the material. A good deal of effort has been expended on investigating these issues, which still remain at the heart of the debate on the origin of the light emission. One of the earliest [2] and most graphic attributes of surface chemistry is the role of hydrogen coverage. After anodization in the HF-based electrolyte, a surface rich in Si-H bonds can be routinely observed using infrared local mode absorption spectroscopy. Bonds involving one (Si-H), two (Si-H2) and three (Si-H3) hydrogen atoms are normally present and both stretching and scissor vibrational modes can be seen [35-381. Figure 12(n) [5] shows a typical absorption spectrum measured for a freshly prepared porous layer (curve (a)). Compared to that of, say, an unprocessed Si wafer, the H bond-related absorption is dramatically increased. Other Vibrational features can be seen which are common to both porous layers and bare wafers; these correspond to SiSi stretching modes and to (probably bulk) Si-0-Si, interstitial 0 asymmetric stretching modes. These and other 0-related modes assume much more significance for oxidized material. The figure shows the evolution of the local mode structure as the sample is annealed in vacuum (curve (b)) and also in nitrogen at 300 "C for 5 (curve (c)) and 10 min (curve (d)). The vacuum anneal completely removes the H-related features, whilst the nitrogen anneals promote 0-related modes, probably due to weak 0 contamination. It was noted above that the process of atmospheric aging has an indeterminate effect on the luminescence, and may cause it to increase. Such aging causes a broadening of the H-related absorption modes and also a growth of 0-related modes. However, by far the most dramatic phenomenon associated with H coverage is observed following desorption on a large scale during vacuum anneal. This causes a complete quenching of the luminescence. Figure 12(b) illustrates the luminescence spectrum for a freshly prepared sample. After vacuum anneal no luminescence can be seen, though some weak recovery is observed if the vacuum anneal is followed by a nitrogen anneal. This recovery is significant, even though no H-related absorption can be measured. By immersing the annealed sample in HF for a few seconds, both the luminescence and surface H bonds measured by absorption are dramatically restored. Very small shifts in peak wavelength are seen due to this cycle, and these are 1195 Figure 11. High-resolution data obtained from TEM measurements of porous p-type silicon. The trend in characteristic sizes of the remaining silicon skeleton as a function of porosity is clear. The associated optical characteristics are described in the text.
  • 11. B Hamilton (4 si.0-si related mods J-defamation I IIIP 1 1 I I I I I 500' lMXl 1500 2m 2500 Wavenumbers cni' Energy (eV) Figure 12. (a) The infrared absorption spectrum of prepared 4550% porosity silicon is rich in Si-H bond-related transitions curve (a), vacuum anneal for 2 min at 400 'C removes these modes completely. Further annealing in nitrogen at 300 'C for 5 (curve c) or 10 min serves only to weakly promote 0-related bonds. (b). The luminescence spectrum of the same sample: as-prepared (full curve), after the vacuum plus the 10 minute nitrogen anneals (dotted line) and finally after immersion in HF (broken curve) not surprising since one expects small changes in the silicon skeleton to Lake place; we don't have precisely the same sample at the end of the sequence. There is no doubt. however. that surface hydrogen coverage plays a key role in the luminescence behaviour of freshly prepared porous Si, and that removal and replacement of the H coverage leads to reversible quenching and restoration of the luminescence. The electronic role of H is not fully understood, but removal of H has been shown to increase the Si dangling 1196 bond density measured by electron spin resonance [39]. Since the dangling bond is known to be a powerful non-radiative recombination centre [40], a straightforward role of H as a passivating centre is suggested. This notion is much connected with the debate surrounding the radiative mechanisms which operate in porous silicoa There is no doubt then that the surface plays an important role and that H bonding is necessw to sustain the luminescence. Furthermore the vibrational assignments suggest that simple Si-H, bonds account for some, possibly most, of the surface hydrogen. It is unrealistic, however, to expect that this bonding arrangement accounts for all of the surface chemistry and some considerable effort has been devoted to probing for other surface constituents which might bear on electronic processes. There are several important candidates for surface bonding, based simply on the processing environment of the wafer; to date most reported work has been aimed at probing the involvement of oxygen, fluorine, or organic radicals of varying complexity. Although surface Si-F bonds play an important role in the dynamics of pore evolution, it seems that they are not stable on the free surface after processing. Probably they are replaced via a hydrolysis reaction, by Si-OH bonds which themselves can dissociate into Si-H or Si-0-Si bonds by reaction with the atmosphere [41]. The role of oxygen is important. Simple exposure to air causes surface oxidation of all silicon, and the effect is enhanced for porous silicon. After all, this fact led directly to the development of the FIPOS process mentioned above. The luminescence aging of porous silicon is connected the incorporation of 0 into the surface bonding arrangement. The detailed form of 0-modified surface bonding has been analysed by Kat0 et al [36]. Low-temperature ( d o 0 "C, for times of less than 50 min) oxidation was used; this might be regarded as a sort of accelerated aging process. Using IR local mode absorption, the Si-H transitions are seen to broaden and shift somewhat. These spectral changes were attributed to the incorporation of 0 into the Si back bond(s) associated with the S-H, atomic arrangement. For example if 0 is incorporated into one of the three back bonds of S-H, the Si-H stretching mode transition was calculated to shift from 2090 cm-l to 2127 cm-', and by considering all possible sites for 0 incorporation into back bonds the general changes which occur in 'lightly' oxidized material are accounted for. This picture of 0 incorporation leaves all surfaces terminated with an Si-H bond; only the back bonds are broken. The spectral deconvolution leading which led to this picture rests on the assumption that the Si-H stretching mode transition is located at 2090 cm-', but it should be noted that some debate exists regarding thc precise vibrational nature of this transition. Aging, and non-aggressive oxidation also cause increased absorption in all peaks that relate to Si-0 vibrational modes. The effect of further increasing the 0 content of porous Si causes yet more changes, and is currently being used as a modification process to stabilize the material. Oxidation is also potentially useful in helping to evaluate the luminescence mechanisms and the role of the dangling bond in quenching luminescence; these issues are discussed further in section 7.
  • 12. Porous silicon 5.1. The quantum confmement mechanism The idea that the surviving silicon skeleton contains within it structures small enough to exhibit quantum confinement effects such as opening up of the bandgap was the first proposal for a mechanism for porous silicon luminescence. This suggestion represents a simple explanation based on a well established attribute of the material, i.e. the existence of nanoparticles in the layer. Figure 13 is an example of the way in which the quantum confinement mechanism is often viewed. As we have seen, nanometre sizes for crystalline Si particles are amply proven from TEM data and are also confirmed by an analysis of the optic phonon Raman lineshape [45]. Energy shifts due to confinement on a scale comparable to the particle size are a universal feature of quantum mechanics, but proof that such a mechanism is correct must rest on a direct observation of luminescence from the nanoparticles and supporting evidence on the interconnection between the geometly of the nanoparticle and the emission wavelength. Such direct evidence does not exist for porous silicon. However, a direct observation of red luminescence from oxidized isolated Si nanoparticles with characteristic sizes of below 5 nm has been reported [46, 471. Such observations demonstrate that isolated particles can luminesce, but the detailed chemical arrangement of the oxidized nanoparticles complicates the interpretation of the luminescence process It remains the case that the blue shift of the visible luminescence with increasing porosity in p+ material is one of the key observations linking the light output to nanoparticle size. However, porosity in itself is a quantitative measurement of the fractional mass removed, but is not a quantitative measure of nanopaaicle size. So. for example one can imagine crudely that a film of a particular porosity might consist of large Si particles and extremely large voids, or of very smal1.nanopartick.s and moderate size voids. In p+ material, in which the morphology is known to exhibit size reduction of the nanoparticles as porosity increases, the increased sensitivity of the luminescence blue shift to porosity is precisely what would be expected from quantum size effects provided that the light originated from recombination within the Si nanoparticles. The slow decay rate of the visible band is really what one might expect from an indirect semiconductor; and if the interior of the residual silicon nanoparticles were perfectly crystalhe and therefore presented a shunt-free environment with no non-radiative recombination centres and with completely passivated surfaces, they might offer the perfect environment for light emission. However, the non-exponential decay indicates that such an idealized notion is unlikely. In fact this behaviour is not unlike that observed in amorphous silicon, in which carrier trapping in the tail states plays a dominant role. The photoluminescence attributes are rather different for n-type material. In general, for such material there is no systematic variation of blue shift of the visible band with porosity. For example, marginally porous layers, e.g. less than 40%, have been shown to exhibit strong visible luminescence with the same general character as that seen 1197 Although IR absorption has emerged as a powerful tool in the surface analysis of porous silicon, other techniques, principally x-ray photoelectron spectroscopy (XPS) and SIMS, have been used. SIMS analysis has confirmed the presence of hydrogen and Ruorine as the major surface species of freshly anodized material, whilst oxygen, carbon and nitrogen were detected at lower concentrations [42]. The SIMS data also c o n h that F is not stable but reduces with atmospheric exposure, presumably due to the hydrolysis reaction mentioned above, and indeed an increase in surface hydrogen to be expected from this process is also observed using SIMS. The other important changes revealed by SIMS measurements are a build-up of carbon and oxygen with prolonged exposure to the atmosphere. The XPS technique has revealed fluorine, carbon and oxygen on porous silicon surfaces [43], in broad agreement with the SIhlS data. Evidence in support of a fluorine-admixed Si02 surface phase has also been claimed, based on XPS analysis [44]. This possibility, of surface layers with rather complex chemistry, for example an Si-0-F-H system, though more difficult to analyse experimentally than simple Si-H bonding arrangements, is the basis for one of the models suggested for the luminescence, to which we now turn. 5. The light emission process Having reviewed the basic features of the material it is possible. by looking in a little more detail at particular pieces of experimental evidence, to try to glean what is cumently understood about the basic light emission mechanism. This of course is the pivotal question surrounding porous silicon. Until it is answered progress towards any technological goals will be limited. It must be readily acknowledged that the complexity of the material provides fertile ground for the proposal of differing models for the light emission process. Porous silicon is a richly interconnected system of small particles and intricate surface topology. This fact alone leads naturally on to the expectation of electronic disorder with its associated defect and interface states; furthermore the large surface area, generated in a chemically varied environment adds the possibility of partial surface coverage with complex molecular films. All of these attributes have formed the basis for hypothesis regarding the light emission, and at the time of writing all of these generic schemes receive support, often zealous. In keeping with the spirit of this lively debate, this section is presented by analysing some of the data which either support or undermine current models. In the current literature, by far the largest attention has been paid to investigations which have been designed around the hypothesis that quantum confinement plays a key role; by far the largest number of reports discuss this issue. Accordingly, this aspect is given more emphasis here, though the correctness of the hypothesis is not proven.
  • 13. B Hamilton Figure 14. The optical transmission spectrum of free-standing porous silicon films. to the PLE measurements reported above, there have been successful attempts to measure absorption in free-standing films, free of any complications associated with the substrate. The absorption measurements often appear to be rather more sensitive to the low-energy tail, near and even below the three-dimensional gap of silicon. Several such measurements seem to indicate that the threshold for absorption in porous silicon is higher in energy than in bulk silicon and that the threshold moves to higher energies with increasing porosity. Figure 14 demonstrates this for films originally processed from both p and p+ substrates [27]. The up-shift in energy was more marked for the p+ material which was found to have smaller nanoparticle sizes than the p material. These observations are consistent with the opening up of the gap due to confinement and would strongly support it if it were to be confirmed that silicon nanoparticles and not some other phase of material were dominating the absorption spectrum. Looked at in more detail, the absorption edge of porous silicon does not wholly support a simple quantum confinement model. Photothermal deflection spectroscopy has shown [49] that the absorption strength increases roughly exponentially above the luminescence peak. Whilst it might be argued that a size distribution of nanoparticles might partly explain such data, the same experiments show that absorption occurs significantly below the gap of three-from Figure 13. Schematic diagrams illustrating the some of the structures envisaged for the optically active material, according to the quantum confinement hypothesis. (a) The transition from quantum wire through oxidized nanoparticles to porous glass [7].(b )T he aligned nanocrystalline or wire structures consistent with EPR data. (From Harvey J F et a/ 1993 NATO AS/ Series voI244, p 179.) (c) An electronic view of how an exciton localized in a nanoparticle might suffer three possible fates for radiative decay giving rise to three luminescence bands. (From Koch F 1993 Mat. Res. Soc. Symp. Proc. 298 319.) P-tyPe material [481. This is an imPortant Point. and dimensional silicon. The existence of significant Urhach is further highlighted by the fact that the pore morphology tails in the density of states is of course a qualitative is much more macroscopic in nature with tYPicallY large measure of departure from crystallinity and is reminiscent widely spaced and crystallographically oriented pores. This of the behaviour of amorphous silicon. Whatever the picture is at variance with the quantum confinement model. origin of the density of states low-energy tails, its existence However, the fact that the macropores have much smaller implies strongly some significant degree of electronic structure on the sidewalls is a further complication which localization. means that we cannot rule out a nanoparticle explanation Falling back on the evidence relating to the for the luminescence. luminescence spectrum, the modification of the emission A key test of low dimensionality in any electronic characteristics by post-anodization processing has been system is a measure of the density of states functions for used to variously support or oppose confinement models, electrons and holes, and it was noted above that, in the The blue shift of the luminescence peak as a result of case of porous silicon, optical ahsorption is in principle oxidation followed by HF dipping was first suggested as a fundamentally better measurement of these (or more supporting evidence, since the consumption (by oxidation) precisely of the joint optical density of states). In addition and subsequent removal of silicon is expected to reduce the 1198
  • 14. Porous silicon ..;- i 10000 c 3 E 8000 .. P - 5 - 6000 - 5 h 4000 - c c 2000 850 800 750 700 650 Wavelength lnml Figure 15. An example of one effect of immersing porous silicon in ethanol: (1) is the ‘as-prepared spectrum’, (2) is for 1 min of immersion, (3) for 3 min, (4) for 10 min and (5) for 60 min. The luminescence was measured in situ overall size of all silicon components in the porous skeleton. To counterbalance this, reports of red shifts with hydrogen loss [SO] would not be expected to affect the particle size, though very small effects due to strain might be expected to produce small wavelength shifts. The influence of low-energy ‘processing’, essentially immersion in a variety of organic fluids, has been shown to have large and nearly reversible effects on the emission spectrum. Effects due to acetic acid, propanol and ethanol have been reported [SI, 521. Figure IS [SI] shows the rather dramatic effect of immersion in ethanol for times of up to 1 h. The detailed chemical interaction with the porous skeleton has not been analysed for these organic treatments, but it seems reasonable to assume that they involve surface or near-surface effects, and such effects are expected to impinge only weakly on optical transitions with energies determined mainly by size quantization. The debate on quantum confinement has been underpinned by attempts to calculate the electronic structure of silicon nanoparticles, and hence to predict optical properties, in particular the transition energies and matrix elements. Effective mass theory (EMT), which has been so successful in predicting the properties of epitaxially grown low-dimensional smctures, has been applied to small silicon structures typical of those know to exist in porous silicon. One such calculation [53] was based on the notion that for cubic structures with sides greater than IO atoms long, bounded by (100) planes, EMT represents a plausible approximation for the description of wavefunctions. Simple envelope functions and confinement energy shifts result. The infinite barrier approximation at the cube boundary leads to an optical matrix element which is an oscillatory function of the cube size. This is difficult to test in a real system because the ensemble of sizes present in a given film inevitably smears the effect. However, the overall trend for radiative lifetime variation with confinement energy shift for the optical transition shift, which of course relates to cube size, is predicted by EMT to vary rapidly: approximately as the inverse cube of confinement shift. Some workers have noted that the measured lifetime of the visible band can vary with 1 + + 0.5 1.5 2.5 3.5 4.5 Olnml Figure 16. The optical gap predicted by LCAO theory for nanometre-size silicon crystallites, as a function of size. porosity and therefore with peak photon energy [54], and there is rough agreement with the EMT prediction and the experimental data. Of course, the very small sizes of some crystallites observed in porous siiicon must eventually limit the applicability of EMT, and point to the need for first principles calculations. One such calculation [5S] has been performed for wire structures, spanning wire thicknesses which vary from the thii, molecular, limit of polysilane to structures which are essentially bulk-like. The calculation was performed for wires with axes in the [OOI] direction, bounded by (110) surfaces which were assumed to be fully terminated with hydrogen atoms. A supercell approach was used with the basic unit cell of the wire repeating in space in order to retain three-dimensional periodicity. Such a calculation is far removed from the effective mass approach, using a first-principles pseudopotential for the Si ions and a bare Coulomb potential for the H ions; the exchange-correlation energy and potential were included using a local density approximation. It is interesting to compare the results of such a calculation with EMT. For the confinement up-shift, agreement was good for wire diameters of greater than 23 A; above this value the EMT prediction appears to be an overestimate. The calculation also yielded a radiative lifetime of around 380 ps for a wire with 72 atoms in the unit cell, i.e. in broad agreement with experiment based on the notion that small crystallites yield the photon output of the visible band. The authors noted, however. that such a long radiative lifetime implies that the high quantum efficiency of this band is largely a consequence of the small non-radiative competition rather than the lifting of the momentum selection rule. A recent calculation using the linear combination of atomic orbitals (LCAO) technique has been used 1561 to calculate the optical properties of wires and crystallites (cylindrical and spherical shapes). This form calculation should yield information on both conduction and valence band properties of the structures. A key result, the calculated optical gap as a function of diameter, is shown in figure 16. The crystallites show the greatest sensitivity to size; this is the intuitive result based on the fact that confinement is in three dimensions. The authors also illustrate the Coulomb electron-hole interaction energy 1199
  • 15. B Hamilton which makes only a small difference to the total calculated energy gap. An interesting comparison is made with the experimentally measured photoluminescence energy measured not from porous silicon but from hydrogen-passivated silicon crystallites produced by nucleation from the gas phase [57]. The result for the wires is a little more complex, showing anisotropy between different wire directions. The authors note that the visible band, which is tunable between 1.4 and 2.2 V, would be consistent with characteristic structure sizes of between 2.5 and 4.5 nm. The exponent relationship between gap and diameter, in the visible band energy window. was found to follow D-',39 rather than D-2 predicted by EMT. However, the calculation predicted an inverse square law at larger D values where EMT is valid. The LCAO calculation also dealt with recombination and optical absorption. It was concluded that the strong confinement in silicon (43 A) induces band mixing and dipole allowed transitions. The optical matrix elements, though, remain small and the radiative decay rates as function of transition energy show strong scatter. Partly this results from the oscillatory behaviour induced by the dependence of the matrix element on the overlap in k space of the electron and hole wavefunctions; this was a point which emerged also from the EMT formalism. For the case of the crystallites, the LCAO calculation also demonstrated that the radiative rate was also sensitive to the symmetry representation of the Td point group which varies greatly as the size of the crystallite is varied. This effect was shown to be more sensitive at lower temperatures. The optical absorption coefficient based on the above calculation, for a crystallite of 3.86 nm diameter is shown in figure 17(a) [%I. This shows that the major absorption strength is in the ultraviolet, with an absorption 'edge' near to 3.5 eV, i.e. close to the direct edge of bulk silicon. The spectral shape is also very structured and bears a superficial resemblance to what might be expected from a molecular system. When viewed on a more sensitive scale. figure 17(b), the calculated absorption coefficient for this crystallite does show that the transition is allowed down to the calculated gap energy, but with small oscillator strength. The absorption coefficient shows a quadratic dependence on photon energy above threshold, unlike that of bulk silicon which shows a linear dependence. The blue shift of absorption edge with porosity (assuming an attendant reduction of crystallite size) reported above is then generally predicted by the LCAO calculation, and the predicted non linear shape has also been recorded experimentally [58]. These three illustrations serve only to review the trend in calculations of small structures, and are by no means exhaustive. They underline the point that in general there is no fundamental disagreement between theory and the quantum confinement model for the main emission band observed from porous silicon. They also highlight the fact that the regime of solid at the heart of the debate is tantalizingly poised between one which is comfortably crystalline and populated with electrons in Bloch states, and one which is better described by a molecular framework This notion is very much the theme of the surface film and defect models reviewed next. 1200 E lev1 Figure 17. LCAO prediction for the optical absorption coefficient of a 3.86 nm crystallite. 5.2. Molecular films, interfaces and defects Since it is clear that surface hydrogen coverage is an important criterion for light emission, at least in unprocessed porous silicon, several groups have explored the possibility that the hydrogen does not simply play a passivating role (i.e. dangling bond saturation), but is somehow involved directly in the radiative process. Two main candidates have emerged; surface hydride species and a class of compounds known generically as siloxenes. Although other variations have been suggested, these two examples are illustrative of the key ideas. The idea that surface hydride species of the form S a I are directly involved in the luminescence process stems largely from the fact that particle size distribution, and in particular size reduction attempts, are not universally consistent with the quantum confinement model. Luminescence from only moderately porous p+ silicon (20%) has been reported (591 which did not show a blue shift with repeated HF dipping, and with increased porosity. On the other hand, this material did show all of the well known attributes of surface hydride coverage, i.e. luminescence could be quenched and restored by hydride removal and replacement. A further report of luminescence peaking at 1.7 eV measured for n-type samples sample of less than 10% porosity bas been made [60]. Particle sizes of around 200 nm were found in this material and it was claimed that side pores did.not exist. Such a size distribution is not appropriate for
  • 16. Porous silicon and clearly show that .the optical gap shrinkage of the amorphous silicon matches well to the red shift of the porous silicon sample. A somewhat more complex model for the involvement surface molecular species in the form of siloxene related compounds has been suggested 1641. Siloxene in its simplest form has the chemical composition SiaO& and can be prepared from Cash via the reaction 3CaSiz + 6HCI + 3Hz0 = Si6O& + 3CaC12 + 3H2. The existence of such compounds has been known for some time, and their fluorescence in the green region of the visible spectrum is also well known in the chemical literature [64]. The initial suggestion was that siloxene or closely related compounds are a by-product of the electrochemical processing of silicon, which is rich in Si. H and 0 atoms. The tuning of the luminescence was suggested tentatively to result from chemical variations to the basic structure, for example by substituting other ligands for the H-terminated Si bonds in the sixfold Si ring of the isolated molecule. Many of the features of porous silicon luminescence were also seen in the fluorescence of siloxene: tunability of wavelength, electroluminescence during anodic oxidation (see section 7), luminescence fatigue and non-exponential decay. The tunability of the chemical structure of the siloxenes and its link to emission wavelength were the main question marks which militated against the siloxene explanation soon after it was suggested. In part this shortcoming was due to a lack of understanding of the physical chemistry of these materials. More recently [65] it has been demonstrated how crystalline films on silicon substrates can in principle be produced by evaporation of calcium followed by reaction with HCI, i.e. a potential planar technology. Perhaps more importantly for the present debate, recent quantum chemical simulations of siloxene [66] crystals have led to a better understanding of the stability of the system and the way in which modification by oxygen incorporation can change the electronic properties. The idealized crystalline siloxene structure is shown in figure 19(n) [65], and consists of a silicon plane, terminated by OH and H radicals on opposite side of the plane. Calculations suggest that this form is metastable; insertion of 0 into the bonds of the Si plane gains 1 eV per Si- @Si bond. Therefore annealing the structure is likely to transform it into that shown in figure 19(b) [651. The stoichiometry remains the same, but now that all the 0 atoms have been incorporated into what was the Si plane notice that isolated Si6 rings begin to appear. The optical properties of the metastable and annealed structures are quite different. The metastable structure fluoresces near 2.6 eV, and has a relatively sharp absorption edge at only a slightly higher energy, i.e. a fairly small Stokes shift, The quantum chemical calculations predict that the Si plane present in the metastable form is a direct-gap semiconductor with a gap of 2.7 eV at the point, broadly consistent with the experimental data. The annealed structure fluoresces in the red, near 2 eV, and shows a much broader absorption edge and a very much larger Stokes shift. This is much less like the properties 1201 "18.705 7 1.70t v = T 1,501 1 ohm-cm porous Si 1.45 T a6:H (Yamasakl et al.) 1.40 100 200 300 400 500 600 Temperature ["Cl Figure 18. A comparison of the measured shift of the luminescence spectra of amorphous and porous silicon as a result of annealing. In both cases the loss of hydrogen from the system is implicated. producing quantum size effects. The same report also detailed measurements of p-type material which was subject to cyclic (atmospheric) oxidation and HF dipping. This is without doubt an obvious way to thin the microstructure, and pore enlargement with an associated reduction in average particle size is to be expected. However, what was in fact observed was a cyclic shift of peak wavelength from 720 nm to 680 nm, the shorter wavelength reappearing after the HF dip. A futther observation [60] relating to size distribution is that high-temperature (up to 1200 "C) annealing of porous silicon under UHV conditions causes a collapse of the microstructure, to .the point where the material resembles a collection roughly spherical particles having a dimension of a few hundred nanometres. The luminescence is also quenched. When such material is dipped in HF, luminescence is restored, even though the particle size distribution is unchanged. These inconsistencies' with the quantum confinement model appear to leave the presence of the surface hydride species as the only completely consistent factor in determining whether or not luminescence is present. The plausibility of this idea gains support from work on hydrogenated amorphous silicon [61], deposited from the vapour phase, with high H content. Luminescence from such material is in the range 1.3 to 2.08 eV, and blue shifts with increasing H content. This was explained in terms of polysilane complexes: (SiHz)" or hydride complexes. Wavelength variation may be a feature of the luminescence of both entities because the 'gap' of the polysilanes depends on the chain length, and the SiH, species, according to tight binding calculations, produce bonding states deep in the silicon at energies which depend on the H content of the molecule [62]. On a more practical note, an interesting comparison has been made [60] between the red shift of the luminescence of porous silicon induced by annealing, (in an argon atmosphere), compared with the red shift induced in the luminescence of hydrogenated amorphous silicon by similar processing [63]. The data are shown in figure 18 [60],
  • 17. B Hamilton expected of a direct gap semiconductor, and is similar to the measured properties of porous silicon. Another similarity between annealed siloxene and porous silicon lies in the involvement of the triplet exciton in the low-temperature luminescence, probed by optically detected magnetic resonance [66]. These experiments also support the idea that the exciton is strongly localized, on a scale compatible with the size of the sixfold Si ring expected to be present in annealed siloxene. It has been proposed that the sixfold silicon ring is the basic luminescence 'centre' for both annealed siloxene and porous silicon [65]. The spectral properties of properties of both are similar and the EPR measurements have identified the Si dangling bond as the key non-radiative shunt for both. It has been argued that coalescing pores will produce fragmentation of monolayer silicon that could lead to the formation of the ring structure. Furthermore it was noted by the authors of [65] that such a process might be a simple explanation for the fact that strong luminescence has been reported from porous amorphous silicon [671, which appears to relax crystallinity as an absolute prerequisite for the luminescence. The siloxene model, like the quantum confinement model, has many appealing features. However, at the present time it seems not to explain in a simple way the smooth shift of wavelength with porosity which is probably the key result. It must he remembered that this band can be reliably tuned down to less than 1.4 eV. A more complete statement seems to be required about the way in which the sixfold silicon ring, or perhaps some perturbation to it, might allow such gross tunability. In contrast, wavelength tunability with size is a natural feature of quantum confinement. 6. Oxidized porous silicon Oxidation of porous silicon has received much attention recently mainly because it produces stable material with additional emission at short wavelengths, often referred to as the blue or fast band. Not only is this luminescence relatively immune to thermal degradation but it exhibits decay times in the nanosecond regime, a potentially useful attribute for devices. It must be stressed, though, that the visible or slow band remains (with some spectral modification) in oxidized material, and rapid oxidation for typically 30 s at 900 "C can give stable material which emits strongly at these longer wavelengths. Higher-temperature processing than this tends to remove the visible band [5, 681, leaving only the fast band, and of course a much more fully oxidized material structure. Figure 20 shows representative spectra of oxidized porous silicon. It has been noted by several groups that the intensity of the visible band drops rapidly with oxidation for oxidizing temperatures up to 600 "C, and then rises with processing temperature until the melting point of silicon is reached. Figure 20(a) [69] shows that the visible band remains, in a broadened blue shifted form, but also that a higher energy band emerges which extends into the blue. Of course the latter band requires pumping with an appropriately short wavelength source. Figure 20(b) shows -,- Figure 19. (a) The idealized structure of the Si planes in as prepared siloxene: the planes are terminated by H or OH radicals on opposite sides. (b) Siloxene after the ordered insertion of 0. isolated SiB rings now appear, and these may be the luminescent centres responsible for emission from modified siloxenes. 1202
  • 18. Energy (eV) Figure 20. Photoluminescence spectra typical of rapidly oxidized porous silicon. (a) The broadening and blue shifting of the visible band and the appearance of the high energy band (from 1691). (b) Besides the visible band (a) measured at 10 K, a lower energy band, the infrared band, is produced. There is strong competition between the infrared and visible bands which depends on temperature; spectra (b), (c) and (d) were measured at 10, 70 and 300 K respectively. another key result; the oxidation also produces a low energy band known as the infrared band. This band which is below the energy of the bulk gap competes with the visible band, but at low temperawes becomes as efficient as, or more efficient than, the visible emission. Electron paramagnetic resonance has been applied to most forms of porous silicon. By far the most important defect to be observed is the dangling bond or Pbo centre. This centre has been known for some time to be present at the SiSi02 interface [70]; it is a [ill] axially symmetric system with the unbonded orbital directed along one of the four equivalent [ 11 I] directions. The density of this centre increases when the visible luminescence is quenched by annealing and it accordingly correlates with hydrogen loss from the surface [711. The centre has also been detected via its influence on the intensity of the visible band as the magnetic field is swept through resonance [72]. The general conclusion is that the dangling bond is a key non-radiative shunt path for the visible band. The optically detected magnetic resonance (ODMR) experiment shows a large effect for the infrared band; the intensity of the hand increases by around 15% at resonance. This compares with values of typically 0.01% variation for the visible band. The data, as shown in figure 21 [73], indicate something slightly more complex than a simple P ~ o Porous silicon 1.22 1.24 1.26 Magnetic field IT1 Figure 21. The strong ODMR signal measured for the low-energy or IR band [73]. These data prove that the PbO centre (together with another centre) is directly involved in the luminescence process responsible for the IR band centre. The lowest-field peak with an isotropic g value of 2.013 does not belong to the dangling bond, but may belong to a localized hole. The two higher peaks are signatures of the Pbo; their g values are 811 = 2.0017 and gl = 2.0085, where the parallel direction of !he magnetic field is along the [ 11 11 direction. This anisotropy is exactly that found in the EPR spectrum of the dangling bond. The strong effect in ODMR may imply that the dangling bond is directly involved in the infrared band. The origin of the blue band is of course of great interest. One obvious question to he asked is that since it is much faster than the visible band [74] could it be the true signature of a direct energy gap in a quantum confined system? This idea has been supported by several groups who point to the similarities which exist between the blue band and emission from direct-gap semiconductors. Unfortunately the blue band is most easily seen in highly porous (oxidized) material, and so its wavelength dependence on porosity cannot be easily explored; i.e. even this rather uncertain size control has not been explored for particles which remain in the system. Besides which, the rather aggressive oxidations used modify the skeleton of the porous layer. Other workers have supported the view that the blue band originates from SO?. For example, it has been reported [69] that the luminescence does not shift spectrally with increasing oxidation time (up to 50 min). It has also been noted that oxidized planar silicon wafers give similar luminescence [75]. In both these references, the suggestion that defects in the Si02 glass might be responsible for the emission was made. 7. Electroluminescence and devices The goal of research into porous silicon is without doubt the realization of an efficient LED. The wavelength range of 1203
  • 19. B Hamilton Figure 22. An indium tin oxidelporous silicon heterojunction LED, showing the device structure and electrical characteristics. any potential device is actually a secondary consideration since any wavelength band has applications. It has become painfully apparent that to fabricate such a device is hugely difficult, and quantum efficiencies of 10-6 are typical; even then large drive voltages, in excess of tens of volts for example are often required. A variety of structures with solid state contacts have been tried, ranging from simple indium tin oxide, to attempts at p n junction technology. Figures 22 and 23 show two devices which are representative of the sort of structures appearing in the literature. Figure 22 demonstrates a simple heterojunction technology using conducting (and transparent) indium tin oxide [761. Rectification is observed for such contacts with light emission occurring only for one polarity of voltage. This particular diode structure produced a rather narrow emission band, about 20 nm wide, centred on 580 nm, significantly narrower than a typical room-temperature photoluminescence spectrum. Although the reasons for this are not clear, it may he that the electroluminescence excitation excites only a subset of the porous layer. A device based on a porous p-n technology is shown in figure 23 [77]. This device is based on an n-type wafer with a surface-implanted pf layer. Anodization was carried out using illumination so that the n-type material as well as the implanted p layer became porous. The vertical porosity profile depends very much on the doping profile, but the active region was thought to be a nanoporous region which straddled the metallurgical p n junction. The quantum efficiency of the device was measured to be lO-4. An interesting attribute of this structure is that the emission hand seen in electroluminescence can he tuned by the wavelength used to excite the layer during the anodization process; the shorter the excitation wavelength, the shorter the emission wavelength. Although the reasons for this are not completely established, the authors point out that in order to supply holes in porous 1204 Figure 23. A porous Si p n ju nction device fabricated by anodization under illumination of a boron implanted n-type wafer. (a) Device structure and I-V characteristics. (b) The electroluminescence output for devices anodized with different wavelengths of light. n-type material, and hence for anodization to proceed, the exciting photon must be strongly absorbed within the small silicon particles characteristic of nanoporous material. Following the qualitative expectations of quantum confinement, very small particles would require short wavelengths for absorption by the opened up energy gaps. Whatever the detail, it was demonstrated that electroluminescence ranging from the infrared to blue could be produced by varying the anodization excitation from infrared to ultraviolet. The electroluminescence data are also shown in figure 23. A variety of device structures have been tried, but the key result, the external quantum efficiency, remains low. It seems an obvious conclusion that the contact technology to date fails because it does not provide volume excitation of the porous film. What is required is a contact technology which transports energy into the whole of the film and then facilitates local minority carrier injection. Such a scheme would also demand a continuous current path to the wafer. Of course optical excitation fulfils all of the required conditions without the need for current continuity; it is the perfect excitation source. It has become increasingly clear
  • 20. Wavelength (nm) Figure 24. (a) The transient (integrated) light emission from porous silicon during anodization. (b) The spectral detail of the light emitted during the transient period. that some innovation in solid state contact philosophy is required in order to achieve better success. Liquid contacts represent a state of suitability which is intermediate between photon beams and solid state contacts, and the study of such systems has yielded some important results for our understanding of luminescence processes in porous silicon. This work has utilized electrolytic solutions in order to transport charge carriers to the interior of the porous film. In 'fact the anodic dissolution process used to create the porous layer is accompanied by some rich luminescence detail 178, 791. The key observations for the light emitted during the anodic oxidation process are shown in figure 24. Firstly there is a time delay before any emission is observed and the authors ascribe this to the fact that the porosity is simply too low in the early stages of anodization for any light to be seen. After this delay, the luminescence builds up to some peak value and then decays to zero. The quenching is always associated with a sharp rise in the required anodic potential (to sustain a constant anodic current). Although it is estimated from the total anodic charge transferred (Qo in figure 24) that the film porosity is only around 50% at the point of quenching, the authors conclude that at this point the electrical connectivity between the small crystallites in the film, which are assumed to be the source of light, is broken. The electrochemical activity then switches to oxidation of the base of the pores; a process which is associated with the increased anodic potential. Actually, light emission is Porous silicon also observed in this high anodic potential regime, but it is of short wavelength and is thought to be associated with processes occurring in the oxide layers 1801. The spectral distribution of the light transiently emitted during the initial stages of anodization, though, shows remarkable similarity to the visible or slow band, and it would be surprising if its origin were not the same as the optically pumped emission. This is shown in figure 24, which also shows that as anodization proceeds the luminescence peak blue shifts. This shift has been analysed by the authors in terms of a tunnelling-limited escape mechanism for excited carriers which leads to non-radiative recombination. A model was described for a tunnelling process occurring through small regions which connect optically active crystallites to non-radiative bulk silicon sinks. The yodel demonstrates that because these regions, which may be even smaller than the crystallites, are particularly sensitive to shape (and hence bandgap) changes, the predicted variation of tunnelling flux with time would produce the observed blue shift. They also commented that the oxidation-induced thinning of nanopdcles mentioned above cannot account for the blue shift: it would be too small. These results then provide some additional insight into the non-radiative processes, and are supported by the qualitative observation that the radiative efficiency of the fluorescent species, perhaps the nanoparticles, seems to increase as the porosity and hence the barriers to tunnelling transport increases. Strictly speaking, many would wish to label these observations of light emission during anodization as chemiluminescence, rather than electroluminescence, and there do remain some problems in understanding the electron injection process in the overall ,scheme of the luminescence. The hole of course is provided directly by the anodic reaction. One possibility for the electron supply channel may lie in the existence of intermediate species being formed during the oxidation process of the silicon wafer; such intermediate species may have energy levels high enough to allow electron injection [SI]. The key point here is that the structure is undergoing chemical transformation during the light emission process. However. luminescence using liquid contacts has also been achieved under cathodic conditions [82], and cathodic injection does not modify the chemistry of the material. The principle of liquid contact electroluminescence is therefore firmly established. 8. Conclusions Research into the physics of light emission from porous silicon is motivated by a very practical desire to extend the functionality of silicon. The fact that enthusiasm remains strong for research into this complex material is a testimony to this huge technological prize and the challenge of bringing together the impressively wide ranging interdisciplinary tools needed to deal with structures of almost fractal complexity. It is the sheer scale of the complexity which has limited the interpretation of data and the success of attempts to make devices. 1205
  • 21. B Hamilton As regards the basic luminescence mechanism. it would be wrong to give the impression that views do not remain divided; quantum structures, surface hydride species, amorphous silicon and more complex molecular arrangements like siloxene derivatives all have their protagonists. It may be that a combination of some or even all these are present in porous silicon, but it is more satisfying to think that there is a single mechanism operating. Such satisfaction is rooted in the traditional approach of physics, which finds elegance in obtaining the most complete solution possible for model systems. This luxury, though, is not available to, say, biologists who deal with systems in which complexity is the intrinsic dominant feature. On balance at the present level of knowledge, the quantum confinement model has most support. It is easy to understand why this should be the case: size confinement and the associated energy shifts are absolutely fundamental features of quantum mechanics. So, whatever additional mechanisms might be invoked in porous silicon, quantum size effects should occur if very small singlecrystal entities are formed. The quantum particle or wire hypothesis for the optical activity, then, is viewed as a development of a straightforward and basic rule of physics. Of course, in practice this intrinsic effect may be overwhelmed by the optical activity of other chemical species, or there may be reasons why quantum sized crystallites simply do not fluoresce with adequate efficiency to explain the obsewed light emission. Such objections, though, are details, especially since the alternative hypotheses are on the whole more complex than quantum confinement involving exotic molecular species for which detailed knowledge of the electronic structure may be missing. The current debate therefore tends support the view that whilst the jury is still out on all of the possible mechanisms, the quantum confinement model remains the simplest explanation: and until better or more innovative experimental techniques prove otherwise, simplicity holds sway. Looking back over the past several years of research into porous silicon, however, it seems that this effort is very much part of a paradigm shift in condensed matter physics, and particularly in materials research. The movement towards understanding materials systems of great complexity is now evidenr One can think of numerous examples, including semicofiductor superlattices, high-T, superconducting materials, hierarchical biological structures characteristic of living matter, and many more. Probably this research will in the end prove to be valuable even if devices do not result, because at the very least it has led to one of the most successful periods interdisciplinary work and experimental development for many years. Acknowledgments The author wishes acknowledge the financial support of the EPSRC and from the ESPRIT programme for maintaining his involvement in porous silicon research. Special thanks are due to his collaborators Ursel Bangert, Phil Dawson, Spyros Gardelis and Robert Pettifer for their unstinting efforts and critical approach. 1206 References 111 Canham L T 1990 A ~ o lP.h vs. Left. 57 1046 [Zi Takagi H, Ogawa H;Yazaki Y, Ishizai A and Nakagiri T 1990 Awl. Phys. Luff. 56 2379 [3] Yablonovilch E ahd Gmitler T 1986 AppL Phys Left. 49 587 [4] Davies G 1989 Phys. Rep. 176 83 [5] Gardelis S and Hamilton B 1994 J. AppL Phys. 76 5328 [6] Gardelis S, Rimmer I S, Dawson P. Hamilton 6, Kubiak R A, Wall T E and Parker E H C 1991 Appl. Phys. Lett. 59 2118 171 Canham L T 1993 Optical Properties of Lmy Dimensional Silicon Structures (NATO AS1 Series, vol 244) (Dordrecht: Kluwer Academic) p 81 Appl. Phys Lett. 60 639 Broomhead D 1993 3. Phys.: Condens. Uutfer 5 L91 [SI llschler M A, Collins R T, Stathis J H and Tsang J C 1992 [9] Calcott P D I, Nash K J, Canham L T, Kane W I and [IO] Skolnick M S , Tapsler P R, Bass S I. Piu A D, Apsley N and Aldred S P 1986 Semicond. Sci. TechnoL 1 1455 [ll] Gardelis S and Hamilton B 1992 Mater. Res. Soc. Symp. Proc. 256 149 [I21 Wang L, Wilson M T, Goorsky M S and Haegel N M Mater. Res. Soc. Symp. Proc. 256 13 1131 Tsu R and Babic C 1993 Oprical Properties of Low DimensionaI Silicon Structures (NATO AS1 Sene& vol 244) (Dordrecht: Kluwer Academic) o 179 [I41 Tumer D R 1958 3. Electrochem. Soc. CO5 402 1151 Gardelis S 1993 PhD Thesis UMIST (161 Memming Rand Schwandt G 1966 Surf: Sci. 4 104 [17] Tumer D R 1960 Suiface Chemistry ofMetuls and Semiconductors ed H C Gatos (New York Wiley) p 82 [181 Tumer D R 1961 Elenrochemistry of Semiconductors ed P J Holmes (New York: Academic) p 161 [19] Unagami T 1980 3. Elecfrochpm. Soc. 127 476 [ZO] Smith R Land Collins S D 1992 J. Appl. Phys. 71 R1 17.11 Dewald I F 1960 The Surface Chemistry ofMefals and Semiconducfors ed H C Gams (New York: Wiley) p 78 [22] Foll H 1991 Appl. Phys. Left. A 53 8 [23] Bang X G and Collins S D 1989 J. Elecfmchem. Soc. 136 1561 1241 Beale MI, Chew N G, Uren M 1, Cu!Jis A G and Benjamin J D 1985 AppL Phys. Lett. 46 86 [25] Beale M 1, Benjamin I D, Uren MI, Chew N G and Cullis A G 1985 J. Crystal Growrh 73 622 [261 Smith R L, Chuang S F and Collins S D 1988 J. Electron. Muter. 17 533 [27l Lehmann V and Gosele U 1991 Appl. Phys. Left. 58 865 [281 Parkhutik V P, Clinenko L K and Labunov V A 1983 Su$ Technol. 20 265 12.91 Parkhutik V P, Martinez-Duart J M and Albella I M 1993 Optical Properties of Low Dimensional Silicon Structures (NATO AS1 Series, ~01244p) ordrecht: Kluwer Academic) p 55 [30] Bomchil G, HaIimaoui A and Herino R 1989 Appl. Su$ Sci. 41/42 604 [31] Hcrino R, Bomchil G. Barla K and Benrand C 1987 J. Electrochem. Soc. 134 1994 (321 Beale M I J, Benjamin I D, Uren M J, Chew N G and Cullis A G 1985 J. Crystal Growrh 76 622 [33] L'ECuyer J D. Lorreto M H. Far J P G, Keen 1 M, Castledine J G and L'Esperance G 1988 Murer. Res. Soc. Symp. Proc. 107 441 [34] Cullis A G 1993 Optical Properfies ofLow Dimensional Silicon Stmctures (NATO AS1 Series, vol 244) (Dordrecht: Kluwer Academic) p 147 [35] Gupla P, Colvin V L and George S M 1988 Phys. Rev. B 37 8234 [36] Kat0 Y, Toshimichi I and Hiraki A 1988 Japan. 3. Appl. Phys. 27 L1046
  • 22. 1.371. C habal Y. Hieashi G S. Raehavachari K and Burrows V A 1989 3. VaE Sci. Technor A 7 2104 1.38.1 V enkateswara R A, Ozanam F and Chazalviel J N 1988 3. Eiecrrochem. Soc. 138 153 2569 1391 Brandt M S and Stutzmann M 1992 Appl. Phys. Lett. 61 [40] Poindexter I H and Caplan P 1 1983 Prog. Sud Sci. 14 201 [41] Konishi T N, Yao T, Tajima M, Ohshima H, It0 H and Hattori T 1991 Japan J. Appi. Phys. 31 L1216 [42] Canham L T, Houlton M R, Leong W Y, Pickering C and Keen I M 1990 J. April. Phvs. 68 2187 I431 VSquez R P, Fathauer'R W, George T, Ksendzov A and [44] Roy A, Chainani, A Sarma D D and Sood A 1990 Appl. Lin T L 1992 Appl. Phys. Lett. 60 1004 Phvs. Lett. 61 2187 [45] Fauchet P M and Campbell I H 1988 Crit. Rev. Solidstate Mater. Sci. 14 7 [46] Littau K A et a1 1993 J. Chem. Phys. 97 1224 [47] BNS L 1991 Appi. Phys. Lett. A53 465 [48] Prokes S M, Glembocki 0 1, Bermudez V M, Kaplan R. Friedersdorf L E and Searson P C 1992 Phys. Rev. B 45 13788 Gavrilenko V 1992 Mater. Res. Soc. Symp. Proc. 283 [49] Koch F, Petrova-Koch T, Muschik T, Nikolov A and 107 .,I (501 Petrova-Koch V, Muschik T. Kux A, Meyer B K and Koch F 1992 Appl. Phys. Lett. 61 943 [SI] Fauchet P M. Ettedgui E;Raisanen A, Brillson L J, Seifirth F. Kurinec S K, Gao Y, Pene C and Tsvbeskov L 1993 Mater. Res. Soc. Symp. Proc.298 271 . 1.521_ Li K H. Choachieh T. Camobell J C. Kovar M and White 1 M 1492 Mater. Res. Soc.'Symp. Pmc. 298 173 [53] Hyberslein M S 1992 Mater. Res. Soc. Symp. Pmc. 256 179 [54] Xie Y H, Wilson W L. Ross F M, Miucha J A, Fitzgerald E A, Macaulay J M and Hams T D 1992 3. Appl. Phys. B 71 2403 [55] Read A 1, Needs R 1, Nash K 1, Canham L T, Calcotl P D J and Qteish A 1994 Phys. Rev. Lett. 69 1232 [56] Delerue C, Allen G and Lannoo M 1993 Phys. Rev. B 48 48 I571 Furakawa S and Miyasato T 1988 Phys. Rev. B 38 5726 [58] Chelikowsky J and Phillips J C 1989 Phys. Rev. Lett. 63 [59] Prokes S M. Glembocki 0 1, Bermudez V M, Kaplan R, I653 Freidersdorf L E and Searson P C 1992 Mater. Res. Soc. Symp. Proc. 256 107 [60] Prokes S M 1993 J. Appl. Phys. 73 407 1611 Wolford D J. Scott B A, Reimer J A and Bradley I A 1983 Physica B 117 + 118 920 Faoaconstantoooulos D A and Economou E N 1981 Phw. Rev. B U j223 Unaeami T and Seki M 1980 J. Electrochem Soc. 125 1339 Brandt M S. Fuchs H D. Stutzman M, Weber J and Cardona M 1992 Solid State Commw. 81 307 Deak P. Rosenbauer M, Stutzman M, Weber J and Brandt M S 1992 Phvs. Rev. Lett. 69 2531 Brandt M S, Rosenbauer M and Stutzman M 1993 Mater. Buslarret E, Ligeon M and Orlega L 1992 Solid State Cullis A G, Canham L T, Williams G M, Smith P W and Prokes S M 1993 Appl. Phys. Lett. 62 3224 Brower K L and Headley T J 1986 Phys. Rev. B 34 3610 Meyer B K, Hofman D M, Stadler W, Petrova-Koch V, Koch F, Omelina P and Emanuelsson P 1993 AD^ Phys. Lett. 63 2120 Brandt M S and Stutzman M 1992A.m. i. Phvs. Lett. 61 2569 Kux A and Hofmann D V 1993 Opticai Properties of Lou, Dimensional Silicon Structures (NATO AS1 Series, vol 244) (Dordrecht: Kluwer Academic) p 197 Kovalev D I, Yaroshetzkii I D, Muschik T, Petrova-Koch V and Koch F 1994 Appi. Phys. Lett, 64 214 Kontkiewicz A I, Konlkiewicz A M, Siejka 1, Sen S, Nowak G, Hoff A M. Sakthivel P, Ahmed K, Mukhejee P, Wtanachchi and Lagowski J 1994 AppL Phys. Left. 65 1436 Kakhoran N M, Namvar F and Maruska H P 1992 Mater. Kozlowski F, Steiner P and Lang W 1993 Optical Res. Soc. Symp. Pmc. 256 89 Properties of Low Dimensional Silicon Structures (NATO AS1 Series. vol 2.44) (Dordrecht: Kluwer Academic) p 123 Halimaoui A. Bomchil G, Odes C, Bsiesy A, Gaspard F, Herino R. Ligeon M and Muller F 1991 Appl Phys. I A t s9 3fl4 I 1 Res. Soc. Symp. Proc. 298 301 Commun. 83 461 Dosser 0 D 1993 Mater. Res. Soc. Symp. Proc. 283 182 -. ... . . . . . Porous silicon .. 1791 Muller F, Herino R, Ligeon M, Billat S, Gaspard F, Romestain R, Vial J C and Bsiesy A 1993 Optical Properties of Low Dimensional Silicon Structum (NATO AS1 Series, vol 244) (Dordrecht: Kluwer Academic) p 101 [801 Billat S. Bsiesy A. Gaspard F, Herino R, Ligeon M, Muller F. Romestain R and Vial J C 1992 Mater. Res. Soc. Symp. Proc. 256 215 J. Electroanal. Chem. 290 229 Taylor L 1992 Appl. Phys. Lett. 61 2563 [811 Peter L M, Borazio A, Levemez H J and Stumpcr J 1990 [82] Canham L T, Leong W Y, Beale M I 1, Cox T I and 1207