Charles University
Faculty of Mathematics and Physics
Doctoral thesis
Silicon Nanocrystals: Narrowing Down Size Distribution, Organic Passivation and Novel
Optical Properties
by
RNDr. Kateˇrina K˚ usov´a
Prague, August 2009
Supervisor:
Prof. RNDr. Ivan Pelant, DrSc.
Institute of Physics, Academy of Sciences of the Czech Republic, v.v.i.
Consultant:
doc. RNDr. Jan Valenta, PhD.
Department of Optics and Chemical Physics, Faculty of Mathematics and Physics, Czech Republic
Prague, August 2009
Acknowledgments
First of all, I would like to express my sincere appreciation to my supervisor, Prof. Ivan Pelant, for his guidance and support throughout my work on this thesis. Without his patience and devotion to the topic, this thesis would have never been put together in the present form. My thanks also belong to Dr. Kateˇrina Herynkov´a, who was my supervisor during the first year.
Apart from them, I am greatly indebted for significant contributions to this work to Prof. Pavel Matˇejka, who made it possible to perform the FTIR measurements, Dr. Jan Lang, who carried out the NMR measurements with enormous enthusiasm, and Prof. Jan Valenta, who opened the door of his laboratory for the quantum-efficiency and single-nanocrystal experiments. I am also grateful to Dr. Snejana Bakardjieva for performing the HRTEM measurements and Dr. Petr ˇStˇep´anek, who carried out the DLS experiments.
My deepest thanks belong to Dr. Fabrice Charra, who made it possible for me to join his group for a research stay, during which he introduced me to a complex measurement technique of LE-STM and during which I gained invaluable scientific experience.
Furthermore, I owe my thanks to all the members of the Department of Thin Films and Nanostructures, especially to Dr. Jan Koˇcka, the head of the Department, for cre- ating helpful and friendly atmosphere. Without hoping to be complete, I would like to mention Dr. Kateˇrina Dohnalov´a for sharing a room with me and for fruitful and insightful both scientific and non-scientific discussions, Dr. Ondˇrej Cibulka for under- taking all the tedious tasks that needed to be carried out, Karel ˇZ´ıdek for performing the femtosecond measurements and putting up with the two of us sharing the etching laboratory, Anna Fuˇc´ıkov´a for helping me with optical measurements, Jan ˇCerm´ak and Alex Vetusha for spending their time with AFM measurements (and Alex once again for unhesitating help with any computer-related problems), Dr. Martin Ledinsk´y for answering my nosy questions regarding the Raman setup, Ing. Emil ˇS´ıpek for insight- ful technical assistance, Dr. Petr Fojt´ık for helping me with the early LE-STM setups, Ing. Miroslav Buˇcek for adapting and fine-tuning the CEA software for the acquisition of photon maps and Petra ˇSnajdrov´a for taking care of all the thankless administrative tasks unbelievably quickly and effectively and for her motherly ways.
Last but not least, my personal thanks belong to Dr. Ladislav Fekete for sticking by me for through thick and thin.
Acknowledgments
iContents
iiList of Symbols and Abbreviations
viList of Figures
viList of Tables
viiiOrganization of this thesis . . . 1
A Word of Introduction
1Chapter 1: Silicon nanocrystals: properties and applications
3 1.1. Silicon photonics . . . 31.1.1. Approaches to the generation of light in (or on) silicon . . . 4
1.2. Photoluminescence of silicon nanocrystals . . . 6
1.2.1. Photoluminescence spectra of silicon nanocrystals: experimental insight . . 7
1.2.2. Quantum confinement . . . 8
1.2.3. Quasi-direct (no-phonon) optical transitions . . . 8
1.2.4. Auger processes . . . 9
1.2.5. Surface states . . . 10
1.2.6. Emerging image . . . 10
1.2.7. Competing views. . . 11
1.3. Other optical properties of silicon nanocrystals . . . 12
1.3.1. Optical absorption cross-section . . . 13
1.3.2. Implications for stimulated emission and lasing . . . 13
1.3.3. Phonons in silicon nanocrystals . . . 13
1.4. Fabrication techniques and types of samples . . . 14
1.4.1. “Wet” methods . . . 15
1.4.2. “Dry” methods . . . 17
1.4.3. Spotlight on: colloids of silicon nanocrystals . . . 17
1.4.4. Application prospects for colloids of silicon nanocrystals . . . 19
1.5. Beyond inhomogeneous broadening . . . 19
1.5.1. Burrato’s group . . . 20
1.5.2. Cichos’ group . . . 21
1.5.3. Linnros’ group . . . 21
1.5.4. Valenta’s group . . . 22
1.5.5. Korgel’s group . . . 22
1.5.6. Single-nanocrystal spectroscopy in silicon: summary . . . 23
Chapter 2: Preparation of samples
25 2.1. Preparation of silicon nanocrystals . . . 262.1.1. Whiteandbluesamples . . . 27
Contents
2.1.2. Size of nanocrystals . . . 28
2.1.3. Photoluminescence of our powders . . . 28
2.1.4. Hydrogen-terminated and oxidized silicon nanocrystals . . . 29
2.2. Colloids of silicon nanocrystals . . . 31
2.2.1. Effect of long-term stirring in various solvents . . . 32
2.2.2. Preparation of colloidal dispersions of silicon nanocrystals in xylene . . . . 33
2.2.3. Transferring the solute to chloroform . . . 35
2.2.4. Labeling of our colloidal dispersions of silicon nanocrystals . . . 36
2.2.5. Preparation method: pros, cons and limitations . . . 36
2.2.6. Comparison with a reference: laser irradiation of the solvent . . . 37
2.2.7. Toluene and mesitylene as solvents . . . 38
2.3. Samples for STM-based measurements . . . 38
2.3.1. SiNc/Ausample . . . 38
2.3.2. SiNc/Ag-npsample . . . 38
Chapter 3: Silicon colloid: structural, chemical and optical properties
39 3.1. Structural properties (and size distribution) . . . 403.1.1. Dynamic light scattering . . . 40
3.1.2. High-resolution transmission electron microscopy. . . 41
3.1.3. Atomic force microscopy . . . 42
3.1.4. Silicon nanocrystals in the colloid: summary and discussion . . . 42
3.2. Surface chemistry . . . 43
3.2.1. Fourier-transform infrared spectroscopy . . . 43
3.2.2. Nuclear magnetic resonance . . . 46
3.2.3. Surface chemistry: summary and discussion . . . 51
3.3. Ensemble optical properties . . . 52
3.3.1. Photoluminescence excitation spectra . . . 52
3.3.2. Photoluminescence quantum efficiency . . . 54
3.3.3. Photoluminescence at higher excitation fluxes . . . 58
3.3.4. Photoluminescence of silicon nanocrystals redispersed in chloroform . . . 59
3.3.5. Photoluminescence decay . . . 60
3.3.6. Ensemble optical properties: summary and discussion . . . 61
3.4. Single-nanocrystal spectroscopy . . . 63
3.4.1. Single-nanocrystal spectroscopy of the silicon colloid . . . 63
3.4.2. Reference microscopical measurement of irradiated xylene . . . 65
3.4.3. Single-nanocrystal spectra: summary and discussion . . . 66
3.5. Concentration of nanocrystals in the colloid . . . 68
3.5.1. Concentration based on the dose: upper limit . . . 68
3.5.2. Concentration based on absorption . . . 68
3.5.3. Concentration based on single-nanocrystal spectroscopy: lower limit . . . . 69
3.5.4. Resulting concentration of nanocrystals . . . 70
3.6. Outlooks . . . 70
Chapter 4: LE-STM
71 4.1. Scanning tunneling microscopy . . . 724.1.1. The tunneling phenomenon . . . 72
4.1.2. Scanning tunneling microscope . . . 74
4.2. The LE-STM technique . . . 75
4.2.1. The origin of light . . . 76
4.2.2. Measurement modes . . . 77
4.2.3. Experimental setup. . . 78
4.2.4. Comparison with SNOM . . . 79
4.3. LE-STM on nanostructures . . . 79
4.3.1. Experimental setups . . . 79
4.3.2. LE-STM from silver nanoparticles . . . 80
4.3.3. LE-STM from silicon nanocrystals mixed with silver nanoparticles . . . 83
4.3.4. General closing remarks on LE-STM from silicon nanocrystals . . . 85
4.4. LE-STM setup in Prague . . . 85
4.4.1. STM microscope in Prague . . . 85
4.4.2. LE-STM setup: first stage . . . 87
4.4.3. LE-STM setup: second stage . . . 87
4.4.4. LE-STM in Prague: closing remarks . . . 89
Chapter 5: Conclusions
91Appendix
93 I. Experimental setups for the characterization of Si-colls . . . 93II. Plasmons . . . 95
Bibliography
97Publications by the Author
101 A. Articles in impacted journals . . . 101B. Conferences . . . 102
C. Patents. . . 103
List of Symbols and Abbreviations
Greek variables
α absorption coefficient ηd dynamic viscosity
λ wavelength
ω frequency
ωSP surface-plasmon frequency σ optical absorption cross-section τ photoluminescence lifetime τ(non)rad (non)radiative lifetime
Roman variables and constants
A absorbance
c speed of light in vacuum (c= 2.99792458×108 m/s) Dc diffusion coefficient
D diameter of the nanocrystal
d distance
E energy
e charge of electron (e= 1.602×10−19 C)
h Planck constant (h= 6.626×10−34 J·s = 4.135×10−15 eV·s)
¯
h reduced Planck constant (¯h= 1.054×10−34 J·s = 6.582×10−16 eV·s) It tunneling current
I intensity of light
kB Boltzmann constant (kB = 1.380×10−23 J/K= 8.617×10−5 eV/K) l inner dimension of a cuvette
m electron mass (m = 9.109×10−31 kg) n volume concentration
T temperature
U voltage
Abbreviations
1D one-dimensional 1Nc single-nanocrystal
1NcS single-nanocrystals spectroscopy 2D two-dimensional
3D three-dimensional AFM atomic force microscopy Ag-np silver nanoparticles
ATR attenuated total reflection CdSeNc nanocrystalline CdSe
CMOS complementary metal oxide semiconductor DLS dynamic light scattering
F-band fast (blue) PL band in SiNc FCA free carrier absorption
FTIR Fourier-transform infrared spectroscopy FWHM full width at half maximum
HOPG highly oriented pyrolytic graphite
HRTEM high resolution transmission electron microscopy HSQC heteronuclear single quantum coherence
(i)CCD (intensified) charge-coupled device
LE-STM light emission induced with the tip of STM NMR nuclear magnetic resonance
PL photoluminescence porSi porous silicon
QE (internal) quantum efficiency R6G rhodamine 6G
SiNc silicon nanocrystal
Si-coll colloidal solution of silicon nanocrystals S-band slow (red) PL band in SiNc
SES shifting excitation spot
SNOM scanning near-field optical microscopy SPM scanning probe microscopy
STM scanning tunneling microscopy UV ultraviolet
VSL variable stripe length
Institutes
AV ˇCR Academy of Sciences of the Czech Republic, v.v.i.
CEA Commissariat `a l’´energie atomique, Saclay, France FZ ´U Institute of Physics
KCHFO Department of Optics and Chemical Physics
MFF UK Faculty of Mathematics and Physics, Charles University
List of Figures
Fig. 1.1 Injection of electrical carriers into SiNc-based field-effect light-emitting
diode by Walters et al. . . 4
Fig. 1.2 Raman laser by Rong et al. . . 5
Fig. 1.3 Evanescent laser by Fang et al. . . 6
Fig. 1.4 Tunability and a typical PL spectrum of SiNc . . . 7
Fig. 1.5 Radiative transitions in bulk and nanocrystalline Si . . . 9
Fig. 1.6 Influence of oxidation on PL spectra of SiNc by Wolking et al. . . 11
Fig. 1.7 Optical absorption cross-section of SiNc, comparison to FCA . . . 12
Fig. 1.8 Calculations of phonon modes in nanocrystalline GaP by Fu et al. . . 14
Fig. 1.9 Preparation of SiNc by chemical synthesis . . . 16
Fig. 1.10 Colloidal suspensions of SiNc by Sato and Swihart . . . 19
Fig. 1.11 1NcS of SiNc by Mason et al. and Martin et al. . . 21
Fig. 1.12 1NcS of SiNc by Sychugov et al. and Valenta et al. . . 22
Fig. 1.13 1NcS of SiNc by English et al. . . 22
Fig. 2.1 PL spectra of the SiNc powders prepared in our laboratory . . . 25
List of Figures
Fig. 2.2 Photos of PL of nanocrystalline layers prepared in our laboratory . . . 27
Fig. 2.3 Photo of white powder scraped off from the Si wafer. . . 27
Fig. 2.4 HRTEM of yellow SiNc . . . 28
Fig. 2.5 Map of PL dynamics of the standard SiNc sample, sketch of its origins 29 Fig. 2.6 Influence of oxidation on PL spectra of our samples . . . 30
Fig. 2.7 Spatial PL maps of 1sed in water and xylene . . . 32
Fig. 2.8 Setup for the measurements of spatial PL maps. . . 33
Fig. 2.9 Time evolution of spatial PL maps of white SiNc powder in xylene . . 34
Fig. 2.10 Photos of our Si-colls, PL spectra before and after filtration . . . 35
Fig. 2.11 Evolution of PL intensity during the preparation of our Si-colls . . . . 37
Fig. 3.1 Autocorrelation functions of our DLS measurements . . . 40
Fig. 3.2 Size distribution according to our DLS measurements . . . 41
Fig. 3.3 HRTEM measurements of our colloids . . . 41
Fig. 3.4 AFM measurements of our colloids . . . 42
Fig. 3.5 FTIR measurements of our colloids . . . 44
Fig. 3.6 1H and 13C NMR measurements of our colloids . . . 49
Fig. 3.7 Diffusion1H NMR of 1sedXDin chloroform . . . 50
Fig. 3.8 Excitation and emission spectra of our colloids . . . 53
Fig. 3.9 Comparison of PL of our colloidal and reference samples . . . 54
Fig. 3.10 Absorption spectra of R6G and 1sedX. . . 55
Fig. 3.11 QE measurements of our colloids . . . 57
Fig. 3.12 Integrated PL intensity as a function of excitation intensity of our colloids . . . 58
Fig. 3.13 PL of the colloidal sample 1sedXDafter its redispersion in chloroform. 59 Fig. 3.14 PL decay of our colloids . . . 60
Fig. 3.15 Comparison of absorption, excitation and emission spectra of our col- loid, scheme of excitation and emission processes . . . 61
Fig. 3.16 1NcS of our colloids . . . 64
Fig. 3.17 1NcS of the reference . . . 65
Fig. 3.18 Comparison of 1NcS by different groups, modified scheme of excitation and emission processes . . . 67
Fig. 4.1 The tip and the sample in the Tersoff-Hamann model. . . 73
Fig. 4.2 Examples of LE-STM applications from literature . . . 75
Fig. 4.3 Example of a theoretical calculation of the STM-cavity plasmon mode by Aizpurua et al. . . 76
Fig. 4.4 Examples of processes leading to LE-STM . . . 77
Fig. 4.5 Scheme of LE-STM setup for a simultaneous acquisition of two photon maps. . . 79
Fig. 4.6 Spectrally resolved LE-STM measurements from Ag-np . . . 80
Fig. 4.7 LE-STM of Ag-np with increasing tunneling bias . . . 81
Fig. 4.8 STM measurements of ncSi on Au(111) . . . 83
Fig. 4.9 LE-STM measurements on SiNc mixed with Ag-np . . . 84
Fig. 4.11 STM measurements of HOPG . . . 86
Fig. 4.10 Photo of the head of air STM microscope at FZ ´U AV ˇCR . . . 86
Fig. 4.13 Photos of LE-STM setups at FZ ´U AV ˇCR . . . 87
Fig. 4.12 Scheme of LE-STM setup. . . 87
Fig. 4.14 LE-STM measurements of Ag-np performed at the setup at FZ ´U AV ˇCR 88
List of Tables
Tab. 1.1 Comparison of CdSe and Si nanocrystals. . . 6Tab. 1.2 Absorption and emission energies calculated for SiNc by Wang et al. . 11
Tab. 1.3 Classification of porous silicon, relative polarity of the most common solvents . . . 15
Tab. 1.4 Classification of solid/liquid mixtures . . . 18
Tab. 1.5 1NcS of SiNc as reported by various groups. . . 20
Tab. 2.1 Preparation of SiNc in our laboratory . . . 26
Tab. 2.2 Effect of long-term stirring in various solvents (PL is excited with a 325-nm laser). . . 31
Tab. 2.3 Overview of Si-coll samples. . . 36
Tab. 3.1 FTIR vibrations, selected chemical terminology . . . 45
Tab. 3.2 Chemical composition of the solvent and Si-coll . . . 48
Tab. 3.3 Losses in samples used for QE measurements . . . 56
Tab. 3.4 Energy distance of hole levels for SiNc by Moskalenko et al. . . 67
Tab. 3.5 Estimates of the dose of nanocrystals and their concentration from absorption . . . 68
Tab. 3.6 Estimate of the concentration of nanocrystals in 1sedX from 1NcS. . . 69
Tab. 4.1 STM timelines. . . 71
Tab. 4.2 Interaction between the tip and the sample as a function of distance . 72 Tab. 4.3 The most common configurations of an LE-STM setup. . . 78
Tab. 4.4 LE-STM reports on silver from literature. . . 82
Tab. I Overview of experimental setups used to characterize Si-colls. . . 94
A Word of Introduction
This thesis is aimed at the investigation of a relatively new type of material, silicon nanocrystals. These tiny structures only a few nanometers in diameter, though still crystalline, represent a transition between two worlds, being neither in a broad sense macroscopic nor on a molecular level, from both theoretic and experimental point of view.
Although “nano” has become a trendy word, frequently misused for touting or even as hype, silicon nanocrystals actually fulfill the true meaning of this word. Despite consisting of several hundreds of atoms, they are small enough for the laws of quantum physics to apply, being thus endowed with unique properties. Especially, bright lumi- nescence, which appears in silicon nanocrystals in contrast to their bulk counterpart, deserves to be closely investigated.
Their size, comparable with smaller proteins, makes them unique, however, it also causes their investigation to be somewhat tricky. Probably most significant difficulties are encountered in theoretical calculations, which are unfortunately still unable to predict and describe their properties, partly also since the properties result from the influence of more than one factor.
We try to tackle the problem experimentally, focusing mostly on the emission of light and its underlying physical origins. Generally, two approaches to this problem can be adopted, using either a chemist’s or a physicist’s point of view. The primary point of this thesis is to try to fill the gap between these two approaches, exploiting the best each of them can offer.
More specifically, the aim in the long run is, consistently with predictions regarding the observation of optical gain in a silicon-based material described further on, to both prepare and study silicon nanocrystals in a sample of high optical quality with bright, stable and preferably yellow or green photoluminescence and high concentration of nanocrystals. This goal has not been reached in any laboratory worldwide though it is subject to intense research. The aim of this thesis is to push this concept further in the set course.
Organization of this thesis
The major part of this work, mostly the preparation of samples and basic optical charac- terization, was carried out at the Institute of Physics, Academy of Sciences of the Czech Republic, v.v.i. Experiments requiring blue nanosecond excitation (quantum efficiency measurements) and single-nanocrystal-spectroscopy measurements were performed at the Department of Optics and Chemical Physics at the Faculty of Mathematics and Physics at Charles University, while the setups of other characterization techniques used in this thesis are located in various collaborating laboratories in Prague, ranging from the Department of Low Temperature Physics (nuclear magnetic resonance) of the Faculty of Mathematics and Physics through the Institute of Chemical Technol- ogy (Fourier transform infrared spectroscopy) to several institutes of the Academy of Sciences (dynamic light scattering, high-resolution transmission electron microscopy).
Experiments involving light emission induced with a tip of a scanning tunneling mi- croscope were mostly carried out at the Nanophotonics laboratory of Commissariat `a
l’´energie atomique in Saclay, France.
Chapter 1 of this thesis both gives a motivation for the study of optical properties of silicon nanocrystals and summarizes the results from the literature concerning their behavior and physics which will be relevant in the following parts.
Chapter 2 then describes the preparation of silicon nanocrystals as is carried out in our laboratory and, for the sake of clarity, also several experimental results shedding light on the physical origins of their photoluminescence. Moreover, several types of samples derived (and prepared) from these silicon nanocrystals are included.
Chapter 3 focuses mainly on the investigation of structure, chemistry and optical properties of one type of silicon-nanocrystals-based sample, namely a colloidal solution of silicon nanocrystals.
Chapter 4 then concentrates on a specialized type of measurement combining the investigation of optical properties with very high spatial resolution.
In order to simplify orientation in the text, all the names of samples are printed in sans-serif font and the sample applies for Figures and Tables. Figures, tables and equations are numbered sequentially, separately for every chapter. Superscripts in square brackets[1], numbered sequentially, are used for citations, which are all placed at the end of the text. A special form of citations based on arabic numerals prefixed a letter[A1] (A for manuscripts, B for conferences, C for patent) is used for the results which the author of the thesis authored or co-authored and a recapitulation of which can again be found at the end of the thesis.
Since building the experimental setups used for the measurements in Chapters 2 and 3 was not a part of the thesis, they are not described in detail in the text. Instead, their overview is included in Tab. I in the Appendix. An overview of labeling of the colloidal samples, essential uncomplicated orientation in this thesis, is summed up in Tab. 2.3.
Chapter 1
Silicon nanocrystals: properties and applications
S
ilicon nanocrystals (SiNc) have received substantial attention ever since their potential for being an active material in efficient light-emitting devices was discov- ered by Canham[ 1 ] in 1990 . Taking into account the fact that silicon is currently the cornerstone in microelectronics, an introduction of Si-based (CMOS-compatible) pho- tonic devices into microelectronic circuitry would be more than welcome and is eagerly awaited. Unfortunately, monocrystalline bulk silicon’s band structure possesses a con- duction band maximum in the center of the Brillouin zone while the six symmetric conduction band minima lie near the X point 1.12 eV above the top of the valence band at room temperature. Such an indirect band-gap structure strongly hampers bulk silicon’s ability to generate light, requiring the participation of a phonon during an electron-hole recombination. A wide range of techniques have been and are being employed to circumvent the indirect nature of silicon’s band-gap, as discussed in sev- eral recent topical monographs[2–4]. SiNcs are one of promising and brightly emitting silicon-based materials, excelling thanks to their spectral tunability.1.1. Silicon photonics
In principle, silicon-based photonics can take two different routes depending on whether the wavelength of the used light is above or below silicon’s band-gap.
Generally, the essential parts of any photonic circuit comprise a source, a modulator, a waveguide and a detector. The latter two are quite well established on the silicon platform; they might consist in a combination[5]of (i) a silica (or silica-based) waveguide and a silicon photodetector in the visible region, which is situated above the band-gap of silicon and where the absorption is sufficient to allow for the use of a silicon detector or (ii) a silicon waveguide in the mid-infrared region below the band-gap of silicon where silicon is transparent and therefore a different type of detector (based on Si/SiGe) is necessary. As for fast modulators, electro-optic (Pockels) effect is usually exploited.
Even though this effect can be utilized in silica waveguides, i.e. in the former case,
(a) (b) (c)
Figure 1.1: Scheme of the injection of electrical carriers into SiNc-based field-effect light-emitting diode (after[15]).
it is not present in silicon due to the centrosymmetry of its lattice, implying that a different approach is necessary in the latter case. Although fast modulation in silicon was considered tricky, a 1-GHz-bandwidth modulation threshold in two metal-oxide- semiconductor silicon-based structures placed in a Mach-Zehnder interferometer has already been exceeded[6].
So far, the greatest challenge in silicon photonics remains to be an electrically driven silicon laser or at least a light-emitting diode with high-efficiency, although coherent light is preferable due to easier manipulation. Let alone its indirect band-gap, the second problem of a silicon-based laser is a high free-carrier absorption coefficient of silicon, which holds back population inversion, one of the prerequisites for lasing.
1.1.1. Approaches to the generation of light in (or on) silicon
Optically active defects and impurities: One way of squeezing light out of silicon is the in- troduction of optically active point defects. These include such defects as e.g. a G center∗, but the most prominent one is definitely erbium-doped silicon (Si:Er3+). In this material, the excitation energy is transferred to the Er ion and the radiative de- excitation takes place through a transition in its 4f-electron shell (4I13
2 →4 I15
2 ). The fact that this transition happens to fall in the absorption minimum of silica (1.54 µm) is very appealing for telecommunication applications.
Generally speaking, the luminescence via optically active point defects favorably produces very narrow lines, but tends to be quenched at “higher” (usually above- nitrogen) temperatures with the onset of back transfer, which is a severe drawback.
However, in the case of erbium, a way to get around this problem has been found in the form of the incorporation of Er ions coupled with SiNcs into SiO2waveguides. In such a structure, the deexcitation processes which occur in bulk Si no longer pose a problem and, moreover, the SiNcs efficiently both absorb the excitation energy and transfer it to the optically active ions. As a result, erbium-doped silicon-rich SiO2 emits 1.54-µm- centered narrow-line luminescence with decay time in the millisecond range and can be used as an amplifier. In addition, both optical gain[7] and electroluminescence[8] have been reported in this material.
∗a point defect constituted by two carbon atoms and one interstitial silicon atom
1.1. Silicon photonics
(a) (b)
Figure 1.2: Schematic of a Raman laser (after[17]), top view (a) and a cross-section of the coupler (b).
Another, quite original use of defects was demonstrated in a traditional Si p-n junction light-emitting diode[9] into which dislocation loops were introduced. These dislocation loops are believed to be agglomerates of interstitial Si atoms (formed by the implantation of boron to produce the p-n junction and subsequent annealing) and create a potential barrier trapping the injected carriers close to the junction region and thus curbing the diffusion to non-radiative recombination centers. The external quantum efficiency of the room-temperature electroluminescence of this device, situated at 1160 nm, reaches 2×10−4.
Quantum confined structures: The band structure of Si can be dramatically altered also by putting strict limitation on the size of the structure, producing superlattices, nanowires or the most commonly studied nanocrystals (crystalline quantum dots) and nanoclus- ters (quantum dots without crystalline structure). The usually visible room tempera- ture luminescence, decaying in microseconds, coming out of these structures is believed to be due to an interplay between quantum confinement and surface-related effects[10]. However, general consensus on the overall picture underlying the emission of light or optical gain, observed by many groups[11–14], in SiNc has still not been reached within the scientific community, which might be hindering further improvements.
An interesting approach to obtaining efficient carrier injection to SiNcs, clearly in- dispensable for any electrically driven device, have been introduced by Walters et al.[ 15 ]. In this study, the SiNcs are embedded in the oxide layer of a metal-oxide-semiconductor structure and the injection of electrons is temporally separated from the injection of holes. Once inside a nanocrystal, the electron-hole pair forms an exciton and recom- bines radiatively, resulting in field-effect electroluminescence (the injection process is sketched in Fig. 1.1).
Stimulated Raman scattering: The recent exploitation of a stimulated Raman-scattering effect[16] is claimed to be the first demonstration of coherent light generation in Si. In a Raman laser, a silicon waveguide is optically pumped with energy below its band- gap (1.55 µm → 0.8 eV) inducing Raman scattering†, whose cross-section is high in silicon. If a resonator-like arrangement is used for the waveguide (e.g. by coating one of the waveguide’s facets with a high-reflectivity layer), stimulated Raman scattering becomes dominant due to the introduction of a feedback loop. Although pumping above the band-gap is avoided, free carriers are still created by two-photon absorption‡ and consequently free-carrier absorption persists. Losses due to free-carrier absorption
†giving out the Stokes-shifted branch at 1.686µm
‡both the Raman scattering and two-photon absorption are third-order non-linear processes
bulk nanocrystals: photoluminescence
band- exciton typical max
gap Bohr radius peak position peak width decay QE QE CdSe 1.75 eV 5.0 nm 490–620 nm 150 meV ns 50% 85%
Si 1.12 eV 4.9 nm 620–900 nm 500 meV µs 1–5% 60%
Table 1.1: Comparison of CdSe and Si nanocrystals.
can be reduced by applying external voltage[17] (25 V) to a p-i-n structure across the waveguide to sweep out the free carriers. Such a device, sketched in Fig. 1.2, then exhibits a lasing threshold of 20 mW and output power of 50 mW.
Drawbacks of Raman lasers lie firstly in the inherent necessity of external optical pumping to induce Raman scattering and secondly in the physical limitation of the efficiency of a third-order non-linear process, placing the required waveguide length to be in the order of centimeters. Nevertheless, if not as an electrically pumped integrated device, Raman laser may find its use as an all-optical amplifier, wavelength converter or sensor.
Figure 1.3: Example of an eva- nescent laser (after[18]).
Hybrid approaches: An obvious way of getting light from a silicon chip is its combination with existing and effi- cient III-V lasers. The most problematic issues of this approach are connected with the induction of defects at the III-V/Si interface due to a lattice mismatch.
The most promising strategy in this field seems to be the production of so-called evanescent lasers, i.e. a III-V quantum-well laser placed on top of a silicon-on-insulator
waveguide[18](Fig. 1.3). The silicon waveguide and the III-V laser (on an InP substrate) are produced separately and bonded via plasma-assisted low-temperature wafer bond- ing, which is a key procedure in the fabrication. These evanescent lasers then lase at the wavelength of 1.5 or 1.3 µm, require the current of tens of milliamperes and have operating temperatures between 10 and 100 .
1.2. Photoluminescence of silicon nanocrystals
Semiconductor nanocrystals are structures containing hundreds to thousands of atoms, whose optical and electronic properties vary somewhere between those of molecules and bulk semiconductors. Of particular interest is the so-called strong-confinement regime, in which the radius of the nanocrystal is smaller than the bulk-exciton Bohr radius and consequently the confining potential substantially exceeds the electron-hole Coulomb interaction.
A “model” semiconductor nanocrystalline material is considered to be direct band- gap CdSe, whose energy levels and basic optical properties can be described with effective-mass calculations[19]. With proper surface passivation (ZnS capping), its pho- toluminescence (PL) quantum efficiency (QE) was found to be as high as 60–85%[20].
1.2. Photoluminescence of silicon nanocrystals
(a) (b)
Figure 1.4: Illustration of the spectral tunability of the PL from SiNc (a) (after[22]) and an example of the most common spectrum containing an S- and an F-band (b) with PL as typically seen by the naked eye in the inset.
Moreover, its PL peak position is tunable in the visible region from blue to red and CdSe nanocrystals are even commercially available§.
On the other hand, the PL emission from SiNcs is much less understood and they might seem to fall behind its rival (see Tab. 1.1). Theoretical calculations of their be- havior during the absorption and emission processes, which obviously involves complex non-equilibrium states, are still scarce and do not reflect realistic conditions. Calcu- lated structures are only tiny nanoclusters composed of at most a few hundreds of Si atoms and either completely lacking the influence of surface passivants (i.e. hydrogen atoms are supposed to passivate the dangling bond) or having just one e.g. oxygen pas- sivating atom “inserted” in between the hydrogen atoms. However, progress in both preparation techniques, largely suitable surface passivation, and understanding the PL emission is still being made and PL quantum yield as high as (62±11)%¶ has been reported[21].
1.2.1. Photoluminescence spectra of silicon nanocrystals: experimental insight
SiNc ensembles exhibit broad (500 meV) PL spectra tunable over the energies above the bulk Si’s band-gap throughout the most of the visible range, as is illustrated in Fig. 1.4a. However, this whole region is often regarded as two separate “bands”, a red (or slow) band (S-band, for short) and a blue (or fast) F-band. The former usually spans wavelengths over 600 nm while the latter is typically situated around 450 nm.
The green-yellow region in between these bands is generally difficult to access and therefore far less studied. Consequently, the most commonly presented PL spectrum comprises an S-band located between 650 and 800 nm (1.55–1.9 eV) and a less intense F-band around 450 nm (2.75 eV) and emits PL which is dark orange to the naked eye due to the broadness of the spectrum (Fig. 1.4b).
The fundamental reason for the distinction of the two bands lies not so much in the difference of their spectral positions but in their different dynamics, as is reflected in the
§colloidal dispersion of CdSe can be obtained fromEvident Technologies under the name ofEviDot
¶although not stable with time
alternative names. The S-band PL dynamics typically follows a stretched-exponential curve with time constants between 1–100 µs, often decreasing with shorter PL wave- lengths[23]. Besides the microsecond decay, an additional faster nanosecond component can also be observed within the S-band, typically at its shorter wavelengths[23]. On the hand, the blue F-band commonly exhibits fast nanosecond decay.
1.2.2. Quantum confinement
When the size of a structure shrinks substantially, the carriers start to be affected by the confining potential. The simplest model illustrating what happens due to this influence is a particle in a potential well. In such a system, discrete energy levels, which cause band-gap opening (blue-shift), arisek:
Egapnc =Egapbulk+Econ =Egapbulk+ ¯h2π2 2mr
1
R2, 1
mr
= 1 m + 1
mh
, (1.1)
¯
h being the reduced Planck constant, R the radius of the nanocrystal and mr the re- duced mass of the electron-hole pair. The difference in the band-gaps Econis sometimes referred to as the confinement energy.
This effect of size on the energy of the band-gap is both very well documented in various systems of semiconductor nanocrystals and explains the spectral tunability of SiNc illustrated in Fig. 1.4a. The prediction of the band-gap size dependence by this particular study[22] reads (Dstands for the diameter of the nanocrystal)
Egapnc (eV) = 1.16 + 11.8
D(nm)2. (1.2)
In general, depending on the model, the value of the exponent in the expression for the band-gap varies between 1.2 and 2.
1.2.3. Quasi-direct (no-phonon) optical transitions
Another consequence of the dramatic reduction in size for SiNc, which, however, has no parallel in direct band-gap semiconductor nanocrystals such as CdSeNc, is the break- down of the quasi-momentum conservation rule. In brief, the spatial localization of an electron and a hole increases the uncertainty of their crystal momentum, which in turn boosts the no-phonon-assisted radiative recombination (Fig. 1.5).
This phenomenon was theoretically developed by Hybertsen[ 25 ] using an envelope function approach by expanding the electron states to the crystal basis states, which includes the electron-phonon coupling. Consequently, he was able to quantify the probabilities of zero-phonon (quasi-direct) and TO-phonon-assisted transitions both of which increase considerably (over several orders of magnitude) with the decrease in the size of a nanocrystal. According to these calculations, quasi-direct transitions should dominate for sizes under 2.0–1.5 nm.
Experimental studies shedding more light on this subject were performed using res- onant PL experiments at low-energy PL tail at low temperatures (4 K)[24]. Narrow
kfor the sake of simplicity, the model of an infinite square potential well is mentioned
1.2. Photoluminescence of silicon nanocrystals
(a) (b) (c)
Figure 1.5: Illustration of radiative transitions in bulk (a) and confined (b) silicon. The ratios between phonon-assisted and no-phonon transitions were also determined experimentally (after[24]) as is shown in (c).
lines revealed in PL spectra were attributed to transitions involving various phonons and allowed for the estimation of the ratios of no-phonon and phonon-assisted transi- tion probabilities (Fig. 1.5c). For band-gap energies over 1.8 eV (690 nm), no-phonon transitions were shown to prevail.
Thus, as a direct result of quantum confinement, the oscillator strength, which is indirectly proportional to lifetime, of both phonon-assisted and no-phonon optical transitions rises, enhancing PL compared to classic bulk silicon although the indirect nature of the band-gap still holds.
1.2.4. Auger processes
Unfortunately, apart from the enhancement of the radiative recombination of excitons in a quantum-confined system, the competing non-radiative Auger recombination pro- cesses also increase. However, two contending effects are present: even if the enhanced Coulomb interaction and the increase in the electron-hole wave function overlap boost the quantum-confined Auger rates, they are hindered by the reduced availability of final states satisfying conservation laws.
Very few studies actually address the topic of Auger processes in SiNc. Probably the only (partly) theoretical work by Delerue et al.[ 26 ] suggests that Auger processes are fast (faster than 1 ns) and, consequently, if more than one exciton is present in the nanocrystal, the “excess” exciton will decay non-radiatively. This effect would, obviously, severely limit the PL intensity at higher excitation levels since the number of excitons that could decay radiatively in one nanocrystal would be strongly limited.
Although studies performed on other quantum-confined systems apart from SiNc cannot be straightforwardly transfered to SiNc, they can give interesting hints about the influence of quantum confinement. For example, Klimov et al.[ 27 ] performed ex- periments on colloidal CdSeNc and deduced that the corresponding relaxation times scale as R3 for quantum dots with sizes between 1 and 4 nm. On the other hand, in elongated CdSe nanorods (diameter ∼5 nm, length ∼30 nm), the Auger recombi- nation occurs in a bimolecular, two-particle, fashion, i.e. the energy of one exciton
is transferred to the other exciton[28]. While in traditional cubic Auger processes the lifetimes and corresponding recombination rates scale as a square of excitation flux n (1/τAugercubic ∼n2), in a bimolecular process Auger recombination has much less influence as it scales linearly (1/τAugerbim ∼n).
Besides the non-radiative Auger recombination itself, a nanocrystal can also be ion- ized through Auger ionization processes: when two excitons are present in a single nanocrystal, one of them can recombine via an Auger process and transfer its energy to one of the remaining carriers, which can consequently escape from the nanocrys- tal. Such ionization suppresses radiative recombination until the nanocrystal becomes neutral again.
1.2.5. Surface states
The effect of the surface in nanocrystalline materials starts to be far from negligible as their surface-to-volume ratio grows significantly with the reduction in size∗∗.
In reality, since silicon under ambient atmosphere readily oxidizes and forms SiO2, the surface of SiNc can never be kept “pure”, i.e. hydrogen passivated††. Theoret- ical calculations by Puzder et al.[ 29 ] of the impact of surface passivation on optical properties‡‡ show that especially double-bonded passivants (Si=O) induce dramatic reductions of band-gap in small nanocrystals by introducing passivant-related states inside the band-gap. Another oxygen configuration, bridged oxygen (Si-O-Si), can also alter the band-gap, though less than in the double-bonded case (the higher the induced distortion, the larger the influence). On the other hand, single-bonded passivants (e.g.
-OH) exhibited much smaller impact on the band-gap energy, and they were able to boost radiative recombination rates.
Qualitatively the very same behavior was also observed experimentally by Wolkin et al.[ 10 ], who prepared two identical sets of samples and kept them separately in argon and air atmospheres. Although the set in argon exhibited band-gap blueshift as predicted by the confinement theory (Fig. 1.6a), PL from the second set was limited to orange (Fig. 1.6b), which agrees with the onset of a limitation on the size of the band-gap by the presence of surface oxide-related states.
Another, this time solely beneficial consequence of the presence of surface states lies in the reduction of non-radiative Auger recombination as spatial localization of at least one of the carriers in a surface state diminishes its interaction with a third carrier[30].
1.2.6. Emerging image
When all the previously mentioned phenomena are taken into account, the physics underlying PL from SiNc emerges: although non-radiative recombination processes
∗∗e.g. the nanoclusters studied in[29] were Si35H36 (D ≈ 1 nm), Si66H65 (D ≈ 1.4 nm), Si87H75
(D≈1.5 nm), (D=a3 q3
4πN = 0.337√3
N (nm),abeing the lattice constant andN the number of Si atoms)
††hydrogen passivation if far more unstable in SiNc than in bulk Si, this effect is most probably caused by the strongly curved surface of SiNcs, which enhances chemical reactions
‡‡carried out by the exchange of one or two hydrogen surface atoms for the studied passivant
1.2. Photoluminescence of silicon nanocrystals
(a) (b) (c)
Figure 1.6: PL of a set of samples prepared under the same conditions kept in argon atmosphere (a) and exposed to air (b) (the color bars indicate the peak wavelength of emission) and a model explaining this behavior (c). All published by Wolkin et al.[ 10 ].
persist, the radiative recombination rates increase to the magnitude comparable with non-radiative rates. With proper surface passivation, the effect of dangling bonds, acting as non-radiative traps, can be diminished. When the size of the nanocrystal shrinks, the band-gap opens and PL shifts to the visible range∗. Moreover, quasi- direct radiative transitions become reasonably probable. However, if oxygen is present, at a certain size, which is usually predicted to be around two or three nanometers, radiative oxide- (either SiO2 or bridged Si-O-Si) -related states inside the band-gap set in and start to limit further changes in the optical band-gap. This situation is illustrated inFig. 1.6c.
When the size of the nanocrystal falls into zone I in Fig. 1.6c, every nanocrystal luminescence at its own wavelength range, giving rise to (inhomogeneously) broadened featureless spectral bands.
1.2.7. Competing views
particle D Eabs Eemi (nm) (eV) (eV) Si5H12 0.45 6.40 2.29 Si35H36 1.09 4.37 2.89 Si59H60 1.36 3.72 3.18 Si87H76 1.48 3.47 3.25 Si123H100 1.74 3.11 3.04 Si199H140 2.00 2.81 2.76
Table 1.2: Absorption and emission ener- gies calculated for SiNc by Wang et al.[ 31 ].
The above-pictured image of SiNc light emission is still not accepted by every member of scientific community and several competing explanations, including PL from siloxenes[32], oxide-related de- fects, an Si-Si dimer[33] or a self-trapped exci- ton[34], have been suggested. Especially con- troversial is the case of blue-emitting SiNcs (F- band), whose PL is often claimed to come solely from SiO2 defects, while sometimes it is at- tributed to the quantum-confined emission from very small nanoclusters, and even a mechanism of direct band-gap recombination in the Γ-point
of the band structure, enabled due to direct band-gap shrinkage as a result of quantum confinement, was proposed.
∗weak PL of bulk silicon is situated in the infrared region: EgSi= 1.12 eV→1.1µm
(a) (b)
Figure 1.7: Optical absorption cross-section of SiNc for various excitation and emission energies as measured by Kovalev et al.[ 37 ] (a). An extrapolation of the emission cross-section σem from (a) is compared to the cross-section of a free-carriers absorption process (σFCA) in (b).
As for the less controversial S-band, an eclectic but insightful point has been raised that in such a tiny system as a 3-nanometer-sized nanocrystal, the distinction between core and surface states is no longer valid, but these two types of states couple and their properties merge[35,36]. However, no theoretical model, up to now, describes this type of behavior neither does it draw general attention.
Another interesting observation has been made in a theoretical work by Wang et al.[ 31 ]. He calculated the limitation of the size-shrinkage-related band-gap open- ing based on the fact that the optical excitation of a nanocrystal causes a geometrical distortion in its crystal lattice (an Si-Si bond is elongated), which in turn changes the nanocrystal’s energy levels. As a consequence, absorption and emission energies start to differ for smaller nanocrystals and the emission energy slightly shifts back to longer wavelengths (see Tab. 1.2).
1.3. Other optical properties of silicon nanocrystals
Naturally, virtually all optical properties are influenced in the confinement regime.
A phenomenon attracting most attention is, obviously, the onset of stimulated emission in contrast to bulk silicon.
Optical gain in SiNcs has been observed by many groups all over the world[11–14], usually based on the measurements performed with the variable-stripe-length (VSL) method, in which the dependence between the exciting stripe length and the detected signal is observed† and should follow an exponential curve. In spite of that, a laser on SiNc has not yet been realized.
†although a “shifting-excitation-spot” (SES) method was proposed as a control measurement[38]
to distinguish small gain from artifacts always present in a real experiment, some of the reports on optical gain in SiNc do not use it; this control method is based on the excitation of the sample with a small spot that is shifted along the whole length of the stripe and an integrated SES signal is then compared to the VSL measurement
1.3. Other optical properties of silicon nanocrystals
1.3.1. Optical absorption cross-section
Being a product of the oscillator strength of the optical transition and the electronic density of states, the optical absorption cross-sectionσ directly reflects the probability of the optical transition. Consequently, if its absolute value is known, it can be used to determine the concentration of nanocrystals in a studied sample directly from the measurements of the optical absorption coefficientα:
α=nσ, (1.3)
(n being the volume concentration).
Although extremely difficult to quantify, σ of silicon nanocrystals was experimen- tally measured by Kovalev et al.[ 37 ] by a pump-probe technique. He studied the PL saturation regime and was able to model his experiments with the absorption cross- section as a single fitting parameter (see Fig. 1.7a). Based on these experiments, σ of SiNcs (for a nanocrystals emitting at about 600 nm) is generally considered to be in the order of 10−17 cm−2.
1.3.2. Implications for stimulated emission and lasing
Considering a simple “two-level” model, in which electrons can be excited from a valence band edge to the conduction band edge and photons are then emitted via the reverse process, the wavelength of the emitting photon depends on the band-gap and therefore the size of the given nanocrystal. The measured absorption cross-section shown in Fig. 1.7a can thus be carefully extrapolated to these “resonant” conditions. Within this model, the absorption and emission cross-sections have the same value and this value can be compared with the cross-section of free-carrier absorption‡, one of the most undesirable processes when it comes to stimulated emission in silicon.
Such a comparison is made in Fig. 1.7b. While emission cross-section rises, free- carrier absorption cross-section decreases with decreasing emission wavelength as it behaves asαFCA ∼λ2 according to the Drude model. As a result, for sufficiently small nanocrystals, the emission cross-section outweighs the free-carrier absorption cross- section and stimulated emission should become observable in this region.
In order to realize a laser, however, the amplifying medium needs to be combined with an optical cavity. An attempt in this direction has been made in our laboratory by Dohnalov´a et al.[39,A8] using an optically induced distributed-feedback laser cavity.
Although true lasing was not achieved in this work, both stimulated emission and the influence of the cavity on the emission were observed, the bottom line being the fact that homogeneous samples with a high density of nanocrystals but also high optical quality (i.e. low light scattering) are necessary for optical-gain experiments.
1.3.3. Phonons in silicon nanocrystals
Similarly to other optical properties, phonons are also altered in the strong confinement regime. This phenomenon is clearly evident from the measurements of Raman spectra,
‡in the case of nanocrystals, this process is sometimes also referred to as “confined-carrier absorp- tion”
in which the position of the peak at 520 cm−1 (corresponding to the transversal optical phonon mode in the Γ point) of monocrystalline silicon shifts for small nanocrystals toward lower energies up to 480 cm−1, which marks the position of the transversal optical phonon in amorphous silicon[40].
Figure 1.8: Calculation of phonon modes in nanocrystalline GaP[41]. Besides the shift in bulk frequencies, new frequencies arise in the “gap” between the bulk phonons (gray shaded areas).
The calculations performed on nanocrys- talline GaP[41], a material, whose bulk counter- part also exhibits indirect band-gap, qualita- tively corroborate these measurements. How- ever, besides a downshift in energies of the
“bulk-like” modes, new modes appear in the region between the acoustic and optical bulk phonons. These new modes were found to be spatially localized near the surface of the dot, and are therefore referred to as “surface-like”
(see Fig. 1.8).
Another phenomenon connected to a quan- tum confined system is the so-called phonon bottleneck. It is claimed to degrade lumines- cence performance as due to the quantization of electron energy levels the non-radiative re- laxation into the “first” excited state cannot
occur unless the energy spacing matches phonon energy[42]. Such an effect is likely to produce slow relaxation and hinder luminescence efficiency. However, other processes such as multiphonon effects and collisional broadening have been proposed to help get around this problem.
1.4. Fabrication techniques and types of samples
Generally, samples containing SiNcs can be of three “forms”: either powder-like SiNc agglomerates, interconnected together by another material (e.g. SiO2)§, or incorporated into a solid matrix, most frequently SiO2, or dispersed in a liquid, forming a suspension or even a colloid.
Since the powder-like material is usually built into a solid matrix or dispersed in a liquid, SiNc samples, and correspondingly their fabrication methods, can be roughly looked at as if they belonged to two groups: those prepared by dry (“physical”), or by wet (“chemical”) approach. Although this distinction might not seem significant, in re- ality there is hardly any overlap between these two groups as physicists are usually able to study optical properties thoroughly, but the range of possible samples, particularly different types of surface passivation, is somewhat limited, whereas chemists tend to be able to prepare particles with diverse surface chemistry, but they investigate optical properties, which strongly depend on the passivation, only superficially.
In addition, another point of view from which the fabrication methods can be distinguished is the classical division into a top-down or bottom-up approach.
§single SiNcs are too small to be stored, they behave like smoke
1.4. Fabrication techniques and types of samples
classification diameter of pores surface area (nm) (m2/cm3)
macroporous >50 10–100
mesoporous 2 to 50 100–300
microporous <2 300–800
nanoporous
refers to the nanometric size of the Si grains, not the pores; often used in connection with light-emitting porSi
(a)
solvent polarity
water 1.000
ethanol 0.654 acetone 0.355 chloroform 0.259 benzene 0.111 toluene 0.099 cyclohexane 0.006
(b)
Table 1.3: Classification of porous silicon (a) (after[43]) and the relative polarity (with respect to water) of the most common solvents (b).
1.4.1. “Wet” methods
Electrochemical etching: The very first method of preparing SiNc was electrochemical etching of bulk silicon, which is a typical example of the top-down approach. A mono- crystalline Si wafer is etched in a hydrofluoric-acid-based solution with current flowing through the electrolyte. By this process, a sponge-like structure of porous silicon (porSi), consisting of interconneted SiNc and silicon nanowires, is formed. The obser- vation of efficient room-temperature PL situated in the orange-red visible region by Canham[ 1 ], which aroused keen interest in the emission of light from silicon nanostruc- tures, dates back to as late as 1990 , although the first porSi was prepared as early as in 1956 by A. Uhlir[ 44 ].
The fabrication process itself usually lies in the electrochemical etching (anodiza- tion) of a P-type silicon wafer using a platinum counter-cathode, which is resistant to the highly corrosive electrolyte. The electrolytic solution contains hydrofluoric acid and ethanol, added in order to ensure good wettability. The silicon dissolution into hexafluorosilicate ions proceeds according to the following scheme[45]
e− H2
H ↑ H F ↑ F SiF4+ 2 F− −→SiF−6
@ ¡ @ ¡ ↑
h+1 Si ←2 F− −→ Si ←2 HF −→ H H,
¡ @ ¡ @ @ ¡ @ ¡
Si Si Si Si Si Si
¡ @ ¡ @
(1.4)
leaving hydrogen-terminated pores in between silicon nanostructures. Obviously, holes are indispensable for the etching process. While during the etching of a P-type wafer holes are provided by the acceptor atoms, the etching of an N-type wafer requires ultraviolet irradiation for their generation¶.
By adjusting the etching current density, layers of different porosities can be fab- ricatedk; the usual classification is summarized in Tab. 1.3a, which also points out the rapidly growing surface-to-volume ratio, typical for any nanostructure.
¶the need for irradiation in N-type wafers is often exploited to create patterned porSi structures
khigh current densities, however, lead to an electro-polishing regime
(a)
(b) (c)
Figure 1.9: Sketch of a tetraoctylammonium bromide ([CH3(CH2)7]4NBr) molecule (a) and a micelle (b). Sketch (c) describes the reduction ofSiCl4 with lithium aluminium hydride to fabricate SiNc.
Although freshly etched porSi can emit from red through orange to green, this PL is unstable since the surface of SiNcs is terminated with an unstable silicon-hydrogen bond. SiNcs can then either oxidize in air, which leads to the introduction of surface- oxide-related states, usually limiting the PL peak maximum to the red region (650–
700 nm), or can be passivated in a controlled atmosphere with other chemical species and various molecules. However, the impact of the passivant on the PL of SiNcs is not yet well investigated and custom-made nanocrystals luminescing in a chosen spectral region are not routinely prepared in laboratories yet.
Colloidal synthesis: Colloidal synthesis is, on the other hand, a bottom-up approach.
Various chemical reactions, usually taking place in a liquid, lead to the formation of SiNc dispersed in a colloidal dispersion. These reactions always take place in controlled atmosphere (e.g. nitrogen), resulting in hydrogen-terminated nanocrystals, and very often also at high pressures and temperatures.
One example of a colloidal synthesis of SiNc is the reduction of SiCl4 with lithium aluminium hydride published by Tilley et al.[ 46 ]. The whole process is depicted in Fig. 1.9c. It takes place, unlike many other syntheses, at room temperature and ambient pressure, however, a nitrogen glove box is still necessary.
The synthesis starts with the dispersion of tetraoctylammonium bromide (TOAB, Fig. 1.9a) in oxygen-free anhydrous toluene. The TOAB molecule is a surfactant molecule composed of a hydrophilic (polar) head part and a hydrophobic (non-polar) tail. When dispersed in an aqueous (polar) solution, for example, such molecules would aggregate to form a micelle (a sketch of which is shown in Fig. 1.9b) screening off the hydrophobic tails from water∗∗, whereas in non-polar solutions inverse micelles, with hydrophobic tails sticking out from an aggregate of hydrophilic heads, would appear.
When silicon tetrachloride (SiCl4) is added to non-polar toluene (the polarities of the most common solvents can be found in Tab. 1.3b), an SiCl4 aggregate in an inverse TOAB micelle forms. Subsequently, a strong reducing agent (LiAlH4) is added to pro- duce SiNc from SiCl4 and finally the reducing agent is quenched by oxidation with methanol. This procedure yields hydrogen-terminated SiNc, which can be then passi- vated with various organic molecules. In addition, the colloid is usually filtered, dried and the powder can be redissolved in another solvent.
∗∗according to the chemists’ rule of thumb: like dissolves like
1.4. Fabrication techniques and types of samples This method provides small SiNcs with negligible size distribution. However, the PL of the resulting dispersion is usually situated in the blue or, surprisingly, even the UV region†† and tends to be, sometimes quite hastily based only on the measured PL spectrum, ascribed to the direct electron-hole recombination in SiNcs.
1.4.2. “Dry” methods
Ion implantation: Ion implantation belongs to bottom-up approaches and consists in the implantation of SiO2 glass matrices with high-dose (1016–1017 ions/cm2) excess Si ions and subsequent annealing (∼ 1000 °C) of the sample. After annealing, a thin film (several hundreds of nanometers) of SiNcs is created inside the glass slab.
Ion-implantation-prepared samples are of very high optical quality and their PL spans the red and the beginning the of infrared region (approximately between 750 and 1000 nm). However, the density of nanocrystals in the matrix is limited and the size of SiNcs, driven mainly by the spontaneous assemblage of Si atoms, cannot be well controlled.
Annealing of silicon-rich films: Annealing of silicon-rich films is based on a similar idea.
This time, however, a silicon-rich thin film (sub-stochiometric SiOx, nitride silicon- rich film) is deposited directly onto a substrate by means of various chemical vapor deposition techniques or magnetron sputtering and after annealing an SiNc layer is formed.
Laser ablation: Laser ablation takes advantage of a high-energy laser focused onto a tar- get monocrystalline Si wafer in inert-gas atmosphere. The laser pulse ejects a plume of material from the target wafer, which then settles down onto a substrate (Si wafer, quartz).
1.4.3. Spotlight on: colloids of silicon nanocrystals
SiNcs in liquid environment are of particular interest since various organic solvents provide, at least in theory, an easy way to influence, or even tailor, surface chemistry of nanocrystals, affecting both PL and the solubility of nanocrystals in the liquid.
The dissolution of SiNc in a liquid would provide a highly convenient sample for the study of optical properties. Moreover, transparent, homogeneous mixtures are an ideal system for the measurements of optical gain, since the losses induced by the scat- tering of light are eliminated. Naturally, one of the most important requirements for the solid/liquid mixture (seeTab. 1.4 for their classification) to become transparent is the sufficiently small size of the solid particles. Unfortunately, oxide-passivated SiNc powders are made up of large agglomerates and when mixed with a solvent yield turbid suspensions. Even if sonication is applied, the agglomerates cannot be reduced in size under about 1 µm. The necessary subsequent filtration, implemented in order to get
††e.g. in the paper by Tilley et al.[ 46 ], the PL maximum was situated at 335 nm and HRTEM measurement confirmed the size of (1.8±0.2) nm, while the core was still crystalline
