Erbium doped silicon nanoclusters for ... - Semantic Scholar

Erbium doped silicon nanoclusters for Microphotonics

Domenico Pacifici

` DEGLI STUDI DI CATANIA UNIVERSITA DOTTORATO DI RICERCA IN FISICA — XVI CICLO

Domenico Pacifici

Erbium doped silicon nanoclusters for Microphotonics

Tutor: Prof. Francesco Priolo Coordinatore: Prof. Antonio Insolia

Tesi per il conseguimento del titolo

Ai miei genitori To my parents

Cover: contour plot showing the simulated number of excited Er ions per Si nanocluster, as a function of Er concentration and pump power. Brighter gray tones represent higher values. More details can be found in Fig. 4.26 within this thesis, on page 102.

Erbium doped silicon nanoclusters for Microphotonics Domenico Pacifici Ph. D. Thesis - University of Catania Printed the 10th of December 2003

When you find your limit, get over it.

Tutor: Prof. F. Priolo Facoltà di Scienze Matematiche Fisiche e Naturali Università degli Studi di Catania

The work described in this thesis was mainly performed at the Laboratory for MAterials and Technologies for the Information and communication Science (MATIS), within the National Institute for the Physics of Matter (INFM), at the Department of Physics and Astronomy, University of Catania and at the Institute for Microelectronics and Microsystems (IMM), within the Italian National Research Council (CNR). Electroluminescent devices have been realized in collaboration with STMicroelectronics, sited in Catania.

Contents 1 General Introduction 1.1 The future of Microelectronics and Photonics . . . . . . . . . 1.2 Towards a silicon based Microphotonics . . . . . . . . . . . . . 1.2.1 Quantum confinement in low dimensional Si structures 1.2.2 Defect engineering: Er3+ in silicon . . . . . . . . . . . . 1.3 Contents of this thesis . . . . . . . . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

1 . 2 . 6 . 7 . 8 . 11

2 Si nanocrystals 2.1 Introduction . . . . . . . . . . . . . . . . . 2.2 Sample preparation . . . . . . . . . . . . . 2.3 Structural properties . . . . . . . . . . . . 2.4 Optical properties . . . . . . . . . . . . . . 2.5 Excitation cross section . . . . . . . . . . . 2.6 Photoluminescence excitation spectroscopy 2.6.1 Experiment . . . . . . . . . . . . . 2.6.2 PLE measurements . . . . . . . . . 2.7 Conclusions . . . . . . . . . . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

13 13 14 15 16 20 23 23 25 30

. . . . . . . . . . . .

31 31 32 32 33 33 39 46 49 49 55 58 60

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

3 Ion-irradiated Si nanocrystals 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Sample preparation . . . . . . . . . . . . . . . . . . . . 3.3 The experiment . . . . . . . . . . . . . . . . . . . . . . 3.4 Defect production in Si nanocrystals . . . . . . . . . . 3.4.1 Damage accumulation in Si nanocrystals . . . . 3.4.2 Phenomenological modeling . . . . . . . . . . . 3.4.3 Theory versus experiment . . . . . . . . . . . . 3.5 Defect annealing in Si nanoclusters . . . . . . . . . . . 3.5.1 Recovery of slightly damaged Si nanocrystals . . 3.5.2 Recrystallization of amorphized Si nanoclusters 3.6 Amorphized Si nanocrystals versus amorphous bulk Si 3.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

4 Er-doped Si nanoclusters 63 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

ii

CONTENTS 4.2 4.3 4.4 4.5 4.6 4.7

Sample Preparation . . . . . . . . . . . . . . . . . . . . . . Experimental . . . . . . . . . . . . . . . . . . . . . . . . . Structural Properties . . . . . . . . . . . . . . . . . . . . . Optical Properties . . . . . . . . . . . . . . . . . . . . . . Quantum efficiency . . . . . . . . . . . . . . . . . . . . . . Excitation mechanisms . . . . . . . . . . . . . . . . . . . . 4.7.1 Photoluminescence excitation spectroscopy . . . . . 4.7.2 Effective excitation cross section . . . . . . . . . . . 4.8 De-excitation mechanisms . . . . . . . . . . . . . . . . . . 4.8.1 Temperature dependence . . . . . . . . . . . . . . . 4.8.2 Concentration quenching effect . . . . . . . . . . . 4.9 Experimental quantum efficiency . . . . . . . . . . . . . . 4.10 Modeling the Si nanoclusters-Er interaction . . . . . . . . 4.10.1 Energy levels scheme . . . . . . . . . . . . . . . . . 4.10.2 Up-conversion mechanism . . . . . . . . . . . . . . 4.10.3 Effects of up-conversion on the excitation rate of Er 4.10.4 Transfer-time estimate . . . . . . . . . . . . . . . . 4.10.5 Implication of the finite transfer time . . . . . . . . 4.10.6 Number of excited Er ions per nanocluster . . . . . 4.10.7 Erbium-induced optical gain . . . . . . . . . . . . . 4.10.8 Confined carrier absorption . . . . . . . . . . . . . 4.10.9 Net optical gain at 1.54 µm . . . . . . . . . . . . . 4.11 Increasing the luminescence yield . . . . . . . . . . . . . . 4.12 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Devices based on Er-doped Si nanoclusters 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . 5.2 Experimental . . . . . . . . . . . . . . . . . . . . 5.3 Electroluminescence measurements . . . . . . . . 5.4 Excitation cross section under electrical pumping 5.5 Conclusions . . . . . . . . . . . . . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . .

65 68 69 70 71 73 73 74 76 77 79 83 84 84 88 93 95 97 101 103 106 107 109 114

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

117 117 120 122 125 128

Summary and Future Perspectives

129

Bibliography

135

Acknowledgements

143

Curriculum Vitae

147

List of Publications

149

Chapter 1 General Introduction

In this chapter, I’m going to draw a general introduction concerning the main topics and problems relative to the research fields this thesis is mainly focused on. Indeed, the two most important research branches which have lead to the Information Era we are living on nowadays, namely Microelectronics and Photonics, are both going to face the so called electronic bottleneck. In fact, on one hand integrated circuits are obediently experiencing the exponential decrease in feature sizes predicted by Moore’s Law, which in a few years would end up generating much faster microelectronics circuits, at the cost of tightly spaced (tens of nanometers) devices and very long (few kilometers) wires interconnecting them over an area of about 1 cm2 . Hence, problems such as the crosstalk among adjacent chips, the increase in power dissipation, inductive effects and delays in signal transmission limiting the overall performances will become of great concerns. On the other hand, on the highways of the world’s optical-fiber telecommunication networks, while data travel at the fastest velocity physically achievable, i.e. light speed, and brand new fibers are conceived to carry more and more information, once light signals reach the central hubs, they need still to be converted into electronic signals by bulky and expensive electronic switches, which are physically unable to handle with more complex optical signals and as a result slow down the overall communication process. A way of overcoming the performance limitation of traditional microelectronic interconnects and to skip the step of converting light into electrons, is to use optical devices able to manipulate directly photons, and which could serve as optical interconnects.

2

CHAPTER 1. General Introduction Microphotonics, a term proposed by Lionel C. Kimerling to name the next revolutionary technology, deals with the integration of optical functionality on an electronic circuit. Indeed photons, unlike electrons, do not dissipate heat, are immune to crosstalk and electromagnetic interference, have very low transmission losses, and, moreover, have practically unlimited information capacity, being characterized by very broad bandwidths. Silicon, due to its mature manufacturing industry and to its peculiar physical properties, would be the material of choice. Unfortunately, up to now the main limitation to achieve monolithically integrated photonics and electronics to share one chip has been the lack of an efficient silicon based light emitter or even a silicon based laser. In this thesis, the feasibility and the realization of efficient light emitters based on Er-doped Si nanoclusters is addressed and electrical pumping demonstrated.

1.1

The future of Microelectronics and Photonics

The invention of the first transistor at Bell Laboratories in 1947 and the realization of the first integrated circuit (IC) at both Texas Instruments and Fairchild Company twelve years later have revolutionized the electronics industry, and determined the rise of a new research field, i.e. microelectronics, devoted to the manipulation and elaboration of information by using electrons at the microscopic level. In 1965 Intel co-founder Gordon Moore published an article about cramming more components onto integrated circuits [1]. By analyzing some statistical data about the evolution of the microelectronics industry, Moore was able to discover a natural trend, i.e. that the number of transistors that could be fit onto a square inch of silicon doubled every 12 months. Actually he was really interested in the cost/integration curve, since he noticed that electronic circuits of all types would have lower costs if they could be integrated in higher numbers onto thiny sliver of silicon. Clearly higher and more dense number of integrated circuits meant also higher computational powers and wider sphere of application. It was this intuition which lead him to extend the trend in the future, and allowed him to predict the strong impact and enormous modification in everyday life caused by large scale integration in microelectronics. In Fig. 1.1 it is possible to appreciate a modern version of the Moore’s Law [2]. The number of transistors on a microprocessor and of bits on a DRAM is shown to double every 18 months over a chip area of 1 cm2 . Starting from values of about 1K in 1970, we arrive nowadays to roughly 1 billion bits on a single memory device. In order to achieve these results, however, new technologies, such as lithography, able to define smaller and smaller wires for connection and microelectronic devices had to be developed. On the right hand scale of Fig. 1.1, indeed, the minimum lithographic feature size is shown to exponentially decrease over time. Starting from values of tens of micrometers, nowadays the minimum size that can be patterned on a chip has been reduced to ∼ 100 nm. And people operating in the field are working hard in order to let Moore’s Law remain true. As a consequence of the size reduction, faster and faster microelectronic components have been generated. Indeed, it is possible

1.1 The future of Microelectronics and Photonics

3

4G 1G 256M 64M 16M 4M 1M 256K 64K 16K 4K 1K

10

1

0.1 DRAM Logic & Microprocessor

1970 1975 1980 1985 1990 1995 2000 2005

0.01

Minimum lithographic feature size (mm)

Transistors or bits on a chip

Moore's Law

Year Figure 1.1. - Left hand scale: number of transistors (open squares) on a Micropocessor (MPU) and number of bits (solid triangles) on a Dynamic Random Access Memory (DRAM) as a function of production year. Right hand scale: minimum lithographic feature size (open circles). Data over 2002 are near term prevision taken from International Technology Roadmap for Semiconductors (ITRS), Interconnect 2002 Update (http://public.itrs.net).

to demonstrate, for example, that in a standard Metal Oxide Semiconductor Field Effect Transistor (MOSFET), the response time of the device is proportional, in a first approximation, to the time constant of an RC circuit equivalent to the device, having resistance R and capacitance C. This capacitance is proportional to the channel width W and length L. Therefore, reducing the characteristic lengths by a certain scale factor, the response time of the device is decreased by the same amount. However, since more numerous electronic components are integrated on the same chip, longer interconnection wires are needed in order to distribute clock and other signals to all of them. For example, in Fig. 1.2 the trend of total interconnect length over an active area of 1 cm2 is shown as a function of production year. Nowadays, kilometers of tiny wires run under a 1 cm2 chip area. Given a constant resistivity, the increase in the total interconnect lengths produces an enormous increase in resistance. Therefore problems such as overheating, caused by the enhanced power dissipation, and delay in the signal transmission, which increases with the square of wire length, become of great concern, since they affect the overall device performances. As an example, in Fig. 1.3 the delay time for 0.8 µm thick and 43 µm long wires is shown as a function

CHAPTER 1. General Introduction

Total Interconnect length (km/cm2)

4

10

1

2002

2004

2006

2008

Year Figure 1.2. - Total interconnect length inside a chip, for active wiring only, per square centimeter of active area. Near term prevision data taken from ITRS, Interconnect 2002 Update (http://public.itrs.net).

of the minimum feature size. While the gate delay is shown to decrease at a constant rate as a result of the shrink in feature size (solid circles), the delay time related to interconnects based on Al wires and SiO2 insulator exponentially increase every time a new device generation is developed (open triangles). In particular, a point in the figure exists where the time response of the integrated circuit is no more limited by the gate switching of the single electronic component but by the delay of local wires interconnecting them. The cross over point at 350 nm represent indeed the start of the so called interconnection bottleneck. A short term solution to this problem, adopted in 1998, has been to replace the standard Al wires (resistivity ρ=3.0 µΩ cm) with a metal element having smaller resistivity, such as Cu (ρ=1.7 µΩ cm), and use low-dielectric-constant materials as insulators, instead of SiO2 (dielectric constant κ= 4.0). In the same Fig. 1.3, the delay times for Cu/low κ interconnects are shown as open squares. The introduction of this materials simply postpones the problem by shifting the cross over point towards smaller feature sizes. Sooner or later, the problem will be faced again. In particular, by 2005 microelectronics is predicted to face the 100 nm wall, since a further scaling down could be blocked. Thus, in order for integrated circuit technology to continue along the Moore’s Law curve, new and alternative approaches beyond copper and low-κ materials need to be conceived for

1.1 The future of Microelectronics and Photonics

5

Speed/Performance Issue

Delay (ps)

40

Gate delay Interconnect delay (Al & SiO2) Interconnect delay (Cu & low k)

30 20 10 0 700 600 500 400 300 200 100

Minimum feature size (nm) Figure 1.3. - Calculated gate and wire delay as a function of the minimum feature size, i.e. device generation. The cross over points represent the start of the interconnection bottleneck. Wires are 0.8 µm thick and 43 µm long. Al and Cu resistivities are respectively 3.0 µΩ cm and 1.7 µΩ cm. The dielectric constant of SiO2 is κ= 4.0, while low κ stands for κ= 2.0. Data are taken from ITRS, Interconnect 2000 Update (http://public.itrs.net).

the long term period. Among the different solutions, the seek for unconventional interconnects is one of the most promising. In particular, optical interconnects are considered a primary option for replacing the conductor/dielectric system [3]. This would signify a drastic change in the mean used to carry information. No more electrons, but photons will transport the information in the form of optical signals. The advantages in using photons are enormous. First of all they have unlimited information capacity, due to their very broad bandwidth. They do not dissipate heat and have very low transmission losses. Moreover they travel at the fastest reachable velocity, i.e. the speed of light, thus strongly reducing the problem of delays. Another advantage of optical interconnects is the application of wavelength and frequency multiplexing of signals simultaneously on the very same interconnect. Photonics will also benefit from optical interconnects, since they could efficiently replace bulky and expensive electro-optic switches and modulator nowadays necessary to convert the optical signals in electrical ones, and viceversa, causing huge delays in the transmission of information which characterizes our modern era. The dream of realizing devices able to directly manipulate and transmit light signals at a microscopic integration level is the first research goal of microphotonics, the

6


optical equivalent of microelectronics for integrated circuits. The long term will see the integration of both optical and electronic fuctionalities on the very same chip. Critical issues for the realization of such a dream are the design and fabrication of low-loss waveguides, of detectors and of light emitters monolithically integrated on silicon. The problem of silicon based light emitters will be the subject of this thesis.

1.2

Towards a silicon based Microphotonics

Silicon microphotonics is a rapidly growing technology which tries to merge on a single chip photonics and silicon microelectronic components [4]. Silicon is by far the most suitable material for microelectronics. Indeed, it is an abundant element, its crystallinity is unparalled by III-V composite materials, and it has very good electrical, thermal and mechanical stability. Moreover, its oxide (SiO2 ) is a very good insulator, which can well-passivate surfaces and acts as an effective diffusion barrier. All of these properties, made silicon the leading material in microelectronic applications. Silicon can do well even in Photonics. Indeed its high refractive index (n=3.5) makes it a good medium through which light can be guided and transmitted. Indeed, several low loss waveguides based on silicon have been developed, using different approaches [5–7]. Moreover, silicon based detectors have been realized, able to convert a light signal into an electrical signal [8–11]. What silicon is still lacking, is an efficient light emitter or even a laser. For a long time, silicon has been considered unsuitable for optical application, due to the indirect nature of its electronic bandgap. Indeed, once an electron-hole pair is created inside bulk silicon, the radiative recombination producing a 1.1 eV photon necessitates of a phonon in order to conserve crystal momentum. Hence the emission of a photon is a three-particles process (electron, hole and photon) and is characterized by very small rates, bringing to lifetimes of the order of ∼ 1 ms. This lifetime is too long if compared to the recombination times of the order of µs or even ns characterizing some non-radiative transitions, such as the ShockleyHall-Read (SHR) recombination by deep levels introduced in the gap of silicon by defects and metallic impurities [12, 13], or Auger recombination with the promotion of a free carrier (electron or hole) in higher lying levels within the conduction or valence band respectively. All of these processes, being orders of magnitude more probable than radiative recombination, make photon emission from silicon a very weak process, characterized by quantum efficiencies of the order of 10−7 -10−6 , four orders of magnitude smaller than for direct bandgap semiconductors. In order to solve this problem, many routes have been followed. Among these, silicon based nanostructures and defect engineering of silicon through insertion of efficient light emitters in the crystal have shown great potentialities. In the following, both of the approaches will be discussed, and the main milestones evidenced.

1.2 Towards a silicon based Microphotonics

1.2.1

7

Quantum confinement in low dimensional Si structures

Once a semiconductor structure is reduced to nanometric sizes (1-10 nm), the electron and the hole become spatially confined, and cannot be described anymore by planar wavefunctions, but as a superposition of them, their wavefunctions becoming wave-packets. Moreover, due to the boundary conditions, only certain wavefunctions with precise wavelengths can be supported by the material. Therefore the energy of the confined particles become quantized. In addition, since the position of the particles can be determined with less indetermination with respect to the bulk, as a result of the Heisenberg’s uncertainty principle, the momentum of the particles suffers a greater indetermination. This determines an increase in the minimum energy level available to the particles. Indeed, in nanostructured silicon, both the electron and the hole are characterized by higher energies, thus producing an increase of the nanocrystal energy gap with respect to bulk silicon. A qualitative schematic of the quantum confinement effects on the energy gap of silicon produced by decreasing the spatial dimension is reported in Fig. 1.4. From a theoretical point of view, the determination of the actual energy gap of a Si nanocrystal depends on the method used, and many corrective factors, such as the electrostatic energy interaction between electron and hole forming the exciton (∼ 1/L), the exchange interaction (∼ 1/L3 ) among them, the spin-orbit interaction, and so on [14]. However, all these methods agree fairly well in predicting the general behavior of the energy gap as a function of nanocrystal size [14–24]. Another effect of quantum confinement is the increase in the probability associated with an optical transition. Indeed, due to the reduced size, the translational symmetry of the system is no more satisfied and, as a consequence, the crystal momentum is no more a good quantum number. This allows for vertical, direct band to band, no-phonon transition to become more probable [25]. The total radiative probability per unit time can be expressed as the product of the oscillator strength times the density of states involved in the transition. It has been shown that the oscillator strength exponentially increases as the number of silicon atoms in the nanocrystal decreases [26], thus leading to an increase of the total radiative recombination probability. Moreover, the density of states for system of decreasing dimensionality becomes more and more atomic-like as far as we increase the number of dimensions involved in the confinement. Going from threedimensional to zero-dimensional (quantum dots) systems, the density of states tends to resemble the Dirac’s delta function, producing again an increase in the probability of a photon to be emitted in a exciton recombination process. For this aspect, Si nanocrystals seem to be the most promising canditates for optical application. Moreover, in Si nanocrystals non-radiative recombination paths such as SHR and Auger with free carriers should be strongly reduced, since on one hand the carriers, being confined, have a reduced mobility which prevents them to reach deep levels in the matrix, on the other hand, due to the increased band gap energy, thermally excited carriers are practically absent, strongly reducing Auger effects.

8


8 eV

1.1 eV

Eg

L

Eg

L

Figure 1.4. - Effects of quantum confinement on the energy gap of silicon. By decreasing the characteristic dimensions L, the energy gap increases (as 1/L2 ).

From an experimental point of view, the first evidence of an enhancement in luminescence yield as a result of quantum confinement was given by Leigh Canham in 1990 [27] through porous silicon, which is obtained by electrochemically etching a silicon wafer. The process induces the formation of nanometric crystalline structures in the form of arrays of linear wires, and strong light emission occurs by exciting the system with UV lamps. The emission can be tuned in the visible by simply varying the porosity of the film. It has been demonstrated that with increasing porosity, the characteristic mean sizes of the crystallites decrease, and contemporarily the emission spectrum is shifted towards the blue, due to quantum confinement effects [27, 28]. Other methods have been developed for preparing nanostructured silicon. Among these, ion implantation of high silicon excess in SiO2 is of great importance due to the strong compatibility with microelectronic processing. A high temperature annealing process following the sample preparation step determines the nucleation of the high Si excess and the growth of Si nanocrystals. Also in this case, efficient light emission mainly due to quantum confinement can be achieved [29–31].

1.2.2

Defect engineering: Er3+ in silicon

Another way of overcoming the intrinsic inability of silicon to work as a light emitter is to introduce deep luminescent centers in the electronic band gap which could efficiently capture the exciton energy and afterwards emit part of it in the form of photons [32]. In this approach, the emission of light becomes a sequential process consisting of the following steps: (i) an electron-hole pair is generated within the semiconductor, (ii) the electron-hole pair, in its free wandering all over the bulk material, is captured by the deep center, (iii) the electron-hole energy is transferred to the center which is therefore excited, (iv) the center emits light. Clearly, in order for the overall process to be efficient, the capture of the electron-hole pair from

1.2 Towards a silicon based Microphotonics 2

9

Energy (eV)

H11/2 4 S 3/2

2.0

4

F9/2

4

I9/2

4

I11/2

Eg 1.0

4

I13/2 1.54 µ m

4

I15/2 Er 3+ Free Ion

0.0

Er 3+ in c−Si

Figure 1.5. - Levels scheme for Er3+ as a free ion and in a solid matrix. Once Er is introduced in crystalline silicon, a splitting of the energy levels occur, due to the Stark effect produced by the crystalline field. The value of the silicon energy gap is also reported on the right energy scale.

the luminescent center has to be more probable than from other quenching centers. Moreover, the electron-hole energy has to match one of the excited levels of the luminescent center, for the energy transfer to occur. The energy transfer process has to be very efficient. Eventually, once the luminescent center is excited, it should be able to de-excite mainly by emitting photons. One of the most promising centers is erbium. Erbium is a rare earth element characterized by an electronic configuration [Xe]-4f11 5d1 6s2 . Once introduced in a solid matrix, erbium is brought in a 3+ state of charge, loosing the three outer electrons. Therefore, 11 electrons remain in the 4f shell. Due to spin-spin and spinorbit interactions, many levels with different electronic configurations and energies are formed [33]. Within the Russel-Saunders approximation, the fundamental level (or term) of erbium is 4 I15/2 . The first excited level is 4 I13/2 and so on for the higher levels. In Fig. 1.5 a scheme of the energy levels of Er3+ is reported, both in free space and in crystalline silicon. For a free Er3+ ion, dipole-dipole optical transition

10


within different levels are parity forbidden, because the levels, originating from the same f shell, have the same parity, while the dipole-dipole operator has odd parity, thus allowing for transitions to occur only for levels characterized by opposite parity. Instead, the crystalline field produced by the atoms of a solid matrix surrounding Er causes a Stark-splitting of the degenerate levels, which as a consequence determines a partial removal of the level degeneracy, and a mixing of levels with different parities. Therefore, the optical transitions become weakly allowed, with excitation and emission cross sections having values of the order of 10−21 cm2 in glass hosts [34]. Among the others, the optical transition occurring from the first excited level (4 I13/2 ) to the ground state (4 I15/2 ) of Er3+ is of particular interest since it produces a photon at 1.54 µm, a strategic wavelength for telecommunications, since optical fibers based on silica have a minimum loss at that wavelength. Erbium in crystalline silicon has been the subject of many studies in the past years [35–44]. It is known that Er3+ in crystalline silicon can be excited by an electron-hole pair which is at first trapped at an erbium-related center (probably an Er-O complex) and then transfers its energy to erbium [37–40], which is hence brought in its first excited level. Usually Er can be introduced in the silicon matrix by ion implantation. Er has a low solid solubility in crystalline silicon (∼ 1×1018 /cm3 , [45]), therefore introduction of high Er concentrations, needed to achieve reasonable gain values given the very low emission cross section, is prohibitive since the thermal process needed to activate Er and eliminate the damage left over by ion implantation would precipitate Er into a silicide phase, which is optically inactive. This problem has been solved by Polman and collaborators [46] who performed erbium implants at 77K in order to amorphize silicon. The following thermal treatment at 600◦ C determines a planar solid phase epitaxial recrystallization of the amorphous layer, thus permitting non equilibrium Er contents to be incorporated in crystalline silicon. However, during the recrystallization process, a fraction of Er segregates at the moving amorphous/crystal planar interface, becoming optically inactive. Further improvements have been obtained by coimplanting oxygen with erbium. The Er-O complex which forms after a suitable thermal treatment is able to block the erbium diffusion across the interface, making it possible to incorporate high concentrations of optically active Er ions (>1020 /cm3 ) in the matrix [47]. The excitation of Er through electron-hole pair recombination is characterized by an effective cross section of 3×10−15 cm2 [48], which is 5 to 6 orders of magnitude higher than direct-resonant optical absorption. However, despite this high excitation cross section, poor luminescence yield are obtained at room temperature. Indeed, strong non radiative processes compete with the radiative recombination of the excited erbium in silicon, namely the Auger effect with free carriers [48–51], in which an excited Er ion gives up its energy to a free electron or hole, promoting it to higher lying levels in the conduction or valence band, respectively, and the energy back-transfer to silicon, with the generation of an exciton [41,48,52,53]. Both processes are thermally activated and determine a strong drop of the lifetime of the

1.3 Contents of this thesis

11

first excited level of erbium, and as a consequence a strong quenching of the 1.54 µm luminescence for temperatures higher than 77 K. Indeed, at room temperature, a feeble emission is observed. Nevertheless, one of the most important result in this research field has been the realization of room temperature operating erbium-doped silicon based light emitting diodes emitting at 1.54 µm [54,55]. Indeed, under reverse biasing an Er-doped p+ -n+ diode, Er is efficiently excited by impact with hot carriers. Moreover the temperature quenching is much less pronounced than in photoluminescence, due to the fact that Er is excited within the depletion region of the device where Auger processes with free carriers are suppressed. This made it possible to obtain internal quantum efficiencies of ∼ 0.01 % at room temperature and under reverse bias. However, there is still room for further improving the Er luminescence efficiency.

1.3

Contents of this thesis

Aim of this thesis is to study in details the optical properties of Er doped Si nanoclusters embedded in SiO2 . This system offers the possibility to merge the positive aspects of the efficient electron-hole mediated excitation of Er in crystalline silicon with the lack of strong non-radiative paths typical for Er in SiO2 . This thesis is organized as follows: In Chapter 2 the optical properties of undoped Si nanocrystals are discussed. It is shown that in densely spaced nanocrystals, a strong interaction sets in, producing an energy transfer among nearby nanocrystals and a consequent energy migration all over the sample. This energy transfer can be suppressed arranging the nanoclusters in multilayers separated by oxide layers. Through risetime and photoluminescence excitation (PLE) measurements, the excitation cross section under optical pumping has been determined over a wide excitation wavelength range (290-800 nm), and values in between 10−17 cm2 and 10−14 cm2 have been found. A comparison between the absorption properties of bulk Si and Si nanocrystals is made in the end. Chapter 3 reports a detailed study on the damaging effects of energetic ion beams on the structural and optical properties of Si nanocrystals. It is shown that the nanocrystal luminescence can be quenched by a single defect inside the nanocrystal volume. A phenomenological model is developed, which relates the fraction of damaged, i.e. quenched, nanocrystals to the total defect concentration left over by the ion beam in the film containing the Si clusters. Moreover, it is shown that due to dimensionality effects, Si nanocrystals can be amorphized by fluences much smaller than those used for crystalline bulk silicon, but higher annealing temperature and longer times are needed in order to completely recover their luminescence properties. This study is important since Si nanocrystals will be doped with Er through ion beam implantation, which is a ULSI-compatible technique.

12


Chapter 4 is entirely devoted to the physics of light emission from Er doped Si nanoclusters. It is shown that the ion implantation completely amorphizes the Si nanocrystals present in the film. Nevertheless, even in their amorphous phase, Si nanoclusters are shown to be efficient sensitizers for the Er luminescence. Indeed, luminescence efficiencies two orders of magnitude higher than for Er in SiO2 can be obtained. This is due to the fact that the effective excitation cross section for Er in presence of Si nanoclusters is enhanced with respect to Er in SiO2 , while the strong non-radiative processes typically limiting the Er luminescence in silicon are here suppressed due to the enlarged energy gap of Si nanoclusters. The excitation and de-excitation processes will be studied in details, and a fit of the experimental data through a model describing the energy transfer from a nanocluster to the surrounding Er ions will give us information about important, yet unknown, physical parameters, such as the nanocluster-Er coupling constant, the energy transfer time, the up-conversion coefficient, and the number of excitable Er ions per nanocluster. Moreover, the possibility of obtaining positive gain values at 1.54 µm from Er doped Si nanoclusters will be addressed, with a particular attention to the main processes, such as up-conversion between excited Er-Er pairs, and confined carrier absorption (CCA) from an excited Si nanocluster, limiting the amplification process. In Chapter 5 the feasibility and promising optical properties of light emitting devices based on Er doped Si nanoclusters are demonstrated. The presence of Si nanoclusters dispersed in the insulating oxide matrix allows for a good carriers injection in the film, thus stabilizing the electrical properties of the device. On the other hand, it is demonstrated that even in this case Si nanoclusters are acting as efficient sensitizers for the Er luminescence, being excited by impact of hot electrons and transferring their energy to the nearby Er ions. The excitation cross section for electrical pumping of Er has a value of ∼1×10−14 cm2 , which makes it possible to reach internal quantum efficiencies of the order of 1 %. Eventually a summary and an outlook on the future trends of this continuously expanding field will be given.

Chapter 2 Si nanocrystals

In this chapter, both the structural and the optical properties of Si nanocrystals produced by plasma enhanced chemical vapor deposition are discussed in details. In particular it is shown that under certain conditions a strong interaction among neighboring Si nanocsrystals sets in, determining a decrease of the luminescence lifetime and a characteristic stretching of the decay-time curves. This interaction can be reduced by increasing the mean distance between nanocrystals. The excitation cross section of Si nanocrystals is determined at an excitation wavelength of 488 nm by measuring the risetime of their luminescence signal at different laser pump powers. Through photoluminescence excitation spectroscopy the values of the excitation cross section are explored in a wide range of excitation wavelengths, between 290 and 800 nm.

2.1

Introduction

Since the discovery of porous silicon as an efficient light emitter [27], a lot of work has been done in order to understand the origin of the peculiar luminescence properties of silicon nanostructures. Nowadays it is well understood that quantum confinement is responsible for the blue-shifted emission of these nanostructures and that the Si=O double bond [56, 57] at the interface between a Si nanocrystal and the oxide host plays a crucial role in explaining the strong Stokes shift between the emission and the absorption observed in porous-Si, producing carrier related states in the band gap from which the electron-hole (e-h) pair can recombine after being quickly trapped in. Moreover the spatial confinement of carriers determines both a greater overlap of the e-h wave functions in k -space and a strong suppression of non radiative recombination processes, thus producing an increase of the luminescence efficiency. Among all, porous silicon [27, 56, 58, 59], silicon nanocrystals embedded within SiO2 [60–69] and Si/SiO2 superlattices [70–73] have been widely studied both from an optical and a structural point of view. Important issues for

14

CHAPTER 2. Si nanocrystals

the optical properties of Si nanocrystals are the absorption cross section and the energy transfer between nanocrystals. These parameters describe the efficiency in nanocrystals excitation and how the energy is redistributed afterwards within the sample. Recently, absorption cross sections have been measured in porous Si [74] while migration and trapping of excitons within nanostructures have been studied in both porous Si and Si nanocrystals formed by ion implantation [68,75]. In the following, it will be demonstrated that Si nanocrystals act as interacting systems within the oxide matrix, with the energy preferentially transferred from the smaller (i.e. larger band-gaps) to the bigger ones (i.e. smaller band gaps). This interaction can be strongly suppressed by ordering the Si nanocrystals in nanometric periodic layers, spaced by an oxide layer [60]. In this way the energy hopping between nanocrystals is less probable, due to the larger mean distance among Si nanocrystals.

2.2

Sample preparation

Si nanocrystals have been produced by high temperature annealing of substoichiometric SiOx (x < 2) thin films prepared by using a parallel plate plasma enhanced chemical vapour deposition (PECVD) system. The deposition system consists of a vacuum chamber (base pressure 1×10−9 torr) and a radio frequency (RF) generator (13.56 MHz), connected through a matching network to the top electrode of the reactor. The bottom electrode is grounded and acts also as a sample holder. Deposition processes have been performed by using 50 W of input power. The substrates, consisting of 5” (100) Czochralski silicon wafers, have been heated at 300◦ C during the deposition. The source gases used were high purity (99.99% or higher) SiH4 and N2 O. The N2 O/SiH4 flow ratio γ has been varied between 6 and 15, while keeping constant the total gas flow (about 140 sccm) and the total pressure (6×10−2 torr). With this procedure the Si concentration of the SiOx films has been varied between 35 and 44 at.% and samples with different excess Si content were obtained. After deposition, the SiOx films were annealed for 1 hour at 1250◦ C in ultra-pure nitrogen atmosphere. The high temperature annealing induces the phase separation between Si and SiO2 , and Si nanocrystals, whose size depends on the excess Si amount, were formed. More details can be found in [69]. An alternative procedure to obtain almost completely isolated Si nanocrystals in SiO2 is through high temperature annealing of PECVD grown Si/SiO2 multilayers (or superlattices, SLs), consisting of 11 SiO2 layers (8.5 nm thick) alternated with 10 ultra thin Si layers (0.9 nm thick). SiO2 layers were deposited by using a N2 O/SiH4 flow ratio of 30, with a total pressure of 3.5×10−2 torr and 30 W of RF power. Si layers were deposited by using a SiH4 flow rate of 20 sccm, with a total pressure of 1.5×10−2 torr and 25 W of RF power. After deposition, the SLs were annealed for 1 hour at 1200◦ C in ultra-pure nitrogen atmosphere to induce the breaking and balling up of the ultra thin Si film, with the formation of Si nanocrystals totally embedded within SiO2 .

2.3 Structural properties

15

Figure 2.1. - High resolution TEM of a SiOx sample after high temperature treatments. The presence of Si nanocrystals is cleraly demonstrated.

2.3

Structural properties

The structural properties of all of the samples have been studied by transmission electron microscopy (TEM). Dark field plan view TEM analyses were carried out with a 200 kV Jeol 2010 FX microscope to determine the nanocrystal size distribution in the annealed samples. Fig. 2.1 reports a high resolution transmission electron microscopy image performed on a SiOx sample containing 42 at.% Si annealed at 1250◦ C for 1 hour. The formation of crystalline Si precipitates having nanometric dimensions is clearly evident. In order to acquire quantitative information about the size distribution of the Si nanocrystals dispersed in the film, many plan-view TEM micrographs have been analized. As an example, in Fig. 2.2 dark-field plan-view TEM images as well as the size distributions relative to the Si nanocrystals are shown for three SiOx samples containing different amount of Si excess and annealed at the very same temperature of 1250◦ C for 1 hour. It is worth noticing that the mean size as well as the density of the nanocrystals formed after the annealing process increases with the excess Si amount present in the as deposited SiOx film. Indeed, this is in agreement with a model proposed by Nesbit [76], where the Si nanocrystals formation is thought of in terms of a first nucleation of small Si clusters inside the matrix, followed by an Ostwald ripening

16


T = 1250 °C (a) 37 at.% Si 30 r = 1.1 nm 20 10

20 10

(c) 42 at.% Si r = 1.7 nm

0 30

Number of crystals (%)

0 (b) 39 at.% Si 30 r = 1.5 nm

20 10 50 nm

1

2

3

0

Radius (nm) Figure 2.2. - Dark-field plan-view TEM micrographs and size distribution for three SiOx samples containing different amount of Si excess and annealed at the very same temperature of 1250◦ C for 1 hour [69].

process which determines the growth of larger nanocrystals at the expenses of smaller ones. According to this model, at a fixed annealing time and temperature, the mean nanocrystal radius depends only on the total amount of Si in excess in the SiOx matrix.

2.4

Optical properties

Optical properties of all of the samples were studied through photoluminescence (PL) measurements performed by pumping with the 488 nm line of an Ar laser. The pump power was varied between 10−2 and 60 mW over a circular area of ∼0.6 mm in diameter and the laser beam was mechanically chopped at a frequency of 55

2.4 Optical properties

PL Intensity (a.u.)

1.2

17

mean radius

1.0 0.8

35 at.Si 37 at.Si 39 at.Si 42 at.Si 44 at.Si

1250 °C 0.6 0.4 0.2 0.0 600

700

800

900 1000 1100 1200

Wavelength (nm) Figure 2.3. - Normalized PL spectra of SiOx thin films with different silicon concentrations annealed at 1250◦ C for 1 hour. Spectra were measured at room temperature, with a laser pump power of 10 mW.

Hz. The luminescence signal was analyzed by a single grating monochromator and it was detected either by a photomultiplier tube or by a liquid nitrogen cooled Ge detector. Spectra were recorded with a lock-in amplifier using the chopper frequency as a reference. All spectra have been measured at room temperature and have been corrected for the detector response. Luminescence lifetime measurements were performed by chopping the laser beam with an acousto-optic modulator, detecting the luminescence signal with a near-infrared photomultiplier tube having an almost flat response from 300 to 1600 nm and analyzing it with a multichannel scaler having the signal from the modulator as a trigger. The overall time resolution of our system is of 30 ns. All of the produced Si nanocrystals emit light at room temperature in the range 700-1100 nm. As an example in Fig. 2.3 the normalized PL spectra of Si nanocrystals obtained by annealing at 1250◦ C SiOx samples having different Si contents are reported. The luminescence signal clearly shows a marked blue shift with decreasing Si content as a result of the smaller size of the Si nanocrystals. Indeed, the average nanocrystals radius (as observed by TEM) increases from 1.1 to 2.1 nm by increasing the Si content from 37 at.% to 44 at.% Si. These data are consistent with a large number of observations supporting the

18


10

Normalized PL Intensity

multilayer 35 at. Si 37 at. Si 39 at. Si 42 at. Si 44 at. Si

= 700 nm

0

-1

10

-2

10

0

100

200

300

Time (s) Figure 2.4. - Measurements of the time-decay of the room temperature PL signal at 700 nm for a Si/SiO2 multilayer (with a Si layer thickness of 0.9 nm) after annealing at 1200◦ C for 1 hour (dotted line) and for SiOx samples with different Si contents annealed at 1250◦ C for 1 hour. Data were taken at room temperature and with a laser pump power of 10 mW.

quantum confinement model. The decay time of the emitted radiation is of particular interest since it reflects the confinement properties of the nanocrystals. Figures 2.4 and 2.5 show the decay-time of the PL intensity at two fixed detection wavelengths, 700 nm and 950 nm respectively, for different samples after switching off a 10 mW 488 nm laser beam at t=0. It is quite interesting to note that the decay time at 700 nm (Fig. 2.4) (i) increases with decreasing Si content, (ii) is characterized by a stretched exponential shape which becomes more and more similar to a single exponential with decreasing Si content. Extremely interesting is the behavior of the 950 nm signal decay time reported in Fig. 2.5, which is characterized by almost single exponentials with the same lifetime of 175 µs for all the silicon contents. In Fig. 2.4, moreover, the decay time of the PL intensity at 700 nm in the nanocrystalline multilayer is also reported. This decay time, under the same experimental conditions and at the same wavelength, is much longer than that of all of the other samples and in particular of the 35 at.% Si sample which has similar nanocrystal size distribution. Stretched exponential functions have been widely observed in the literature in the decay time of both porous Si and Si nanocrystals [68,75]. In a stretched exponential the decay line shape is given by:


2.4 Optical properties

19

= 950 nm

0

10

-1

10

35 at. Si 37 at. Si 39 at. Si 42 at. Si 44 at. Si

-2

10

0

100

200

300

Time (s)

400

500

Figure 2.5. - Measurements of the time-decay of the PL signal at 950 nm for samples with different Si nanocrystals concentration and mean size.

I (t) = I0 e−(t/τ )

β

(2.1)

where I(t) and I0 are the intensity as a function of time and at t=0, τ is the decay time and β is a dispersion factor. In general it is β < 1. The smaller is β the more stretched is the exponential. The factor β decreases from 0.85 to 0.63 and τ from 65 µs to 10 µs on increasing the Si content from 35 at.% to 44 at.%. For the multilayer, the time-decay curve resembles a single exponential being fitted with τ =80 µs and β=0.9. A decrease in β has been associated [68] with a redistribution of the energy within the sample caused by an energy transfer from smaller nanocrystals (having larger gaps) to larger nanocrystals (having smaller gaps). This picture is consistent with the reported data, showing smaller β and τ values in Si-rich samples in which the nanocrystal concentration is higher and hence the energy transfer is more probable. It should be stressed that, since all measurements are performed at the same wavelength they reveal the properties of the same class of nanocrystals (i.e. having the same size) embedded within different samples. The markedly different behavior observed demonstrates that the environment of the nanocrystal plays a quite important role in determining its decay time. The more the nanocrystals are isolated (larger β) the larger is the decay time τ (since energy transfer becomes less

20


700 nm

950 nm

a)

b)

Figure 2.6. - Schematic picture of the energy transfer mechanism among Si nanocrystals. a) Energy levels and b) space domain representation of the energy migration.

probable). Moreover, larger nanocrystals (those emitting at 950 nm, see Fig. 2.5) cannot transfer their energy to the surrounding nanocrystals since their energy gap is smaller. Therefore they act as ’isolated’ nanocrystals in all systems in the sense that once excited they will re-emit the energy only radiatively. This explains the identical lifetime with all surroundings and the single exponential-like behavior. A schematic picture of the energy transfer mechanism occurring among Si nanocrystals is reported in Fig. 2.6. In particular, Fig. 2.6a) evidences that an energy transfer from a nanocrystal to another one having higher or equal dimensions can occur due to the difference in the energy gaps, while the reverse mechanism is strongly forbidden by the energy conservation rule. In Fig. 2.6b) a scheme of the energy migration all over the sample produced by the energy transfer mechanism is reported. Since the energy transfer mechanism is likely due to a dipole-dipole interaction among two neighboring Si nanocrystals represented as atomic-like systems, it is easy to understand that this mechanism can be controlled only by a fine tuning of the mean distance among nanocrystals, and can be suppressed only if very far apart nanostructures are formed.

2.5

Excitation cross section

An important issue to be investigated is the excitation cross section of these nanocrystals and its dependence on the nanocrystals density and on the excitation and detection wavelength. In order to do this we have studied the risetime of the Si nanocrystals PL intensity. PL intensity I is in general given by: N∗ (2.2) τR being N ∗ the concentration of excited centers and τR the radiative lifetime. The I∝

2.5 Excitation cross section

= 850 nm

1.2


21

1.0 0.8 0.6 60 mW 30 mW 10 mW 2 mW 0.4 mW

0.4 0.2 0.0

0

100

200

300

400

500

Time (s) Figure 2.7. - Time resolved PL intensity of Si nanocrystals emitting at 850 nm when switching on the pumping laser at t=0 for a SiOx film annealed at 1250◦ C for 1 hour. The excitation wavelength was 488 nm. Data were taken at room temperature and at different pump powers and are normalized to the maximum intensity, corresponding to the steady state equilibrium conditions.

rate equation for nanocrystals excitation will be: dN ∗ N∗ = σφ (N − N ∗ ) − (2.3) dt τ being σ the excitation cross section, φ the photon flux, N the total concentration of optically active nanocrystals and τ the decay time, taking into account both radiative and non-radiative processes. If a pumping laser pulse is turned on at t=0, the PL intensity, according to eq. (2.2) and (2.3), will increase with the following law: I (t) = I0 1 − e−t/τon

(2.4)

with I0 being the steady state PL intensity and τon the characteristic risetime, given by the following expression: 1 1 (2.5) = σφ + τon τ A measure of the risetime as a function of photon flux φ will therefore give direct information on the excitation cross section. We have performed such measurements

22


4

5x10

= 850 nm 4

1/on (s-1)

4x10

4

3x10

= 5.8x10 cm -16

2

2x104 = 73 s 4

1x10 0.0

19

2.0x10

4.0x1019 -2 -1

6.0x1019

Photon flux (cm s ) Figure 2.8. - Reciprocal of the risetime τon as obtained from the data in Fig. 2.7 as a function of the pump laser photon flux. The slope of the straight line which best fit the data is the excitation cross section for the Si nanocrystals emitting at 850 nm.

for several samples at different detection wavelengths and at room temperature. As an example in Fig. 2.7, the room temperature PL risetime of Si nanocrystals emitting at 850 nm, within a SiOx sample annealed at 1250◦ C for 1 hour, is reported. The excitation wavelength was 488 nm, with pump powers ranging from 0.4 mW to 60 mW. As predicted by eqs. (2.4) and (2.5) the risetime becomes shorter and shorter with increasing pump power. By fitting these risetime curves through eq. (2.4), the values of τon are obtained for each pump power. The reciprocal of τon is reported in Fig. 2.8 as a function of the photon flux. The data follow a straight line according to eq. (2.5), with a slope σ = 5.8×10−16 cm2 . The intercept of the fit straight line with the vertical axis gives the lifetime τ of the Si nanocrystals in the system at the detection wavelength. The obtained value (τ = 73 µs) is consistent with decay time measurements performed at 850 nm on the same sample. In this way we have been able to obtain a direct measurement of the absorption cross section of Si nanocrystals emitting at a certain wavelength and excited at a fixed wavelength. We have also performed such measurements by varying the detection wavelength, and the noteworthy feature we have discovered is that the cross section seems to be independent of the detection wavelength (at least in the explored range of 700-900 nm) [60]. This could probably be the result of the energy transfer mechanism itself. Indeed the system of interacting Si nanocrystals act as a whole as far as the excitation process is concerned.

2.6 Photoluminescence excitation spectroscopy

23

A different behavior is observed in systems in which Si nanocrystals are almost isolated (as in the nanocrystalline multilayer and in the sample with 35 at.% Si). By using the very same method, a lower value of about 2×10−16 cm2 has been obtained for the excitation cross section of both samples. The increase in the cross section observed in samples characterized by higher nanocrystals concentration is probably due to the fact that, in presence of energy transfer, since the excitation of each nanocrystal can occur not only through direct photon absorption but also through energy transfer from a nearby nanocrystal, the effective excitation cross section is now the sum of the cross section for the absorption of a photon plus the cross sections related to the energy transfer mechanism itself, i.e. it is higher than the mere photon absorption cross section. In conclusion we may assert that the increase in cross section and decrease in time decay as a function of the nanocrystals concentration are different aspects of the same physical process, related to the energy transfer mechanism among neighboring luminescent nanoclusters.

2.6 2.6.1

PL excitation spectroscopy Experiment

The photoluminescence excitation (PLE) spectroscopy is a very powerful technique able to determine the absorption properties of optically active centers dispersed within a thin film sample. In Fig. 2.9 a sketch of the photoluminescence excitation experiment is reported. A flux φ of photons having a wavelength λexc impinges on a thin film sample containing a density N of optically active centers, which after being excited by the incoming photons, decay with a characteristic lifetime τ , emitting photons with wavelengths generally larger than the exciting one. Then through a system of lenses, a part of the emitted light is collected and its intensity IP L and wavelength dispersion are measured. The excitation rate of the centers, i.e. the number of times each center is excited per second, is simply given by the product of the flux φ times the absorption cross section σ at the excitation wavelength. If the flux of incoming photons is very feeble, i.e. the excitation rate σφ is much smaller than the de-excitation rate 1/τ , the photoluminescence intensity can be approximated by the following formula: IP L ∝ σφ

τ τrad

N

(2.6)

being τrad the radiative lifetime of the emitting centers. If we now assume that both the lifetime and radiative lifetime are independent of the excitation wavelength, which is indeed quite reasonable since they depend mostly on the de-excitation mechanisms related to that particular center, and we devote our attention to a particular emission wavelength λrev , so that the number of emitting centers N is fixed too, from eq. (2.6) we get:

24


Figure 2.9. - Scheme of the photoluminescence excitation spectroscopy experiment.

σ (λexc ) ∝

IP L (λexc ) φ (λexc )

(2.7)

where the dependence on the excitation wavelength has been explicited. Equation (2.7) is quite interesting, since it tells us that by measuring the luminescence intensity at a particular emission wavelength λrev as a function of the excitation wavelength λexc , and dividing it for the flux of the incoming photons, it is possible to give a quantitative estimate of the absorption cross section for the optically active centers emitting at λrev , and for different excitation wavelengths, i.e. it is possible to determine the spectral shape of the absorption cross section. It has to be noticed that the PLE technique can be more powerful than absorbance measurements. Indeed, in order to measure the absorbance of a film, we should perform reflectance R and transmittance T measurements. Indeed, it can be shown, applying energy conservation, that the expression for absorbance A takes the form: A = 1 − R − T . The sample is usually deposited on top of a transparent substrate, to avoid contributes from the substrate in the transmitted beam. However, the absorbance value we obtain in this way refers to the overall film/substrate structure, and could be affected by absorption features of the substrate, since no ideal transparent substrate exists. Moreover, through absorbance measurements we are looking indistinctly at all the absorbing centers, whether they are able or not to emit light. On the contrary, PLE measurements are much less affected by the substrate and they are more selective, since, by looking only at photoluminescence intensity as a function of excitation wavelength, they reveal exclusively the absorption properties of the optically acive centers dispersed in the film.


exc: 290 nm 450 nm 600 nm 750 nm

8

PL Intensity (arb. units)

25

10

107 6

10

105 104 103 600

800

1000

1200

Wavelength (nm) Figure 2.10. - Room temperature PL spectra for a sample containing Si nanocrystals, for different excitation wavelengths λexc . The sample has been obtained by annealing a 0.2 µm thick SiOx film (42 at.% Si) deposited on a transparent quartz substrate, in order to reduce the interferences inside the film. The radiation source is a 150W Xenon lamp (ORIEL) coupled with a CORNERSTONE monocromator. The spectra have been measured at a constant flux of ∼ 1×1016 cm−2 s−1 .

2.6.2

PLE measurements

PLE measurements have been performed by using a 150W ORIEL Xenon lamp coupled with a CORNERSTONE monochromator. The excitation wavelengths used in the present work span a range between 290 and 800 nm. The full width at half maximum of the radiation falling onto the sample is ∼ 4 nm, so that the resolution of our PLE spectra is of about 2 nm. The average photon flux depends on the particular excitation wavelength, but it doesn’t exceed the value of 5×1016 cm−2 s−1 . The coherence length of the radiation source has been estimated to be of the order of 1 µm. In Fig. 2.10 the room temperature photoluminescence spectra, for a sample containing Si nanocrystals, are reported for different excitation wavelengths. The sample has been obtained by annealing a 0.2 µm thick SiOx film (42 at.% Si) deposited onto a transparent quartz substrate, in order to reduce the interference inside the film. It is worth noticing how, with increasing the excitation wavelength, the PL intensity decreases following a logarithmic trend.

26



1.2 exc:

1.0 0.8

290 nm 450 nm 600 nm 750 nm

0.6 0.4 0.2 0.0 600

800

1000

1200

Wavelength (nm) Figure 2.11. - PL spectra as taken from Fig. 2.10, normalized for the maximum intensity value.

In order to better study the spectral shape, we plot in Fig. 2.11 the normalized PL spectra taken from Fig. 2.10. The full width at half maximum (FWHM) slightly increases with increasing the excitation wavelength, from a value of 170 nm to 194 nm, but all of the spectra are peaked at around 900 nm. Still, a slight red-shift of ∼ 15 nm can be noticed when exciting the sample at 750 nm. This could be due to the fact that at that excitation wavelength, some of the smaller nanocrystals, i.e. those characterized by band-gap energies higher than the photon, are indeed transparent to the incident photons, and, since they can’t be excited, their contribution to the PL spectrum is therefore removed. Another interesting feature is the small peak appearing at 825 nm in the spectrum obtained by exciting the sample at 750 nm. The energy difference between the exciting photon and the spectral feature at 825 nm is about 150 meV, and suggests that the origin of the peak could be due to a radiative exciton de-excitation involving the emission of phonons, which at the moment are not yet well identified. In Fig. 2.12, the photoluminescence intensities measured at two different emission wavelengths (850 nm and 980 nm) are reported as a function of the excitation wavelength, normalized to the incoming photon flux values. The photon flux at 488 nm is φ = 4×1016 cm−2 s−1 . The excitation cross section at both the detection wavelengths has been measured for the very same sample by using an excitation wavelength of 488 nm from an Ar laser. The experimental value is σ 5.8×10−16


27

cm2 at 850 nm, and is almost independent of detection wavelength. We have also measured the lifetime at 850 nm, and it has a value τ = 70 µs, independent of the excitation wavelength. From these numbers, we get σφ = 2.32×101 s−1 and 1/τ = 1.43 ×104 s−1 , which means σφ 1/τ , i.e. the excitation rate of a nanocrystal emitting at 850 nm is much smaller than its de-excitation rate, meaning that for those photon fluxes we are in a linear regime of excitation. In this regime, through eq. (2.7), we know the excitation cross section to be proportional to the PL intensity divided by the photon flux. The proportionality factor can be fixed at once since we have independently measured the excitation cross section at a certain excitation wavelength (488 nm), through risetime measurements. This permits us to easily convert PLE data in excitation cross sections all over the range of excitation wavelengths used in the experiment. The results of such an exercise are reported in the right-hand scale of Fig. 2.12, where indeed the PLE data have been rescaled to obtain the excitation cross section σ (λexc ). The same operation has been performed the 980 nm emission wavelength, as reported in Fig. 2.12. It is worth noticing that σ varies between 10−17 and 10−14 cm2 . Moreover, σ has the same trend for two very different emission wavelengths (850 and 980 nm). Since different emission wavelengths correspond to different classes of nanocrystals characterized by different sizes, this result is somewhat surprising. This can be explained if we assume that the system of interacting Si nanocrystals acts as a whole as far as the photon emission process is concerned. Indeed, very recent theoretical calculation have shown that with increasing the Si nanocrystal concentration, i.e. decreasing the mean distance among them, a new kind of delocalized electronic state seems to arise just under the conduction band of the isolated nanocrystal [77]. This state, being delocalized, can be shared by many nanocrystals, which therefore become a network of interacting centers. It is interesting to compare in Fig. 2.13 the excitation cross section so far obtained for Si nanocrystals by PLE measruments with the value obtained through absorbance measurements performed on the very same samples, and with the experimental values of the absorption coefficient obtained for bulk Si, by using numerical tables from [78]. The first thing to be noticed is that the cross section values determined by PLE (open circles) and by absorbance measurements (open triangles) are equal up to an excitation wavelength of 540 nm. For higher wavelengths, their trends start to differ strongly. In particular, the cross section determined from absorbance measurements shows a saturation which is mainly due to the fact that the quartz substrate over which the Si nanocrystals containing film is deposited is starting to absorb by itself, thus increasing the total (i.e. film/substrate) absorbance values in that region. It is worth noticing that this behavior is not observed at all in PLE measurements, since we are now looking only at the absorption properties of the optically active centers, i.e. of Si nanocrystals. This demonstrates the superiority of PLE spectroscopy with respect to absorption measurements in determining the cross section of Si

28


rev = 850 nm rev = 980 nm

10

-14

10

10-10

10

10-11

10

-15

(cm2)

IPLE/ (arb. units)

-9

-16

-12

10

10-17 300

400

500

600

700

800

Excitation Wavelength (nm) Figure 2.12. - PLE intensity at 850 nm and excitation cross section as a function of the excitation wavelength. In the inset, the PLE intensities for two different detection wavelengths (850 and 980 nm, respectively) are reported, for the very same sample. It is worth noticing that indeed the excitation cross section is comparable for both the detection wavelengths, all over the excitation wavelength range spanned in the experiment.

nanocrystals over a wide wavelength range. In Fig. 2.13 it is also possible to make an important comparison between the Si nanocrystals cross section measured at 850 nm by PLE and the absorption coefficient of bulk crystalline silicon, as extracted from [78]. A wide range between 400-650 nm exists in which the two curves have a similar behavior, suggesting that indeed the absorption properties of Si nanocrystals in the film and of bulk Si have the same physical origin. We may be tempted to attribute this similarity to the Si atoms forming the nanocrystals and bulk silicon. In order to verify this hypothesis, we define an atomic cross section σat for bulk Si as the ratio between the absorption coefficient α and the atomic density ρ of silicon: σat =

α ρ

(2.8)

The atomic density of bulk Si is a well known quantity, i.e. ρ 5 × 1022 at./cm3 . Just to fix the ideas, we can perform all of the calculations at 488 nm. At this wavelength, for bulk silicon it is α 2×104 cm−1 , hence from eq. (2.8) we get σat 4× 10−19 cm2 , which indeed is the absorption cross section for a single atom in the crystal. It is therefore straightforward to calculate the number of atoms Nat


29

10-13 6

PLE - Si nc Absorbance - Si nc bulk Si

5

10 -15

10

4

10

(cm-1)

(cm2)

10-14

10

10-16 103 10-17 300

400

500

600

700

800

Excitation Wavelength (nm) Figure 2.13. - Excitation cross section for Si nanocrystals (left-hand scale) as determined by PLE measurements (open circles) or by absorbance measurements (open triangles), and absorption coefficient for bulk Si (solid circles, right-hand scale) as a function of the excitation wavelength.

in an absorbing nanocrystals, by simply dividing the experimental cross section σ for a nanocrystal by the atomic cross section σat : Nat =

σ σat

(2.9)

substituting the values we get Nat = 1.45×103 atoms per nanocrystal. Dividing this number by the atomic density of Si, we get an estimate of the volume of the absorbing nanocrystal, i.e. Vnc = 2.9×10−20 cm3 . Assuming a sperical shape for the nanocrystal volume, we can calculate the mean nanocrystal radius, i.e. rnc = 1.9×10−7 cm, or 1.9 nm. The agreement between this value estimated through optical measurements and the value measured by TEM analyses is not casual, and definitely demonstrates that the absorption properties of a Si nanocrystal are mainly determined by the Si atoms contained in its volume and constituting it. However, for excitation wavelengths smaller than 400 nm, the curve relative to the excitation cross section of Si nanocrystals starts departing from the absorption of bulk Si, as reported in Fig. 2.13. This is somewhat to be expected, since it is well known that quantum confinement in Si nanostructures produces a strong modification both on the energy gap and in the density of states. This modification becomes more evident, and observable, at the shortest excitation wavelengths, since they are indeed able to probe the higher lying electronic states of the material.

30

2.7


Conclusions

We have shown that thermal annealing of PECVD prepared SiOx films or Si/SiO2 multilayers is a quite powerful method to produce Si nanocrystals embedded within SiO2 with a large variety of nanocrystals sizes and densities. Si nanocrystals have been demonstrated to interact among themselves with a net energy transfer occurring from smaller nanocrystals (larger gaps) to larger ones (smaller gaps). This interaction is evidenced by a stretched exponential decay-time of the PL intensity which tends towards a single exponential in almost isolated nanocrystals. This energy transfer will play an important role when discussing the PL quenching mechanisms occurring in ion-irradiated Si nanocrystals (Chapter 3). Moreover a similar energy transfer mechanism will be demonstrated to set in once Er ions are put in the vicinity of a Si nanocrystal, causing an efficient sensitizing action for the Er luminescence at 1.54µm (Chapter 4). The absorption cross section of PECVD prepared Si nanocrystals has been determined through PL risetime measurements. A value of 2×10−16 cm2 has been found at an excitation wavelength of 488 nm for isolated nanocrystals, while it increases to 5.8×10−16 cm2 in interacting nanocrystals. This has been attributed to the fact that the interaction among neighboring Si nanocrystals creates additional excitation paths for the single nanocrystal other than the mere photon absorption, thus increasing its effective excitation probability. However, this does not result in any efficiency enhancement, due to the simultaneous reduction of lifetime produced by the non-radiative energy transfer to neighbors nanocrystals. Through PLE measurements, the excitation cross section of Si nanocrystals has been determined in a wide range of excitation wavelengths (290-800 nm), and it has been shown to be independent of the detection energy, probably beacause the system of Si nanocrystals is acting as a whole as far as the absorption and emission properties are concerned, due to the strong interaction existing among neighboring nanocrystals. In conclusion it has been shown that a strong similarity exists between the absorption cross section of samples containing Si nanocrystals and the absorption coefficient of bulk Si, extracted from optical data, at least in the range 400-650 nm. Assuming that absorption is mainly caused by the Si atoms forming the nanocrystal, we estimated that about 103 Si atoms have to form an absorbing nanocrystal. This lead to nanocrystal sizes of about 1.9 nm, in agreement with the value extracted by TEM analyses. Therefore, in that wavelength region, the absorption properties of Si nanocrystals are similar to bulk Si, while they start to differ for smaller excitation wavelengths, probably because quantum confinement effects are strongly affecting the electronic structure of Si nanocrystals.

Chapter 3 Ion-irradiated Si nanocrystals

In this chapter the formation and annihilation of defects produced in Si nanocrystals by ion beam irradiation are investigated in details. The luminescence properties of Si nanocrystals embedded within SiO2 are used as a probe of the damaging effects generated by high energy ion beam irradiation. Samples have been irradiated with 2 MeV He, Si, Ge and Au ions at different fluences, in the range between 109 /cm2 and 1016 /cm2 . The nanocrystals related photoluminescence and decay-time strongly decrease for fluences higher than a critical value, which depends on the ion mass. The total number of emitting, i.e. optically active, centers has to diminish too. By assuming that a Si nanocrystal is damaged, i.e. become optically inactive, when it contains at least one defect inside its volume, a model can be developed, which relates the fraction of quenched nanocrystals to the total defect concentration in the film and to the nanocrystal volume. Moreover, the recovery of the damaged Si nanocrystals is studied as a function of both isochronal and isothermal annealings. It is demonstrated that in slightly damaged Si nanocrystals, a large variety of defects characterized by activation energies between 1 and 3 eV exists. On the contrary, the recovery of the PL properties of completely amorphized Si nanocrystals is characterized by a single activation energy, whose value is 3.4 eV. Actually, this energy has been associated to the phase transition between the amorphous and the crystalline structure of each Si grain.

3.1

Introduction

The damage and recrystallization of crystalline bulk silicon is a well known and widely studied subject [79–85]. Only recently the recrystallization properties of amorphous Si layers with thicknesses larger than 2 nm have been studied in greater details, and the most important result is that the crystallization temperature increases with decreasing layer thickness [86–88]. In contrast, a part from a few

32

CHAPTER 3. Ion-irradiated Si nanocrystals

papers [89, 90], the role of ion beam induced damage in the structural and optical properties of Si nanocrystals embedded in SiO2 is not yet clear, and a comprehensive picture of the overall damaging mechanisms in Si nanocrystals is still lacking. This problem, apart from being interesting from a mere physical point of view, is of great importance for the understanding of the damage produced in Si nanocrystals upon introduction of Er by ion implantation. This step is crucial for the realization of the Er-doped Si nanoclusters system, which is the subject of the following chapters. In this chapter, the luminescence properties of Si nanocrystals produced by PECVD are used as a probe of the defect generation under irradiation through energetic ion beams [91]. By varying both the fluence and the mass of the incident ions, the main physical parameters responsible for the quenching of the photoluminescence are determined. Moreover the defect annihilation is studied by observing the luminescence recovery after both isochronal and isothermal annealings. It is shown that a large variety of defects, having different activation energies, exists in partially damaged Si nanocrystals, while the recrystallization of fully amorphized Si nanoclusters is a thermally activated process, characterized by a single activation energy of 3.4 eV. A phenomenological model able to explain the overall experimental picture is presented.

3.2

Sample preparation

Samples were prepared by PECVD, using high purity SiH4 and N2 O as precursors. Two different kinds of samples containing Si nanocrystals were produced. The first one is a 0.1 µm thick substoichiometric SiOx (x < 2) thin film with 39 at.% Si. After deposition, the SiOx films were annealed for 1 hour at 1250◦ C in ultra-pure nitrogen atmosphere. The high temperature annealing induces the phase separation between Si and SiO2 and a high density of small Si clusters is obtained [60][5], with a mean radius of about 1.5 nm (± 0.4 nm). The second kind of sample is a Si/SiO2 superlattice (SL), consisting of 11 SiO2 layers (6.5 nm thick) alternated with 10 ultra thin Si layers (0.9 nm thick). After deposition, the SL were annealed for 1 hour at 1100◦ C in ultra-pure nitrogen atmosphere to induce the breaking and balling up of the ultra thin Si films, with the formation of Si nanocrystals totally embedded within SiO2 and almost completely isolated, as demonstrated in the previous chapter and in [60, 72].

3.3

The experiment

In order to study the defect formation in Si nanocrystals, the samples were irradiated at room temperature with 2 MeV He+ , Si+ , Ge+ or Au+ ions at different fluences in the range between 1×109 /cm2 and 1×1016 /cm2 . A schematic of the experiment is reported in Fig. 3.1. In particular the film containing the Si nanocrystals and having thickness t is represented, together with the bulk Si substrate. The beam

3.4 Defect production in Si nanocrystals

33

Rp

Ions

t Figure 3.1. - Scheme of the ion beam irradiation experiment. Rp is the projected range for the ions sent onto the sample, and t is the thickness of the film containing Si nanocrystals. Note that Rp t, in order to avoid doping the nanocrystals.

energy was chosen in such a way to locate the projected ion range Rp far beyond the film containing Si nanocrystals, in order to avoid doping the sample. Indeed, the ion beam is simply releasing part of its energy in electronic and nuclear collisions with the atoms of the matrix. For all the ions used in the experiment, the calculated projected range varies between 0.47 µm for Au and 6.84 µm for He. By changing the ion mass, the nuclear energy loss was varied over four orders of magA for He and 2.6×102 eV/˚ A for Au. nitude, in the range between 2.5×10−2 eV/˚ Through a detailed damage calculation by using the Transport of Ions in Matter (TRIM) Monte Carlo Simulations code [92,93], we calculated the mean area of each collision cascade. Moreover, the average number of displaced atoms per incident ion and per unit trajectory length has been evaluated within each cascade, and it A for He and 2.4 at/ion ˚ A for Au. varies in the range between 2.8×10−4 at/ion ˚ Damage recovery of the as-implanted samples was performed by isochronal thermal annealings in the temperature range between 100◦ C and 1150◦ C in vacuum or in N2 atmosphere. Moreover, in order to understand the kinetics of the damage recovery, we also performed isothermal annealings at different temperatures and for times in the range between 5 s and 20 hours. The structural and optical properties of all the samples have been studied by transmission electron microscopy (TEM) and photoluminescence (PL), respectively.

3.4 3.4.1

Defect production in Si nanocrystals Damage accumulation in Si nanocrystals

Before ion irradiation, all the produced samples show a strong photoluminescence even at room temperature. As an example, in Fig. 3.2 the PL spectrum of a sample

34



2 MeV Si 1.0 0.8

Reference 10

-2

11

-2

11

-2

12

-2

1x10 cm 1x10 cm 5x10 cm 1x10 cm

0.6

12

-2

3.5x10 cm 12

-2

5x10 cm

0.4

12

-2

7.5x10 cm 13

-2

13

-2

1x10 cm

0.2 0.0 600

2x10 cm

650

700

750

800

850

900

Wavelength (nm) Figure 3.2. - PL spectra of a SiOx film with 39 at.% Si annealed at 1250◦ C for 1 hour (Reference sample) and of the very same sample irradiated with a 2 MeV Si+ ion beam at different fluences in the range between 1×1010 and 2×1013 /cm2 . The ions are not implanted in the film, since their projected range is 2.08 µm, i.e. deep in the bulk Si substrate. Spectra were measured at room temperature, with the 488 nm line of an Ar-Kr laser as the excitation source and at a pump power of 10 mW.

containing Si nanocrystals (reference sample) obtained by annealing at 1250◦ C a SiOx film with 39 at.% of Si content is reported, together with the PL spectra of the same sample irradiated with 2 MeV Si+ ions at different fluences, in the range 1×1010 - 2×1013 /cm2 . The projected range is 2.08 µm. The PL spectrum of the reference sample is peaked at 820 nm and has a full width at half maximum (FWHM) of about 140 nm. With increasing the ion fluence, the PL intensity remains unaffected up to a value of 1×1010 /cm2 . For higher fluences, the PL signal decreases until it completely disappears for fluences higher than ∼1×1013 /cm2 . Neither shifts nor changes in the peak shape have been observed, for all the ion fluences. Interesting information can be extracted by studying the decay time of the luminescence signal recorded at 800 nm after shutting off the pumping laser beam through the acousto-optic modulator, as reported in Fig. 3.3 for different ion fluences. The luminescence signal decay is characterized by a stretched exponential shape. The lifetime values, obtained by fitting the experimental curves reported in Fig. 3.3 through eq. 2.1 (page 19), remain constant up to a fluence of 1×1011 /cm2 ,


35


Reference

rev= 800 nm

0

10

11

-2

11

-2

12

-2

1 x 10 cm 5 x 10 cm 1 x 10 cm 12

-2

3.5 x 10 cm 12

-2

5 x 10 cm 12

-2

7.5 x 10 cm

-1

10

0

50

100

150

200

Time (s) Figure 3.3. - Decay-time measurements of the photoluminescence signal measured at 800 nm for a SiOx sample with 39 at.% Si, annealed at 1250◦ C for 1 h (Reference sample) and subsequently irradiated with 2 MeV Si+ at different fluences. Data were taken at room temperature and at a laser pump power of 10 mW. The measured lifetime for the reference sample is 54 µs.

while they start to decrease for higher ion fluences, suggesting that new non radiative paths related to the presence of defects left over by the ion beam are influencing the decay dynamics of the emitting centers. On the other hand, β is characterized by a value of 0.75, which remains constant for all the ion fluences. All the spectra and the decay time curves have been obtained with 10 mW of pumping power, resulting in a photon flux φ of about 8.7×1018 cm−2 s−1 . The Si nanocrystals excitation cross section σ is almost independent of the detection wavelength, i.e. of the Si nanocrystals size, and has a value of ∼ 10−16 cm2 , as shown in Chapter 2. Thus, for the excitation rate of each Si nanocrystal we obtain a value σφ ∼ 8.7×102 s−1 , while the de-excitation rate for the reference sample is 1/τ ∼ 1.85×104 s−1 (being τ the luminescence lifetime of the reference sample, measured at 800 nm). Being σφ 1/τ , we are in the low pump power regime. In this regime, which is valid for all the samples used, the PL intensity of the emitting centers can be approximated by eq. (2.6). This equation is extremely important for the understanding of the damaging mechanisms. In fact it tells us that if the excitation conditions remain unchanged (i.e. φ is constant), assuming that the excitation cross section and the radiative lifetime remain constant, which is indeed plausible, a variation of

36


the luminescence yield can be caused only by a change in the lifetime τ or in the total number of emitting centers N , or in both of them. In Fig. 3.2 it is possible to observe that at a fluence of 1×1012 /cm2 the residual luminescence is 60% of the reference one, while for the same fluence the lifetime, extracted from Fig. 3.3, is still 80% of its original value. This means that in order to quantitatively explain the decrease of PL intensity, the total number of emitting centers has to diminish too. In particular it has to be about 75% of the total number of emitting centers present in the unirradiated sample. Thus we have an experimental evidence of the fact that even at very small fluences (much smaller than the critical fluence needed for the amorphization of bulk crystalline Si) we can damage Si nanocrystals in such a way that a fraction of them becomes dark from a luminescence point of view. These nanocrystals, however, though damaged, are not necessarily amorphized. In fact, from transmission electron microscopy (TEM) analyses (not shown) it appears that even at a Si fluence of 5×1012 /cm2 , a huge number of Si nanocrystals with the same size distribution as the reference sample is still visible in dark field configuration, while the PL signal reported in Fig. 3.2 is almost completely quenched, demonstrating that a Si nanocrystals can become dark without necessarily being amorphous. We repeated the irradiation experiment on the very same sample by using several ions, at the same energy of 2 MeV. The results are shown in Fig. 3.4, where both the intensity (a) and the lifetime (b) of the PL signal at 800 nm coming from the irradiated sample are reported for different incident ions and versus the ion fluence, normalized to the respective reference value. As can be seen, even if the range of fluences varies over several orders of magnitude, due to the large spreading in the nuclear energy loss of the different ions, the general trend remains unchanged, with both the intensity and the lifetime strongly decreasing for fluences greater than a critical value, which depends on the particular ion used. In particular, the critical dose needed to quench the PL intensity and the lifetime strongly decreases as the ion mass is increased. It is worth noticing again that for all the ions the luminescence intensity decreases more than the lifetime does with increasing the ion fluence, confirming the fact that the total number of emitting Si nanocrystals is decreasing too. The picture that arises up to now is therefore the following: i) with increasing the ion fluence the total amount of damage left over by the ion beam in the matrix increases, the amount of damage being higher at a fixed fluence for heavier ions; ii) the effect of damage is twofold, determining both the rising of new non radiative paths, as attested by the decrease of the luminescence lifetime, and the reduction of the total fraction of emitting Si nanocrystals , both effects leading to a quench of the luminescence intensity in the irradiated samples. Thus for a fixed ion fluence, we can think at the sample as composed by two different classes of Si nanocrystals , i.e. the first made of Si nanocrystals which, being damaged by the ion beam, become optically inactive and cannot emit anymore, the second class formed by Si nanocrystals which can still emit but whose lifetime is affected by new non radiative processes, which become active for fluences higher than the critical value.


37

rev = 800 nm

I / I Ref

1.0 0.8 0.6 0.4 0.2

(a)

0.0 1.0

/Ref

0.8 0.6 Ref = 54 s

0.4

2 MeV He 2 MeV Si 2 MeV Ge 2 MeV Au

0.2

(b)

0.0 9

10

10

10

11

10

12

13

14

10 10 10 Fluence (cm-2)

15

10

16

10

Figure 3.4. - Normalized PL intensities (a) and lifetimes (b) at 800 nm for a SiOx film with 39 at.% Si annealed at 1250◦ C for 1 hour and irradiated with 2 MeV He+ , Si+ , Ge+ , Au+ at different fluences. The lines are guides to the eye.

In order to investigate the nature of these non radiative phenomena, we produced a sample with largely spaced Si nanocrystals by annealing a Si/SiO2 superlattice at 1100◦ C for 1 hour, as described previously within this thesis (see also [60, 72]). The Si nanocrystals are located in parallel and far apart planes instead of being randomly and uniformly distributed all over the oxide host, and particular care was taken in order to obtain similar Si nanocrystals size distributions. The sample was then irradiated with a 2 MeV Si ion beam, under the same experimental conditions. The PL spectra and the luminescence lifetime at 800 nm are respectively reported in Fig. 3.5 and Fig. 3.6. From Fig. 3.5 it can be noticed that, in spite of the similar size distribution, the PL spectrum of the reference sample is blue-shifted with respect to the randomly distributed Si nanocrystals sample, being peaked now at around

38


2 MeV Si in SL Reference


1.0

1x1010 cm-2 11

-2

1x10 cm

1x1012 cm-2

0.8

3.5x1012 cm-2 13

-2

1x10 cm

0.6 0.4 0.2 0.0 600

650

700

750

800

850

900

Wavelength (nm) Figure 3.5. - PL spectra of a Si/SiO2 superlattice (with a Si layer thickness of ∼0.9 nm) annealed at 1100◦ C for 1 hour (Reference sample) and after being irradiated with a 2 MeV Si+ ion beam at different fluences in the range between 1×1010 and 1×1013 /cm2 . Spectra were measured at room temperature and at a laser pump power of 10 mW.

730 nm. Moreover the lifetime of the same class of Si nanocrystals, i.e. those emitting at 800 nm, is different in the two reference samples, being longer and less stretched in the sample containing far apart Si nanocrystals (Fig. 3.6). Both of these experimental evidences are a proof of a strongly suppressed interaction among Si nanocrystals, which now act as almost isolated. Indeed, in this sample the smallest nanocrystals have no bigger nanocrystals nearby to whom transfer their energy. Hence they can emit radiatively their energy and contribute to the PL spectrum, thus producing a blue shift with respect to the system of interacting nanocrystals, where instead the larger nanocrystals act as efficient sinks of energy, capturing the energy from smaller nanocrystals and quenching their luminescence. It is interesting to note that in this sample, though the PL intensity decreases with increasing fluence (Fig. 3.5) the time decay remains constant, at least up to 1×1012 /cm2 . This demonstrates that, while the total number of luminescent centers is decreasing, no introduction of new non-radiative de-excitation paths is observed this time. The Si nanocrystals which can still emit seem to be unable to interact with the defects which are present in the oxide host after ion irradiation. The same must be true in the annealed SiOx sample. So how does we explain the strong lifetime



39

Reference

0

10

1x1010 cm-2 1x1012 cm-2

rev = 800 nm

10-1 0

50

100

150

200

Time (s) Figure 3.6. - Decay-time measurements of the PL signal at 800 nm for a Si/SiO2 superlattice (with a Si layer thickness of ∼0.9 nm) annealed at 1100◦ C for 1 hour (Reference sample) and subsequently irradiated with 2 MeV Si+ at different fluences. Data were taken at room temperature and at a laser pump power of 10 mW. The measured lifetime for the reference sample is 63 µs.

quenching in that sample? The Si nanocrystals concentration is higher, so that the mean distance between Si nanocrystals decreases, and a strong interaction among Si nanocrystals sets in. After ion irradiation, a non damaged Si nanocrystal might be surrounded by a nearby damaged nanocrystal with whom it can interact. The damaged nanocrystal will therefore behave as a quenching center for the luminescing one, thus generating a stronger non radiative path for it. With increasing the ion fluence, the mean number of damaged nanocrystals surrounding a luminescent center increases, and so does the probability of a non radiative decay from the same nanocrystal, thus producing the lifetime quenching reported in Fig. 3.3.

3.4.2

Phenomenological modeling

In the following we are going to develop a model able to explain the experimental data reported so far, by correlating the luminescence measurements with complete simulations of the damage produced in the film by the energetic ion beams we used. First of all we focus our attention to the way a Si nanocrystal is damaged by an ion beam. It is well known that an energetic ion passing through a solid substrate can in-

40


Au in SiOx - 2 MeV 100

50

50

0

0

Z Axis (nm)

100

-50

-50

Z Y

-100 0

X

20

40

60

X Axis (nm)

80

100

-50

0

50

-100 100

Y Axis (nm)

Figure 3.7. - Example of collision cascades generated by 2 MeV Au ions onto a SiOx film.

teract with a host atom by loosing energy in electronic or nuclear collisions, being eventually stopped in the material. In the electronic stopping regime, the fast ion looses energy interacting with electrons in the solid, i.e. producing electronic excitations or ionizations of target atoms, while in nuclear stopping regime the ion looses much of its energy in an elastic collision with a host atom, eventually displacing it from its original lattice position if the lost energy is higher than a critical value, characteristic of the matrix and known as the displacement energy. If the displaced atom has enough energy, it can become itself a source of other displacements, thus producing a collision cascade, which is the spatial envelope of all the atoms displaced by both the primary and secondary ions. In our experiment we can assume that the energy of the incident ion is practically constant with depth, being the film thickness small (∼0.1 µm) with respect to the projected range Rp . This also means that within the film, each cascade can be approximated by a cylinder, whose height is the film thickness and whose area Ac can be defined as the area of a section perpendicular to the ion trajectory within which it is most probable to find a displaced atom. In order to determine the important parameters for the different collision cascades, we performed TRIM’90 Monte Carlo simulations [92, 93] for the different ions, at the irradiation energy of 2 MeV. As an example, in Fig. 3.7 the simulated collision cascades for 2 MeV Au on SiOx film are reported. The solid circles represent the displaced atoms produced both by primary and secondary ions. In particular, the ion beam travels along the X axis, starting from Z=0. The left panel represents the displaced atoms in the X-Z plane, while on the right panel the cross section distribution of displaced atoms is reported in the Y-Z plane, which is perpendicular to the propagation axis of the ion beam.


41

Cumulative Counts

Displaced Atoms 90 80 70 60 50 40 30 20 10 5 2 1

Z

Ac

rc 66.7 %

NAu-ions= 100

5000

Y

rc

Counts

4000

rc = 28.0 ± 0.1 nm

3000 2000

rc

1000 0

0

10

20

30

40

50

Radial distance (nm) Figure 3.8. - Bottom panel: number of displaced atoms (in counts) as a function of the radial distance from the ion beam propagation axis, as determined through a Monte Carlo simulation using 100 Au ions. Top panel: cumulative counts, i.e. the integral of the bottom panel curve, as a function of radial distance. The radius of the collision cascade, comprising both primary and secondary ions, is defined as the distance within which the 66.7% of the total displaced atoms is generated.

The simulation give us the coordinates of all of the displaced atoms, therefore we can calculate the distribution of defects all along the beam axis, at a given distance 1/2 r (x) = (y 2 + z 2 ) from the beam axis, for each x. Since the energy of the incident ion is only slightly affected in traversing the film, we can assume that r (x) is almost constant with depth x. In particular, in Fig. 3.8 we reported both the number (bottom panel) and the cumulative number, i.e the integral from zero to the given radius (top panel), of

42


displaced atoms as a function of their radial distance r from the beam axis, obtained by simulating the full collision cascades generated by 100 Au ions. We can now define a mean collision cascade area Ac as the circular surface centered at the ion beam axis and containing the 66.7% of all of the displaced atoms. Indeed, for Au we get the 66.7% of the cumulative counts at a radial distance rc = 28 nm, as reported in the top panel of Fig. 3.8. Moreover, from the same simulations we can obtain the mean number of displaced atoms Dc per incident ion within each cascade and for each ion mass, and the area Ac of each collision cascade. Once these parameters are known, it is easy to calculate the mean defect concentration nc inside each collision cascade, defined as the ratio between the total number of displaced atoms Dc inside the cascade and the volume Vc = Ac t of the cascade, being t the film thickness: Dc Dc d = (3.1) = Vc Ac t Ac where d ≡ Dc /t is defined as the number of displaced atoms per incident ion per unit trajectory length within each collision cascade. Thus, from eq. (3.1) the defect concentration in each collision cascade is just the number of displaced atoms per incident ion per unit length (d), divided by the mean cascade area (Ac ). In Table 3.1 the most important parameters, extracted from the simulations for all of the ions used in the experiment are reported. Now we want to focus our attention on the defect formation within the Si nanocrystals dispersed in the SiO2 matrix. We will assume that a single defect inside the volume of a nanocrystal is enough to quench its luminescence. This is plausible since, for instance, it has been demonstrated for porous Si that even a single Si dangling bond at the surface of a nanocrystallite acts as a very efficient non radiative center, causing the complete quenching of the nanocrystal luminescence [59]. The fraction of quenched nanocrystals is therefore equal to the fraction of damaged nanocrystals, which is equal to the probability of a nanocrystal being damaged. In other words, for a fixed ion fluence φ, a certain concentration of defects is left over the sample. Thus we need to calculate the probability of finding at least nc =

Parameters Dc (displ./ion) d (displ./ion˚ A) (dE/dx)e (eV/˚ A) (dE/dx)n (eV/˚ A) Rp (µm) rc (nm) Ac (cm2 ) φc ≡ 1/Ac (cm−2 ) nc ≡ Dc /Ac t (cm−3 )

4

He 0.284 2.84×10−4 27.13 2.5×10−2 6.84 1.0 3.0×10−14 3.4×1013 9.6×1017

28

Si 59.6 5.96×10−2 168.8 4.99 2.08 9.4 2.8×10−12 3.6×1011 2.1×1018

74

Ge 459 4.59×10−1 129.2 4.19×101 1.19 19.4 1.2×10−11 8.5×1010 3.9×1018

197

Au 2430 2.43 171.1 2.59×102 0.474 28.0 2.5×10−11 4.0×1010 9.8×1018

Table 3.1. - Parameters extracted from the complete collision cascade simulations within the film thickness of 0.1 µm. The parameters (dE/dx)e and (dE/dx)n are the electronic and nuclear energy loss, respectively. The other symbols are defined within the text.


Overlapping Cascades

Sparse Cascades c

S=1/

c

Pc=Ac/S

Ac=1/c

43

a)

Pc=1

b)

Figure 3.9. - Scheme of the phenomenological damage model in the two collision cascade regimes. The circles represent the mean area Ac of the collision cascade generated by a single ion impinging onto the sample. The spheres are for Si nanocrystals. Pc is the probability of a Si nanocrystal being located within the volume of a collision cascade. S is the mean surface, over the sample, containing just one collision cascade.

one of those defects inside the nanocrystal volume. It is possible to distinguish two different regimes for the ion fluence, depending on the area of the collision cascade Ac . In fact we can define a critical fluence φc ≡ 1/Ac , such that for φ φc we are in a first regime, characterized by very far apart collision cascades, while for φ φc a second regime arises, in which the collision cascades are completely overlapping. A scheme of both regimes is reported in Fig. 3.9. Sparse cascades: φ φc In the first case, schematized in Fig. 3.9a), the defects are totally concentrated within the collision cascades which occupy only a small volume within the film. Hence, a Si nanocrystal can be damaged only if it is located inside a cascade. But even if the nanocrystal is inside a cascade, this doesn’t necessarily mean that it is damaged, being the location of a defect inside a cascade a stochastic phenomenon. Thus the probability Pd of a Si nanocrystal being damaged after ion irradiation at fluences much smaller than the critical value must be the product of the probability Pc of finding the nanocrystal inside a cascade, times the probability P1 that, being inside the cascade, at least one defect is contained in the nanocrystal volume Vnc . At a fixed ion fluence φ, we can determine a surface of the film S ≡ 1/φ containing only one collision cascade, as drawn if Fig. 3.9a). Since we are interested in the fraction of quenched nanocrystals, we can concentrate our analysis over the volume

44


under this area S, which is representative of the overall film. The probability of a nanocrystal being in the volume of the collision cascade is simply given by: Ac (3.2) = φAc S no matter at which depth it is located. Now we need to calculate P1 , i.e. the probability that at least one defect is inside a nanocrystal, being the nanocrystal located inside the cascade. First of all it is necessary to calculate the probability of having exactly m defects inside a nanocrystal. The total number M of defects inside the cascade is given by M = nc Vc where nc is the concentration of displaced atoms and Vc = Ac t is the volume of the cascade. The probability p of generating a defect exactly within the nanocrystal is the chance a defect has of being located within the nanocrystal volume after the ion beam irradiation. This chance increases with increasing the nanocrystal volume, being equal to 1 when the nanocrystal and the cascade volumes are the same. Therefore p is simply the fraction of the cascade volume Vc occupied by a nanocrystal having volume Vnc ), i.e.: Pc =

Vnc (3.3) Vc Whilst the probability q of a defect being located outside the nanocrystal is given by the fraction of the cascade volume which is not occupied by a nanocrystal, i.e. (Vc − Vnc ) /Vc , or: p=

Vnc ≡1−p (3.4) Vc Thus, taking into account all the possible configurations of defects inside the cascade (whose volume can be divided in many empty cells each having the nanocrystal volume), the probability of having exactly m defects within the nanocrystal and the remaining M − m outside its volume is given by the binomial distribution: q =1−

M! (3.5) pm q M −m m! (M − m)! It is worth noticing that the mean number of defects contained in a nanocrystal is equal to: Pm =

Vnc nc Vc = nc Vnc (3.6) Vc which is the defect density inside a cascade times the nanocrystal volume, as one would expect. Since Vnc Vc , it results that p q and also m M . Thus eq. (3.5) can be approximated by the Poisson’s distribution: m = pM =

mm −m e (3.7) m! In order to calculate the probability of having at least one defect inside the nanocrystal, we need to sum eq. (3.7) over m in the range between 1 and the total Pm =


45

number of defects M . Being in general M m, we make a small error in summing between 1 and ∞. Thus we have: P1 =

∞

Pm =

m=1

∞

Pm − P0

(3.8)

m=0

where we have used the normalization condition: ∞

Pm = 1

(3.9)

m=0

P0 is the probability that the nanocrystal has no defects at all, being therefore still optically active, and is given by: P0 = e−m

(3.10)

P1 = 1 − e−m = 1 − e−nc Vnc

(3.11)

Eventually we get:

Then the total probability for a nanocrystal to be damaged after the ion irradiation and for a fluence much smaller than the critical one, i.e. φ φc , is: Pd = Pc P1 = φAc 1 − e−nc Vnc

(3.12)

It is worth noticing that the mean defect concentration Nd inside the film can be expressed as the total number of displaced atoms within a cascade divided by the volume of the region of interest containing that cascade, having surface area S and thickness t: Dc = dφ (3.13) St where we have used again the definitions S ≡ 1/φ and d ≡ Dc /t. By using eq. (3.13) together with eq. (3.1) it is possible to rewrite eq. (3.12) in the form: Nd =

Pd =

Nd 1 − e−nc Vnc nc

(3.14)

Developing the exponential in terms of nc Vnc 1, eq. (3.14) can be written in a first order approximation as: Pd Nd Vnc

(3.15)

which shows that for φ φc and for poorly dense cascades the probability of damaging a nanocrystal varies linearly with the mean defect concentration Nd in the film and with the nanocrystal volume, being independent from any parameter related to the particular ion used.

46


Overlapping cascades: φ φc If φ φc , the cascades are totally overlapping, thus the film surface is completely covered by the ion beam, as schematized in Fig. 3.9b). Hence the probability Pc of finding a nanocrystal in a cascade is just 1. In this regime the probability Pd of a nanocrystal containing at least one defect is determined only by the defect concentration Nd , which now increases linearly with ion fluence and can be approximated by: N d = nc

φ d φ = = dφ φc Ac φc

(3.16)

where we used the definition φc ≡ 1/Ac . Thus we obtain the same result found in eq. (3.13) for a completely different fluence regime. In conclusion, the probability of a nanocrystal having at least one defect inside its volume can be easily obtained by replacing nc with Nd in eq. (3.11), thus resulting in: Pd = 1 − e−Nd Vnc

(3.17)

which tells us that the damage mechanism depends again only on the defect concentration Nd , i.e. through eq. (3.16) on the number d of displaced atoms per incident ion per unit length, which is fixed by the matrix and the ion mass and energy, on the ion fluence and on the nanocrystal volume. In particular, the smaller the nanocrystal volume, the higher has to be the ion fluence in order to obtain the same fraction of damaged nanocrystals. Moreover, fixed the nanocrystal volume and the ion fluence, the fraction of damaged nanocrystals depends only on d, i.e. on the ion mass and energy, being independent of the nanocrystals concentration.

3.4.3

Theory versus experiment

We previously assumed that a nanocrystal becomes dark, i.e. optically inactive, when it contains at least one defect. For this reason we developed an expression for the probability of a nanocrystal having at least one defect inside its volume, eq. (3.17). Now we want to compare the model previosly developed with the experimental data. By using eq. (2.6) and the luminescence intensity and lifetime data reported in Fig. 3.4 it is possible to experimentally obtain the fraction fq of quenched Si nanocrystals, i.e. of nanocrystals which are unable to emit light after ion irradiation, versus the defect concentration related to the fluence through eq. (3.16), which is valid for all fluences and for all the ions used. If N is the concentration of Si nanocrystals which are still able to emit light after ion irradiation, and if NRef is the concentration of optically active centers in the not irradiated reference sample, the fraction of ’surviving’ nanocrystals can be defined as fs = N/NRef . According to eq. (2.6), N is directly proportional to the luminescence intensity I and inversely proportional to the lifetime τ . Thus we easily obtain the expression for the surviving fraction of Si nanocrystals:


1.0

47

0.8

fq

0.6

Pd

2 MeV He 2 MeV Si 2 MeV Si in SL 2 MeV Ge 2 MeV Au Simulation

0.4 0.2 0.0 16

10

17

10

18

10

19

10

20

-3

10

21

10

22

10

Nd (cm ) Figure 3.10. - Quenched fraction fq of Si nanocrystals versus defect concentration Nd left over by the ion beam for a Si/SiO2 superlattice (with a Si layer thickness of ∼ 0.9 nm) annealed at 1100◦ C for 1 hour and after 2 MeV Si+ implants and for a SiOx film with 39 at.% Si annealed at 1250◦ C for 1 hour after 2 MeV He+ , Si+ , Ge+ , Au+ implants. The continuous line is the probability of a nanocrystal having at least one defect inside its volume, as obtained from eqs. (3.14) and (3.17) with the parameters fixed by the experiment.

fs =

N I τRef = NRef IRef τ

(3.18)

Therefore the fraction of quenched, i.e. optically inactive, Si nanocrystals is simply given by: fq = 1 − fs = 1 −

I τRef IRef τ

(3.19)

in terms of variables which are experimentally measurable. In Fig. 3.10 the experimental fraction of quenched Si nanocrystals as determined through an analysis of the data shown in Fig. 3.4 by using formula (3.19) is reported with different symbols for each ion and versus the defect concentration calculated through eq. (3.16), by using for d the values reported in Tab. 3.1. In the same figure, the simulated curve of the damaged fraction of Si nanocrystals given by eqs. (3.14) and (3.17) is reported as a continuous line (right-hand scale), by using a value of 1.7 nm for the mean Si nanocrystals radius, in good agreement, within

48


60 50

(s)

40 30 20 10 0

He Si Ge Au Annealing data

0.1

fq

50 %

1

Figure 3.11. - Lifetime measured at 800 nm for a SiOx film with 39 at.% Si annealed at 1250◦ C for 1 hour after irradiation with 2 MeV He+ , Si+ , Ge+ , Au+ . Open triangles () are lifetime data for the same sample irradiated with 2 MeV Si+ at a fluence of 5×1012 /cm2 , and afterwards annealed at different temperatures in the range between 100 and 800◦ C. These latter data are discussed in the next section.

the experimental errors, with the value determined by dark field TEM plan views. It is worth noticing that eq. (3.14) and (3.17) perfectly match at Nd = nc . The impressive agreement between the experimental data and the simulated curve, in view of the fact that no adjustable parameters have been involved in the simulation, validate the hypothesis that in order to quench the luminescence of a nanocrystal, it is sufficient to have just one defect inside its volume. It is worth noticing that the superlattice, despite its very different structure, show the same experimental trend as for uniformly distributed Si nanocrystals. We can also gain a greater insight in the lifetime behavior of the irradiated Si nanocrystals formed in the annealed SiOx film by plotting the luminescence lifetime at 800 nm versus the quenched fraction of Si nanocrystals fq , as reported in Fig. 3.11. It is interesting to observe that the lifetime has the same trend for all the ions used. This means that fq is the correct physical parameter to look at in trying to explain the quenching mechanisms occurring with increasing the ion fluence. In particular, it

3.5 Defect annealing in Si nanoclusters

49

is evident from Fig. 3.11 that the trend of the experimental data is characterized by two slopes, one for fq < 0, and the other for 0.5 < fq < 0.9. Indeed, for fq < 0.5 the lifetime decreases only slightly, while in the range 0.5 < fq < 0.9 it decreases much strongly, i.e. at a much faster rate. In order to understand this experimental trend, at least qualitatively, we have first of all to stress the fact that the Si nanocrystals are interacting among each other in the present sample. This interaction is reflected in the stretched exponential behavior shown by the luminescence decay time, which is characterized by a dispersion factor β=0.75 which remains constant after the ion irradiation, demonstrating that the irradiation doesn’t affect at all the energy transfer among neighboring nanocrystals. What the irradiation does is certainly to reduce the number of emitting Si nanocrystals, as previously demonstrated. In fact, with increasing the defect concentration inside the film, more and more nanocrystals begin to contain at least one defect inside their volume, thus becoming damaged and therefore dark. But this doesn’t prevent an optically active nanocrystal to still interact with a damaged one. And in fact for a small fraction of damaged nanocrystals, i.e. small defect concentration Nd , the majority of nanocrystals is surrounded by non damaged nanocrystals with which they can interact. So the energy can wander all around the sample until a damaged nanocrystal, acting as a non radiative center, is met, thus causing the energy to be definitively lost and the luminescence lifetime to be shortened. With increasing the fraction of quenched nanocrystals, the total amount of quenching centers in the sample increases too, and the lifetime gradually decreases. In fact, this slight decrease of the lifetime versus fq is experimentally observed for fq values up to 0.5. When fq is equal to 0.5, 50% of the total number of emitting centers is damaged, so it is very likely to find each emitting nanocrystal surrounded by a damaged one. The vicinity between the emitting and the damaged centers causes the lifetime quenching to become much stronger. And, as a matter of fact, for fq > 0.5 a strong drop of the luminescence lifetime is experimentally observed, as reported in Fig. 3.11. The behavior of τ for fq values in the range between 0.9 and 1, is not experimentally accessible, being the luminescence signal of the irradiated samples too feeble for lifetime measurements to be performed, but it would be reasonable to expect a saturation behavior for fq approaching the value of 1. In fact in this regime the few Si nanocrystals which are still able to emit are very far apart one to each other, being completely surrounded by damaged nanocrystals. Thus adding one more damaged Si nanocrystal in the sample wouldn’t affect the lifetime properties of those emitting nanocrystals.

3.5 3.5.1

Defect annealing in Si nanoclusters Recovery of slightly damaged Si nanocrystals

In order to investigate in greater detail the nature of the irradiation damage on Si nanocrystals, we annealed at different temperatures and for different times the Si nanocrystals, formed in the SiOx film annealed at 1250◦ C for 1h and irradiated with

50


Recovery PL Intensity (arb. units)

1.0 0.8 0.6 0.4

Reference 5x1012 Si/cm2 100°C 200°C 300°C 400°C 500°C 600°C 700°C 800°C

0.2 0.0 600

650

700

750

800

850

900

Wavelength (nm) Figure 3.12. - PL spectra of a SiOx film with 39 at.% Si excess annealed at 1250◦ C for 1 hour (Reference), of the same sample after 2 MeV Si+ irradiation at a fluence of 5×1012 /cm2 (solid line) and of the irradiated sample after annealing processes for 4 h at different temperatures, in the range between 100 and 800◦ C. Spectra were measured at room temperature and at a laser pump power of 10 mW.

2 MeV Si+ at a fluence of 5×1012 /cm2 . At this fluence, while a huge number of Si nanocrystals is still present in TEM dark field views, only 30% of the initial number of nanocrystals are surviving from a luminescence point of view, as can be calculated from Fig. 3.4. Therefore this fluence is enough to quench 70% of the initial number of Si nanocrystals, without amorphizing them. In Fig. 3.12 the PL spectra of the as implanted, of the reference and of the samples annealed at different temperatures for 4h are shown. With increasing the annealing temperature, the luminescence intensity is seen to increase, until it reaches the reference value for a temperature of 800◦ C. No changes in the spectral shape have been observed within the temperature range used. The photoluminescence decay-time curves recorded at 800 nm (not reported) show a complete recovery of the lifetime after annealing the sample at 800◦ C too. This means that after an annealing at 800◦ C for 4h, also the number of emitting nanocrystals recovers its original value, attesting the fact that this thermal treatment is able to recover the damage in all the quenched Si nanocrystals.


51

Isothermal Annealings

1.2

I/IRef

1.0 0.8 0.6 0.4 0.2

(a)

0.0

/Ref

1.0 0.8 0.6

800 °C 700 °C 400 °C

0.4 0.2

(b)

0.0

N/NRef

1.0 0.8 0.6 0.4 0.2 0.0

(c) 0

4000

8000

12000

16000

Annealing Time (s) Figure 3.13. - Normalized PL intensities (a), lifetimes (b) and number of Si nanocrystals (c) emitting at 800 nm for a SiOx film annealed at 1250◦ C for 1 hour, irradiated with 2 MeV Si+ at a fluence of 5×1012 /cm2 and annealed at 400◦ C, 700◦ C and 800◦ C, for different times. The saturation behavior for each annealing temperature suggests the presence of many activation energies in the recovery process of slightly damaged Si nanocrystals.

At this stage a comment needs to be made. In Fig. 3.11 we showed that the lifetime of irradiated nanocrystals depends on the fraction of quenched Si nanocrystals and we explained this behavior in terms of energy migration among Si nanocrystals. If this is the case, a similar behavior should be obtained also after annealing. Indeed, in Fig. 3.11 the open symbols refer to the lifetime measured in annealed samples (at different temperatures) versus the fraction of quenched Si nanocrystals (which decreases with increasing the annealing temperature). The similar trend followed by the lifetime in both the irradiation and annealing experiments definitely demonstrates that the physical parameter which rules the quenching of the lifetime is just the fraction of quenched nanocrystals fq , no matter how this value is obtained, whether through ion irradiation or by annealing of a damaged sample. It is interesting to investigate the kinetics of both the intensity and the lifetime

52


Recovery of Damaged Si nc 1.0

N/NRef

0.8 0.6 0.4 0.2 0.0 0

12

2

5x10 Si/cm 200

400

600

800

1000

Annealing Temperature (°C) Figure 3.14. - Fraction of emitting Si nanocrystals in a SiOx film annealed at 1250◦ C for 1 h, irradiated with 2 MeV Si+ at a fluence of 5×1012 /cm2 , and then annealed at different temperatures. All of the annealing processes were performed for 4h, for which a saturation of both the intensity and the lifetime, i.e. of N/NRef , was observed. It is worth noticing that at 800◦ C the complete recovery of the number of emitting Si nanocrystals occurs, thus fixing an upper limit for the activation energies characterizing the defects in the damaged nanocrystals.

recovery, and eventually of the fraction of Si nanocrystals which have recovered their optical activity. Thus we performed isothermal annealing processes for times in the range between 5s and 4h, and for temperatures up to 1150◦ C. Just as examples, in Fig. 3.13 the results obtained for three different temperatures (400◦ C, 700◦ C and 800◦ C) are reported. Both the intensity and the lifetime measured at 800 nm tend to quickly saturate for annealing times greater than ∼10 s. The same trend is followed by the fraction of emitting nanocrystals, obtained through eq. (3.18). It is worth noticing that for the two temperatures of 400◦ C and 700◦ C the saturation values are different, in particular the higher is the temperature, the higher is the saturation value, but, more important, despite the long annealing times it is impossible to regain the initial value of emitting Si nanocrystals. This trend is typical for processes in which a continuous range of activation energies can take place. At each of the used annealing temperatures, a different set of annihilation processes is probed. Indeed for low temperatures, only those defects characterized by a low activation energy can annihilate at a measurable rate and when they are all removed a saturation is reached. At higher temperatures the processes with low activation energies have


53

long since been completed, and only defects with progressively higher activation energies annihilate. Thus, the fact that at 800◦ C after 4h it is possible to obtain a total recovery of both the PL intensity (Fig. 3.13a) and the lifetime (Fig. 3.13b), i.e. of the fraction of emitting Si nanocrystals (Fig. 3.13c), means that an upper limit for the activation energies of the damaged sample exists. It is possible to obtain the saturation value for the fraction of emitting nanocrystals vs the annealing temperature, by deducing it from the data reported in Fig. 3.13 and from similar data performed at several different temperatures. The result is shown in Fig. 3.14. We want to stress the fact that the annealing time of 4h is much longer than the typical saturation time of ∼10 s. This guarantees that for each annealing temperature we are in an equilibrium regime. This means that in Fig. 3.14 we are looking at the total fraction of Si nanocrystals that can be recovered for each particular annealing temperature. At room temperature the surviving fraction of nanocrystals after the ion irradiation with a fluence of 5×1012 /cm2 is about 30% of the total number of emitting nanocrystals in the reference sample. Up to 200◦ C, the population of damaged nanocrystals seems to be unaffected by the annealing. With increasing the temperature, the fraction of emitting nanocrystals starts to increase linearly, meaning that a certain number of defects are beginning to be annealed out. This linear regime ends at 800◦ C, for which the damaged fraction of nanocrystals has totally been recovered. Indeed, annealings performed at higher temperatures (≤ 1000◦ C) are not able to increase anymore the number of optically active Si nanocrystals. More information can be obtained by a detailed analysis of the data shown in Fig. 3.14. In fact, from the data shown in Fig. 3.14, we can now obtain the activation energy spectrum of the defects present in Si nanocrystals. By fixing the annealing temperature we are selecting all those recovery events which involve the annealing of those defects with activation energies lower than a critical value Ea , related to that particular temperature by the well known relation [94, 95]: Ea = kT ln (νt)

(3.20)

where k is Boltzmann’s constant, T is the annealing temperature in Kelvin, t∼10 s is the saturation value of the annealing time (as determined from Fig. 3.13), and ν is an attempt frequency. Since we are involved in the annealing of defects in a crystalline matrix, we can assume that ν is the vibration frequency of an atom at that particular temperature, which is equal to ∼kT /h, being h the Planck’s constant. We can look at the distribution of recovery events by plotting the derivative of the emitting fraction of Si nanocrystals versus the activation energy given by eq. (3.20), as reported in Fig. 3.15. This derivative, indicated as ρ (Ea ), represents the density of defect recovery events occurring per unit energy range. Hence ρ (Ea ) ∆E gives the fraction of defects having activation energies in the range between Ea and Ea + ∆E. The trend of ρ (Ea ) versus Ea , reported in Fig. 3.15, is very interesting, as it demonstrates the presence of a population of defects in Si nanocrystals characterized by a continuous spectrum of activation energies, ranging between ∼ 1.5

54


0.8

Activation Energy Spectrum 12

2

(Ea) (eV-1)

5x10 Si/cm 0.6 0.4 0.2

0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Ea (eV) Figure 3.15. - Activation energy spectrum obtained for a SiOx film with 39 at.% Si content annealed at 1250◦ C for 1 hour and irradiated with 2 MeV Si+ at a fluence of 5×1012 /cm2 . ρ (Ea ) is the derivative of N/NRef , shown in Fig. 3.14, with respect to the activation energy Ea related to the annealing temperature T through eq. (3.20), and is proportional to the number of recovery events occurring per unit energy interval. The damaged Si nanocrystals are therefore characterized by a variety of defects with activation energies in the range between 1.5 and 3.0 eV.

and 3 eV. In particular, for Ea < 1.5 eV, i.e. temperatures lower than ∼ 300◦ C, practically no recovery events occur, while in the energy range between 1.5 and 3.0 eV, corresponding to the temperature range within 300◦ C and 800◦ C, a variety of defects which are annealed out at a constant rate exists in the low-dose-damaged nanocrystals. A question could arise on the nature of the defects that we are producing and annihilating. At a Si fluence of 5×1012 /cm2 the collision cascades are overlapping, being φ φc (Table 3.1). Thus, through eq. (3.6) in which we use for the defect concentration the value given by eq. (3.13) instead of the defect concentration inside each collision cascade nc , we can estimate the mean number of defects inside a nanocrystals to be ∼ 1.2. In fact for this value the surviving fraction of Si nanocrystals calculated by putting m = 0 in eq. (3.7) is just 30%. Moreover, by using eq. (3.7) we can calculate the probability of a nanocrystals having exactly m defects, or, which is the same, the fraction of nanocrystals containing in their volume exactly m defects, left over by the ion beam. In particular the probability of having 1, 2, 3 or 4 defects in a nanocrystal is respectively equal to 0.36, 0.22, 0.09, or 0.03, the


55

probability of having more than 5 defects being infinitesimal. Therefore 36% of the damaged nanocrystals contains only one defect, while the remaining 34% have more than one defect inside. It is tempting to assume that the recovery events occurring with activation energies between 1 and 3 eV are related to the annihilation of defect structures whose variety and complexity increase with the number of defects actually present in each Si nanocrystal.

3.5.2

Recrystallization of amorphized Si nanoclusters

A totally different scenario is met by dealing with a Si nanocrystals sample irradiated with 2 MeV Si+ ions at a fluence of 2×1013 /cm2 . In fact at this particular fluence the PL intensity is completely quenched, as can be seen in Fig. 3.2. This means that all the nanocrystals have been quenched. At the same time, in TEM dark field views (not shown) the signs of a crystalline phase in the film are missing, meaning that only amorphous nanoclusters are left in the sample. Therefore Si fluences as low as 2×1013 /cm2 are enough to totally amorphize the Si nanocrystals present in the matrix.

1.2


1.0 0.8

Reference 2 MeV 2x1013 Si/cm2 1000°C - 30 s 1000°C - 1 min 1000°C - 5 min 1000°C - 10 min 1000°C - 20 min

0.6 0.4 0.2 0.0 600

650

700

750

800

850

900

Wavelength (nm) Figure 3.16. - PL spectra of a SiOx film with 39 at.% Si excess annealed at 1250◦ C for 1 hour (Reference), of the same sample after 2 MeV Si+ irradiation at a fluence of 2×1013 /cm2 and of the irradiated sample after annealing processes at 1000◦ C for different times. Spectra were measured at room temperature and at a laser pump power of 10 mW.

56


Recrystallization of amorphized Si nc 800°C 900°C 1000°C 1100°C

I/IRef

1.0 0.8 0.6 0.4 0.2

(a)

/Ref

0.0 1.0 0.8 0.6 0.4

N/NRef

0.2

(b)

0.0 1.0 0.8 0.6 0.4 0.2 0.0

(c) 0

10

1

10

2

10

3

10

4

10

5

10

Annealing Time (s) Figure 3.17. - Normalized PL intensities (a), lifetimes (b) and number of Si nanocrystals (c) emitting at 800 nm for a SiOx film annealed at 1250◦ C for 1 hour, irradiated with 2 MeV Si+ at a fluence of 2×1013 /cm2 , and afterwards annealed at different temperatures in the range between 800◦ C and 1100◦ C and for different times. Dashed lines are guides to the eye, while continuous lines are fit to the data through eq. (3.21). The PL properties as well as the crystallinity of the amorphized Si nc can be recovered at each of the annealing temperatures used, after an average annealing time τc which strongly depends on the particular temperature.

In order to study the recovery behavior of this system we performed thermal treatments of the as implanted sample at temperatures up to 1150◦ C and for annealing times in the range between 5s and 20h. It has been observed that for temperatures up to 700◦ C practically no recovery of the PL intensity is observed. In addition, no recrystallization occurred as seen by TEM. On the contrary, for temperatures higher than 700◦ C something different occurs. As an example, in Fig 3.16, PL spectra of a SiOx film with 39 at.% Si excess annealed at 1250◦ C for 1 hour (Reference), of the same sample after 2 MeV Si+ irradiation at a fluence of 2×1013 /cm2 (open triangles) and of the irradiated sample after annealing processes at 1000◦ C for different times are shown. Spectra were


57

T (°C)

Recrystallization Time (s)

1200 1100

105

1000

900

800

Ea = 3.4 ± 0.3 eV

4

10

Si nanograins Bulk Si

3

10

102 101 0.70

0.75

0.80

0.85

-1

0.90

0.95

1000/T (K ) Figure 3.18. - The characteristic recrystallization time τc needed to completely recover the initial number of emitting Si nanocrystals, as extracted from the data in Fig. 3.17c), is reported versus the reciprocal of the annealing temperature. An activation energy of 3.4 eV can be estimated for the recrystallization of amorphized Si nanocrystals.

measured in the same conditions, at room temperature and at a laser pump power of 10 mW. It is worth noticing that with increasing the annealing time, the PL intensity increases, and it is completely recovered after 20 minutes of annealing time. In Fig. 3.17 the PL intensity, the lifetime and the number of nanocrystals emitting at 800 nm, as obtained through eq. (3.18), normalized to the respective values of the reference sample are reported versus the annealing time and for different annealing temperatures. As can be clearly seen, at each temperature, the PL intensity as well as the luminescence lifetime, measured at 800 nm, increase with increasing the annealing time, until they completely recover the reference values after a certain time, which strongly depends on the particular annealing temperature. This behavior is also followed by the fraction of Si nanocrystals which are able to emit light again after the thermal treatments. Actually this fraction is equal to the fraction of recrystallized Si grains. In fact, after annealing at these temperatures, nanocrystals do appear again as observed by dark-field TEM views. Thus by studying the recovery of the luminescence properties of the amorphized Si nanocrystals, we are indeed monitoring the transition between the amorphous and the crystalline phases of the Si grains present in the irradiated sample. The solid lines in Fig. 3.17c) are fit to the data by using the expression:

58


N = 1 − e−t/τc NRef

(3.21)

being τc the characteristic crystallization time, strongly temperature dependent. As a matter of fact, by reporting in an Arrhenius plot (Fig. 3.18) this characteristic time needed to attain the recovery of the Si grains, versus the reciprocal of the annealing temperature, it results that the recovery process is thermally activated, with a single activation energy of 3.4 eV. Actually this is a direct measurements of the activation energy for the recrystallization process of an amorphized Si nanocrystal.

3.6

Amorphized Si nanocrystals versus amorphous bulk Si

A few comments need to be made at this stage. First of all, from the picture drawn up to now, it emerges that the complete amorphization of a Si nanocrystal can occur at Si fluences as low as 2×1013 /cm2 , at which it is estimated that only a few defects are left by the ion beam inside each nanocrystal. Since these fluences are far away from the amorphization threshold of Si single crystal [83,94], this demonstrates that amorphization of Si nanocrystals is by far easier than amorphization of bulk Si. This is not surprising. In fact, it is well known that the surface of a single crystal Si wafer (i.e. its interface with the native oxide) represents a preferential nucleation site for amorphization [85] which starts there under ion bombardment well before than in the bulk. In the present case, each nanocrystal has a large surface area acting as a strong nucleation site for the amorphous phase. Moreover, our data demonstrate that, while amorphization is easier in Si nanocrystals, crystallization of all the amorphized Si nanocrystals is much more difficult with respect to bulk amorphous Si. In fact, temperatures higher than 800◦ C and long annealing times are needed to recrystallize the amorphized nanocrystal, the amorphous to crystalline phase transition being characterized by an activation energy of 3.4 eV. This behavior is in agreement with recent data by Zacharias et al. [88] showing that the crystallization temperature of very thin amorphous Si layers increases with decreasing layer thickness. Indeed, we can think at the recrystallization of an amorphous Si nanocluster as a transformation process occurring through nucleation and subsequent growth of the crystalline phase inside each nanocluster, with the nucleation being the limiting step. Usually the homogeneous nucleation of the crystalline phase in an amorphous Si film is characterized by an activation energy of 5.3 eV, as reported in [96]. In our case, due to the large surface/volume ratio of each nanocrystal, the nucleation is probably heterogeneous, the surface being a preferential nucleation site. The activation energy of the nucleation process can be expressed as [96]: Ea,nucl = ∆G∗ + Ekinetic

(3.22)

3.6 Amorphized Si nanocrystals versus amorphous bulk Si

Amorphous bulk Si

1 m

59

Amorphized Si nanograins

a)

1 nm

b)

Figure 3.19. - Schematic of the main recrystallization mechanisms in bulk amorphous Si a) and in a sample of dipersed amorphized Si nanocrystals b). Even if the activation energies for the two processes are identical, the kinetic of Si nanoclusters recrystallization proceeds much more slowly, since in order to attain the complete recrystallization, at least one crystalline nucleus has to be produced in every nanocluster. This increases enormously the number of nucleus that need to be formed, i.e. the time it takes to complete the recrystallization process, given a certain crystallization rate.

being ∆G∗ the free energy difference at the critical nucleus and Ekinetic the kinetic contribution, due to the formation and migration of the defects responsible of the phase transformation. In thermal nucleation ∆G∗ is ∼ 2 eV [96], its value being determined by a balance among a free energy reduction term due to the free energy difference among the crystalline and the amorphous phases times the transformed volume, and a free energy increase term due to the surface tension times the new surface formed as a result of the phase transformation. Since the nucleation occurs very likely at the surface of the amorphous zone the new surface formed is much smaller and this results in a reduction in the surface free energy term and, in turn, in a dramatic reduction in ∆G∗ . For instance a surface reduction by a factor of 2 produces a difference in ∆G∗ by a factor of 8. We can then state that ∆G∗ is probably small in the present case and most of the activation energy for the nucleation is due to the kinetic term. In fact Ekinetic ∼ 3.4 eV for amorphous Si [96]. This value actually equals the value we obtained for the activation energy associated to the recrystallization of amorphized nanocrystals. Thus we would expect the crystallization of the amorphized nanocrystals to occur through nucleation followed by a rapid growth of the crystal grain. In fact, at 800◦ C a few seconds are enough to totally grow a 1.5 nm Si nanocrystal using the literature growth velocities for small crystal grains [96]. But, since the activation energy for nucleation is lower in amorphized nanocrystals than in bulk Si, why does the recrystallization of all the nanocrystals

60


requires such long times and high annealing temperatures? Actually, in order to attain the total crystallization of an amorphous bulk Si, only a few crystal nuclei need to be formed. In fact subsequent growth of the crystalline phase is so fast that the total crystallization occurs in characteristic times which are much smaller than those required for the formation of new crystal nuclei. Indeed, in Fig. 3.18 it is shown (solid triangle) that at 800◦ C bulk amorphous Si recrystallizes in 20 s [96], i.e. more than three orders of magnitude faster than the recrystallization of all the amorphized nanocrystals in our samples. The final structure consists of a few large grains, as schematically represented in Fig. 3.19a). On the other hand, in order to observe the total recovery of the PL properties of the amorphized nanocrystals, all of them have to recrystallize. But the crystal growth is now spatially limited in the single grain volume, see Fig. 3.19b), ending at the interface between the nanocrystal and the oxide matrix. Thus in order to observe the complete recrystallization, we have to wait enough time for the nucleation to occur in all of the amorphous grains. As an example, since at 800◦ C the time needed for the total crystallization of the amorphous Si film is ∼20 s [96], and since the measured nucleation rate is ∼1013 /cm3 s, we get that only 2×1014 /cm3 nucleation sites are formed, randomly distributed in the amorphous film. In our sample, the concentration of nanocrystals is roughly ∼1019 /cm3 , while the characteristic crystallization time at that temperature is 5×104 s, bringing to a mean nucleation rate of more than 2×1014 /cm3 s, which is still one order of magnitude higher than for bulk amorphous Si. This increased value is a clear dimensionality effect present in our sample, arising from the high surface/volume ratio which characterizes the amorphized Si nanocrystals and which favors the heterogeneous nucleation at the amorphized nanocrystal/SiO2 interface. However, despite the increased nucleation rate, the characteristic recrystallization times are much longer than for bulk amorphous Si, since we need to form a greater number of crystalline nuclei, i.e. at least equal to the total number of amorphous nanoclusters, and moreover in a well defined region, i.e. the Si nanocluster volume.

3.7

Conclusions

In the present work, the damaging effects of ion beam irradiation and subsequent annealing on Si nanocrystals have been reported. In particular, the luminescence properties of the irradiated and annealed nanocrystals have been used as a probe of the damaging mechanisms occurring in Si nanocrystals. We have demonstrated that by increasing the ion fluence, the luminescence and the lifetime recorded at 800 nm begin to decrease after a critical fluence depending on ion mass, the luminescence quenching being much stronger than the lifetime one. Thus we have experimentally demonstrated that the number of emitting centers has to diminish too. This means that even at fluences much lower than those needed to amorphize bulk Si, a Si nanocrystal can be damaged in such a way to become dark from a luminescence point of view, not necessarily being amorphous. Anyway, by increasing the ion fluence, a Si nanocrystal can even be amorphized at fluences well below those needed

3.7 Conclusions

61

to amorphize bulk crystalline silicon. By assuming that a Si nanocrystal remains damaged when it contains at least one defect, we developed a model in which the quenched fraction of nanocrystals depends exponentially on the nanocrystal volume and on the total defect concentration left over by the ion beam. The agreement with the data for the quenched fraction of nanocrystals, extracted by the luminescence and lifetime measurements, demonstrates that it is sufficient to have just one defect inside the nanocrystal in order to quench its luminescence. Moreover by performing isochronal and isothermal annealings of Si-irradiated samples, we were able to study in detail the recovery mechanisms in different fluence regimes. We have demonstrated that at a Si fluence of 5×1012 /cm2 , corresponding to ∼1 defect per nanocrystal on the average, defects are characterized by a large variety of activation energies in the range between 1.5 and 3 eV. In the case of amorphized nanocrystals, obtained at a Si fluence of 2×1013 /cm2 , the recovery process and consequently the transition between the amorphous and the crystalline phases of a Si nanocluster, have been demonstrated to be thermally activated, with a single activation energy of 3.4 eV, which has been directly associated to the nucleation process starting at the nanocluster surface. Anyway, despite it has been shown that nucleation rates one order of magnitude higher can be obtained due to surface effects, temperatures higher than 800◦ C and longer annealing times are required with respect to bulk amorphous Si, in order for the recrystallization process to become measurable. In fact, while for amorphous Si only a few nuclei need to be formed, in order to observe the recrystallization of all the amorphized Si nanocrystals it is necessary to form a much larger number of nuclei, i.e. equal to the number of amorphous nanoclusters, and moreover in a well defined region, i.e. the nanocluster volume. Thus the characteristic crystallization times become much longer than for bulk amorphous Si. These data demonstrate that the amorphization of low dimensional Si structures is much easier while crystallization is much more difficult with respect to bulk silicon.

Chapter 4 Er-doped Si nanoclusters

In the present chapter, a detailed study of the optical properties of Erdoped Si nanoclusters, obtained through ion implantation of Er in samples containing Si nanocrystals formed by PECVD, is reported. It is shown that all of the Si nanocrystals are amorphized by the ion beam, and the subsequent annealing at 900◦ for 1 hour is not enough to recrystallize all of them. Nevertheless, it is shown that Si nanoclusters even in their amorphous phase are still able to absorb the incoming photons and to transfer efficiently their energy to nearby Er ions. The principal effects of this sensitizing action are the following: i) the effective excitation cross section for Er ions is enhanced by more than two orders of magnitude with respect to the direct absorption of photons, ii) the excitation cross section of Er mimics the broadband features of the Si nanoclusters absorption band, therefore allowing the use of photons in a range of wavelengths much wider than the resonant ones to excite Er. A phenomenological model able to fit the experimental data in a wide range of Er concentrations (between 3×1017 /cm3 and 1.4×1021 /cm3 ) and excitation pump powers (in the range 1-103 mW) is presented. Important physical parameters, such as the coupling constant, the concentration quenching and up-conversion coefficients, and the energy transfer time are extracted. The implication of the finite transfer time are addressed. Eventually, the role of Si nanoclusters and of strong gain limiting processes, such as cooperative up-conversion and confined carrier absorption from an excited nanocluster, in determining positive gain at 1.54 µm are investigated in details.

4.1

Introduction

Among the different approaches developed to overcome the intrinsic low efficiency of silicon as a light emitter, rare earth doping of silicon is dominating, together with

64

CHAPTER 4. Er-doped Si nanoclusters

nanostructured silicon systems, the scientific scenario of silicon-based microphotonics. Indeed, studies on Er doped crystalline Si have demonstrated that Er can be efficiently excited through electron-hole pair recombination or through impact of energetic carriers [48, 54] in silicon and, despite the presence of efficient non radiative de-excitation processes like Auger with free carriers and energy back transfer [48], Er-doped Si devices operating at room temperature have been developed [54, 97]. Recently, Er doping of Si nanocrystals has been recognized as an interesting way of combining the promising features of both the previous methods [60,98–104]. Indeed, it has been demonstrated that Si nanocrystals in presence of Er act as efficient sensitizers for the rare earth [60, 104–107]. In particular, the nanocrystal, once excited, promptly transfers quasi-resonantly [108] its energy to the nearby Er ion, which then decays emitting a photon at 1.54 µm. The effective excitation cross section for Er in presence of Si nanocrystals is more than two orders of magnitude higher with respect to the resonant absorption of a photon in a silica matrix [60]. Moreover the non radiative de-excitation processes are strongly suppressed, the Er lifetime being almost constant in the temperature range between 11 and 300 K ( [109]). The recent demonstration of efficient room temperature electroluminescence from Er-Si nanoclusters devices [110] and the determination of net optical gain at 1.54 µm in Er-doped Si nanocluster sensitized waveguides [111–113] opened the route towards the future fabrication of electrically driven optical amplifiers based on this system. Several basic issues remain however to be first addressed, namely the role of amorphous Si nanoclusters in the Er excitation, the strength of the various competitive non radiative processes, the coupling strength between Si nanoclusters and Er ions and the microscopic details of the interaction. In the present chapter, it is demonstrated that amorphous Si nanoclusters, as well as crystalline ones, can act as efficient sensitizers for the Er luminescence. Indeed, a rate equation model describing the time evolution of the coupled Er and Si nanoclusters level populations is presented and an estimate of the efficiency of the energy transfer reported. Moreover, the occurrence of non radiative phenomena involving Er-Er interacting pairs, such as concentration quenching and cooperative up-conversion, is demonstrated. Through a comparison of simulated and measured photoluminescence data in a wide range of Er concentrations and pump powers, both the coupling and the up-conversion coefficients will be determined. Important issues such as the implications of the finite energy transfer time in the overall luminescence efficiency, the maximum number of excitable Er ions per nanoclusters, and the role of up-conversion and carrier absorption from an excited nanocluster in limiting the possibility of obtaining positive gain at the 1.54 µm Er related emission are addressed. These data are reported, a comparison with recent results [111–115] is performed and future perspectives towards the achievement of optical amplifiers are discussed.

4.2 Sample Preparation

65

Energy (MeV) 30

0.5 O in film

1.0

1.5

Normalized Yield

25 20 15

Si in film

10

Er in film

5 0

100

200

300

400

500

600

Channel Figure 4.1. - Rutherford backscattering spectrometry (RBS) measurement at a scattering angle of 165◦ with respect to the incident ion beam direction, in random configuration, of a 0.2µm SiOx film (with 42 at.% Si) deposited on top of a Si substrate, annealed at 1250◦ C for 1 hour and multiplyimplanted with 170 keV, 300 keV, 500 keV Er ions at the respective doses of 0.5×1015 cm−2 , 0.8×1015 cm−2 and 1.6×1015 cm−2 . As the result of the multiple Er implants, a concentration of 2.2×1020 cm−3 is obtained almost all over the film thickness, as evidenced in the inset reporting the Er profile inside the sample, as extracted by the RBS data.

4.2

Sample Preparation

Si nanocrystals were produced by high temperature annealing of a 0.2 µm thick substoichiometric SiOx film (with 42 at.% Si) grown by PECVD on top of a Si substrate. Thermal treatments were performed at 1250◦ C for 1 hour in ultrapure nitrogen atmosphere. The annealing process induces the separation of the Si and SiO2 phases, leading to the formation of uniformly distributed Si nanocrystals, as previously discussed in this thesis. The actual concentration of Si nanocrystals is a very difficult parameter to determine by TEM analyses. The estimated Si nanocrystal density in the sample has a value which varies in the range ∼ 1018 1019 /cm3 . Afterwards, Er ions were then implanted at different energies (in the range between 170-500 keV) and fluences in order to produce an almost constant Er concentration (in the range 3×1017 -1.4×1021 /cm3 ) all over the film thickness. Other Er implants have been performed on a sample containing Si nanoclusters obtained by annealing the SiOx film at lower temperatures, in the range 0◦ C-1100◦ C for 1h. The very same Er implants were also performed in SiO2 layers not containing Si nanocrystals in order to have reference samples. All of the samples were eventually

66


20

Er Concentration (cm-3)

3x10

Exp. profile Simulated profiles Sum

2x1020

1x1020

0 0

50

100

150

200

250

Depth (nm) Figure 4.2. - Comparison between the experimental Er profile (solid line) as determined by an analysis of Fig. 4.1 and the simulated one (solid circles), obtained through TRIM Monte Carlo simulations. The dash-dotted lines represent the simulated Er profiles corresponding to the multiple Er implants at 170 keV, 300 keV, 500 keV Er ions at the respective fluences of 0.5×1015 /cm2 , 0.8×1015 /cm2 and 1.6×1015 /cm2 . Each profile has been obtained through a detailed TRIM calculation with full damage cascades and by using 104 incident ions. The sum of the three simulated profiles, giving the total Er concentration, is reported in solid circles and it is in quite good agreement with the experimental data. In particular, as the result of the multiple Er implants, a concentration of 2.2×1020 /cm3 is obtained over 150 nm.

annealed at 900◦ C for 1 hour in ultrapure nitrogen atmosphere in order to activate Er, preventing it from clustering. In Fig. 4.1, the Rutherford backscattering spectrometry (RBS) measurement at a scattering angle of 165◦ of a 0.2 µm SiOx film deposited on top of a Si substrate, annealed at 1250◦ C for 1 hour and afterwards multiply-implanted with Er ions is E (keV) 170 300 500

Dc 2603 4201 4397

rc (nm) 13.4 20.0 29.3

Conc./Fluence (cm−1 ) 2.65×105 1.84×105 1.16×105

Rp (nm) 66.6 102.7 157.3

Straggling (nm) 14.6 22.4 33.9

Table 4.1. - Parameters extracted from the complete collision cascade simulations for Er ions implanted within a 0.2 µm thick SiOx film. In particular, the number Dc of displaced atoms per incident ion, the mean collision cascade radius rc , the Er peak concentration per fluence, the projected range Rp and the longitudinal straggling are reported for three different ion beam energies E.

4.2 Sample Preparation

67

Conc. of Displ./ Fluence (x 108 cm-1)

166

Er

5

170 keV 300 keV 500 keV

4 3 2 1 0

0

50

100

150

200

250

Depth (nm) Figure 4.3. - Concentration of displaced atoms (at./cm3 ) divided by the ion beam fluence (at./cm2 ) produced in a 0.2 µm thick SiOx film by 166 Er at different ion beam energies (170, 300 and 500 keV), as obtained by a TRIM Monte Carlo simulation with full damage calculation.

reported, after an annealing of 900◦ C for 1 h. A fit of the RBS measurement gives a value of 42 at.% for the total amount of Si in the film. Figure 4.2 shows a comparison between the experimental Er profile (solid line), as determined by an analysis of Fig. 4.1, and the simulated one (solid circles), obtained through TRIM Monte Carlo simulations. The dash-dotted lines represent the simulated Er profiles corresponding to the multiple Er implants at 170 keV, 300 keV, 500 keV Er ions at the respective fluences of 0.5×1015 /cm2 , 0.8×1015 /cm2 and 1.6×1015 /cm2 . Each profile has been obtained through a detailed TRIM calculation with full damage cascades and by using 104 incident ions. The sum of the three simulated profiles, giving the simulated total Er concentration, is reported in solid circles and it is in quite good agreement with the experimental data. In particular, as the result of the multiple Er implants, a constant concentration of 2.2×1020 /cm3 has been obtained almost all over the film thickness. A summary of the most significant parameters extracted by performing a detailed analysis of the collision cascades produced by TRIM can be found in Tab. 4.1, for the different energies used in the multiple implants. It is interesting to quantify the damage left over in the film by the Er ion beam. In Fig. 4.3, the depth distribution of displaced atoms generated by Er implants is

68


shown for three different energies. In particular, the left hand scale refers to the concentration of displaced atoms per ion fluence. The data have been obtained by a TRIM Monte Carlo simulation with full damage calculation. The maxima of the three curves have a value of about 4×108 /cm, but they are located at different depths, i.e. the higher the beam energy the deeper in the film. If we now recall eq. (3.17) developed within the phenomenological damage model, we get that for a critical defect concentration Nd,c such that Nd,c Vnc = 1, the 63% of the existing nanocrystals are optically quenched, containing at least one defect inside their volume. Since we know the nanocrystal volume from TEM images, we can estimate a critical defect concentration of ∼5×1019 /cm3 . For higher defect concentration, all of the Si nanocrystals are therefore optically quenched. By recurring to Fig. 4.3, the critical Er fluence after which all the nanocrystals are optically quenched is ∼1.25×1011 /cm2 corresponding, through Tab. 4.1, to a peak Er concentration of roughly 2×1016 /cm3 , which is much lower than the nanocrystal concentration indeed. Moreover, for an Er concentration ten times higher than this, we expect all of the Si nanocrystals to be amorphized, as observed in the previous chapter. In our experiments the Er concentration is always > 1017 /cm3 , hence we expect all of the nanocrystals to be in their amorphous phase after all the implants. In addition, we expect the majority of them to remain amorphous even after the annealing at 900◦ C for 1 h, necessary to activate Er preventing it from clustering, is performed. Indeed at a temperature of 900◦ C, longer annealing times, of about 3 hours, would be necessary to completely recrystallize the amorphized nanocrystals, as seen in the previous chapter (Fig. 3.17).

4.3

Experimental

Photoluminescence (PL) measurements were performed, when not differently stated, by pumping with the 488 nm line of an Ar laser. The pump power was varied in a wide range, between 1-103 mW, and focused over a circular area of ∼0.3 mm in radius. The laser beam was chopped through an acousto-optic modulator at a frequency of 11 Hz. The luminescence signal was analyzed by a single grating monochromator and detected by a photomultiplier tube for the visible range (0.40.9 µm) or by a Ge detector for the infra-red (0.8-1.7 µm spectral region). Spectra were recorded with a lock-in amplifier using the chopping frequency as a reference. All the spectra have been measured at room temperature and corrected for the spectral system response. Time resolved PL measurements were performed by first detecting the modulated luminescence signal with a Hamamatsu photomultiplier tube (model R5509-72) having an almost constant spectral response in the range 0.4-1.7 µm. The signal was hence analyzed with a multichannel scaler, triggered by the acousto-optic modulator. The structural characterization was performed by using a 200 kV energy filtered transmission electron microscope (EFTEM) Jeol Jem 2010F with Gatan Image Filter. This system consists of a conventional TEM coupled with an electron energy

4.4 Structural Properties

69

(a)

(b)

20 nm

20 nm

(c)

20 nm

(d)

10 nm

Figure 4.4. - (a) EFTEM image of a SiOx sample annealed at 1250◦ C for 1h and (b) corresponding dark field image. (c) EFTEM image of a SiOx sample annealed at 1250◦ C for 1h, implanted with Er and subsequently annealed at 900◦ C for 1h. Dark-field TEM images show no signal from crystalline Si nanograins. (d) EFTEM image of the sample pre-annealed at 800◦ C for 1h, implanted with Er and finally annealed at 900◦ C for 1h. Note that the magnification of this last image is doubled with respect to the other ones.

loss spectrometer (EELS). Through EFTEM it is possible to create an image of the sample by using only electrons that have lost a specific amount of energy. This is a very suitable method to detect silicon nanograins (both crystalline and amorphous and independently of the crystal orientation) dispersed in a silica matrix, in fact the plasmon loss energy in silicon is 16.7 eV, that is well separated from the plasmon loss energy in silica (which has a value of 23.2 eV).

4.4

Structural Properties

In Fig. 4.4 TEM plan view images of different samples are shown. Fig. 4.4a reports the EFTEM image of the unimplanted SiOx sample annealed at 1250◦ C for one hour. This image was obtained by selecting an energy window centered at 16 eV (corresponding to the value of the Si plasmon loss) and 5 eV wide. Therefore bright spots correspond to Si nanograins present in the film. A large number of Si grains appears. The clusters are well separated and are characterized by a mean radius of about 2.2 nm. A dark field TEM image, reported in Fig. 4.4b, was obtained from the same sample by selecting a small portion of the diffraction ring of the 111 Si planes. The dark field image confirms that the Si

70


clusters are crystalline; also in this case their mean radius is about 2.2 nm. The huge difference between the number of Si clusters detected by the two techniques is due to the fact that EFTEM is able to detect all the Si agglomerates present in the layer, independently of their orientation and/or crystalline structure. Fig. 4.4c reports an EFTEM plan view image of the sample annealed at 1250◦ C, implanted with Er and subsequently annealed at 900◦ C for 1h. The image shows a Si nanoclusters distribution very similar to the one reported in Fig. 4.4a for the unimplanted sample. Therefore, the implantation process is not producing a mixing of the Si clusters with the SiO2 matrix. However, for this sample it was impossible to obtain a dark field image, since the electron diffraction pattern reveals the absence of a crystalline phase. This demonstrates that the nanocrystals are amorphized by the Er ion beam, and after implantation the grains preserve their shape but remain amorphous even if a thermal process at 900◦ C for 1h is performed. Indeed, in order to completely induce the recrystallization of amorphized Si nanocrystals it is necessary to perform thermal processes at higher temperatures than those needed to crystallize bulk amorphous silicon, and for longer times, as shown in the previos chapter. Finally, for comparison, Fig. 4.4d reports an EFTEM image of a sample annealed at 800◦ C for 1h. Only some initial stages of phase separation with extremely small amorphous Si nanograins are visible in this case. We will discuss in more details the implications of these structural data in Section 4.11.

4.5

Optical Properties

In Fig. 4.5, the room temperature PL spectra of three different samples consisting of Er in presence of Si nanoclusters (solid lines), Er in SiO2 (dashed line) and Si nanocrystals before Er implantation (dash-dotted line) are shown. The Er concentration in the two implanted samples is 2.2x1020 /cm3 and, due to the multiple implants, it is constant almost all over the film thickness. It is worth noticing that all of the spectra have been obtained in the very same conditions, i.e. exciting the systems with the 488 nm Ar-laser line at a pump power of 10 mW and at a chopper frequency of 11 Hz. Moreover all of the spectra have been corrected for the spectral system response, as obtained through a calibrated lamp. Hence the ratio of the intensities at the various wavelengths is an absolute value. A lot of information can be gained from Fig. 4.5. First of all, it is worth noticing that the efficient luminescence at around 0.8 µm due to Si nanocrystals alone completely disappears when introducing Er in the sample. Indeed, an intense peak at 1.54 µm from Er3+ ions appears at the expenses of the Si nanocrystal related emission. Moreover, the 1.54 µm PL intensity due to Er ions in presence of Si nanoclusters is almost two orders of magnitude higher with respect to the sample with Er in SiO2 . As can also be seen in Fig. 4.5, the 1.54 µm luminescence intensity coming from Er in presence of Si nanoclusters is even more intense than the 0.8 µm luminescence signal related to Si nanocrystals without Er. Another important feature shown in

4.6 Quantum efficiency

71

20

5


3

2.2x10 Er /cm

488 nm, 10 mW 11 Hz

4 3

Er and Si nc Er in SiO2 Si nc without Er

2 1

x5

x5

0

0.6

0.8

1.0

1.2

1.4

1.6

Wavelength (m) Figure 4.5. - Room temperature photoluminescence (PL) spectra for unimplanted Si nanocrystals (dash-dotted line), Er in SiO2 (dashed line) and Er in presence of Si nanoclusters (continuous lines). The Er concentration in the two implanted samples is 2.2x1020 /cm3 . All of the spectra have been normalized to the maximum PL intensity of Si nanocrystals.

Fig. 4.5 is the presence of the 0.98 µm line in the spectrum related to the Er-doped Si nanoclusters sample, due to the radiative transition from the second excited state 4 I11/2 to the ground state 4 I15/2 of Er. This is another effect of the presence of Si nanocluster in the matrix, indeed this line is still visible at room temperature even at laser pump powers as low as a few mW, being in that range totally undetectable for the Er-doped silicon dioxide sample, in which Er can be excited only by direct absorption of a resonant photon.

4.6

Quantum efficiency

In order to understand the increase of the Er luminescence intensity caused by the presence of Si nanoclusters, we should first of all find out the main physical parameters affecting the luminescence yield. The luminescence yield y is proportional to the collection efficiency of the optical system, which is independent of the sample and can therefore be skipped in this analysis, to the extraction efficiency, given by the probability an emitted photon has to escape the active material, which is roughly

72


equal to 1/2n2 with n the refraction index, and to the internal quantum efficiency η of the center, which comprises both the probability of an incoming photon being absorbed by the center and the probability of a photon being emitted in the subsequent de-excitation of the same center. Therefore: 1 η (4.1) 2n2 We can define the internal quantum efficiency η of an emitting center as the ratio between the number of internally photons Nem emitted per unit time divided by the number of exciting photons Nex per unit time, i.e.: y∝

Nem (4.2) Nex If A is the area hit by the incoming laser beam, we can focus our attention on a volume V = At of the sample, being t the film thickness, containing a uniform concentration ρ of optically active centers. If the film thickness is much smaller than the typical extinction length of the absorbing material at the excitation wavelength, i.e. t α−1 , being α the absorption coefficient, the flux φ of incoming photons can very reasonably be considered constant. Therefore the number of exciting photons passing through the volume V per unit time equals the number of photons incident on the area A per unit time, i.e. Nex = φA. On the other hand, the number of photons emitted in the volume V per unit time is simply given by the number of excited centers NV∗ in that volume times the probability of a radiative de-excitation 1/τR , being τR the radiative lifetime for the excited center. Therefore: η=

NV∗ /τR (4.3) φA In the linear excitation regime NV∗ σφτ NV , being σ the absorption cross section, τ is the total lifetime, comprising the radiative and non radiative processes, and NV is the total number of optically active centers, which is equal to the concentration ρ of centers in the film times the volume V . Hence eq. (4.3) becomes: η=

η=

στ ρAt τ σφτ ρV = = σρt φAτR φAτR τR

(4.4)

Eventually we can define: ΦEr ≡ ρt

(4.5)

as the total dose of absorbing centers, integrated all over the film thickness. Therefore we get the interesting result: τ (4.6) τR which tell us that fixing the amount of emitting centers, the internal quantum efficiency depends only on the strength of the excitation mechanism, through σ, η = σΦEr

4.7 Excitation mechanisms

73

and on the de-excitation mechanism, through the lifetime of the system τ . As far as the Er ion in presence of Si nanoclusters is concerned, this arguments hold true within the two levels approximation only by using an effective excitation cross section σef f which takes into account excitation mechanisms other than the simple direct absorption of a photon. Nonetheless, eq. (4.6) strongly suggest that the increase in the luminescence yield caused by the presence of Si nanoclusters (see Fig. 4.5) could be attributed either to an enhancement of the absorption cross section or to a suppression of non radiative paths, or, more likely, it could be the result of an interplay between both processes. In the following, we are going to explore in great details both the excitation mechanisms and the non radiative de-excitation processes active in the Er-doped Si nanocrystals system.

4.7

Excitation mechanisms

In the following we are going to investigate the excitation mechanisms occurring in Er doped Si nanoclusters. In particular we will show that the absorption features of Er in presence of Si nanoclusters are very similar to the absorption cross section of Si nanocrystals. This indicates that indeed the excitation of Er is mediated by Si nanoclusters which, after absorbing the incoming photon, are excited and subsequently transfer its energy to the nearby Er ion. This transfer mechanism has to be quite efficient, since the effective absorption cross section for Er in presence of Si nanocrystals has a value very similar to the one found for Si nanocrystals alone.

4.7.1

Photoluminescence excitation spectroscopy

In Fig. 4.6 the PLE spectra of three different samples, normalized to the incoming photon fluxes, are reported in a semilogarithmic scale as a function of the excitation wavelength. The solid and the dashed lines refer to the 1.54 µm line of Er in presence of Si nanoclusters and in SiO2 , respectively. The Er concentration is 2.2×1020 /cm3 for both of the samples, which have been annealed at 900◦ C for 1h after the multiple Er implants. It is worth noticing that for Er in SiO2 , photon emission can occur only at the resonant wavelengths of 378 nm, 488 nm, and 520 nm, corresponding to higher lying excited levels related to Er3+ . Since σ ∝ IP LE /φ, we can deduce that the excitation cross section at 520 nm for Er in SiO2 is ∼10 times higher than the absorption cross section at 488 nm. The presence of Si nanoclusters produces a quite interesting modification in the absorption spectrum of Er. Indeed (solid line) now Er can be excited in a much broader wavelengths range. If we now compare the PLE spectrum of Er in presence of nanoclusters with that obtained for Si nanocrystals emitting at 0.98 µm, we find a strong similarity in the trend over a wide range of excitation wavelengths. This suggests indeed that the Er excitation is occurring via the Si nanoclusters which may absorb the incoming photons over a broad wavelength range and then transfer the energy to a nearby Er ion, by coupling with one of the Er excited levels. A striking difference

74


1.54 m (Er + Si nanoclusters) 1.54 m (Er in SiO2) x 4 0.98 m (Si nanocrystals)

-9

IPLE/ (arb. units)

10

-10

10

-11

10

x4

-12

10

200

300

400

500

600

700

Excitation Wavelength (nm) Figure 4.6. - PLE spectra for three different samples, normalized from the photon flux at each excitation wavelength. The data for Er in quartz have been multiplied by 4.

however exists in the wavelength range between 290 and 420 nm. In this range, indeed, while for undoped Si nanocrystals the absorption cross section increases by one order of magnitude going from 420 to 290 nm, the absorption cross section for Er in presence of Si nanoclusters is seen to saturate in that range. The origin of this difference may reside on the fact that the Si clusters absorbing the incoming light are in an amorphous phase, and could therefore be characterized by an energy level structure quite different from that of crystalline Si clusters. The main point we want to stress here is that through PLE measurements, shown in Fig. 4.6, we have been able to clearly demonstrate that the Er excitation in samples containing amorphous Si nanoclusters (i) does not occur by a direct photon absorption, as is the case for Er in SiO2 , and (ii) mimic very well the features of the absorption cross section for Si nanocrystals alone, suggesting indeed that the main excitation path is an energy transfer from an excited nanocluster, acting as a sensitizer center, to a nearby Er ion.

4.7.2

Effective excitation cross section

In order to quantify the strength of the excitation mechanism, we can perform risetime measurements similar to those performed for Si nanocrystals. Fig. 4.7

4.7 Excitation mechanisms

75


20 3 1.0 2.2x10 Er/cm in Si nc

0.8 300 K 0.6

exc = 488 nm

0.4 0.2 0.0 0

10 mW 2 mW 1 mW 0.2 mW

Increasing pump power

1

2

3

4

5

Time (ms) Figure 4.7. - Risetime measurements at 1.54 µm for a sample of Er-doped Si nanoclusters (nc) and at different pump powers. All the curves are normalized to their steady state value.

shows the risetime of the 1.54 µm PL intensity for different pump powers at an excitation wavelength of 488 nm, for the sample containing 2.2×1020 Er/cm3 . Even in this case, as for Si nanocrystals, the risetime decreases with increasing pump power according to eq. (2.5), where σ now stands for an effective excitation cross section, comprising the absorption of a photon by a Si nanocluster and the subsequent energy transfer to the Er ion. In fact, the reciprocal of the risetime can be reported as a function of photon flux (Fig. 4.8) and shows a linear behavior with a slope given by the effective excitation cross section. From a fit to these data we obtain an excitation cross section σ ∼ 1.1×10−16 cm2 . The intercept of the fit line with the vertical axis corresponds to the Er lifetime (∼ 2 ms). It is now interesting to compare this excitation cross section with the value found for undoped Si nanocrystals. For these Si nanocrystals we have measured an effective cross section of ∼8×10−16 cm2 . However, it should be noted that this value is enhanced with respect to the direct absorption of isolated Si nanocrystals (σ=1.8×10−16 cm2 ) by the energy transfer within the sample. When Er is introduced in the sample the nanocluster-Er interaction is particularly strong. In this case the excited nanocluster preferentially transfers its energy to the Er ion. The Er ion is excited in a high energy state and decays very rapidly in the first excited state (4 I13/2 ). At this

76


2 mW 20

1/on (s-1)

800 700

3

2.2x10 Er/cm in Si nc = (1.1 ± 0.1) x 10 cm -16

2

600 = 2 ms

500 400 0.0

17

5.0x10

18

1.0x10

18

1.5x10 -2 -1

18

2.0x10

Photon flux (cm s ) Figure 4.8. - Reciprocal of the mean risetime as a function of the incoming photon flux. The slope of the solid line fitting the data is a direct measurements of the effective excitation cross section for Er.

stage its energy is too small to be transferred back to the nanocluster and the Er remains excited until the radiative emission occurs. Therefore, in presence of Er ions a transfer of the energy among Si nanocrystals is thought to be less probable. In this case, as we have seen, the excitation cross section of Si nanocrystals is only ∼ 1.8×10−16 cm2 . It is quite interesting to note that the excitation cross section of Si nanocrystals and that of Er in presence of Si nanoclusters are almost identical. This supports the idea that energy transfer from a nanocrystal to Er is extremely efficient. We have also measured in a similar fashion the cross section for Er excitation in pure SiO2 , exciting it with the 488 nm line (which excites Er into the 4 F7/2 level) and we obtained a value of ∼ 1×10−19 cm2 . Hence the presence of nanoclusters enhances the Er excitation efficiency by three orders of magnitude.

4.8

De-excitation mechanisms

In the following we are going to study the main de-excitation mechanisms occurring in the Er doped Si nanoclusters system at very low pump powers, i.e. when only a few Er ions are excited. In particular, we have measured the lifetime and the luminescence intensity at 1.54 µm as a function of the sample temperature, and they are

4.8 De-excitation mechanisms

77

Figure 4.9. - Temperature behavior of the 1.54 µm luminescence intensity for Er in presence of Si nanoclusters, for Er in SiO2 and for Er in crystalline silicon.

shown to vary very little, showing that thermally activated non radiative processes are almost absent, and in any case much less important than in Er doped crystalline Si. However, a non radiative de-excitation process dependent on Er concentration is shown to become active for Er concentrations higher than ∼ 2×1020 /cm3 . This process is well known in other hosts, and it is named concentration quenching effect.

4.8.1

Temperature dependence

In Fig. 4.9 the PL intensity at a pump power of 10 mW versus 1000/T , being T the sample temperature, is reported for Er in SiO2 , and in presence of Si nanoclusters, and for Er in crystalline silicon, at a fixed Er concentration and under the same excitation conditions. The Er luminescence intensity in presence of Si nanoclusters is two orders of magnitude higher than that of Er in SiO2 demonstrating that indeed we are in the presence of an enhanced excitation efficiency (nanoclusters act as efficient sensitizers). Moreover it remains almost constant all over the temperature range spanned in the experiment, decreasing by less than a factor of 4 at room temperature. On the contrary, the luminescence from Er in silicon strongly decrease as the temperature increases, due to Auger with free carriers and to the energy

78


0


10

-1

10

Er doped Si nanocrystals Er doped SiO2

-2

10

0

2

4

6

8

10

Time (ms) Figure 4.10. - Comparison between the decay-time curves of Er in presence of Si nanoclusters and in SiO2 , at room temperature.

back-transfer mechanisms [41, 48–53]. In Fig. 4.10 a comparison between the luminescence decay times at 1.54 µm for Er in SiO2 and in presence of Si nanoclusters is reported. It is shown that the presence of Si nanoclusters produces a reduction of the Er lifetime, from a value of about 10 ms for Er in SiO2 to a value of 2 ms. This reduction is partially due to a change in refractive index in the insulating host, produced by the presence of an excess Si in the matrix. Indeed, the refractive index is seen to increase from a value of 1.45 to 1.7, due to the presence of Si nanoclusters. Since the radiative lifetime is inversely proportional to the refractive index, we get that the radiative lifetime for Er in presence of Si nanoclusters should be reduced to a value of ∼ 8.5 ms, assuming that the lifetime measured for Er in SiO2 is totally radiative. Therefore, any further decrease in the measured lifetime has to be ascribed to non-radiative de-excitation channels available in the matrix containing Si nanoclusters. This leads to a reduction of the Er-related PL intensity by a factor of ∼ 4, with respect to Er in SiO2 . Figure 4.11 shows the time decay data of the 1.54 µm luminescence for Er in presence of Si nanoclusters at two different temperatures, 11 and 300 K. The absence of a marked temperature dependence demonstrates that the processes typically limiting Er luminescence in crystalline Si [48] are strongly reduced in this case, showing


79

0


10

11 K 300 K

-1

10

0

1

2

3

4

5

6

7

8

Time (ms) Figure 4.11. - Er related lifetime at 1.54 µm for two different temperatures.

that in Er doped Si nanoclusters a suppression of non-radiative decay processes occurs. How can we explain this reduction? Since the energy gap of a Si nanocluster is higher with respect to silicon, thermally generated carriers (electron and holes) are hardly produced in the conduction and valence bands, strongly suppressing the Auger effect which can occur more easily in bulk silicon. For the same reason, energy back-transfer from the first excited level of Er to a Si nanocluster is much more difficult to occur with respect to silicon, since now the energy mismatch existing between the Si nanocluster gap and the first excited level of Er is enhanced, due to the enlargement of the Si nanoclusters band gap produced by quantum confinement. The efficiency of the Er doped Si nanoclusters system is therefore the result of both an increase in the effective excitation cross section caused by the enhanced absorption cross section of Si nanoclusters, and a suppression of the non radiative processes.

4.8.2

Concentration quenching effect

Even if the strong non-radiative processes typical for Er in crystalline silicon are suppressed in Er-doped Si nanoclusters, some peculiar de-excitation mechanisms do exist in this system. As an example, Fig. 4.12 reports the time-decay measurements obtained by recording the 1.54 µm luminescence intensity versus time after switching



80

5.4x1019 Er/cm3

0.2 mW

100

20

3

2.2x10 Er/cm

6.6x1020 Er/cm3 1.3x1021 Er/cm3

10-1

0

1

2

3

4

5

6

7

8

9 10

Time (ms) Figure 4.12. - Decay time measurements for the 1.54 µm Er luminescence for samples containing the very same Si nanoclusters and different Er concentrations. The pump power was 0.2 mW.

off the laser beam at time t=0, for samples containing the very same Si nanoclusters and different Er concentrations, after the thermal treatment at 900◦ C. The pump power used is 0.2 mW, and is low enough to guarantee that only a small fraction of Er ions is in the first excited level. The decay time is almost constant up to an Er concentration of 2.2×1020 /cm3 . After this value, the decay time quickly drops down, showing that new non-radiative decay paths, depending on Er concentration, are becoming active in the sample. Actually this phenomenon is well known in standard insulating hosts doped with Er, and it is named concentration quenching effect. A scheme of the process can be found in Fig. 4.13. In this figure, Er in its 3+ optically active state is represented as a three level system. Concentration quenching occurs when two Er ions are so near, see Fig. 4.13, that a dipole-dipole interaction sets in, producing an energy transfer from an excited Er ion to another one in the ground state which is then promoted to the first excited level, Fig. 4.13b). This latter process can occur again and again, thus producing an energy migration all over the sample. This migration ends either when a 1.54 µm photon is emitted, Fig. 4.13c), or when the last excited Er ion non-radiatively transfers its energy to a quenching center, then causing the energy to be completely lost, Fig. 4.13d). This non radiative decay channel is active only at high Er concentration, since the energy transfer among Er ions becomes more probable by reducing the mean Er-Er


81

a)

b)

c)

d)

4

I11/2

4

I13/2

4

I15/2

Figure 4.13. - Energy level scheme representing the concentration quenching effect. (a) The process involves two closeby Er3+ ions, one in the first excited level the other one in the ground state. (b) At a certain distance, a dipole-dipole interaction sets in among the two ions, producing an energy flow from the excited ions to the other one. This process can accur again and again, until the latter excited Er3+ ion emits a photon (c) or non-radiatively transfer its energy to a nearby quenching center (d). The overall process produces a shortening of the first-excited-level decay time.

distance. In particular we can say that while for low Er concentrations only those Er ions directly coupled to the quenching center can see the non radiative path, by increasing the Er concentration over a certain critical value, more and more Er ions start to see the same quenching center due to energy migration. This produces an additional non-radiative recombination path and an increase in the total deexcitation rate, i.e. a quenching of the 1.54 µm luminescence decay time. In our experiment, we found a critical Er concentration of 2.2×1020 /cm3 . If we assume that Er is uniformly and randomly distributed all over the sample, the mean Er-Er distance corresponding to that concentration is ∼ 1.7 nm. Within the concentration quenching model, the probability per unit time wEr for a recombination to occur from the first excited level of Er can be written in the following form: wEr = w0 + 8πCEr−Er Nq NEr

(4.7)

where w0 takes into account both the radiative and non radiative processes when the concentration quenching can be omitted, i.e. at low Er concentration, CEr−Er

82


Er = 2.9 ± 0.2 ms 3.0x10-3

-20

3

CEr-Er Nq = (3 ± 0.5) x 10 cm /s Fit

(s)

-3

2.0x10

-3

1.0x10

0.2 mW

0.0 20

10

21

10 -3

Er Concentration (cm ) Figure 4.14. - Experimental lifetimes of the first excited level of Er3+ as extracted by Fig. 4.12 and as a function of Er concentration, for a sample containing Er-doped Si nanoclusters. Solid line is a fit of the data within the concentration quenching model schematized in Fig. 4.13.

is a constant describing the strength of the Er-Er coupling, and Nq , NEr are the concentrations of quenching centers and of optically active Er ions in the matrix, respectively. We can try to fit the experimental lifetime data extracted from Fig. 4.12 and reported in Fig. 4.14 by using eq. (4.7) with wEr = 1/τ , being τ the experimental lifetime, and w0 = 1/τEr with τEr the lifetime of Er in absence of concentration quenching. The solid line reported in Fig. 4.14 represent the fit to the data, which permits the determination of the following parameters: CEr−Er Nq = 3×10−20 cm3 s−1 and τEr = 2.9 ms. For Er in SiO2 this process has been characterized and a constant CEr−Er = 2.3×10−39 cm6 s−1 has been found [116]. While in SiO2 the quenching centers have been identified with -OH groups, in our case we do not know neither the kind nor the concentration of quenching centers. However, by assuming a constant CEr−Er for Er in Si nanoclusters similar to that of Er in SiO2 we obtain that the concentration of quenching centers is Nq ∼ 1×1019 /cm3 . This value should serve as a mere estimate. Indeed, the matrix could strongly influence the Er-Er interaction strength, thus affecting the concentration quenching coefficient, and the concentration of quenching centers as well. As a brief comment, we would like to add here that the presence of concentration quenching is an indirect, even if qualitative, proof of the fact that a large fraction of

4.9 Experimental quantum efficiency

83

the implanted Er ions is in the optically active 3+ state after the thermal process. Indeed, if this was not the case, i.e. if many Er ions were optically inactive, the energy migration among Er3+ ions, which resides on a resonant energy transfer, would have been impossible, since the energy level scheme reported in Fig. 4.13 wouldn’t be valid for Er which are not in the 3+ state. As a consequence we would have obtained an almost constant lifetime as a function of the total Er content. Kik et al. [106] have indeed observed a constant decay time of 2 ms in samples containing different Er concentrations, up to 1.8 at.%. This findings strongly suggest that the concentration of optically active Er centers in their samples should be quite low, probably because of the sample preparation conditions. This behavior could also explain why the same authors find that Si nanocrystals are able to excite at maximum the 0.02% of the total Er amount, which corresponds to just one Er ion per nanocrystal [105, 106]. In contrast to this, we are going to demonstrate in the following sections of this thesis that under continuos wavelength excitation many Er ions can be sequentially excited by the very same Si nanocluster.

4.9

Experimental quantum efficiency

As a first summary of the experimental findings described up to now, I’d like to recall the arguments treated in Section 4.6, and to draw some preliminary conclusions. In particular we left that paragraph by asking whether the photoluminescence enhancement by two orders of magnitude had to be attributed to an increase in the excitation cross section or an increase in lifetime, or both of them. From the data reported above, we now are aware that Si nanoclusters produce two important effects: (i) on one hand they act as efficient sensitizers for the Er luminescence, determining (at an excitation wavelength of 488 nm) an increase in the effective excitation cross section of Er by almost three orders of magnitude with respect to Er in SiO2 , (ii) on the other hand the presence of Si nanoclusters in the matrix increases the effective refractive index of the material and introduces new non radiative paths in the Er de-excitation, with the result that the ratio τ /τR decreases by a factor of ∼ 4. By using eq. (4.6), we get that the internal quantum efficiency for Er-doped Si nanoclusters should be enhanced by a factor ∼250 with respect to Er in SiO2 . However, since the refractive index is somewhat enhanced, the extraction efficiency 1/2n2 is reduced by a factor (1.7/1.45)2 . Hence the overall luminescence yield y for Er in presence on Si nanoclusters, given by eq. (4.1), is ∼ 180 times higher than for the reference sample of Er in SiO2 . This value can explain the two orders of magnitude increase in photoluminescence intensity caused by the presence of Si nanoclusters. Another comment need to be made here. Indeed, since we have seen that for Er concentrations higher than 2.2×1020 /cm3 a quick drop of the lifetime occurs, due to the concentration quenching effect, the total luminescence yield decreases too. Moreover for higher Er concentrations, another mechanism will be demonstrated to reduce the effective excitation rate of Er at high pump powers. This process, named up-conversion mechanism, is one of the subject of the following sections.

84


m 4

Cb3

0.66

ab

1.54

2

Cb2

Cb1

b

0.80

C3

w54

I9/ 2 4 4 I11/ 2 3 4

w43 Cup

w32

wb wEr

w31

-OH

a Si nc

F7 / 2 H11/ 2 5 4 S3 / 2

4

CA

0.52

0.98

F3 / 2

Cbt

4

I13/ 2 2

4

I15/ 2 1

w21

Er3 +

Figure 4.15. - Energy levels scheme for the system of interacting Si nanoclusters and Er ions. Each Si nanocluster, whether it is amorphous or crystalline, can be represented by a two-effective energy level diagram. Er3+ ions can be thought of as a five-energy-level system. Si nanoclusters having different sizes can be coupled with both the 4 I9/2 and 4 I11/2 levels.

4.10

Modeling the Si nanoclusters-Er interaction

4.10.1

Energy levels scheme

Up to now, the experimental facts can be summarized as follows: i) in presence of Si nanoclusters, Er ions are excited through the nanocluster themselves which act as the absorbing system; ii) the energy transfer occurs quasi-resonantly [108] from the Si nanoclusters to the 4 I9/2 Er manifold for Si nanoclusters emitting at ∼0.8 µm; iii) the effective excitation cross section for Er is enhanced by ∼3 orders of magnitude by the sensitizing action of Si nanoclusters [60], with respect to Er in insulating hosts; iv) the effective excitation cross section for Er in presence of Si nanoclusters is comparable to the absorption cross section of isolated Si nanocrystals, thus attesting the strong coupling between Si nanoclusters and Er; v) the decay channels, typically limiting the efficiency of Er emission in crystalline Si (Ref. [48]), are absent in this case [102, 104, 109]. In order to explain the overall experimental picture, we developed a model for the Si nanoclusters-Er interaction, based on the schematic energy levels diagram reported in Fig. 4.15 [117]. Since from a theoretical point of view it has been shown that the electronic structures of amorphous and crystalline Si nanoclusters are

4.10 Modeling the Si nanoclusters-Er interaction

85

very similar [118], the same arguments valid for Si nanocrystals can be extended to amorphous Si nanoclusters. In this scheme, therefore, a Si nanocluster is represented by a three level system, consisting of two band edge levels and of an interfacial level. Since the trapping is assumed to be a very fast process if compared to the typical decay times, we commit a small error for our purposes in considering the Si nanocluster as an effective two levels system, where the ground state is represented by level a and the excited state by level b. Thus σab is the effective excitation cross section describing the creation of an exciton and its subsequent fast trapping at the interfacial level, following the absorption of a 488 nm photon. wb is the total recombination rate of an exciton for the isolated Si nanocluster, comprising both radiative and non radiative recombination rates. Er3+ is schematized as a five levels system, where the fifth level is the sum of 4 S3/2 , 2 H11/2 and 4 F7/2 , which are strongly overlapping, due to the Stark splitting caused by the matrix field on each multiplet, and can therefore be treated as a single level. With wij we indicate the total transition rate from level i to level j, where i, j = 1, 2, ..., 5 and i > j. Many of the parameters shown in Fig. 4.15 can be determined easily from experiments or theoretically, other deserve a detailed study in order to be determined. One of the most important coefficients is Cb1 , since it describes the coupling between the excited Si nanocluster level and the ground state of Er, and it is therefore responsible for the energy transfer between a Si nanocluster and the rare earths surrounding it. This energy transfer is indeed represented by the two white arrows, depicting the non radiative de-excitation of an excited Si nanocluster (down arrow) and the following Er3+ excitation to the 4 I9/2 level (up arrow). Cbi (i 2) describes the excited state excitation (ESE) from level i. This process is very similar to the excited state absorption of a photon from an excited level of Er in insulating matrices. In this case, ESE describes the re-excitation of an excited Er level through the energy transfer from an excited Si nanocluster. After Er is excited to the 4 I9/2 level, an energy back transfer from this level back to the first excited level of a nanocluster could occur and the strength of this process is given by Cbt . However, the back transfer mechanism is quite ineffective, since level 4 I9/2 is efficiently depleted by a very fast relaxation to level 4 I11/2 . Another effect which we should take into account is very similar to the Auger effect occurring in Er-doped crystalline Si (Ref. [48]), where the energy of an excited level of Er can be given up to a carrier (electron or hole) which is free in the conduction or valence band of bulk Si, promoting it to a higher lying level. In the Er-Si nanoclusters system, it could happen that the energy can be transferred from an excited Er level back to a confined exciton, thus promoting it to a higher energy level. This process is represented by the constant CA , taking into account the Auger process with both the electron or the hole forming the exciton. Clearly, this Auger process, in order to occur, requires that a new exciton is formed in the nanocluster while an excited Er level is still filled. Therefore it becomes effective only at high pump powers and it mainly involves the long lived 4 I13/2 level. We also take into account the concentration quenching effect, which is due to

86


the energy migration all over the sample caused by the energy transfer between two nearby Er ions, one in the first excited level and the other in the ground state. Concentration quenching has been well characterized in a previous section, and therefore the value of the total rate of de-excitation from the first excited Er level, i.e. w21 + wEr , where wEr takes into account the concentration quenching effect, is well known. Eventually, the constants Cup and C3 are the cooperative up-conversion coefficients describing the interaction of two nearby Er ions which are both in the first or in the second excited states, respectively. In the first case one of the two ions will return to the ground state giving its energy to the other which will be excited to the 4 I9/2 level. In the second case the interaction will bring one ion to the 2 H11/2 level and the other to the ground state. In the model we take also into account the possibility of a direct absorption of a 488 nm photon by Er leading to a transition from the ground state to the 4 F7/2 level. This process is characterized by an excitation cross section of ∼1x10−19 cm2 , as determined in an oxide reference sample. Within this scheme, we are able to write down the set of first order rate equations describing the time evolution of the concentration of Si nanoclusters and Er ions in each level: dnb = σab φna − wb nb − Cbi nb Ni dt i=1 3

dna = −σab φna + wb nb + Cbi nb Ni dt i=1 3

dN5 Cbi nb Ni + C3 N32 + = σφN1 + dt i=2 3

− (w51 + w54 ) N5 dN4 = Cb1 nb N1 + Cup N22 + w54 N5 − w43 N4 dt dN3 = w43 N4 − (w32 + w31 ) N3 − Cb3 nb N3 + dt −2C3 N32 dN2 = w32 N3 − (w21 + wEr ) N2 − 2Cup N22 + dt −Cb2 nb N2 − CA nb N2 dN1 = (w21 + wEr ) N2 + Cup N22 + CA nb N2 + w31 N3 + dt +C3 N32 + w51 N5 − σφN1 − Cb1 nb N1

(4.8)

where φ is the flux of photons incident onto the sample, na,b and Ni , with i 1, are the density level populations of Si nanoclusters and of Er ions, as represented in Fig. 4.15. In particular, we can define na +nb = n0 and i Ni = N0 , being n0 and N0


87

the total concentrations of excitable Si nanoclusters and Er ions, respectively. Here we assume that each Si nanocluster contains at maximum one exciton at a time, since the Auger effect between two excitons would lead to a quick (100 µs, as shown in Fig. 4.18. But due to the strong interaction between a Si nanocluster and the nearby Er ions, the lifetime of the exciton in the Si nanocluster has to be much lower than the exciton lifetime in absence of Er, which is ∼ 50 µs. Therefore, once the excitation is switched off, an exciton in presence of Er ions can survive in the nanocluster for times 50 µs, and it is only in this narrow temporal window that the interaction leading to an Auger quenching process with the first excited level of Er would occur. Therefore, an Auger process would have explained a luminescence quenching at 1.54 µm for times 50 µs but cannot explain the quenching reported in Fig. 4.18, since it can be observed even for times >100 µs, i.e. much longer than the typical lifetime of an exciton interacting with Er. Another direct evidence of the presence of cooperative up-conversion can be envisaged in Fig. 4.20, where a decay-time measurement at the 0.98 µm Er emission is reported, for the very same sample containing 6.5x1020 Er/cm3 and by using a pump power of 66 mW. The 0.98 µm line coming from the radiative recombination


= 0.98 m

0

10


93

66 mW 10-1

-2

10

0

50

Time (s) Figure 4.20. - Time-decay curve of the 0.98 µm emission for a sample containing 6.5x1020 Er/cm3 in presence of Si nanoclusters. The tail which extends for times longer than 10 µs is due to the re-filling of the 4 I11/2 level caused by cooperative upconversion between two Er ions excited both in the long lived 4 I13/2 level. The dashed line is a single exponential fit of the first part of the curve, from which a value of 2.5 µs is found.

of level 4 I11/2 is characterized by a lifetime of ∼ 2.5 µs, attesting that most of the de-excitations from that level are non radiative. The decay-time curve reported in Fig. 4.20 is composed by a first straight part, showing a lifetime of just ∼ 2.5 µs, and of a very long tail extending over 10 µs and characterized by a much longer lifetime. Indeed, this long tail is due to the refilling of the 4 I11/2 at the expenses of level 4 I13/2 determined by the cooperative up-conversion involving two Er ions in the first excited level. Since this level is long-lived, two excited Er ions can interact through up-conversion even at times longer than a few µs, as reported in Fig. 4.18, thus repopulating level 4 I11/2 , which otherwise would have been depleted in a mean time of 2.5 µs.

4.10.3

Effects of up-conversion on the excitation rate of Er

Since we are eventually interested in the emission of a 1.54 µm photon, we want to better investigate the real effect of up-conversion in the dynamics of the 4 I13/2 level

94


= 1.54 m


1.0 0.8 0.6

0.06 mW 0.7 mW 6.4 mW 66 mW 618 mW

0.4 0.2 0.0

0

1

2

Time (ms)

3

Figure 4.21. - Time resolved PL intensity at 1.54 µm as a function of the excitation pump power, normalized to the steady state values.

population. Fig. 4.21 shows the normalized luminescence intensities recorded at 1.54 µm as a function of time after switching on the laser beam at t = 0, and for different pump powers. As can be observed, the luminescence signal reaches the saturation value in a time which is shorter the higher is the excitation power. This is now shown for a sample containing a higher Er concentration (i.e. 6.5×1020 /cm3 ) than in Fig. 4.7, in which the Er-Er interaction is more effective, and in a pumping power regime which is orders of magnitude higher. We define the typical experimental risetime τon as the time it takes the luminescence signal to reach the 63% (i.e. 1 − e−1 ) of the saturation value. In Fig. 4.22, the reciprocal of the experimental risetime (solid circles) extracted from Fig. 4.21 is plotted as a function of pump power. The experimental trend is linear up to 10 mW. Indeed, within this low power regime the reciprocal of the risetime τon follows the law: 1 1 = σeff φ + τon τ

(4.10)

and from a linear fit of the experimental data through eq. (4.10) a value of ∼2x10−16 cm2 can be estimated for the Er excitation cross section of this sample. At higher pump powers, the linear approximation is no more valid, and indeed the


95

= 1.54 m

9000

= 2x10-16 cm2

8000

1/on (s-1)

7000 -17

6000

3

Cup= 7x10 cm /s

5000 4000 3000 2000

1000 0

0

Simulation Experiment

100 200 300 400 500 600

488 nm Pump Power (mW) Figure 4.22. - The reciprocals of the experimental risetimes measured at 1.54 µm are reported as a function of pump power, as determined from Fig. 4.21. From a linear fit of the data in the low pumping power regime (up to 10 mW) it is possible to estimate an excitation cross section of ∼ 2x10−16 cm2 for the Er excitation mediated by Si nanoclusters. At higher powers, a strong saturation is observed, due to cooperative up-conversion which limits the excitation rate of Er.

trend of the experimental data in Fig. 4.22 shows a strong saturation, which can be attributed to the up-conversion mechanism, as indicated by the good agreement of the simulated data (open circles) obtained by solving eq. (4.8). Another mechanism which could in principle limit the excitation rate of Er is clearly the transfer mechanism itself. Indeed, given a certain Er concentration, the total number of excitable Er ions depends clearly on the excitation power, but at very high pump powers the effective energy transfer time between a single Si nanocluster and each Er ion becomes the physical limiting factor.

4.10.4

Transfer-time estimate

In Fig. 4.23, the luminescence intensity at 0.98 µm is reported as a function of time, normalized to the steady state values, for two very different excitation powers. The mean risetime is clearly independent of the pump power, being equal to 2.4 µs, which is comparable to the decay time of the 4 I11/2 level. Since this level is directly pumped by a Si nanocluster through a fast decay from level 4 I9/2 , we can conclude that the time it takes a Si nanocluster to transfer its energy to a nearby Er ion must have a

96



1.2

= 0.98 m

1.0 0.8 0.6

6.6 mW 618 mW

0.4 0.2 0.0

0

5

10

Time (s)

15

20

Figure 4.23. - Photoluminescence intensities at 0.98 µm as a function of time, after switching on the laser excitation at t=0 for the sample containing 6.5x1020 Er/cm3 in presence of Si nanocluster. The mean risetime is independent of the pump power excitation, suggesting that the energy transfer time between Si nanocluster and Er is actually comparable to the lifetime of level 4 I11/2 , i.e. of the order of ∼1 µs. The time resolution of our set up is ∼5 ns.

typical value of τtr ∼ 1µs. This value is in agreement with measurements performed by Fujii and collaborators on co-sputtered Er-doped Si nanocrystals samples [108]. Since the transfer time is finite, this means that even if the excitation power is such that more than one exciton is created in each nanocluster in a time interval of 1 µs, only one of these excitons can transfer its energy to the nearby Er ion in that time interval. Therefore increasing the excitation power over a critical value Pc cannot produce a corresponding increase in the excitation rate, thus contributing to the saturation observed in Fig. 4.22. In order to estimate the value of the critical power Pc , we need to equal the excitation rate σab φ of a nanocluster and the transfer rate wtr = 1/τtr . By using an excitation cross section of 2x10−16 cm2 and a transfer time of 1 µs, we get for the critical photon flux a value φc = (σab τtr )−1 =5x1021 cm−2 s−1 . From this critical photon flux value, it is possible to estimate a critical pump power of Pc = 5.7 W. It is evident from Fig. 4.22 that the pump power used is well below the critical value Pc . Therefore we conclude that the strong saturation in the


97

excitation rate deduced from Fig. 4.22 is mainly due to the upconversion mechanism. At this stage, a few comments need to be made, as far as the efficiency of the energy transfer mechanism itself is concerned. Indeed we can define the efficiency ηtr of the transfer mechanism as the ratio between the transfer probability wtr , i.e. the probability per unit time that an exciton generated inside a Si nanocluster gives out its energy to a nearby Er ion in the ground state, and the total recombination probability of the same exciton, comprising both the recombination rate wb for the isolated nanocluster and the transfer probability. Therefore we get: ηtr =

wtr wtr + wb

(4.11)

By using the experimentally estimated value of ∼1 µs for τtr , and of 2×104 s−1 for wb we get a value of ∼ 98 % for the energy transfer efficiency, which indeed confirms the strong coupling existing between Si nanoclusters and Er ions.

4.10.5

Implication of the finite transfer time

In this paragraph we are going to show some important implications of the finite transfer time through a stringent comparison between pulsed and continuous excitation. The PL measurements were carried out at room temperature under pulsed and continuous wavelength (cw) pumping. The pulsed measurements were performed at the Van der Waals-Zeeman Institute of the University of Amsterdam by the group of Prof. Tom Gregorkiewicz. As a pulsed source, a tunable Optical Parametric Oscillator (OPO) was used, producing pulses of 5 ns duration at a 20 Hz repetition rate. The cw measurements were performed in Catania. For cw pumping, the 514.5 nm and 476.5 nm lines of an Ar+ laser were used. Figure 4.24 presents the dependence of PL intensity (at λ = 1.53 µm) under pulsed excitation as a function of excitation flux. Displayed sets of points correspond to: (a) Er-doped SiO2 excited at λexc = 520 nm; (b) Er-doped Si nanoclusters excited at λexc = 520 nm, where both indirect and direct excitation channels are possible, and (c) Er-doped Si nanoclusters excited at λexc = 510 nm (indirect excitation is only possible). The measurements were performed with the same experimental settings, thus the PL intensity scale is common for all the data points. As can be seen, the indirectly excited Er3+ emission from Er in presence of Si nanoclusters (c) saturates at the highest photon fluxes used in the experiments. This saturation level can be exceeded when also the direct excitation channel of Er3+ ions is enabled, by setting the OPO to λexc = 520 nm (b). For smaller fluxes, much before saturation, both samples show an approximately linear dependence, as shown in the inset to Fig. 4.24, with the emission from Erdoped Si nanoclusters being higher than the Er-doped SiO2 sample. From Fig. 4.24 we conclude that the Er doped SiO2 emission shows a linear dependence over the whole investigated flux range. Hence, the Er-doped SiO2 system can be used to attribute the PL intensity to a particular concentration of excited Er3+ ions, as given by the upper right hand scale in Fig. 4.24. Since the Er ab-

98


Figure 4.24. - Integrated photoluminescence intensity at λ = 1.53 µm as function of the pulsed excitation density: (a) Er-doped SiO2 excited at λexc = 520 nm; (b) Er-doped Sinanoclusters excited at λexc =520 nm; (c) Er-doped Si nanoclusters excited at λexc = 510 nm. The right hand scale shows the concentration of excited Er3+ ions in Er-doped SiO2 (upper part) and Er-doped Si nanoclusters (lower part). The error in the values obtained by such a procedure is a factor of 10. The inset presents a detail for the low flux regime. Work performed in collaboration with the group of prof Prof. Tom Gregorkiewicz, Van der Waals-Zeeman Institute, Amsterdam.

sorption cross section for the process of direct absorption of a photon is very low, the experimental value we found is affected by errors. Moreover, even the photon flux is known with some imprecision, due to the error in determining the spot size of the laser beam. Hence, the overall procedure, which permits us to convert PL data in concentration of excited Er ions, is affected by a total error which is taken into account by the variability of the factor multiplying the right hand scale. Although the PL intensity scale is common for all the data points in Fig. 4.24, the Er3+ excited state population should be corrected for the Si-nanoclusters-induced change in radiative and total lifetimes, which reduces the τ /τrad ratio by a factor of ∼ 4. Consequently, the excited state population in the Er-doped Si nanoclusters system has to be 4 times higher in order to give PL intensity equal to that of the Er-doped SiO2 sample, in the same excitation conditions. This correction is included in the lower part of the right hand scale in Fig. 4.24. We now conclude that the PL saturation observed for the Er-doped Si nanoclusters sample under indirect ex in the range 2×1018 excitation corresponds to an excited Er3+ concentration NEr 2×1019 /cm3 , which is only a minor part of the total Er3+ amount present in the

99

3

10

1021 x 0.82

102

1020

101

(c)

100

(b)

1019 1018

cw Pumping

Er-SiO = 1x10 cm -19

-1

10

10-2

2

2

1017

(a) 17

10

18

10

19

10

20

-2

10 -1

Photon Flux (cm s )

Excited Er Concentration (cm-3)

1.53 m PL Intensity (arb. units)


21

10

Figure 4.25. - Photoluminescence intensity at λ = 1.53 µm as a function of the continuous excitation density: (a) Er-doped SiO2 system excited at λexc = 514.5 nm; (b) Er-doped Si nanoclusters excited at λexc = 514.5 nm (in resonance); (c) Er-doped Si nanoclusters excited at λexc = 476.5 nm (out of resonance); the right hand scale shows the concentration of Er3+ ions in the first excited state attributed to Er-doped SiO2 . Due to the radiative and total lifetime reduction in presence of Si nanoclusters, the concentration scale for Er-doped Si nanoclusters should be corrected by a factor ∼0.82, as indicated in the figure.

sample (2.2×1020 /cm3 ). It should be noted that the obtained saturation level is comparable with the estimated concentration of Si nanoclusters in the Er-doped Si nanoclusters system. This is a quite important point, since, while we have previously demonstrated that the excitation proceeds via Si nanoclusters, it could be possible that the saturation of the luminescence emission is directly due to Si nanoclusters rather than Er3+ ions. Indeed, with the photon flux of φ = 1 × 1025 cm−2 s−1 we get for the excitation rate of a nanocluster a value of σφ = 109 s−1 , where we have used the measured value σ = 1 × 10−16 cm2 for the excitation cross section of a Si nanocluster. Therefore, during the 5 ns of the laser pulse each nanocluster is excited 5 times. It cannot, however, accumulate generated excitons due to a strong Auger effect which rapidly (in times of the order of ∼1 ns) reduces their number. We note that although the nanocluster can transfer an exciton to a nearby Er3+ ion, the transfer time is finite and of the order of a microsecond, as determined in this thesis [117] or in other works [108], i.e. much longer than the Auger time constant. This makes the Er3+ excitation process noncompetitive with Auger quenching. At the end of the laser

100


pulse, only one exciton per nanocluster is left to transfer its energy to a nearby Er3+ ion. Hence, due to this mechanism, under high fluxes and pulsed pumping conditions the concentration of excited Er ions cannot exceed the concentration of Si nanoclusters. In that way the present measurements provide a unique possibility of direct experimental determination of the concentration of sensitizer centers. The above outlined PL saturation mechanism will not occur under cw pumping, where temporal limitations are of no importance and each nanocluster can undergo multiple excitations during the illumination time which is much longer than the Er lifetime (∼2 ms) and the finite transfer time (∼1 µs). We hence performed cw measurements in and out of resonance using an excitation wavelength of 514.5 or 476.5 nm of an Ar+ ion laser, respectively. The data reported in Fig. 4.25 show that the 1.53 µm PL intensity in the Er-doped Si nanoclusters system is two orders of magnitude higher than for Er in SiO2 at very low photon fluxes, and increases linearly with photon fluxes up to φ ≈ 5 × 1018 cm−2 s−1 . For higher flux values, a sublinear dependence is observed, due to the introduction of new non radiative processes, such as cooperative up-conversion, but no saturation is reached within the investigated range, up to φ = 1021 cm−2 s−1 . The higher PL intensity at λexc = 476.5 nm is ascribed to the larger absorption coefficient of Si nanoclusters at shorter excitation wavelengths, as can be seen in Fig. 4.6 (solid line). Since the PL intensity from Er-doped SiO2 is a linear function of the photon flux, similarly as for the pulsed experiment, we can use this dependence to determine the concentration of the excited Er3+ ions in Er-doped Si nanoclusters. Indeed, under cw excitation, in steady state conditions and in linear regime, the fraction of excited Er ions is simply given by the number of excitations per unit time (σφ) divided by the number of de-excitation per unit time (1/τ ). In order to get the concentration of excited Er ions we have simply to multiply this fraction for the total Er concentration. For example, by using for Er in SiO2 an experimentally determined cross section of 1×10−19 cm2 for direct absorption at 514.5 nm, and a measured lifetime τ = 10 ms, at an excitation photon flux of 1018 cm−2 s−1 the fraction of excited Er ions is 1×10−3 . Since the total Er concentration is 2.2×1020 /cm3 , the excited Er concentration is therefore 2.2×1017 /cm3 . The same procedure can be repeated at each photon flux, and the right scale of Fig. 4.25 is readily obtained. By comparing the PL intensities of the two samples (which are proportional to the number of excited Er ions divided by the respective radiative lifetime), and knowing the decrease in radiative lifetime produced by the increase in refractive index, we can obtain the excited Er3+ ions in Er-doped Si nanoclusters too (the right hand scale of Fig. 4.25 has to be multiplied by a factor ∼0.82, in order to take into account the variation in radiative lifetime). In that way we conclude that the concentration of the Er3+ ions which can be excited in the Er-doped Si nanoclusters under cw pumping exceeds the saturation level realized under pulsed pumping, by more than an order of magnitude. Actually, this demonstrates that under cw pumping many Er ions can be excited, even under off-resonance excitation conditions (by exciting at 476.5 nm), demonstrating that


101

indeed each Si nanocluster can excite more than one Er ion. The fact that very high concentration of Er3+ ions can be excited through Si nanoclusters is also confirmed by the presence of up-conversion, which has been demonstrated to be active in this material in a previous paragraph. Indeed, up-conversion couldn’t occur if only a few Er ions were excited [117] under high pumping conditions. These findings provide the most direct illustration of the sequential character of the Si nanoclusters sensitized Er ions. Under pulsed excitation and high-power pumping each sensitizer can transfer energy only to one Er3+ ion, and the excited state population stabilizes at a level equal to the concentration of Si nanoclusters acting as sensitizers. Such a measurement allows direct experimental determination of the total concentration of sensitizing nanoclusters. On the other hand, under intense cw pumping, several Er3+ ions can be excited by the same Si nanocluster.

4.10.6

Number of excited Er ions per nanocluster

In this and following paragraphs we want to better investigate problems and perspectives related to optical gain in the Er-doped Si nanoclusters system, by simply extending the simulations obtained through eq. (4.8). Indeed, in previous paragraphs, the overall experimental data have been fitted through eq. (4.8), accurately determining all of the main physical parameters involved. The physical variables that have been varied in the investigation are the Er concentration (ranging between 3×1017 /cm3 and 1.4×1021 /cm3 ) and the excitation pump power (in the range 1-103 mW). The density of Si nanoclusters in the sample has been fixed to ∼1019 /cm3 . Many parameters can determine and influence a possible optical gain measurement at 1.54 µm in Er-doped Si nanoclusters. Among these, the number of excitable Er ions per nanocluster clearly deserves great attention. Indeed, in order to reach high material gain values, high concentrations of Er need to be inserted in the sample and moreover in their optically active state. Clearly, if only a small fraction of Er ions is coupled with nanoclusters, all the Er ions which do not benefit from the sensitizing action of nanoclusters become strongly absorbing. Under these conditions, measuring a positive net gain at 1.54 µm could become a quite difficult task to achieve, since other processes such as photon absorption from a ground state Er3+ ion, or excited state absorption from the first level of Er will dominate. From an experimental point of view, at least two facts demonstrates that more than one Er ion can be excited by successive excitation of a single nanocluster: (i) the presence of up-conversion mechanism in steady state cw pumping of the Erdoped Si nanoclusters sample, (ii) a quantitative estimate of the concentration of excited Er ions reported in Fig. 4.25 by converting photoluminescence intensity data, and demonstrating indeed that at the maximum pump powers, even using out of resonance excitation, i.e. through the energy transfer mechanism from nanoclusters, almost all of the Er ions present in the sample can be excited. In Fig. 4.26 a contour plot reports in gray scale the simulated number of Er ions excited in the 4 I13/2 level per Si nanocluster, in steady state conditions, as a function of both the Er concentration and the power density, obtained by dividing the pump

102


488 nm Power Density (W/cm2)

Number of Excited Er ions per nc 3

10

31 -- 45 21 -- 31 14 -- 21 9.8 -- 14 6.7 -- 9.8 4.6 -- 6.7 3.1 -- 4.6 2.1 -- 3.1 1.5 -- 2.1 1.0 -- 1.5

2

10

1

10

10, as indicated in Fig. 4.30 through an arrow. In Fig. 4.31a), the Er induced gain g (up-triangles), the Si nanocluster induced losses αcca (down-triangles) and the net gain gnet = g − αcca (close circles) are shown as a function of the excitation power, for an Er concentration of 2.4x1020 /cm3 . For very high pump powers, the simulated net gain is 0.24 cm−1 . This positive gain is a result of the sensitizing action played by Si nanoclusters in exciting Er ions. Indeed, since each nanocluster is able to excite more than one Er ion, the material gain can be so high to overcome the Si nanoclusters induced losses. Moreover, due to the strong interaction between nanoclusters and Er, the energy transfer time can be so fast to strongly reduce the fraction of excited Si nanoclusters, thus lowering the nanoclusters induced losses. Higher gain values could be obtained increasing the coupling coefficient between a nanocluster and Er and reducing the up-conversion mechanism, by playing both with the material used and with the preparation conditions. However, an accurate determination of the confined carriers absorption cross section is still necessary, since as reported in Fig. 4.31b) the actual value of this parameter strongly influences the possibility of reaching net gain in Er doped Si nanoclusters. Indeed, from the figure it can be noticed that the system can have quite high gain values (∼ 10 cm−1 ) if σcca =1x10−19 cm2 , while it turns to be strongly absorbing if the CCA cross section tends towards the bulk-Si value of 1x10−17 cm2 . The recent experimental observation [111–113] of net optical gain in the system is a clear proof that CCA should be quite reduced with respect to the bulk Si value.

4.11

Increasing the luminescence yield

Even if there’s no complete agreement in literature about the exact law describing the energy transfer mechanism between Si nanoclusters and Er ions, i.e. whether it is resonant dipole-dipole interaction [120] or direct electron exchange [121], it is experimentally known that the interaction range has to be of the order of 1 nm [122]. This poses a limit on the minimum concentration of sensitizers and emitting centers that have to be inserted in the sample for the interaction to become active. Moreover, it is known that, within a certain size distribution which guarantees the resonance condition for the energy transfer, the smaller is a nanograin, the stronger is the interaction mechanism, i.e. the energy transfer efficiency [123]. In the following we are going to explore a new processing of the material, able to increase the light emission yield of Er in presence of Si nano-aggregates, while avoiding the high temperature (low ULSI compatible) annealing step at 1250◦ C needed to form well formed and crystalline Si nanoclusters.

110


35

as deposited Pre-annealed at 500 °C Pre-annealed at 800 °C Pre-annealed at 1000 °C Pre-annealed at 1250 °C

PL Intensity (a.u.)

30 25 20 15

Post-annealing 900 °C 1 h

10 5 0 1450

1500

1550

1600

1650

Wavelength (nm) Figure 4.32. - Room temperature PL spectra taken by exciting with a 10 mW laser SiOx samples pre-annealed at different temperatures and then implanted with 5×1014 Er/cm2 . The post-implantation thermal process was 900◦ C for 1 h for all of the samples.

The substoichiometric SiOx layers with 42 at.% Si, 0.2 µm thick, deposited by PECVD, after deposition were annealed at different temperatures in the range from 0-1250◦ C for 1 h under N2 atmosphere. This first thermal step will be referred to as the pre-annealing process. All of these samples were then implanted with 300 keV Er ions to a dose of 5×1014 /cm2 . After implantation, a final thermal treatment at 900◦ C for 1 h under N2 atmosphere, referred to as the post-annealing process, was performed in order to activate Er. Figure 4.32 reports the room temperature PL spectra of Er-doped SiOx layers preannealed at different temperatures in the range 0-1250◦ C prior to Er implantation. The temperature of the post-annealing process was 900◦ C for all of the samples. PL intensity at 1.54 µm is identical for the SiOx samples as-deposited and preannealed at 500◦ C, increases for a preannealing of 800◦ C, and decreases again for higher pre-annealing temperatures. In particular, for a pre-annealing process of 1250◦ C, the intensity at 1.54 µm is about a factor of 5 lower with respect to the maximum PL signal observed. The reasons for the observed enhancement can be envisaged in the excitation and/or the de-excitation processes present in the samples.

1.54 m Normalized PL Intensity

4.11 Increasing the luminescence yield

111

Pre-annealig: 800°C 1250°C

0

10

-1

10

post annealing: 900°C for 1h 0

1

2

3

4

5

6

7

8

9 10

Time (ms) Figure 4.33. - Decay time measurements for samples pre-annealed at different temperatures.

We have measured the Er luminescence decay lifetime at 1.54 µm in samples preannealed at different temperatures by exciting Er at steady state and recording the PL intensity as a function of time after switching off the laser beam at time t=0. The time-decay curves, shown in Fig. 4.33, are identical in all of the samples, independently of the preannealing treatment, with a lifetime of about 2 ms, indicating that the difference in the PL signals cannot be attributed to a different contribution of nonradiative decay centers. As far as the excitation is concerned, we have measured the Er excitation cross section in the sample pre-annealed at 800◦ C, for which the PL intensity at 1.54 µm has a maximum, and in the sample pre-annealed at 1250◦ C, the one with the lower PL intensity. In this case the measured values, obtained in Fig. 4.34, are very similar. We found a value of 2.5×10−16 cm2 for the sample pre-annealed at 800◦ C and 1.4×10−16 cm2 for the sample pre-annealed at 1250◦ C, and therefore can only partially account for the observed difference in the PL signals. In order to understand the origin of the higher luminescence yields in the Erdoped SiOx samples pre-annealed at lower temperatures, we performed PLE measurements on the samples by using the 150W Xe lamp at a detection wavelength of 1.54 µm. The results are reported in Fig. 4.35. It is worth noticing that all the curves show a monotonic increase of the PL intensity as the excitation wavelength

112


1/on (s-1)

700

= 2.5x10 cm -16

Pre-annealing: 800°C 1250°C

2

600

= 1.4x10 cm -16

2

500

1/on = + 1/ 400 0.0

17

5.0x10

18

-2 -1

1.0x10

Photon Flux (cm s ) Figure 4.34. - Determination of the excitation cross section at 488 nm of excitation wavelength.

is decreased, up to 340 nm, where a maximum occurs. This maximum is determined by an interference effect caused by the high refractive index mismatch between the film (n=1.7) and the Si substrate (n=3.5). The fact that resonant features are absent for all the curves, ensures that Er excitation is mediated by sensitizing centers, independently of the pre-annealing temperature. However, the pre-annealing temperature does affect the absolute intensity scale. Indeed, the PL intensities for the standard sample of Er-doped Si nanocrystals obtained by a pre-annealing temperature of 1250◦ C, are the lowest all over the excitation wavelengths range. Moreover, the shapes of the PLE spectra are very similar for all the samples pre-annealed at temperature lower than 1000◦ C. We can enhance the main differences by plotting in Fig. 4.36 the PLE spectra of Fig. 4.35 normalizing them to the curve relative to the sample pre-annealed at 1250◦ C. Indeed this last sample is somehow a reference sample, since it has been well characterized and since we know that the Si/SiO2 phase separation is complete, as attested by the presence of well formed high density Si nanocrystals after the pre-annealing process, see Fig. 4.4a). The striking difference between the samples pre-annealed at lower temperatures and the 1250◦ C-pre-annealed sample emerges clearly in Fig. 4.36. Indeed a broad absorption peak centered at around 450 nm and characterized by a FWHM of ∼ 100 nm is present in all of the samples preannealed at temperatures ≤1000◦ C. The maximum peak intensity is obtained for a pre-annealing temperature of 800◦ C. To understand the origin of this peak, we

4.11 Increasing the luminescence yield

-12

IPLE/ (arb. units)

10

rev = 1534 nm

10-13

113

Pre-annealing: as dep. 500°C 1h 800°C 1h 1000°C 1h 1250°C 1h

-14

10

Post annealing: 900°C

300

400

500

600

Excitation Wavelength (nm) Figure 4.35. - Semi-logarithmic plot of the PLE intensity measured at 1.54 µm, normalized for the incoming photon fluxes, for the SiOx films pre-annealed at different temperatures, implanted with 5×1014 Er/cm2 and eventually annealed at 900◦ C for 1 h. The feature at 340 nm, characterizing all of the PLE spectra, is an interference effect caused by the high refractive index mismatch existing between the film and the Si substrate.

should at first focus our attention on the structural properties of this sample. Figure 4.4d) reports an EFTEM plan view image of the sample pre-annealed at 800◦ C (this is the sample for which the luminescence signal at 1.54 µm is higher). It is possible to distinguish a certain degree of phase separation with very small Si grains. Note that the magnification of this image is doubled with respect to the other ones. A careful analysis of the image reveals that the clusters are not very well separated, but they seem to be inter-connected in a Si network. This effect is due to the fact that the phase separation process is not completed, as a result of the low temperature used for the thermal process. Note that the same kind of structure is not present in as deposited SiOx films, whose EFTEM images are characterized by the absence of any relevant contrast. Moreover, the diffraction pattern (not shown) provides no evidence for the presence of the crystalline phase. The feature at around 450 nm in Fig. 4.36, could therefore be attributed to some kind of absorbing centers which could be identified with the Si atoms forming the connected network, or to other defects in the heterogeneous matrix, which anyway dissolve at higher annealing temperatures. Hence, the PL intensity is stronger in samples pre-annealed at lower

114


IPLE/ IPLE,1250°C

6

rev = 1.54 m

Pre-annealing: as dep. 500°C 1h 800°C 1h 1000°C 1h 1250°C 1h

5 4 3 2

Post annealing: 900°C

1 300

400

500

600

700

Excitation Wavelength (nm) Figure 4.36. - PLE Intensity normalized to the sample pre-annealed at 1250◦ C.

temperatures and for excitation wavelengths in the range 400-500 nm because in this range there are more numerous absorbing centers that can absorb the energy from the impinging laser beam and transfer it to a nearby Er ion.

4.12

Conclusions

In conclusion, we have presented a detailed experimental study of the Er-doped Si nanoclusters system. A phenomenological description of the Er-doped Si nanoclusters system able to quantitatively describe the measured optical properties has been developed. By introducing a rate equation based formalism, describing the density populations of interacting Si nanoclusters and Er levels, we were able to determine both the nanocluster-Er coupling constant and the up-conversion coefficient, through a fit of the overall experimental data. An energy transfer time of ∼1 µs has been experimentally estimated. It has been demonstrated that each Si nanocluster can be coupled with more than one Er ion, the maximum number of excitable Er ions per nanocluster being only limited by the Si nanocluster excitation rate under cw excitation, by the total Er concentration, given a fixed nanocluster

4.12 Conclusions

115

concentration, and eventually by the finite transfer time under pulsed excitation. The possibility of observing positive gain at 1.54 µm in such a system has been extensively discussed, with a particular attention to gain limiting effects, such as cooperative up-conversion and confined carriers absorption (CCA) induced by excited nanoclusters. Indeed we have shown that, as far as the optical gain at 1.54 µm from Er-doped Si nanoclusters is concerned, Si nanoclusters can have positive effects, i.e. the increase of the effective excitation cross section of Er and the lowering of the pump threshold for population inversion to occur, as well as negative aspects, i.e. the confined carrier absorption of 1.54 µm signal photons. Nevertheless, under appropriate conditions net optical gain can be reached. In the end we have reported a new material processing, essentially based on lower temperatures, i.e. more ULSI compatible, pre-annealing processes of the SiOx matrix, able to produce a higher concentration of sensitizing centers in the visible range (at around 450 nm) and generating, as a consequence, a 5-times increase in the overall Er luminescence yield in that range, which in definitive leads to quantum efficiencies 500 times higher than for standard samples of Er-doped SiO2 .

Chapter 5 Devices based on Er-doped Si nanoclusters

The electroluminescence (EL) properties of Er-doped Si nanoclusters embedded in metal-oxide-semiconductor devices are investigated. It is shown that, due to the presence of Si nanoclusters dispersed in the SiO2 matrix, an efficient carrier injection occurs and Er is excited producing an intense 1.54 µm room temperature electroluminescence. The EL properties as a function of the current density, temperature and time have been studied in details, elucidating the radiative and non-radiative de-excitation properties of the system. We have also estimated the excitation cross section for Er under electrical pumping finding a value of 1×10−14 cm2 . High quantum efficiencies can therefore be obtained from these room temperature devices. These data will be presented and the impact on future applications discussed.

5.1

Introduction

Silicon has been the leading material for microelectronics since the first 60s, due to its unparalled electrical properties, and to the ease in production and manifacturing. On the other hand, Si was soon labeled as a poor light emitter, due to the indirect nature of its band structure which leads to very low efficiencies for the band to band radiative recombination, assisted by a phonon. For this reason, for more than 50 years, the realm of optoelectronics has been dominated by III-V semiconductors (e.g. GaAs), due to their unequalled luminescence efficiency determined by the direct nature of their band-gap electronic structure. However, after the discovery of porous silicon by Canham in 1990, with luminescence efficiencies of the order of 10%, the scientific community has redirected new efforts and interests towards the achievement of efficient light emission from Si, interests which are continuously increasing.

118

CHAPTER 5. Devices based on Er-doped Si nanoclusters

Since then, one of the most important results has been the demonstration of optical amplification by Si nanocrystals embedded in SiO2 [124]. Clearly, after this result, achieving electroluminescence from this system has been considered the natural step towards the development of a Si-based optoelectronics. Before, this task had been somewhat neglected by the scientific community, mainly because of the difficulty in obtaining an efficient carrier injection in the semi-insulating material. Indeed, in spite of the large amount of experimental data present in the literature on the photoluminescence properties of Si nanocrystals, only a few papers have reported their electroluminescence (EL) characteristics [125–135]. Among this, in particular, (i) the first evidences of electroluminescence from porous silicon (under 200 Volts polarization) by Richter et al. [125] and the following increase of the external quantum efficiency up to 1% by using liquid contacts [126], (ii) the realization of 0.1% efficient light emitting diodes with a 5 Volts operating voltage [127] and the increase of the device stability by suitable thermal treatments [128], are noteworthy as far as porous silicon is concerned. Unfortunately, despite its promising high efficiency, porous silicon suffers from many problems which strongly limit its full applicability as the active material in light emitting devices, namely: (i) the low compatibility of the electrochemical process needed to generate the porous structure with standard ULSI processes, (ii) the fragility of the active material, due to its high porosity, (iii) the aging effects which determine the degradation of the optical and electrical properties of the active material, and which don’t guarantee the necessary long-term stability of spectral shape and emission intensity, (iv) the difficulty to obtain stable, uniform, and good quality electrical contacts over all the porous surface. In order to avoid these problems, new preparation techniques have been developed. The first one is the ion implantation technique, through which an excess of silicon can be created inside a SiO2 film. After high temperature annealing, the nucleation of Si nanocrystals embedded in the insulating SiO2 matrix occurs. By using this technique, which is strongly ULSI-compatible, the group of Linnros [132], for example, succeded in fabricating a light emitting device which, even if more stable due to the good passivation of the nanocrystals in the insulating matrix, nevertheless was characterized by electroluminescence intensity and external quantum efficiency (∼ 0.003%) much smaller than for porous silicon. Another approach leading to intense room temperature electroluminescence from silicon nanocrystals has been followed by Franzò and coworkers [134], who obtained the excess of silicon through plasma enhanced chemical vapor deposition, another ULSI-compatible technique. In their work, it is shown that the main parameter ruling the electroluminescence intensity is the current density passing through the device, from which the authors conclude the origin of electroluminescence to be the impact of hot electrons on each nanocrystal, which generates an electron-hole pair inside its volume, whose subsequent radiative recombination produces a 900 nm photon. Moreover, it has been definitely demonstrated that the overall electroluminescence process is the result of a delicate balance between a good current injection, i.e. high silicon excess in the SiO2 matrix, and a good quantum confinement, i.e.

5.1 Introduction

119

small Si nanocrystals. Indeed, the best performances have been obtained for active layers containing high silicon excess (i.e. 46 at.%) annealed at a temperature of 1100◦ C for 1h, which guarantees the formation of very small (less than 1 nm in radius) Si nanocrystals. The absolute values of electroluminescence efficiency for Si nanocrystals embedded in SiO2 are of the order of ∼1%, i.e. comparable to the values found for porous silicon. The additional advantage of this system resides on the complete lack of the limitation affecting the performances of porous silicon. Among the dominant approaches followed for obtaining a light emitting source, rare-earths doping hold an important role. In particular Er-doping has long been used for the realization of optical devices able to emit, guide and even amplify light at 1.54 µm, one of the low-loss windows for telecommunication applications. One of the outstanding result has been the realization (in the late 80s) of erbium doped fiber amplifiers (EDFA) which are commonly used in long-distance communications through optical fibers, and which serve to amplify and regenerate the signal. However, EDFAs have some problems related to the fact that in order to obtain gain values of some dB, high Er concentration over long distances are needed, due to the intrinsic low absorption cross sections (10−21 -10−20 cm2 ) of the excited levels of Er. Moreover, high pump powers are needed in order to attain the population inversion necessary for amplification. In this thesis, in Chapter 4, it has been shown that the presence of Si nanoclusters enhances by three orders of magnitude the effective excitation cross section of Er. This is a quite promising result, since it would permit to obtain the population inversion with lower threshold powers. Moreover, even if not definitely understood, it seems that the emission cross section of the first excited level of Er, responsible for the 1.54 µm emission, increases too, thus making it possible to achieve the same gain performances over much smaller distances. This makes Er-doped Si nanoclusters a very promising system for optoelectronics, where the large scale integration would impose the shortening of the mean distances, without changing the amplifying or emission properties of the system. However, even in this case like for Si nanocrystals, the demonstration and feasibility of electrical excitation had for long time remained a dream, if not a very hard task to achieve, due to the insulating nature of the active layer. Indeed, while Er-doped crystalline Si has been extensively studied and room temperature operating devices with efficiencies around 0.05% have already been achieved [54], only a few papers have reported the electroluminescence properties of Er in various insulating hosts (i.e. SiO2 or SiOx , with x < 2). The first proof of the feasibility of silicon dioxide films as a host for obtaining high efficient EL from rare earths was suggested and experimentally realized by Buchal and co-workers [136–139]. They based their concept on the fact that the wide bandgap material is favored for the sustaining of carrier distribution with electrons of high kinetic energy. In the case of a semiconductor with a lower bandgap, the onset of the avalanche process limits the carrier energy. The emission efficiency also appears to increase with the bandgap value. Recently this group developed an Er

120

CHAPTER 5. Devices based on Er-doped Si nanoclusters

implanted MOS diodes with external quantum efficiencies of 10%. However, the devices are short-lived and very poorly stable. Indeed, Er pumping occurs through impact excitation of hot electrons that are injected in the conduction band of the oxide from the n -doped poly-Si layer. These electrons reach high average energies (up to 5 eV), and are able to rapidly deteriorate the oxide characteristics, thus limiting the reliability of the devices. Another approach is to use Si nanoclusters dispersed in the matrix and doped with Er through ion implantation [140, 141]. Indeed, on one hand Si nanoclusters are shown to improve the conduction and the stability properties of the film, and, on the other, they have already been shown to act as efficient sensitizers for the Er ions. In this chapter, we are going to demonstrate that indeed an efficient electroluminescence can be obtained from Er-doped Si nanoclusters devices operating at room temperature. A study of the radiative and non-radiative decay processes occurring in these devices will be also reported, together with an estimate of the Er excitation cross section under electrical pumping. The value found for the excitation cross section of Er under electrical pumping (∼ 1×10−14 cm2 ) is comparable to the value for Si nanocrystals alone, and it can account for the high internal efficiencies (up to 1%) obtained.

5.2

Experimental

Electroluminescent devices emitting at 1.54 µm have been fabricated by Er ion implantation in a SiOx matrix. The dielectric layer of the MOS device was a substoichiometric SiOx (x < 2) film, 70 nm thick, deposited by PECVD. The Si concentration in the film was 46 at.%. After deposition, the SiOx film was implanted with Er ions to a fluence of 7×1014 cm−2 . The energy was chosen in order to locate the Er profile in the middle of the dielectric layer. After the implantation step, the sample was annealed at 900◦ C for 1 h in N2 atmosphere in order to remove the implantation damage, to activate Er and to induce the separation of the Si and SiO2 phases. Si substrate Metal contacts

Active layer Active area

n+ poly Si

SiO2

0.3 mm

Figure 5.1. - Schematic cross section of an Er-doped Si nanoclusters based light emitting device. The active layer and the poly-Si layer are respectively 70 nm and 100 nm thick (drawing not in scale). The thickness of the SiO2 layer is ∼300 nm. The thickness of the metal contacts, measured from the SiO2 surface, is about 1 µm.

5.2 Experimental

121

The low temperature chosen for the annealing process, needed to prevent Er clustering and precipitation, determines the nucleation of Si clusters with dimensions 1.7, in which Si nanoclusters and Er ions are simultaneously present. They have been

133

400

Microcavity Reference

Er-doped Si nanocrystals

300

300 K 200

PL Intensity (a.u.)

PL Intensity (a.u.)

1.0 0.8 0.6 0.4 0.2 0.0

100

0

10 20 30

Angle (deg.)

x2 0

1450

1500

1550

1600

1650

Wavelength (nm) Figure 5.10. - Room temperature PL spectra for an Er-doped Si nanoclusters microcavity (continuous line). The spectrum of Er-doped Si nanoclusters outside the cavity is also shown for comparison (dashed line). The inset reports the integrated PL intensity emitted by the cavity as a function of detection angle with respect to the microcavity vertical axis.

grown by sequential plasma enhanced chemical vapor deposition on top of a Si substrate. The thicknesses of the layers and of the active region were varied in order to match λ0 /4n and λ0 /2n, respectively, being λ0 the intended value of the resonance in free space and n the refractive index of the medium. As an example in Fig. 5.9 a cross sectional TEM image of one of the grown microcavities is shown. The top and bottom DBRs are clearly visible, together with the half-lambda cavity. In Fig. 5.10 the spectra for Er doped Si nanoclusters inside (continuous line) and outside (dashed line) the cavity obtained by optically pumping at 488 nm are shown, respectively. The spectrum is much sharper (∆λ = 5.5 nm) in the microcavity if compared with ∆λ = 40 nm in Er-doped Si nanoclusters without DBRs. Moreover, its intensity is enhanced by over an order of magnitude. Since the room temperature luminescence intensity of Er-doped Si nanoclusters is already 2 orders of magnitude above that of Er-doped SiO2 , the presence of both Si nanoclusters and a microcavity gives a net enhancement of more than 3 orders of magnitude.

134

Summary and Future Perspectives poly Si SiO2 Metal contacts

Active area

Active layer

DBR

Figure 5.11. - Scheme of an electrically pumped Vertical Cavity Surface Emitting Laser (VCSEL) having an Er doped Si nanoclusters layer as the active medium, in between two Si/SiO2 Distributed Bragg Reflectors (DBRs) acting as the mirrors of the optical cavity providing the necessary optical feedback.

In addition, the emission is now strongly directional, as shown in the inset of Fig. 5.10. Having demonstrated an efficient electrical pumping and having realized a silicon based microcavity, the following logical step is to put together the two approaches and realize an electrically driven VCSEL where the active layer (Er doped Si nanoclusters) is electrically pumped and the microcavity furnishes the necessary optical feedback for laser action to occur. A schematic picture of a possible silicon-based VCSEL is reported in Fig. 5.11. Such a structure, if proven, would be important for the realization of free-space interconnects able to link with light two or more chips. In conclusion, this thesis yields new interesting results on the realization of efficient silicon-based light emitters. Having demonstrated high luminescence efficiencies, electrical pumping, and spectral purity and directionality for the 1.54 µm emission of Er doped Si nanoclusters, we are only few steps away the realization of a full silicon based Microphotonics. Hence, silicon is not yet at the end of the rainbow, but at the beginning of its shining future.

Bibliography [1] G. E. Moore, Electronics 38, (1965) [2] International Technology http://public.itrs.net

Roadmap

for

Microelectronics

(ITRS),

[3] D. A. Miller, Proc. IEEE 88, 728 (2000) [4] L. C. Kimerling, Appl. Surf. Sci. 159/160, 8 (2000) [5] B. P. Pal, Progress in Optics 32, edited by E. Wolf (Amsterdam, Elsevier), p. 1 (1993) [6] U. Hilleringmann and K. Goser, IEEE Trans. Electron. Devices 42, 841 (1995) [7] T. Miya, IEEE J. Sel. Top. Quantum Electron. 6, 38 (2000) [8] H. Zimmermann, Integrated Silicon Optoelectronics (New York, Springer) (2000) [9] L. Colace, G. Masini, G. Assanto, H. C. Luan, K. Wada and L. C. Kimerling, Appl. Phys. Lett. 76, 1231 (2000) [10] S. M. Csutak, J. D. Schaub, W. E. Wu, R. Shimer and J. C. Campbell, J. Lightwave Technol. 20, 1724 (2002) [11] S. Winnerl, D. Buca, S. Lenk, Ch. Buchal, S. Mantl and D. X. Xu, Mater. Sci. Eng. B 89, 73 (2002) [12] W. Shockley and W. T. Read, Jr., Phys. Rev. 87, 835 (1952) [13] R. N. Hall, Phys. Rev. 87, 387 (1952) [14] K. Leung and K. B. Whaley, Phys. Rev. B 56, 7455 (1997) [15] A. Franceschetti and A. Zunger, Phys. Rev. Lett. 78, 915 (1997) [16] T. Takagahara and K. Takeda, Phys. Rev. B 46, 15578 (1992) [17] E. Martin, C. Delerue, G. Allan, and M. Lannoo, Phys. Rev. B 50, 18258 (1994)

136

BIBLIOGRAPHY

[18] T. Takagahara, Phys. Rev. B 47, 4569 (1993) [19] G. W. Bryant, Phys. Rev. B 37, 8763 (1988) [20] T. Takagahara and K. Takeda, Phys. Rev. B 53, R4205 (1996) ¨ ut, J. R. Chelikowsky, and S. G. Louie, Phys. Rev. Lett. 79, 1770 (1997) [21] S. Og¨ [22] A. Franceschetti, L. W. Wang, and A. Zunger, Phys. Rev. Lett. 83, 1269 (1999) [23] F. A. Reboredo, A. Franceschetti, and A. Zunger, Appl. Phys. Lett. 75, 2972 (1999) [24] N. A. Hill and K. B. Whaley, Phys. Rev. Lett. 75, 1130 (1995) [25] D. Kovalev, H. Heckler, M. Ben-Chorin, M. Schwartzkopff, and F. Koch, Phys. Rev. Lett. 81, 2803 (1998) [26] B. Delley and E. F. Steigmeier, Phys. Rev. B 47, 13970 (1993) [27] L. T. Canham, Appl. Phys. Lett. 57, 1046 (1990) [28] A. G. Cullis and L. T. Canham, Nature 353, 335 (1991) [29] M. L. Brongersma, Optical properties of ion beam synthesized Si nanocrystals in SiO2 , Ph. D. Thesis, Utrecht University (1998) [30] K. S. Min, K. V. Shcheglov, C. M. Yang, H. A. Atwater, M. L. Brongersma, and A. Polman Appl. Phys. Lett. 69, 2033 (1996) [31] J. Linnros, N. Lalic, A. Galeckas, and V. Grivickas J. Appl. Phys. 86, 6128 (1999) [32] L. C. Kimerling, K. D. Kolenbrander, J. Michel and J. Palm, Light Emission from Silicon, Academic Press, 333 (1997) [33] S. H¨ ufner, Optical Spectra of Transparent Rare Earth Compounds, Academic Press (1978) [34] W. J. Miniscalco, J. Lightwave Technol. 9, 234 (1991) [35] P. N. Favennec, D. Moutonnet, M. Salvi, and M. Gauneau, Jpn. J. Appl. Phys. 29, L524 (1990) [36] J. Michel, J. L. Benton, R. F. Ferrante, D. C. Jacobson, D. G. Eaglesham, E. A. Fitzgerald, Y. -H. Xie, J. M. Poate, and L. C. Kimerling, J. Appl. Phys. 70, 2672 (1991) [37] S. Schmitt-Rink, C. M. Varna, and A. F. J. Levi, Phys. Rev. Lett. 66, 2782 (1991)

BIBLIOGRAPHY

137

[38] S. Coffa, F. Priolo, G. Franzò, V. Bellani, A. Carnera, and C. Spinella, Phys. Rev. B 48, 11782 (1993) [39] F. Priolo, G. Franzò, S. Coffa, A. Polman, S. Libertino, R. Barklie, and D. Carey, J. Appl. Phys. 78, 3874 (1995) [40] S. Libertino, S. Coffa, G. Franzò, and F. Priolo, J. Appl. Phys. 78, 3867 (1995) [41] P. G. Kik, M. J. A. de Dood, K. Kikoin, and A. Polman, Appl. Phys. Lett. 70, 1721 (1997) [42] I. Tsimperidis, T. Gregorkiewicz, H. H. P. Th. Bekman, and C. J. G. M. Langerak, Phys. Rev. Lett. 81, 4748 (1998) [43] T. Gregorkiewicz, D. T. X. Thao, J. M. Langer, H. H. P. Th. Bekman, M. S. Bresler, J. Michel, and L. C. Kimerling, Phys. Rev. B 61, 5369 (2000) [44] M. Forcales, T. Gregorkiewicz, I. V. Bradley, and J-P. R. Wells, Phys. Rev. B 65, 195208 (2002) [45] D. J. Eaglesham, J. Michel, E. A. Fitzgerald, D. C. Jacobson, J. M. Poate, J. L. Benton, A. Polman, Y. -H. Xie, and L. C. Kimerling, Appl. Phys. Lett. 58, 2797 (1991) [46] A. Polman, J. S. Custer, E. Snoeks, and G. N. Van Den Hoven, Appl. Phys. Lett. 62, 507 (1993) [47] F. Priolo, S. Coffa, G. Franzò, C. Spinella, A. Carnera, and V. Bellani, J. Appl. Phys. 74, 4936 (1993) [48] F. Priolo, G. Franzò, S. Coffa, and A. Carnera, Phys. Rev. B 57, 4443 (1998) [49] J. M. Langer and L. Van Hong, J. Phys. C 17, L923 (1984) [50] J. M. Langer, J. Lumin. 40/41, 589 (1988) [51] A. Suchocki and J. M. Langer, Phys. Rev. B 39, 7905 (1989) [52] S. Coffa, G. Franzò, F. Priolo, A. Polman, and R. Serna, Phys. Rev. B 49, 16313 (1994) [53] J. Palm, F. Gan, B. Zheng, J. Mitchel, and L. C. Kimerling, Phys. Rev. B 54, 17603 (1996) [54] G. Franzò, F. Priolo, S. Coffa, A. Polman, and A. Carnera, Appl. Phys. Lett. 64, 2235 (1994) [55] G. Franzò, S. Coffa, F. Priolo, and C. Spinella, J. Appl. Phys. 81, 2784 (1997)

138

BIBLIOGRAPHY

[56] M.V. Wölkin, J. Jorne, P.M. Fauchet, G. Allan, and C. Phys. Rev. Lett. 82, 197 (1999)

Delerue,

[57] M. Luppi, and S. Ossicini, J. Appl. Phys. 94, 2130 (2003) [58] A. G. Cullis, L. T. Canham, and P. D. J. Calcott, J. Appl. Phys. 82, 909 (1997) and references therein [59] C. Delerue, G. Allan, and M. Lannoo, Phys. Rev. B 48, 11024 (1993) [60] F. Priolo, G. Franzò, D. Pacifici, V. Vinciguerra, F. Iacona, and A. Irrera, J. Appl. Phys. 89, 264 (2001) [61] Y. Kanemitsu, T. Ogawa, K. Shiraishi, and K. Takeda, Phys. Rev. B 48, 4883 (1993) [62] T. Shimizu-Iwayama, K. Fujita, S. Nakao, K. Saitoh, T. Fujita, and N. Itoh, J. Appl. Phys. 75, 7779 (1994) [63] J.G. Zhu, C.W. White, J.D. Budai, S.P. Withrow, and Y. Chen, J. Appl. Phys. 78, 4386 (1995) [64] K.S. Min, K.V. Shcheglov, C.M. Yang, H.A. Atwater, M.L. Brongersma, and A. Polman, Appl. Phys. Lett. 69, 2033 (1996) [65] M.L. Brongersma, A. Polman, K.S. Min, E. Boer, T. Tambo, and H.A. Atwater, Appl. Phys. Lett. 72, 2577 (1998) [66] V.I. Klimov, Ch. Schwarz, D.McBranch, and C.W. White, Appl. Phys. Lett. 73, 2603 (1998) [67] T. Shimizu-Iwayama, N. Kunumado, D.E. Hole, and P. Townsend, J. Appl. Phys. 83, 6018 (1998) [68] J. Linnros, N. Lalic, A. Galeckas, and V. Grivickas, J. Appl. Phys. 86, 6128 (1999) [69] F. Iacona, G. Franzò, and C. Spinella, J. Appl. Phys. 87, 1295 (2000) [70] Z.H. Lu, D.J. Lockwood, and J.-M. Baribeau, Nature 378, 258-260 (1995) [71] D.J. Lockwood, Z.H. Lu, and J.M. Baribeau, Phys. Rev. Lett. 96, 539 (1996) [72] V. Vinciguerra, G. Franzò, F. Priolo, F. Iacona, and C. Spinella, J. Appl. Phys. 87, 8165 (2000) [73] P. Photopoulos, A.G. Nassiopoulou, D.N. Kouvatsos, and A. Travlos, J. Appl. Phys. 76, 3588 (2000)

BIBLIOGRAPHY

139

[74] D. Kovalev, J. Diener, H. Heckler, G. Polisski, N. K¨ unzner and F. Koch, Phys. Rev. B 61, 4485 (2000) [75] L. Pavesi, J. Appl. Phys. 80, 216 (1996) [76] L.A. Nesbit, Appl. Phys. Lett. 46, 38 (1985) [77] S. Ossicini, Private communication [78] D.F. Edwards, Handbook of optical constants of solids, p. 547-569, Academic Press (1985) [79] J. Dennis, and E.B. Hale, J. Appl. Phys. 49, 119 (1978) [80] F.F. Morehead, and B.L. Crowder, Radiat. Eff. 25, 49 (1980) [81] O.W. Holland, S.J. Pennycook, and G.L. Albert, Appl. Phys. Lett. 55, 2503 (1988) [82] D.N. Seidman, R.S. Averback, P.R. Okamodo, and A.C. Baily, Phys. Rev. Lett. 58, 900 (1987) [83] S.U. Campisano, S. Coffa, V. Raineri, F. Priolo, and E. Rimini, Nucl. Instr. And Meth. B 80/81, 514-518 (1993) [84] G.L. Olson and J.A. Roth, Mater. Sci. Rep. 3, 1 (1988) [85] F. Priolo, A. Battaglia, C. Spinella, and E. Rimini, Mat. Res. Soc. Symp. Proc. 201, 345 (1991) [86] P.D. Pensans, A. Ruppert, and B. Abeles, J. Non-Cryst. Solids 102, 130 (1988) [87] M. Zacharias, and P. Streitenberger, Phys. Rev. B 62, 8391 (2000) [88] M. Zacharias, J. Blasing, P. Veit, L. Tsybeskov, K. Hirshman, and P.M. Fauchet, Appl. Phys. Lett. 74, 2614 (1999) [89] G.A. Kachurin, M.O. Ruault, A. K. Gutakovsky, O. Kaitasov, S.G. Yanovskaya, and K.S. Zhuravlev, H. Bernas, Nucl. Instr. and Meth. B 147, 356 (1999) [90] S. Cheylan, N. Langford, and R.G. Elliman, Nucl. Instr. and Meth. B 166/167, 851 (2000) [91] D. Pacifici, E.C. Moreira, G. Franzò, V. Martorino, F. Priolo, and F. Iacona, Phys. Rev. B 65, 144109 (2002) [92] J.P. Biersack, and L.G. Haggmark, Nucl. Instrum. Methods 174, 257 (1980) [93] J.F. Ziegler, J.P. Biersack, and U. Littmark, The Stopping and Range of Ions in Solids (Pergamon, New York, 1985)

140

BIBLIOGRAPHY

[94] S. Roorda, W.C. Sinke, J.M. Poate, D.C. Jacobson, S. Dieker, B.S. Dennis, D.J. Eaglesham, F.Spaepen, and P. Fuoss, Phys. Rev. B 44, 3702 (1991) [95] S. Coffa, F. Priolo, and A. Battaglia, Phys. Rev. Lett. 70, 3756 (1993) [96] C. Spinella, S. Lombardo, and F. Priolo, J. Appl. Phys. 84, 5383 (1998) [97] S. Coffa, G. Franzò, F.Priolo, Appl. Phys. Lett. 69, 2077 (1996) [98] A. J. Kenyon, P. F. Trwoga, M. Federighi, and C. W. Pitt, J. Phys.: Condens. Matter 6, L319 (1994) [99] M. Fujii, M. Yoshida, Y. Kanzawa, S. Hayashi, and K. Yamamoto, Appl. Phys. Lett. 71, 1198 (1997) [100] M. Fujii, M. Yoshida, S. Hayashi, and K. Yamamoto, J. Appl. Phys. 84, 4525 (1998) [101] T. Komuro, T. Katsumata, T. Morikawa, X. Zhao, H. Isshiki, and Y. Aoyagi, Appl. Phys. Lett. 74, 377 (1999) [102] G. Franzò, V. Vinciguerra, and F. Priolo, Appl. Phys. A: Mater. Sci. Process. 69, 3 (1999) [103] G. Franzò, F. Iacona, V. Vinciguerra, and F. Priolo, Mat. Sci. Eng. B 69/70, 338 (1999) [104] G. Franzò, D. Pacifici, V. Vinciguerra, F. Iacona, and F. Priolo, Appl. Phys. Lett. 76, 2167 (2000) [105] P. G. Kik, M. L. Brongersma, and A. Polman, Appl. Phys. Lett. 76, 2325 (2000) [106] P. G. Kik, and A. Polman, J. Appl. Phys. 88, 1992 (2000) [107] S. -Y. Seo, and J. H. Shin, Appl. Phys. Lett. 78, 2709 (2001) [108] K. Watanabe, M. Fujii and S. Hayashi, J. Appl. Phys. 90, 4761 (2001) [109] F. Priolo, G. Franzò, F. Iacona, D. Pacifici, V. Vinciguerra, Mat. Sci. Eng. B 81, 9 (2001) [110] F. Iacona, D. Pacifici, A. Irrera, M. Miritello, G. Franzò, F. Priolo, D. Sanfilippo, G. Di Stefano, and P. G. Fallica, Appl. Phys. Lett. 81, 3242 (2002) [111] H. -S. Han, S. -Y. Seo, and J. H. Shin, Appl. Phys. Lett. 79, 4568 (2001) [112] H. -S. Han, S. -Y. Seo, J. H. Shin, and N. Park, Appl. Phys. Lett. 81, 3720 (2002)

BIBLIOGRAPHY

141

[113] J. H. Shin, H. -S. Han, and S. -Y. Seo, Towards the first silicon laser, edited by L. Pavesi, S. Gaponenko, and L. Dal Negro, NATO Science Series vol. 93 (Kluwer Academic Publishers, Dordrecht 2003), p. 401 [114] P. G. Kik and A. Polman, J. Appl. Phys. 91, 534 (2002) [115] P. G. Kik and A. Polman, Towards the first silicon laser, edited by L. Pavesi, S. Gaponenko, and L. Dal Negro, NATO Science Series vol. 93 (Kluwer Academic Publishers, Dordrecht 2003), p. 383 [116] A. Polman, J. Appl. Phys. 82, 1 (1997) [117] D. Pacifici, G, Franzò, F. Priolo, F. Iacona, and L. Dal Negro, Phys. Rev. B 67, 245301 (2003) [118] G. Allan, C. Delerue, and M. Lannoo, Phys. Rev. Lett. 78, 3161 (1997) [119] G. N. van den Hoven, E. Snoeks, A. Polman, C. van Dam, J. W. M. van Uffelen, M. K. Smit, J. Appl. Phys. 79, 1258 (1996) [120] T. Förster, Ann. Phys. (N.Y.) 2, 55 (1948) [121] D.L. Dexter, J. Chem. Phys. 21, 836 (1953) [122] J. H. Shin, Appl. Phys. Lett. 82, 4489 (2003) [123] M. Fujii, M. Yoshida, S. Hayashi, and K. Yamamoto, J. Appl. Phys. 84, 4525 (1998) [124] L. Pavesi, L. Dal Negro, C. Mazzoleni, G. Franzò, and F. Priolo, Nature 408, 440 (2000) [125] A. Richter, P. Steiner, F. Kozlowski, and W. Lang, IEEE Electr. Dev. Lett. 12, 691 (1991) [126] A. Halimaoui, C. Oules, G. Bomchil, A. Biesy, F. Gaspard, R. Herino, M. Ligeon, and F. Muller, Appl. Phys. Lett. 59, 304 (1991) [127] N. Lalic, and J. Linnros, Thin Solid Films 276, 155 (1996) [128] L. Tsybeskov, S. P. Duttagupta, K. D. Hirschman, and P. Fauchet, Appl. Phys. Lett. 68, 2058 (1996) [129] K.D. Hirschman, L. Tsybeskov, S.P. Duttagupta, and P.M. Fauchet, Nature 384, 338 (1996) [130] L. Rebohle, J. von Borany, R.A. Yankov, W. Skorupa, I.E. Tyschenko, H. Frob, and K. Leo, Appl. Phys. Lett. 71, 2809 (1997) [131] S. Fujita and N. Sugiyama, Appl. Phys. Lett. 74, 308 (1999)

142

BIBLIOGRAPHY

[132] N. Lalic, and J. Linnros, J. Lumin. 80, 75 (1999) [133] P. Photopoulos and A.G. Nassiopoulou, Appl. Phys. Lett. 77, 1816 (2000) [134] G. Franzò, A. Irrera, E.C. Moreira, M. Miritello, F. Iacona, D. Sanfilippo, G. Di Stefano, P.G. Fallica, and F. Priolo, Appl. Phys. A 74, 1 (2002) [135] A. Irrera, D. Pacifici, M. Miritello, G. Franzò, F. Priolo, F. Iacona, D. Sanfilippo, G. Di Stefano, and P. G. Fallica, Appl. Phys. Lett. 81, 1866 (2002) [136] S. Wang, A. Eckau, E. Neufeld, R. Carius, Ch. Buchal, Appl. Phys. Lett. 71, 2824 (1997) [137] S. Wang, H. Amekura, A. Eckau, R. Carius, Ch. Buchal, Nucl. Instr. Meth. Phys. Res. B 148, 481 (1999) [138] S. Wang, S. Coffa, R. Carius, Ch. Buchal, Mater. Sci. Eng. B 81, 102 (2001) [139] Ch. Buchal, S. Wang, F. Lu, R. Carius, S. Coffa, Nucl. Instr. Meth. Phys. Res. B 190, 40 (2002) [140] D. Pacifici, A. Irrera, G. Franzò, M. Miritello, F. Iacona, and F. Priolo, Physica E 16, 331 (2003) [141] F. Iacona, D. Pacifici, A. Irrera, M. Miritello, G. Franzò, F. Priolo, D. Sanfilippo, G. Di Stefano, and P.G. Fallica, Appl. Phys. Lett. 81, 3242 (2002) [142] F. Iacona, G. Franzò, E. C. Moreira, and F. Priolo, J. Appl. Phys. 89, 8354 (2001)

To the reader If you have got to this point and have read all that goes before, I’m very pleased to say that you are the first person worthy of acknowledgement. You have noticed that I’ve used many words to describe scientific and technical stuff of this thesis. However, trying to figure out how to write this section, after having looked for a while at the blank page, I have come to the conclusion that there are not enough and adequate words for me to express the gratitude and the esteem I feel for those persons who have directly or indirectly helped and sustained me all over these intense years. Hence I’ve decided that this section will not be as short as is customary, but on the contrary, I will take all the necessary time to acknowledge almost everybody who deserves it. As it is my concern to help the reader find his/her name, I have divided this section of acknowledgements in categories. Nevertheless, I expect this list of acknowledgements not to be exhaustive at all, and since only who does not write does not commit errors for sure, please don’t be too severe with me if you do not succeed in finding your name. Indeed, if you feel that you have the right to be mentioned somewhere in the following, please let me know, surely your suggestion will be taken into account . . . ;)

Acknowledgements First of all I’d like to acknowledge my father and my mother. Without their continuous supports, all the important goals I’ve reached in my life, and this thesis is among them, wouldn’t have been possible. I’d like to thank them for the patience and understanding they have shown me even in the most difficult periods. University I’d like to especially thank my tutor and mentor Prof. Francesco Priolo, who gave me the possibility of working within his dynamic and productive research group. In the long and deep scientific discussions I had with him, I greatly appreciated his enormous Physical sense and his profound passion for scientific research. He always stimulated me, giving me the best opportunities, and, as a good master does, always

144

Acknowledgements

expecting the best results. Many thanks to Giorgia Franzò who first introduced me to the experimental world, by teaching me the hard job of doing luminescence measurements. Thank you to Simona Boninelli, who performed the EFTEM reported in this thesis, and to Maria Miritello, for her valuable help in the laboratory, during the electroluminescence measuremnts. I’d like to express my esteem to Prof. Emanuele Rimini, whose lessons have approached me for the first time to the stimulating world of materials science and physics of matter. My thoughts go also to the late Prof. Ugo Campisano, whose words and advises are still vivid in my memory. Many thanks to Riccardo Reitano, for teaching me the use of the UV-Visible spectrometer but mostly for his always stimulating questions, to Paolo Musumeci, who raised my mood every time I met him, to Prof. Gaetano Foti, for sharing his eclectic culture and for filling the corridors with his thundering voice, to Tony Terrasi, an ever green man with a strong taste for irony. One of the most important person I’d like to acknowledge is Giovanni Piccitto. He has always been and still is for me a master and a sincere friend. To Maria Grazia I’d say: thank you for the joy spreading out from your person every time we meet. This few lines are for Natale Marino. You are the kindest man I’ve ever had the opportunity to know and to work with. Moreover your great ability in practically and genially solving all the problems we encounter in the lab makes you a valuable man for all the group. Thank you to Orazio Parasole, who kindly taught me the use of the Ion Implanter I’ve extensively used for preparing the sample, and to Salvo Leotta, who helped me with the Van der Graaf I’ve used for RBS measurements. Thank you to Carmelo Percolla and Salvo Tat`ı for their expert technical assistance. To Pasquale Tomasello, thank you for the deep discussions and for the friendship which started a long time ago. CNR-IMM Many thanks to Corrado Spinella for having allowed me to use the machinery of the Institute all over these years. Thank you to Fabio Iacona, for sample preparation by PECVD and for his advice. I’d like to acknowledge Alessia Irrera for the electrical characterization of electroluminescent devices, and also for her sincere friendship. Antonio Marino is a very special person for me. He has not only helped me understanding the main features of the TANDEM accelerator, kindly answering to all my boring questions, but he has also demonstrated to be a sincere friend. Thank you Antonio! To Aldo Spada, the computer-wizard, who taught me many tricks for generating

145 a good computer program. But mainly for the reciprocal friendship and understanding. I’d like to acknowledge Salvo Pannitteri and Corrado Bongiorno for their help in the difficult art of TEM sample preparation and measurements and for their friendly behavior. Many thanks also to Antonino La Magna. STMicroelectronics I’d like to acknowledge D. Sanfilippo, G. Di Stefano, and P.G. Fallica for realizing the light emitting devices whose electroluminescent properties are discussed in this thesis. Friends & Colleagues I’d like to acknowledge the persons with whom I’ve spent most of the time at the University. Many thanks to my dear friends Salvo Mirabella, Alberto Piro, Massimiliano Chiorboli and Andrea Giammanco. They have encouraged me with their continuous presence all over these years. I really hope their friendship will maintain its strength in the years to come. Thank you also: to Lucia, who literally fed me every time I was starving ;), to Daniele, our coach, to Luca and Salvo, my favorite students, to Monia, Elena and Giuliana for their kindness. People around the Globe I’d also like to thanks all the persons with whom I’ve had many fruitful and interesting scientific discussions. First of all I’d like to spend a few words about Luca Dal Negro. He is one of the most ‘stimulating’ physicist I’ve ever known. The time spent together in the lab and in vacation will remain for ever vivid in my memory. Many thanks also to Prof. Lorenzo Pavesi, to Prof. Harry Atwater, to Prof. Tom Gregorkiewicz for the fantastic time I had in their laboratory. When I first entered the world of Physics of Matter, I had the pleasure to read one of the most interesting thesis, which was a reference point in my future studies. I want to acknowledge the author of this thesis, Mark Brongersma, for his curiosity and for the nice chats and discussions we have together every time we meet. Thank you to Pieter Kik, for the exciting discussions we had and which strongly contributed to deepen my understanding of the Er-doped Si nanocrystals system. To Professors Albert Polman, Philippe Fauchet, Minoru Fujii, Jung Shin, and Stefano Ossicini, for their very interesting questions, and for the important contribution they give in this field.

146

Acknowledgements

‘Only’ Friends If I had to write down all the name of the persons I consider friends, who have left a piece of themselves in my soul and to whom I’m grateful, I’d probably have to add a book to this thesis, but it should be enough for them to know that they all have a place in my heart where I jealously guard their faces and smiles. A few of them are worth naming here since they have been indispensable for my personal growth and for my everyday life. Francesco-Chery, Sebastian e Orazio, Giuseppe, Dario e Daniele, Luca, Francesco e Francesca, Paolo, Roberto, Ema, Marco, e Sandro. To all of you, thank you for existing! Relatives This place is reserved for my beloved relatives. I know they will be pleased to find their names in the following.

A Paoletta, la mia cara sorellina, a Chiara e a Fede, che considero come fratelli, a Roby e Antonio, che nonostante la lontananza sono sempre vicini nel mio cuore, a mio zio Pasquale (zio Quaque) e a mia zia Ninetta (zia Nannenne) per avermi cresciuto e considerato sempre come un figlio, a mio Zio Carmelo (‘grande caddozzu’) e a mia Zia Marcella (la zia pi` u dolce e bella), e ai miei nonni tutti: a mio nonno Domenico (nonno Chetto), per la sua bontà e per i suoi consigli (studia!) che sempre mi accompagnano nei ricordi, a mio nonno Nino, nonno tutto fare, a mia nonna Paola, per la sua dolcezza, e a mia nonna Rosina, per avermi sempre dato consigli impagabili. A tutti voi, grazie per essere stati sempre al mio fianco.

Infine vorrei ringraziare Paolo e Ester. Grazie per la serenità che mi avete trasmesso in tutti questi anni, ma soprattutto grazie davvero per avermi sempre considerato come un figlio. E a Myriam, zuzzina che mi fa sempre sorridere. Lovers Questo posto non può che essere riservato all’unica persona che per tutti questi anni mi è stata sempre vicina, fedele compagna e amica.

Alla mia dolce e amata Claudia.

Curriculum Vitae Domenico Pacifici was born on 17th of February 1976 in Conegliano, Treviso, Italy. In 1994 he attained the General Certificate of Education at the Liceo Scientifico Galileo Galilei, with final mark 60/60. In that year he started his university education at the University of Catania, joining the Physics course. In July 2000 he received his Master Degree in Physics (summa cum laude), discussing a thesis titled Interazione tra nanocristalli di silicio e ioni Er3+ . In January 2001 he was awarded by Accademia Gioenia di Catania for his ‘original contributions to the development and diffusion of scientific knowledge’. The 28th of February 2001, he was classed first at the entrance exam for a three-year PhD course in Physics at the University of Catania. The 1st of December 2003 he won a post-doctoral position, as research collaborator, at the University of Catania. Domenico Pacifici received his research formation mainly at the MATIS Laboratory of INFM, under the supervision of Prof. F. Priolo. He has also collaborated with CNR-IMM, and with STMicroelectronics in Catania. Moreover, he spent some periods of his research activity at CALTECH (California Institute of Technology, Pasadena), under the supervision of Prof. H. Atwater, at the Department of Physics of the University of Trento, under the supervision of Prof. L. Pavesi, and at the Van der Waals-Zeeman Institute of the University of Amsterdam, under the supervision of Prof. T. Gregorkiewicz. During his PhD activity, Domenico Pacifici has been mainly involved in the experimental study of low-dimensional silicon structures and their interaction with rare earths for a silicon based Microphotonics. He contributed to the understanding of the peculiar physical properties of these systems, developing a model for the damage induced by energetic ion beams on silicon nanocrystals and a model describing the energy transfer process in erbium doped silicon nanoclusters. He was directly involved in the design and realization of efficient silicon-based optical microcavities and light emitting devices, with silicon nanocrystals and erbium doped silicon nanoclusters as the optically active medium. In this field, he has been able to achieve important results, as attested by the various publications in international scientific journals, and by the numerous oral contributions to international conferences. Among these, he attended as an invited speaker the Annual Meeting of the Italian Physics Society (SIF) in Milan (September 2001), and the Spring Meeting of the European Materials Research Society (e-MRS) in Strasbourg (June 2002). He is among the invited speakers at the Spring Meeting of the Materials Research Society (MRS) to be held in San Francisco (April 2004).


This thesis is based on the following publications: 1. Modeling and perspectives of the Si nanocrystals-Er interaction for optical amplification D. Pacifici, G. Franzò, F. Priolo, F. Iacona, and L. Dal Negro Physical Review B 67, 245301 (2003) 2. Defect production and annealing in ion-irradiated Si nanocrystals D. Pacifici, E. C. Moreira, G. Franzò, V. Martorino, F. Priolo, and F. Iacona Physical Review B 65, 144109 (2002) 3. Electroluminescence at 1.54 µm in Er-doped Si nanocluster-based devices F. Iacona, D. Pacifici, A. Irrera, M. Miritello, G. Franzò, F. Priolo, D. Sanfilippo, G. Di Stefano, and P.G. Fallica Applied Physics Letters 81, 3242 (2002) 4. Excitation and de-excitation properties of silicon quantum dots under electrical pumping Irrera, D. Pacifici, M. Miritello, G. Franzò, F. Priolo, F. Iacona, D. Sanfilippo, G. Di Stefano, and P.G. Fallica Applied Physics Letters 81, 1866 (2002) 5. Er3+ ions - Si nanocrystals interactions and their effects on the luminescence properties G. Franzò, D. Pacifici, V. Vinciguerra, F. Priolo, F. Iacona Applied Physics Letters 76, 2167 (2000) 6. Role of energy transfer in the optical properties of undoped and Er-doped interacting Si nanocrystals F. Priolo, G. Franzò, D. Pacifici, V. Vinciguerra, F. Iacona, A. Irrera Journal of Applied Physics 89, 264 (2001) 7. Er doped Si nanostructures D. Pacifici, G. Franzò, F. Iacona, S. Boninelli, A. Irrera, M. Miritello, and F. Priolo Materials Science and Engineering B 105/1-3, 197 (2003)

150


8. Excitation and non-radiative de-excitation processes in Er-doped Si nanocrystals F. Priolo, G. Franzò, F. Iacona, D. Pacifici, V. Vinciguerra Materials Science and Engineering B 81, 9 (2001) 9. Luminescence properties of Si nanocrystals embedded in optical microcavities F. Iacona, G. Franzò, E.C. Moreira, D. Pacifici, A. Irrera, F. Priolo Materials Science and Engineering C 19, 377 (2002) 10. Erbium-doped Si nanocrystals: optical properties and electroluminescent devices D. Pacifici, A. Irrera, G. Franzò, M. Miritello, F. Iacona, and F. Priolo Physica E 16, 331 (2003) 11. Amorphization and recrystallization of ion implanted Si nanocrystals probed through their luminescence properties D. Pacifici, G. Franzò, F. Iacona, and F. Priolo Physica E 16, 404 (2003) 12. Electroluminescence properties of light emitting devices based on silicon nanocrystals Irrera, D. Pacifici, M. Miritello, G. Franzò, F. Priolo, F. Iacona, D. Sanfilippo, G. Di Stefano, and P.G. Fallica Physica E 16, 395 (2003) 13. Ion beam synthesis of undoped and Er-doped Si nanocrystals G. Franzò, E.C. Moreira, D. Pacifici, F. Priolo, F. Iacona, C. Spinella Nuclear Instruments and Methods B 175-177, 140 (2001) 14. Luminescence from Si nanocrystals and Er ions embedded in resonant cavities F. Iacona, G. Franzò, E.C. Moreira, D. Pacifici, F. Priolo Solid State Phenomena 82-84, 617 (2002) 15. Structural and optical properties of PECVD grown silicon nanocrystals G. V. Prakash, N. Daldosso, E. Degoli, F. Iacona, M. Cazzanelli, Z. Gaburro, G. Pucker, P. Dalba, F. Rocca, E. C. Moreira, G. Franzò, D. Pacifici, F. Priolo, C. Arcangeli, A.B. Filonov, S. Ossicini, L. Pavesi J. Nanosci. Tech. 1, 159 (2001)

Other related publications: 16. Dynamics of stimulated emission in silicon nanocrystals L. Dal Negro, M. Cazzanelli, L. Pavesi, S. Ossicini, D. Pacifici, G. Franzò, and F. Priolo Applied Physics Letters 82, 4636 (2003)

151 17. Transient behavior of the strong violet electroluminescence of Ge-implanted SiO2 layers L. Rebohle, T. Gebel, J. von Borany, W. Skorupa, M. Helm, D. Pacifici , G. Franzò, F. Priolo Applied Physics B 74, 53 (2002) 18. Stimulated emission in Plasma Enhanced Chemical Vapour Deposited silicon nanocrystals L. Dal Negro, M. Cazzanelli, N. Dal Dosso, Z. Gaburro, L. Pavesi, F. Priolo, D. Pacifici, G. Franzò, and F. Iacona Physica E 16, 297 (2003) 19. X-Ray absorption study of light emitting Si nanocrystals N. Daldosso, G. Dalba, R. Grisenti, L. Dal Negro, L. Pavesi, F. Rocca, F. Priolo, G. Franzò, D. Pacifici, and F. Iacona Physica E 16, 321 (2003)

Conference proceedings: 20. Coupling and cooperative up-conversion coefficients in Er-doped Si nanocrystals D. Pacifici, G. Franzò, F. Iacona, and F. Priolo Mat. Res. Soc. Symp. Proc. Vol. 770, I6.8.1 (2003) 21. Light emitting devices based on silicon nanocrystals A. Irrera, D. Pacifici, M. Miritello, G. Franzò, F. Priolo, F. Iacona, D. Sanfilippo, G. Di Stefano, and P.G. Fallica in Towards the first silicon laser edited by L. Pavesi, S. Gaponenko, L. Dal Negro, NATO Science Series vol. 93 (Kluwer Academic Publishers, Dordrecht), pag. 29 (2003) 22. Tuning of the electroluminescence from Si nanocrystals through the control of their structural properties A. Irrera, F. Iacona, D. Pacifici, M. Miritello, G. Franzò, D. Sanfilippo, G. Di Stefano, P.G. Fallica, and F. Priolo Mat. Res. Soc. Symp. Proc. Vol. 737, F11.9 (2003) 23. Electroluminescent devices based on Er-doped Si nanoclusters F. Priolo, F. Iacona, D. Pacifici, A. Irrera, M. Miritello, G. Franzò, D. Sanfilippo, G. Di Stefano, and P.G. Fallica Mat. Res. Soc. Symp. Proc. Vol. 737, F9.3 (2003) 24. Optical gain and stimulated emission in silicon L. Dal Negro, M. Cazzanelli, Z. Gaburro, P. Bettotti, L. Pavesi, D. Pacifici,

152

List of Publications G. Franzò, F. Priolo, and F. Iacona Mat. Res. Soc. Symp. Proc. Vol. 738, G8.8.1 (2003)

25. Time-resolved gain dynamics in silicon nanocrystals L. Dal Negro, M. Cazzanelli, N. Daldosso, L. Pavesi, F. Priolo, G. Franzò, D. Pacifici, and F. Iacona Mat. Res. Soc. Symp. Proc. Vol. 770, I3.4.1 (2003) 26. Stimulated emission in silicon nanocrystals: gain measurement and rate equation modelling L. Dal Negro, M. Cazzanelli, Z. Gaburro, P. Bettotti, L. Pavesi, F. Priolo, G. Franzò, D. Pacifici, and F. Icona in Towards the first silicon laser edited by L. Pavesi, S. Gaponenko, L. Dal Negro, NATO Science Series vol. 93 (Kluwer Academic Publishers, Dordrecht), pag. 145 (2003) 27. Silicon nanocrystal nucleation as a function of the annealing temperature in SiOx films N. Daldosso, G. Das, G. Dalba, S. Larcheri, R. Grisenti, G. Mariotto, L. Pavesi, F. Rocca, F. Priolo, G. Franzò, A. Irrera, M. Miritello, D. Pacifici, and F. Iacona Mat. Res. Soc. Symp. Proc. Vol. 770, I1.3.1 (2003) 28. Optical gain in PECVD grown silicon nanocrystals L. Dal Negro, M. Cazzanelli, Z. Gaburro, L. Pavesi, D. Pacifici, F. Priolo, G. Franzò, and F. Iacona in Optical properties of nanocrystals edited by Z. Gaburro, Proceedings of SPIE vol. 4808, pag. 13 (2002) 29. Will silicon be the photonics material of the third millennium? L. Pavesi, L. Dal Negro, N. Daldosso, Z. Gaburro, M. Cazzanelli, F. Iacona, G. Franzò, D. Pacifici, F. Priolo, S. Ossicini, M. Luppi and E. Degoli Proceedings of the International Conference on the Physics of Semiconductors (Edinburgh 2002)