Characterization of undoped and nitrogen-doped carbon nanodots using spectroscopy

University of Amsterdam

Faculty of nature sciences, mathematics and informatics

Van der Waals-Zeeman institute

Characterization of undoped and

nitrogen-doped carbon nanodots using

spectroscopy

Mischa Hillenius

10361634

Report Bachelor Project Physics and Astronomy, size 12EC conducted between 06-11-2015 and 07-04-2016 submission date: 07-04-2016 supervisor: Tom Gregorkiewicz instructors: Chung Xuan Nguyen, Leyre G´omez Navascu´es Second assessor: Wim Sinke

Submitted in part fulfilment of the requirements for the degree of Bachelor of Science in Beta-Gamma major Physics of the University of Amsterdam, april 2016

Populaire Samenvatting

Koolstof is een van de meest voorkomende elementen op aarde en kent vele toepassingen. Zo is het de bouwsteen van het leven op aarde, zit het als grafiet in je potlood en wordt het als gepoli-jste diamant gebruikt voor decoratie. Koolstof kan ook als nanodeeltje gesynthetiseerd worden. Deze koolstof nanodeeltjes beloven onderdeel te worden van de nieuwe generatie zonnepanelen. Zonnepanelen kunnen licht met ´e´en bepaalde energie opnemen. Licht met een lagere energie wordt niet opgenomen, en licht met een hogere energie wordt voor een deel opgenomen- de extra energie gaat verloren als warmte. Koolstof nanodeeltjes kunnen invallend licht met een te hoge of te lage energie omzetten naar de energie die optimaal is voor het zonnepaneel en kunnen dus een goede oplossing zijn om zonnepanelen efficienter te maken.

Als atomen met elkaar binden tot een molecuul, vormen valentie elektronen samen een paar. Dit heeft als gevolg dat nieuwe energie niveaus ontstaan. Als meer en meer atomen met elkaar binden kunnen de energie niveaus als continu kunnen worden gezien. Bandkloven zijn gaten in deze gradatie van energieniveaus, en wanneer de bandkloof klein genoeg is, spreken we van een halfgeleider. Koolstof nanodeeltjes zijn halfgeleiders. Dat betekent dat ze alleen elektriciteit geleiden als ze extra energie krijgen van buitenaf in de vorm van licht of hitte. Wanneer licht met een energie hoger dan de bandkloof op koolstof nanodeeltjes wordt geschenen, kunnen elektronen het gat overbruggen en kunnen ze wegstromen.

De nanodeeltjes zijn zo klein dat quantum verschijnselen optreden: het quantum confinement effect stelt dat de bandkloof groter wordt als de elektronen in het materiaal samengedrukt worden hoe kleiner het deeltje wordt. Kleine deeltjes kunnen dus licht van een hogere energie uitzenden dan grotere deeltjes.

Als in de deeltjes atomen van een ander element geinjecteerd worden (doping), veranderen de eigenschappen van de deeltjes. Door zowel de kristalgrootte als de hoeveelheid vreemde atomen te veranderen, kunnen veel verschillende materialen gemaakt worden.

In dit onderzoek zijn twee soorten koolstof nanodeeltjes met elkaar vergeleken waarvan er in een stikstof was toegevoegd. De materialen functioneren anders in het licht dat ze uitzenden en hoe lang dat gebeurt. Verder onderzoek kan nieuwe toepassingen van deze materialen mogelijk maken.

Het quantum confinement effect : als elektronen in een materiaal samengedrukt worden, splitsen de energielevels waardoor een grotere bandkloof ontstaat [Chukwuocha et al., 2012]

Abstract

Carbon nanodots (CNDs) hold great potency to access a broader part of the solar spectrum in next-generation solar cells by down- and upconverting light in energy. They have a tunable band gap due to the quantum confinement effect and controllable crystal growth. The PL mechanism in CNDs is still an open debate among researchers, therefore sufficient characterization to illuminate the PL mechanism and systematic comparison of different CNDs can guide the development of effective synthesis routes and novel applications. In this paper two samples of CNDs, one undoped and one nitrogen doped, are characterized in terms of absorption and emission curves and PL dynamics. Results indicate CNDs doped with nitrogen have a lower band gap energy than undoped pure CNDs based on evaluation of the PL peaks. PL decay of doped CNDs is about 40ns slower than PL decay of pure CNDs. Experimental data on upconversion suggests no upconversion is present when exciting CNDs doped with nitrogen, leading to the believe future research is necessary to create clarity in the ongoing debate on upconversion.

Acknowledgements

I would like to express my feelings of appreciation and gratitude to:

• My supervisor, T. Gregorkiewicz,

for helping me with understanding the fundamental physical concepts of research in nano-materials and taking the time to assess the thesis

• My second supervisor, W. Sinke,

for assessing the thesis as independent reviewer

• Leyre G´omez Navascu´es,

who helped me understanding the literature and provided me with many useful comments on the first versions of the manuscript

• Chung Xuan Nguyen,

who successfully guided me in my research and answered the many questions I had re-garding theory as well as spectrometers and data analysis. Furthermore I appreciate the many useful remarks you gave on the manuscript in its earlier versions.

Contents

2.2.1 The origin of the electronic band structure . . . 6

2.2.2 The example of a particle in a box . . . 6

2.2.3 Effective mass approximation model . . . 8

2.3 Doping of CNDs . . . 9

2.4 Corrections of optical emission spectra . . . 9

2.5 Determination of band gap energy - Tauc plots . . . 11

3 Methods 12 3.1 Photoluminescence spectra . . . 12

3.2 PL Dynamics . . . 14

3.3 Excitation dependence of PL peak . . . 14

CONTENTS 1

3.4 Absorption . . . 15

4 Results 16 4.1 Excitation dependence of PL peak . . . 16

4.2 PL Dynamics . . . 17

4.3 PL and Absorption Spectra . . . 18

4.4 PL spectra - obtained data on upconversion . . . 20

5 Discussion 21

6 Conclusion and prospectives 24

Chapter 1

Introduction

Nanomaterials and nanostructures hold promising potency to enhance the performance of solar cells by both improving light trapping and photo-carrier collection [Miao et al., 2015]. The family of nanomaterials has great diversity, and these structures all have different physical properties, such as band-gap, absorption coefficient, surface/bulk recombination rate, etc., as well as different synthesis/fabrication methods. Carbon is one of the most abundant natu-ral elements and as a nanomaterial can be applied in different fields such as optoelectronics, biomedicine, catalysis and sensors.

Carbon nanodots (CNDs) are part of the quantum dot family and have generated enormous excitement because of their remarkable properties in water stability, chemical inertness, low toxicity, ease of functionalization and resistance to photobleaching. One of the most intriguing properties of CNDs is the photoluminescence (PL) emission which is strong in the visible spectral range [Miao et al., 2015]. This property accounts for the believe they hold great potency to access a broader part of the solar spectrum in next-generation solar cells by down-and upconverting incident light in energy [Miao et al., 2015].

CNDs are semiconductor nanoparticles whose excitons are confined in all three spatial dimen-sions [Chukwuocha et al., 2012]. It is essentially a tiny semiconductor crystal with size in the order of nanometers. They are often called ’artificial atoms’ because of its quantum properties [Chukwuocha et al., 2012] and interactions similar to bulk semiconductor materials. Addi-tionally, the electronic characteristics of the CNDs are closely related to the size and shape of every crystal. In general holds that the smaller a CND is, the larger the band gap of the crystal becomes and the more energy is needed to excite the electrons [Chukwuocha et al., 2012]. Limiting the crystal growth in synthesis of CNDs results in a tunable band gap and thus tunable absorption and emission spectra. This is impossible for atoms, but desirable for opti-cal properties. Larger CNDs show a greater spectrum shift towards red compared to smaller dots, while smaller dots show a greater spectrum shift towards blue in photoluminescence (PL) [Chukwuocha et al., 2012].

At present, the actual mechanism of the PL of CNDs is still an open debate among researchers. For new theories to develop and research to be conducted, it is therefore important to summarize the current state of the art knowledge. Four reasonable PL mechanisms have been confirmed: the quantum confinement effect, which is determined by the carbon core, the surface state, which is determined by hybridization of the carbon backbone and the connected chemical

3

groups, the molecule state, which is determined solely by the fluorescent molecules connected on the surface of the CDs, and the crosslink-enhanced emission (CEE) effect [Zhu et al., 2015]. The quantum confinement effect will be discussed in the theoretical background, providing a framework to interpret the results posed in this thesis. For quantum confinement to be understood properly, an analog will be made with the particle in a box model in quantum mechanics, and an introduction to the effective mass approximation model will be given. Since four theories for the PL mechanisms exist, it is useful to gain experimental data on dif-ferent CNDs samples. In this thesis two samples of CNDs with difdif-ferent synthesis methods are characterized and compared against each other. The first sample, from now on denoted by UP, has been doped with nitrogen ions and is expected to show upconversion, while the second sample, from now on denoted by DOWN, has not been doped. The average size of the CNDs in UP is 2.2 ± 0.3nm and in DOWN 3.4 ± 0.4nm, as calculated from the TEM images of the Up sample and DOWN sample in Figure 1.1.

Figure 1.1: TEM image of a CNDs sample doped with nitrogen, UP, (left), and an undoped pure CNDs sample, DOWN, (right)

The samples were synthesized by members of the Nanostructured Films and Particles group (NFP) which belongs to the Institute of Nanoscience of Aragon (INA) from the University of Zaragoza (Spain) [Ortega-Liebana et al., 2015]. These samples are similar to a material they characterized in Figure 1.2 in terms of PL and absorption spectra showing upconversion, but synthesized in a different way. Therefore, these different systems will be treated as completely unknown and will be started from scratch: the characterization made in this research will be in terms of absorption and emission curves, as well as PL dynamics, to complement the data on this similar material and compare the CNDs doped with nitrogen (UP) with undoped CNDs (DOWN).

Aside from downconverting qualities in the PL, the CNDs sample doped with nitrogen is also likely to have notable upconverting qualities (hence the denotation UP). In this research we will provide this claim with more experimental data. This is meaningful especially since Gan et al. [2013] commented very recently that mistakes can be made in interpreting the normal luminescence excited by the second-order diffraction light of wavelength λ/2 as evidence of

4 Chapter 1. Introduction

upconversion. Artifacts in the setup and alignment can occur, and in addition, no in depth research has yet been published on the upconversion mechanism in CNDs [Gan et al., 2013].

Figure 1.2: UV-Vis absorbance spectrum and Photoluminescence spectra as measured by the members of the Nanostructured Films and Particles group (NFP), which belongs to the Institute of Nanoscience of Aragon (INA) from the University of Zaragoza (Spain). The upconverting carbon dots were measured at room temperature in DI-H2O at different excitation wavelengths. Upconversion is visible for excitation at 740 and 800nm (1, 55

to 1, 68eV). Band gap energy is determined from the extrapolation of the absorbance spectrum. Note that the absorption spectrum does not decay totally and a fraction extends throughout the whole visible range. A LS55 Fluorescence Spectrometer (PerkinElmer) equipped with a xenon arc lamp was used as the light source [Ortega-Liebana et al., 2015].

Zhu et al. [2015] state that there are three main challenges in CNDs. In this thesis there will be focused on two of these challenges: sufficient characterization to illuminate the PL center or PL mechanism and the study of different CNDs together for systematic comparison. Therefore the behaviour of two samples of carbon nanodots in colloidal will be characterized with respect to photoluminescence spectra (PL), excitation dependence of the PL peak (PLP), absorption spectra and PL dynamics. This will be done using advanced spectrometers with different high resolution sensors. Furthermore the optical band gap will be determined by extrapolating the absorption spectrum and experimental data on upconversion will be added to the current data available on upconversion in nitrogen doped CNDs. Since a great variety of CNDs exists, researching the role of doping with various elements can guide the development of effective synthesis routes and novel applications.

Chapter 2

Background Theory

The PL mechanism in CNDs is still an open debate among researchers. In this section the different available theories on PL are introduced, and is explained how quantum confinement makes nanoparticles differ from their bulk counterparts.

2.1

Carbon nanodots

’Carbon Dots’ (CDs) is a term comprising different nanosized materials. In a broad sense, all nanosized materials composed mainly of carbon can be called Carbon Dots. Carbon Dots possess at least one dimension less than 10nm and have fluorescence as major property [Zhu et al., 2015]. Their structure consists of sp2_/sp3 _{hybridized carbon. CDs may carry functional}

surface groups as oxygen/nitrogen based groups or polymeric aggregations. These functional groups can also be fluorophores and be excited directly by the laser. We distinguish 3 different kinds of Carbon Dots: Graphene Quantum Dots (GQDs), that possess one or a few layers of graphene and connected chemical groups on the edges, Carbon Nanodots (CNDs), that are always spherical and do not have a crystal lattice and Carbon Quantum Dots (CQDs) that do have an obvious crystal lattice [Zhu et al., 2015]. This research focuses on CNDs with downconverting optical properties.

2.2

Quantum confinement

As a result of their different structures, the PL center varies for the different types of Carbon Dots. This is due to the diversity in crystal lattice, that is present in GQDs and CQDs and is lacked by CNDs. To understand the origin of the photoluminescence phenomenon, the origin of the electronic band structure and quantummechanics’ fundamentals have to be considered.

6 Chapter 2. Background Theory

2.2.1

The origin of the electronic band structure

The electrons of a single, isolated atom are located in atomic orbitals, each of them having a discrete energy level. As multiple atoms join together to form a molecule, the atomic orbitals of two individual atoms are combined, producing bonding and anti-bonding molecular orbitals at different energy levels. Adding more atoms results in molecular orbitals becoming larger and larger, so that the energy levels become as dense they can be considered as a continuum [Chukwuocha et al., 2012]. This theory is called Linear Combination of Atomic Orbital (LCAO) or Molecular Orbital Theory (MO).

Electronic bands are considered equi-energy lines in momentum space, band-gaps are essentially left-over ranges of energy not covered by any band, a result of the finite widths of the energy bands. The band structure of a material originates from solving Schr¨odingers equation for a single electron relative to a certain position (Γ) in the material [Chukwuocha et al., 2012]. In a crystal the band structure is approximately the same for every unit cell in its lattice away from the edges, in an amorphous structure the band structure depends on the uniformity of the amorphous structure.

A band structure can contain multiple band gaps of different sizes at different energy levels. One band gap, referred to as the band gap of a certain material, has the Fermi energy level in between the energy gap [Bera et al., 2010]. This energy level describes which maximum energy state electrons can be in at zero Kelvin. The band gap exists between the valence band and the conduction band. Almost all electrons naturally occupy the energy levels in the valence band below the band gap, and only very few of them are in the conduction band above the band gap. They can jump to the conduction band when they get additional energy from outside (heat, radiation, etc.) leaving a hole in the valence band [Chukwuocha et al., 2012]. The occupied (bonding) molecular orbital quantum states (equivalent to the valence band) are called the highest occupied molecular orbital (HOMO) levels. The unoccupied antibonding orbitals (equivalent to the conduction band) are called the lowest unoccupied molecular orbital (LUMO) levels [Bera et al., 2010]. The presence and size of the band gap defines if a material is a metal (no band gap), a semi-conductor (small band gap) or insulator (large band gap).

2.2.2

The example of a particle in a box

Quantum confinement can be explained using a simple quantum mechanic example called ’the particle in a box’ model. This model describes a free particle to move in a small space sur-rounded by impenetrable barriers, a three-dimensional box. as the box becomes small enough (on the scale of a few nanometers) quantum effects become important. The particle may only occupy certain positive energy levels, and is more probable to be found at one place than another. In quantum mechanics the behaviour of a particle is described by its wavefunction ψ(x, t), which can be found solving the Schr¨odinger equation for the system. For a particle inside a box of lengths L, a free particle wavefunction is appropriate, but since the probability of finding the particle outside the box is zero, the wavefunction must go to zero at the walls. This constrains the form of the solution to

2.2. Quantum confinement 7

which requires kx,y,z = n_Lx,y,z_x,y,zπ, comparable to string nodes. After normalization, the

wavefunc-tion is ψnx,ny,nz = q 8 LxLy Lzsin( nxπ Lxx)sin( ny π Ly y)sin( nz π Lz z) nx,y,z=1,2,3,... (2.2)

describing the probability of finding a particle confined in a box with lengths L having n different energy levels. These energy levels are inversely proportional with it’s dimensions Lx,y,z and described by

Enx,ny,nz = ¯hω = h2π2¯_2m h (nx Lx) 2 +(ny_Ly)2+(_Lznz)2 i (2.3) So the smaller a particle becomes, the higher the energy levels become.

Nanomaterials like CNDs differ from bulk materials due to their quantized energy levels, in analog to the example of a particle in a box [Chukwuocha et al., 2012]. The electrons in the material do not have all space they need to behave as a free particle, because the material ceases to exists at its edges, creating boundaries for the electrons. Thus the electrons are confined, making the probability of finding an electron at the edges decay to zero. As a result, the energy levels become discrete. This can also be explained by LCAO and MO theory: since the linear combination of atomic orbitals is finite in a C-dot relative to bulk material, the molecular orbitals cannot be considered a continuum, hence discrete energy levels. The occurance of discrete energy levels due to the size of the material is denoted by the term ’quantum confinement’ and occurs when the diameter of a material is of the same magnitude as the De Broglie wavelength of the electron wave function [Dickerson, 2005]. This relation, λ = h

p, describes the wavelength associated with a massive particle like an electron and is an

expression of the wave-particle duality of matter. It’s wavelength is related to its momentum, p, through the Planck constant, h.

Aside from discrete energy levels due to a finite number of molecular orbitals, energy levels also stretch if the electrons are confined. As the size of the particle becomes smaller than a critical quantum measurement, the bulk exciton Bohr radius, the energy levels stretch, creating a size-dependent band gap [Dickerson, 2005]. This is explained in the following subsection discussing the exciton Bohr radius.

The exciton Bohr radius - a critical quantum measurement

The band-gap in a material is the energy required to create an electron and a hole at rest (with zero kinetic energy). If one carrier approaches the other, they may form a bound electron-hole pair, an exciton, whose energy is a few meV lower than the band-gap due to Coulomb interaction [Bera et al., 2010]. This exciton behaves alike a hydrogen atom, except that a hole, not a proton, forms the nucleus. The distance between the electron and hole is called the exciton Bohr radius (rB), and is material and size dependent. The exciton Bohr radius is given

by the relation [Bera et al., 2010]

rB = ¯ h2 e2  1 me + 1 mh  (2.4)

in which me and mh are the effective masses of electrons and holes, respectively,  is the size

8 Chapter 2. Background Theory

charge of an electron. For a nanodot to have quantum mechanical properties, the nanodot should be smaller than the bulk-exciton Bohr radius, in which the optical dielectric constant for the bulk material is used in formula 2.4. If the radius (R) of a nanodot approaches rB,

eventually becoming smaller than it, the motions of electrons and holes are confined spatially to the dimension of the nanodot. This causes an increase of the excitonic transition energy, hence the stretching of energy levels and the blueshift in emitted light as the size of the particle decreases [Bera et al., 2010]. This is illustrated in Figure 2.1 where a comparison is made between the energy levels and band gap in nanoparticles and bulk material.

Figure 2.1: quantum confinement causes the energy levels to stretch, resulting in a larger band gap for nanoparticles than for bulk material [Chukwuocha et al., 2012]

Two detailed theoretical approaches exist that better predict the exciton properties, specifically the effective mass approximation (EMA) model and the tight binding method (TB). Below, we discuss the effective mass approximation model that makes use of the ’particle in a box model’, an analog for CNDs.

2.2.3

Effective mass approximation model

This approach is based on the ’Particle in a box model’, discussed earlier in this section. It is the most widely used model to predict quantum confinement. It was first proposed in 1982 by Efros and Efros and later modified by Brus [Bera et al., 2010]. The EMA model comes from solving the Schr¨odinger equation for an isolated electron and then for an isolated hole in a sphere, and assuming that the effective masses of carriers in the quantum dot are the same as in a bulk semiconductor [Dickerson, 2005].

To obtain the confinement energy at various dots radii the Brus equation can be used. Brus gave the first theoretical calculation for semiconductor nanoparticles (using CdS and CdSe as examples) based on ”effective mass approximation” (EMA) [Chukwuocha et al., 2012]. Equa-tion 2.5 describes the relaEqua-tion between the energy (E) and wavevector (k) of a particle free to assume any position in a box:

E = ¯h

2_k2

2.3. Doping of CNDs 9

In the EMA model, this relationship (equation 2.5), is assumed to hold for an electron or hole in the semiconductor, therefore the energy band is parabolic near the band-edge. The shift of band-gap energy (∆Eg) due to confinement of the exciton in a quantum dot with a diameter

R can be described as followed (equation 2.6), where µ is the reduced mass of an electron-hole pair and E_Ry∗ is the Rydberg energy [Bera et al., 2010].

∆Eg = ¯ h2π2 2µR2 − 1.8e2 R = ¯ h2π2 2R2  1 me + 1 mh  −1.78e 2 R − 0.248E ∗ Ry (2.6)

The first term in the right hand size of this equation represents a relation between the ’particle in a box’ confinement energy and the radius of the quantum dot (R). The second subtractive term stands for the Coulombic interaction energy exciton with a R−1 dependence. The Rydberg energy term is size independent and only significant in case of semiconductor materials with small dielectric constant [Bera et al., 2010]. Based on equation 2.6, the band-gap increases as the size of the quantum dot decreases (dependent on R−2 for the confinement term and on R−1 for the Coulombic interaction term). However, the EMA model breaks down for very small nanocrystals where the potential landscape in no longer periodic [Chukwuocha et al., 2012] and the E-k relationship can no longer be approximated as parabolic [Bera et al., 2010].

Thus we have seen discrete energy levels exist if a material is small enough to have a finite number of molecular orbitals relative to the bulk material and the motion of electrons is spatially confined to the dimensions of the material, in analog with the example of the particle in a box. The quantum confinement effect occurs if the size of a carbon nanodot is of the same magnitude as the bulk exciton Bohr radius, causing the discrete energy levels to stretch, creating a size-dependent band gap.

2.3

Doping of CNDs

Embedding impurities in the crystal lattice of an semiconductor with impurities leads to other PL characteristics for doped (extrinsic) semiconductors than for undoped (intrinsic) semicon-ductors. In this research the Carbon nanodots denoted by UP are doped with Nitrogen, having a proton and electron more than carbon. This provides the lattice with an extra electron. The CNDs doped with nitrogen are N-type semiconductors, where the ”N” stands for ’negative’. N-type semiconductors have negative charged ions or in other words have excess electrons. These electrons are donated to the conduction band, making electrons the major carriers in the N-type semiconductor [Xu et al., 2013]. The impurities add extra electron energy levels near the top of the band gap, so that the electrons can easily be excited into the conduction band. This greatly improves the conductivity of the N-type semiconductor [Xu et al., 2013].

2.4

Corrections of optical emission spectra

In the context of corrections of optical emission spectra (PL) obtained from a spectrometer, we must draw attention to an experimental problem. This problem concerns two alternative plotting modes of the emission spectra. Both representations are possible and widely used

10 Chapter 2. Background Theory

(sometimes even both modes are applied to the same graph, one at the bottom and the other at the upper axis or vice versa, see e.g. figure 2.2) [Ivan Pelant, 2012]. However, plotting the emission spectrum using photon energy is more suitable to analyse the results, because the band-gap energy is related to photon-energy rather than photon wavelength.

Figure 2.2: In plots of photoluminescence are both wavelength scale in nm and energy scale in eV commonly used, each having different elementary step interval [Trinkler and Berzina, 2011]

Conversion between wavelengths and photon energies can be done simply by using the trivial relation

E = hν = hc

λ (2.7)

However, such an approach is not always fully justified. One has to take into account what type of wavelength scan was used during the experiment [Ivan Pelant, 2012]. Considering a grating monochromator with a linear scan in wavelengths λ, the experiment is carried out using a constant slit width ∆λ, the result being the actual spectrum I(λ). However, the elemental step (interval) in photon energy is not kept constant during the measurement, as is clearly seen by differentiating formula 2.7 [Ivan Pelant, 2012].

|d(hν)| = hc

λ2|dλ| (2.8)

Obviously, as the wavelength λ is increased, keeping dλ constant, the interval |d(hν)| drops, because it scales like λ−2. Therefore, if one wishes to plot the emission spectrum against the photon energy scale like |I(hν)|, the decrease in the the magnitude of |d(hν)| should be corrected by multiplying each value of I(λ) by λ2_:

I(hν) = λ2I(λ). (2.9)

This multiplication obviously boosts the long-wavelength part of the spectrum, causing a red-shift in the PL intensity peaks. Thus, in principle it is incorrect to convert simply λ into hν while keeping the spectral curve unchanged. However, the redshift of the peaks is small, and significant only for sufficiently broad spectra (full width at half maximum about 100nm or more) [Ivan Pelant, 2012]. Correcting the PL intensity values by formula 2.9 is therefore particularly a matter of completeness in most cases, and only in case of broad emission peaks at long wavelengths a necessity.

2.5. Determination of band gap energy - Tauc plots 11

2.5

Determination of band gap energy - Tauc plots

To determine the optical band gap, or Tauc gap, of a semiconductor, a Tauc plot can be used. This plot is in particular used to characterize the optical properties of amorphous materials [Stenzel, 2005]. Jan Tauc discovered that the optical absorption spectrum of amorphous ger-manium resembled the spectrum of indirect transitions in crystalline gerger-manium (plus a tail due to the localized states at lower energies) [Stenzel, 2005]. In Figure 2.3 is shown that the band structure has not always parabolic bands, as the effective mass approximation (EMA) and the Tauc plot assume, but has besides these extended states also localized states with a lower band-gap energy.

Figure 2.3: The band gap is not always parabolic as the EMA model and the Tauc plot assume (left), but localized states and different transitions also exist [Stenzel, 2005].

Tauc proposed an extrapolation to find the optical gap of the crystalline-like states. First the absorption coefficient a is calculated using the following formula:

Absorption = e−a∗d (2.10) in which d is the thickness of the cuvette [Stenzel, 2005]. Then the direct allowed transitions (r = 1₂) are calculated using

(a ∗ h ∗ e)1r (2.11)

in which a is the absorption coefficient, h is the constant of Planck and e is the photon energy of excitation. The resulting plot has a distinct linear regime which denotes the onset of absorption. Thus, the extrapolation of this linear regime to the excitation energy of the tail formed by the localized transitions yields the energy of the optical band gap of the material. An example of this extrapolation is shown in Figure 2.4 [Stenzel, 2005].

Figure 2.4: An example of a Tauc plot, absorption as (a ∗ h ∗ e)2 _{set out to the quantity of photon energy,}

Chapter 3

Methods

Experimental data on the CNDs was retrieved using various setups to conduct measurements on photoluminescence (PL), excitation dependence of PL peak (PLP), absorption spectra and PL dynamics. In this chapter the different setups, lasers/lamps and spectrometers are explained. Furthermore attention will be given to the use of several applied filters in the excitation and/or detection part of the setup.

3.1

Photoluminescence spectra

Photoluminescence (PL) measurements were conducted using an Optical Parametric Oscillator II (OPO II) of the type LG350 for the excitation part. The process of optical parametric oscillation is the most promising technique for achieving high-power laser output in a very wide spectral range. Neither of the lasers available can ensure such a wide tuning range [Solar Laser Systems]. With OPO, the pump radiation at 3, 6eV is converted to radiation of two lower frequencies: a signal wave and an idler wave, polarized orthogonal relative to each other. A nonlinear Beta Barium Borate (BBO) crystal is used to create different frequencies for the signal and idler wave. The signal wave and the idler waves are polarized orthogonal to each other, which makes it possible to separate the waves using linear polarizers in front of the output of the OPO [Solar Laser Systems].

To obtain data on upconversion of the UP sample, excitation between 800 to 600nm (1, 6 eV to 2, 1 eV) was necessary. The setup allowed us to excite the sample to a maximum of 1, 8eV with sufficient power. An additional setup had been invented to block irradiation of the basic laser in the signal wave when measuring at energies lower than 2, 1eV (see Figure 3.1). The laser first reflected on a mirror and was directed through a prism to refract the beam. Due to the difference in refraction between different wavelengths, resulting from a different refractive index of the material for different colours (dispersion), the basic laser could be separated from the signal wave. Thereafter the light was focused and filtered by a 2, 5eV low pass filter to exclude high energy photons of the OPO to reach the sample and be mistakenly interpreted as upconversion. After passing the filter, the light was directed into an optic fiber, having a strong convex lens at its other end directly in front of the sample to create light of high intensity. Since changing the excitation wavelength means different refractive angles for the prism, this setup allowed us to change the excitation wavelength and excite the sample by adjusting only one

3.1. Photoluminescence spectra 13

mirror instead of several. Two other setups without this key solution had been tried earlier. Using the optical fiber was important, because optimizing the signal by aligning the setup is possible only by using visible light. The region of 2, 5eV light was used to align by emission of the sample, since in this region the output of the laser was powerful enough to assure a short exposure time. In the near infrared region at 1, 8eV the exposure time of the sensor would have been too long to align the setup.

Emission of the sample was collected orthogonal to the direction of the laser using two large focusing lenses. The light was directed into an optical fiber connected to an M266 monochroma-tor & spectrograph from Solar Laser Systems. The incoming light first passed through a filter of choice in the monochromator to exclude scattering of the laser on the sample if necessary. In the monochromator a grating is used to separate the different wavelengths before entering the CCD.

A charge-coupled device (CCD) array (type S10141 − 1108), having 2068 light sensitive sensors in 430nm range (2, 9 eV). The wavelength corresponding to the center of the CCD could be altered by adjusting the center-wavelength using the software of the monochromator to change the angle of the grating. A grating does not only split the signal in different colours, but also creates second- and higher order maximums by diffraction. To avoid measuring second-order maximums, the center-wavelength and the number of grooves of the grating could be adjusted. A schematic sketch of the setup used is included in Figure 3.1.

Figure 3.1: A schematic sketch of the setup used for photoluminescence spectra measurements. The setup included an OPO II, a monochromator and a CCD measurement device and prevented high energy light of the basic laser to reach the sample using a prism.

14 Chapter 3. Methods

3.2

PL Dynamics

Measurements on the PL dynamics of carriers were conducted using the same setup as for PL measurement, aside a photomultiplier tube (PMT) was used to accurately measure the decay process of the signal (type C9143 − 02). A PMT accelerates and multiplies electrons excited by photons using several electric fields. The incident photons first strike a photo cathode, where after electrons are being ejected from the surface as a consequence from the photoelectric effect. These electrons are focused and directed to the electron multiplier. This multiplier consist of several electrodes (called dynodes) each of them held at increasingly positive potential than the preceding one, creating electric fields that increase in strength as the electrons travel. Each initial electron creates a small group of primary electrons that accelerate to the next dynode due to the electric field, which create more groups of electrons, and so on. The electrons are finally collected on an anode and thus an amplified signal is generated. PMT detectors can multiply the current produced by incident light by as much as 10 to 108 times [Hamamatsu, 2007]. A schematic sketch of a PMT is seen in Figure 3.2.

Figure 3.2: A schematic sketch of a PMT in which electrons ejected by the photoelectric effect are multiplied using several increasingly strong electric fields. [Nuclear Security and Safeguards Education Portal , NSSEP].

The samples were measured using laser pulses with energy of 4, 13eV and 3, 18eV to excite the samples. For excitation at 4, 13eV a low pass emission filter of 3, 5eV was applied, while for excitation at 3, 18eV a low pass emission filter of 3, 1eV was applied. Both filters diminished the influence of scattered light of the laser by a vast amount, which was important since emission of the sample produced a weak signal. PL was also measured for these excitation energies using the OPO, CCD and the same filters to complement the PL dynamics measurements.

3.3

Excitation dependence of PL peak

Photoluminescence excitation (PLE) measurements were conducted using Fluorolog-3 Spec-trofluorometer of Horiba. This machine uses a xenon flash lamp that can produce intense ultraviolet, visible and infrared light. An internalized monochromator with two gratings cre-ates monochrome light which is selected by letting it pass through a slit. Thereafter the light is directed into the sample chamber. The emission is collected orthogonal to the incident beam, directed into another monochromator and processed by a Synapse CCD Detection System (S/N M CD-1950BR-3115). The data acquired is a 3D plot that shows for every excitation energy a graph of the PL spectra. The PL spectra is a graph of the emission energy versus emission intensity.

3.4. Absorption 15

A 3, 1eV low pass filter was applied as emission filter to exclude scattered light of the lamp on the sample that ranged from 3, 18 to 4, 6eV, all higher energies. The center wavelength was set at 2, 58eV and the step size for excitation wavelength has been set at 3nm. The sample with nitrogen doped CNDs (UP) was measured two times more between 1, 7 to 1, 9eV to obtain data on the upconversion properties of the sample. In the first measurement no excitation filter was applied, whereas in the second measurement a 2, 5eV low pass filter was applied to ensure no light of the second order maximum of the grating inside the monochromator reached the sample. That way, the sample could not be excited by light of this energy and occurring downconversion be falsely interpreted as proof for upconversion.

3.4

Absorption

A Lambda 950 of PerkinElmer was used to measure transmittance of the samples. This machine uses a D2 and a Tungsten lamp as light source to generate a broad spectrum of wavelengths. A two-grating monochromator is thereafter used to produce a coherent light source. The beam it generates is split by a mirror in a sample beam and reference beam and is directed to a beam selector before it passes to the sample chamber. Both the sample and the reference beam have an entrance to a sphere in the detection part, which is why a beam has to be selected to measure one quantity at a time. The sphere created a diffuse light source with a homogeneous intensity distribution of the incident light. The intensity of the light is then measured by a photomultiplier tube (PMT). A baffle prevented direct light of the D2 and Tungsten lamp to reach the PMT.

Lambda 950 provided data on the transmittance of the samples. The absorption was obtained by the simple formula absorption = 1 − transmission, neglecting light that was reflected by the cuvette of the sample. Influence of emission of the sample was diminished by placing the sample as far away as possible from the sphere: emission of the sample is seen by the detector as higher transmittance and thus perturbs the measurement. Since emission is spread homogeneously over all directions, creating a slit by placing the sample away from the sphere diminishes the influence of the emission on the obtained data.

Corrections on the data were made using a 0% blocked beam baseline and a check with no sample, 100% transmittance, was made to see if the data obtained was to be trusted. Also a separate cuvette with only solvent, H2O, was measured. This makes it possible to extract the

data on the absorption of the CNDs from the measurements of the solvent. The absorption of the CNDs was calculated by subtracting the absorption of the solvent from the absorption of the samples.

Chapter 4

Results

In this chapter results of the research carried out will be presented. The characterization of CNDs comprises measurements of photoluminesence excitation, PL dynamics, absorption and PL spectra of the samples. These results will be displayed using different plots, where attention will be given to the manner in which the data presented has been processed.

4.1

Excitation dependence of PL peak

Data on the excitation dependence of the PL peak was obtained using the Fluorolog-3 spec-trofluorometer. This machine simultaneously measured the photoluminescence intensity and the emission spectrum of the sample for several excitation energies. The raw data can be seen as a 3D-plot, multiple graphs at different excitation energy of PL intensity versus emission energy. These graphs were each smoothed using the function smooth of Matlab which uses a FFT filter. Thereafter the excitation wavelength is converted to energy scale using formula 2.7 and the data columns of photoluminescence intensity are converted to energy scale using formula 2.9 for reasons explained in the paragraph corrections on optical emission spectra in the theoretical section. Then the peaks of PL intensity for excitation energy were calculated, creating a new data set. The plot of this data set is presented below in figure 4.1 and shows us the position of the intensity peaks in the plane of emission energy versus excitation energy- a graph of the excitation dependence of the PL peak.

The emission energy grows with growing excitation energy until a certain point at approximately 3, 9eV, albeit a lower emission energy than excitation energy. Until 3, 9eV UP and DOWN behave in the same manner. After it, both samples start emitting light with decreasing energy for increasing the excitation energy. However, UP emits lower energy light than DOWN for the same excitation energy greater than 4, 0eV.

4.2. PL Dynamics 17

Figure 4.1: Excitation dependence of PL peak (PLP) of a CND sample doped with nitrogen (Up) and an undoped CND sample. UP and DOWN emit lower intensity light than their excitation energies, increasing to 3, 9eV, decreasing after it.

4.2

PL Dynamics

PL decay of the samples was measured using the OPO II and the PMT. The samples were excited with two different excitation energies, one of which in the regime where the PL peaks of the samples are close to the other (3, 5eV), one of which in the regime where the PL peaks are far from the other (4, 13eV) as seen from figure 4.1. Exciting with 3, 5eV also has a strong advantage since this is the output frequency of the basic laser, surpassing the OPOII and gaining stable high density power laser light while exciting the sample. This is the reason why PL and PL decay at 3, 5eV have a much higher signal to noise ratio than the measurements taken at 4, 13eV excitation energy. Detection energies have been chosen at 2, 5eV, 2, 7eV and 2, 9eV, of which PL decay was most clearly visible at 2, 9eV detection energy.

The data has been fitted using the slow decay regime because the decay of the laser was of the same magnitude as the decay of the emission of the sample in the first 600ns. In Figure 4.2 two plots of PL decay are shown for excitation energy of 3, 5eV and 4, 13eV. In the graphs the experimental data and the fit through the data are shown, both normalized by the tail of the fitted curve. The curves of the two samples have been synchronized by emission intensity peak. Using the fitted curve lifetime of carriers could be calculated. The results are shown in Table 4.1. From the table it can be noticed that UP has a longer decay time than DOWN, 45ns when exciting at 3, 5eV and 34ns when exciting at 4, 13eV.

To complement the data on PL dynamics presented in Figure 4.2, two PL measurements were taken at the same excitation energies, 3, 5eV and 4, 13eV using the Fluorolog Spectrofluorom-eter. These are shown in Figure 4.3. The PL of the samples differs most in the region 2, 4eV to 2, 6eV.

18 Chapter 4. Results

Figure 4.2: PL dynamics of nitrogen doped CNDs (UP) and pure CNDs (DOWN), excited with 3, 5 and 4, 13eV and measured in nanoseconds at 2, 9eV emission energy. The experimental data and the fit through the data are visible in the graph. Experimental data is normalized by the tail of the fitted curve.

Lifetime of carriers 3,5 eV 4,13 eV UP 604ns 653ns DOWN 559ns 619ns

Table 4.1: Lifetime of carriers as calculated by fitting of the slow decay regime

Figure 4.3: Photoluminescence intensity of a CND sample doped with nitrogen (Up) and a pure CND sample at 4, 13eV, 3, 5eV excitation energy respectively.

4.3

PL and Absorption Spectra

In Figure 4.4 four PL spectra at different excitation energies and the absorption spectrum with estimated band gap energy for one sample are displayed. The experimental PL spectra are shown for excitation energies between 3, 20eV and 4, 59eV.

4.3. PL and Absorption Spectra 19

Figure 4.4: PL and absorption spectra (Tauc plot) of DOWN (left) and UP (right). The cross section of the extrapolated small dotted line of the linear regime with the x-axis yields the optical band gap [Stenzel, 2005]. For DOWN this is at approximately 2, 43eV, while for UP this is at 2, 8eV. The PL spectra correspond with the left y-axis and the absorption spectra correspond with the right y-axis.

Transmittance was measured using a Lambda 950 from PerkinElmer. Absorption of CNDs was obtained by subtracting the absorption of the solvent from the absorption of the sample using ABS of CN Ds = ABS of sample − ABS of solvent in which ABS stands for absorption. This formula neglects the influence of reflection. The optical band gap can be estimated by plotting absorption as (a ∗ h ∗ e)2 and extrapolating the linear regime to the absorption of the tail, as explained in Tauc Plots in the theoretical section. Formula’s 2.10 and 2.11 have been used to create the plot for absorption in Figure 4.4. Extrapolating the linear regime with the highest slope yields the optical band gap. The extrapolation crosses the x-axis at the band gap energy. For DOWN this is at approximately 2, 43eV and for UP this is at approximately 2, 8eV. In Figure 4.5 a graph of the absorption for both samples can be seen, showing a rise in absorption in the region of 3, 0eV ±0, 5 and showing a distinct absorption peak at excitation of 3, 8eV for nitrogen doped CNDs, which is absent for undoped CNDs. It is remarkable that this peak is at about the same energy as the energy the PL peaks of the samples start to differ (see PLP at 3, 8eV, Figure 4.1).

Figure 4.5: Absorption spectra of DOWN (pure CNDs, red) and UP (N-doped CNDs, black). Absorption of the CNDs was acquired by subtracting the absorption of the solvent from the absorption of the sample. A distinct absorption peak is visible for doped CNDs at 3, 8eV.

20 Chapter 4. Results

4.4

PL spectra - obtained data on upconversion

Using Fluorolog spectrofluorometer UP was excited with 1, 77eV and its PL spectrum was mea-sured. A measurement with and without an excitation filter was made to prevent high energy light of the lamp, light of the second order maximum of the grating in the monochromator, to reach the sample. The results are presented in Figure 4.6. Both signals show a horizontal line at 0, 0 intensity counts with a few sudden peaks. The excitation filter let light with energy lower than 2, 5eV pass, resulting in two measurements under the same conditions for energies lower than 2, 5eV.

Data on Upconversion was also obtained using the OPO II and CCD, again resulting in a horizontal line at 0, 0 intensity counts. This matched perfectly the background signal at which excitation and emission shutter were closed. However, this data is left out because an artifact was present in the setup the day before the measurement (every curve including ’solvent’ had a positive slope for increasing excitation energy). The measurement the day after was performed without changing the setup, but this time the setup did not show the artifact. Below the data from the Fluorolog spectrofluorometer is shown.

Figure 4.6: Photoluminescence spectra of nitrogen doped CNDs (UP) with and without low pass excitation filter excited with 2, 5eV measured using the Fluorolog.

Chapter 5

Discussion

In this chapter obtained experimental data is discussed and placed in context of current scien-tific research. Attention will be given to differences between the samples that are remarkable, leading to a good comparison of the samples. The results for the different measurements will be discussed below.

Figure 4.1 in the results shows the data obtained on PL excitation. Emission energy grows with growing excitation energy until a certain point at approximately 3, 9eV. In this first re-gion downconversion is visible, exciting with 3, 6eV for example results in emission of lower energy, 2, 85eV. Unto a peak in emission energy at 3, 9eV UP and DOWN behave in the same manner. After this peak, both samples start emitting photons with lower energy as excited by higher energy photons, showing increasingly strong downconversion. However, UP emits lower energy light than DOWN for the same excitation energy greater than 4, 0eV and thus UP shows stronger downconversion than the undoped CND sample (DOWN). Both samples have a size distribution that plays a role in the interpretation of Figure 4.1. Intensity peaks at high emission energy are caused by small excited CNDs (large band gap), while intensity peaks at low emission energy are caused by large excited CNDs (small band gap).

In Figure 4.2 the PL dynamic is shown for excitation energies 3, 5 and 4, 13eV measured at 2, 9eV detection energy. The values for the lifetime of carriers extracted from this data are presented in Table 4.1. The values show that UP has a longer decay time than DOWN: 45ns when exciting at 3, 5eV and 34ns when exciting at 4, 13eV. Doping with nitrogen leads to the addition of extra energy levels, which would cost the carriers to spend more time to reach the Lowest Unoccupied Molecular Orbital (LUMO). However, cooling of carriers is typically in the range of picoseconds and cannot be held accountable for this difference in carrier lifetime. A lot of factors play a role in the transition speed of carriers across the band gap, among other things the surroundings, presence of doping atoms or defects and coupling, through which it is difficult to give an adequate explanation for the longer lifetime of carriers in nitrogen doped CNDs. Two PL spectra at the same excitation energies used for PL dynamics are seen in Figure 4.3. Exciting at 3, 5eV results in a better signal to noise ratio then exciting at 4, 13eV since 3, 5eV is emitted by the basic laser, creating a stable light source with high power den-sity. Consistent with the plot of the PL peaks is visible that the PL peak differs if excited at 4, 13eV and is almost equal if excited at 3, 5eV. PL differs mostly in the region between 2, 4

22 Chapter 5. Discussion

and 2, 6eV, where UP emits stronger light than DOWN. This is only a small difference, because no sample can be exactly reproduced during synthesis. So based on the PL spectrum at these excitation/detection energies could be said that both samples behave alike.

PL and absorption spectra are visible in Figure 4.4. Four PL spectra at evenly spaced energy intervals are chosen for both samples, showing the change in the shape of PL spectra when changing the excitation energy. This shows the contribution of small and large CNDs. We can see that for low excitation big CNDs are dominant and when excitation energy rises smaller CNDs with larger band gaps become dominant. However, when excitation energy is higher, we may have energy transfer from small CNDs to big CNDs, which is why at 4, 59eV the PL peak has moved again to lower energy.

In the plot the absorption spectrum in (A ∗ h ∗ e)2 _{is visible with accompanying extrapolation}

to estimate the band gap energy (Tauc). The estimation of the band gap energy for DOWN is 2, 43eV and for UP 2, 8eV, as noted next to the extrapolation in the graphs. There is an error on the determined band gap energies due to not taking reflection into account when calculating absorption using the measured transmittance. Extrapolating the band gap energy from the absorption spectrum is in principle not very accurate, since it involves all kinds of transitions across the band gap, and not only the one from the bottom of the conduction band to the top of the valence band. Nevertheless, the extrapolation of the absorption spectrum suggests UP has CNDs with a larger band gap energy than DOWN, leading to the expectation UP emits higher energy photons than DOWN. Looking at the PL intensity peaks presented in Figure 4.1 this contradicts, since the peaks of UP are almost consequently at lower emission energy than DOWN. Figure 4.1 gives better information about the band gap energy, because the peak of PL spectra shows us only the smallest transitions across the band gap, those between the Highest Occupied Molecular Orbital (HOMO) state and LUMO state. Moreover, extrapolation of the band gap energy using Tauc plots does not take size distribution of the CNDs into account. Theory of CNDs doped with nitrogen suggests that for the same average size nitrogen doped CNDs would have a smaller band gap than pure CNDs. Doping with nitrogen leads to the addition of extra electron energy states near the top of the conduction band, reducing the size of the band gap. On the other hand there is a difference in the average sizes of the CNDs between the samples, affecting the optical band gap as well (quantum confinement). The average size of the CNDs in UP is 2.2 ± 0.3nm and in DOWN 3.4 ± 0.4nm, as calculated from the TEM images of the samples in Figure 1.1. The smaller the CNDs are, the bigger the band gap of the nanocrystal is. Thus two processes are influencing the optical band gap, one enlarging it (smaller sizes) and one reducing it (doping with nitrogen).

It happens UP has been doped with nitrogen and has smaller sizes. Thus quantum confine-ment and doping with nitrogen work against each other in influencing the optical band gap. Therefore we cannot expect one sample to have a larger optical band gap than the other based on theory. Experimental work however shows which process influences the optical band gap to a greater extend. We have seen extrapolation of the band gap using Tauc plots gives contra-dictory information with the plot of the PL peaks, and that a large error is expected on the information of the Tauc plots for reasons stated above. Thus we base our conclusions on the information resulting from the PL peaks. The plot of the PL peaks shows UP emits light of lower energy and thus has a smaller band gap than DOWN. Therefore, doping with nitrogen has a strong influence on the band gap energy.

23

Obtained data on upconversion using Fluorolog Spectrofluorometer as well as OPO II and PMT suggest no upconversion is present, contrary to the results of the similar sample (Figure 1.2). The spectra shown in Figure 4.6 are measured with and without excitation filter to exclude a possible artifact in the setup. The filter prevented high energy photons from the second order maximum of the grating to reach the sample. Figure 4.6 shows that the spectra are similar. Thus the light used for excitation did not contain high energy photons from the second order maximum of the grating. No proof for upconversion can be derived of the experimental results presented in this research.

Chapter 6

Conclusion and prospectives

In this thesis a sample of undoped CNDs and a sample of nitrogen-doped CNDs are char-acterized in terms of photoluminescence spectra (PL), excitation dependence of the PL peak (PLP), absorption spectra and PL Dynamics. With respect to PLP we have seen the samples behave different in the high energy regime (3, 9-4, 8eV). UP emits light of lower energy than DOWN and thus shows stronger downconversion properties. Extrapolation of the absorption spectrum using Tauc plots yields energy values for the band gaps of UP at 2, 8eV and DOWN at a lower band gap energy of 2, 43eV, contradicting experimental data on PLP. This can be explained by strong reduction of band gap energy caused by doping with nitrogen and the fact that evaluation of band gap energy using Tauc plots gives a rough estimation. Investigating the band gap energy at higher significance can be a goal for future research. With respect to PL dynamics we have seen PL of UP decays between 35 and 45ns slower than that of DOWN. Obtained experimental data on upconversion shows no evidence for upconversion. This may be due to too low power density in the excitation part at low energies. Further research on the existence of upconversion is being encouraged. Providing experimental data on nitrogen doped CNDs and making a systematic comparison of these N-doped CNDs with an undoped CND sample can guide the development of effective synthesis routes and novel applications of semiconductor materials.

Bibliography

Debasis Bera, Lei Qian, Teng-Kuan Tseng, and Paul H. Holloway. Quantum dots and their multimodal applications: A review. Materials, 3(4):2260, 2010. ISSN 1996-1944. doi: 10.3390/ma3042260. URL http://www.mdpi.com/1996-1944/3/4/2260.

Chukwuocha et al. Theoretical studies on the effect of confinement on quantum dots using the brus equation. World Journal of Condensed Matter Physics, 2:96–100, 2012.

B.D. Dickerson. Organometallic synthesis kintetics of CdSe Quantum Dots. PhD thesis, 4 2005. Zhixing Gan, Xinglong Wu, Gengxia Zhou, Jiancang Shen, and Paul K. Chu. Is there real upconversion photoluminescence from graphene quantum dots? Advanced Optical Materials, 1(8):554–558, 2013. ISSN 2195-1071. doi: 10.1002/adom.201300152. URL http://dx.doi.org/10.1002/adom.201300152.

Hamamatsu. Photomulitiplier Tubes - basics and applications, 2007. URL https://www.hamamatsu.com/resources/pdf/etd/PMT handbook v3aE.pdf.

Jan Valenta Ivan Pelant. Luminescence Spectroscopy of Semiconductors. OUP Oxford, 2012. Peng Miao, Kun Han, Yuguo Tang, Bidou Wang, Tao Lin, and Wenbo Cheng. Recent advances

in carbon nanodots: synthesis, properties and biomedical applications. Nanoscale, 7:1586– 1595, 2015. doi: 10.1039/C4NR05712K. URL http://dx.doi.org/10.1039/C4NR05712K. Nuclear Security and Texas University Safeguards Education Portal (NSSEP). Scintillation

detectors. URL http://nsspi.tamu.edu/nssep/courses/basic-radiation-detection/ gamma-ray-spectroscopy/scintillation-detectors/scintillation-detectors. [On-line; accessed February 8, 2016].

M. C. Ortega-Liebana, J. L. Hueso, A. Larrea, V. Sebastian, and J. Santamaria. Feroxyhyte nanoflakes coupled to up-converting carbon nanodots: a highly active, magnetically recov-erable, fenton-like photocatalyst in the visible-nir range. Chem. Commun., 51:16625–16628, 2015. doi: 10.1039/C5CC05387K. URL http://dx.doi.org/10.1039/C5CC05387K.

Solar Laser Systems. LG350 Second Harmonic Generator for Optical Parametric Oscillators -User’s manual.

Olaf Stenzel. The Physics of Thin Film Optical Spectra: An Introduction. Springer, 2005. Laima Trinkler and Baiba Berzina. Luminescence properties of aln ceramics and its

potential application for solid state dosimetry. 3, 2011. doi: 10.5772/18658. URL http://www.intechopen.com/books/advances-in-ceramics-characterization-raw-ma terials-processing-properties-degradation-and-healing/luminescence-properties -of-aln-ceramics-and-its-potential-application-for-solid-state-dosimetry.

26 BIBLIOGRAPHY

Yang Xu, Ming Wu, Yang Liu, Xi-Zeng Feng, Xue-Bo Yin, Xi-Wen He, and Yu-Kui Zhang. Nitrogen-doped carbon dots: A facile and general preparation method, pho-toluminescence investigation, and imaging applications. Chemistry A European Jour-nal, 19(7):2276–2283, 2013. ISSN 1521-3765. doi: 10.1002/chem.201203641. URL http://dx.doi.org/10.1002/chem.201203641.

Shoujun Zhu, Yubin Song, Xiaohuan Zhao, Jieren Shao, Junhu Zhang, and Bai Yang. The photoluminescence mechanism in carbon dots (graphene quantum dots, carbon nanodots, and polymer dots): current state and future perspective. Nano Re-search, 8(2):355–381, 2015. ISSN 1998-0124. doi: 10.1007/s12274-014-0644-3. URL http://dx.doi.org/10.1007/s12274-014-0644-3.