FACULTY OF SCIENCE
DEPARTMENT OF PHYSICS
DISSERTATION
Down- and Up-Conversion
in Fluorozirconate-Based Glasses
and Glass Ceramics
for Photovoltaic Application
BERND AHRENS
PADERBORN
2009
Abstract
Sol omnibus lucet.
Satyricon, 100
TITUS PETRONIUS
Mono-gap solar cells, like commercial silicon solar cells, are unable to use the
whole solar spectrum. In particular, photons with high energy have thermal-
ization losses and photons with an energy lower than the bandgap energy can
not be absorbed. Materials, which convert one UV photon into one or more
visible photons, so called down-converters, or which convert two or more
sub-bandgap photons into photons with an energy higher than the bandgap
energy, so called up-converters, are of great interest for photovoltaic applica-
tions.
In this work new materials were investigated for their optical properties and
their potential as down- or up-converters.
For down-conversion applications, barium chloride and barium bromide sin-
gle crystals, as well as fluorozirconate-based (FZ) glasses, were doped with
samarium. The glasses were additionally doped with bromide ions in order
to initiate the formation of barium bromide nanocrystals in the glass upon
thermal processing. Optical measurement techniques determined the divalent
charge state of samarium in both single crystals. The FZ glasses show a differ-
ent behavior: samarium enters the glass matrix either in its divalent or in its
trivalent state. Fluorescence measurements indicate that during the annealing
process Sm2+ions enter the nanocrystals leading to enhanced fluorescence ef-
ficiency and to changes in the fluorescence lifetime.
For up-conversion applications, BaCl2single crystals, as well as FZ-based glas-
ses, were doped with neodymium. Upon excitation at 796 nm, Nd3+-doped
BaCl2single crystals show several up-converted fluorescence bands in the
visible spectral range, with the most intense bands at 530, 590, and 660 nm,
in addition to the typical fluorescence bands in the infrared spectral range.
The power dependence of the infrared fluorescence and of the two-photon
iii
up-conversion fluorescence intensities as well as the corresponding radiative
lifetimes have been investigated. An enhanced up-converted fluorescence in
Nd3+-doped fluorozirconate (FZ) glasses which were additionally doped with
chlorine ions was found. Upon annealing between 240 and 290 ◦C, hexagonal
phase BaCl2nanocrystals between 20 and 180 nm in diameter were formed
in the glass. During thermal processing, some of the Nd3+ions enter the
nanocrystals leading to additional splitting of the up-converted fluorescence
and infrared fluorescence spectra.
Nd-doped glass ceramics are useful as a model system, but are not applicable
due to the excitation of the up-conversion at ∼800 nm, which is light that can
be absorbed by a silicon solar cell itself. However, erbium-doped FZ glasses
were found to be more applicable systems for up-conversion-based silicon so-
lar cells due to their excitation at 1540 nm. To show their potential the external
quantum efficiency (EQE) of a commercial monocrystalline silicon solar cell
with an Er-doped FZ glass on top of it was determined. For an excitation
power of 18 mW at a wavelength of 1540 nm an EQE of almost 1.5 % was
found for a 5 mol % Er-doped FZ glass.
iv
Contents
1 Introduction 1
2 Basics - Physical Background 3
2.1 PhotonConversion.......................... 3
2.1.1 Down-Conversion and Quantum Cutting . . . . . . . . . 3
2.1.2 Up-Conversion ........................ 5
2.2 Glasses and Glass Ceramics . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Phonons............................ 9
2.2.2 ZBLANglasses ........................ 10
2.2.3 Thermal Processing and Nano-Particles . . . . . . . . . . 14
2.3 RareEarthIons ............................ 14
2.4 AnalyzingMethods.......................... 16
2.4.1 Photoluminescence . . . . . . . . . . . . . . . . . . . . . . 16
2.4.2 X-Ray Diffraction . . . . . . . . . . . . . . . . . . . . . . . 18
3 Down-Conversion based on Sm 21
3.1 Simulations .............................. 21
3.2 Dopant:Samarium .......................... 23
3.2.1 Motivation........................... 23
3.2.2 Barium Chloride and Barium Bromide . . . . . . . . . . . 24
3.2.3 Zirconium Fluoride Glass and Glass Ceramics . . . . . . 28
3.3 Discussion............................... 36
4 Up-Conversion based on Nd and Er 39
4.1 Simulations .............................. 39
4.2 Dopant:Neodymium......................... 40
4.2.1 Motivation........................... 40
4.2.2 Barium Chloride . . . . . . . . . . . . . . . . . . . . . . . 41
4.2.3 Fluorozirconate Glasses and Glass Ceramics . . . . . . . 49
4.2.4 Discussion........................... 70
4.3 Dopant:Erbium............................ 74
4.3.1 Motivation........................... 74
4.3.2 Fluorozirconate Glasses . . . . . . . . . . . . . . . . . . . 74
4.3.3 Discussion........................... 86
5 Conclusion 89
v
1 Introduction
One of today’s major challenges is the energy managing a supply of power for
an growing population. Due to the limited supply of energy sources, like coal,
oil, gas, or uranium, a replacement of most currently used power supplies with
renewable energy sources is paramount. Even in the case of an infinite amount
of these fossil fuels, using them is harmful to the earth due to the creation of
global warming. Today, about 28 ·1012 kg of CO2per year [1], a daily output
of 76 million tonnes, are released in the atmosphere. “First effects of global
warming are already visible and determined action is required to restrict the
negative consequences for mankind, the environment and subsequent gener-
ations. Renewable energies are essential contributors to the energy supply
portfolio as they provide opportunities for mitigating greenhouse gases” [2].
The renewable energy sources include water, geothermal, wind, and solar [1].
Solar cells convert solar radiation into electrical energy, by the photovoltaic
effect, first reported by Becquerel [3] in 1839. Today a commercial silicon so-
lar cell has an efficiency of around 22 % [4]; ratio of incident to useable energy.
The two major loss mechanisms (see figure 1.1) are thermalization losses due to
high energy photons and transmission losses of sub band gap photons. Ther-
Figure 1.1: Solar spectrum AM1.5. The red area can be used by a commercial silicon
solar cell to generate current. The number of photons per square-meter, second and
nanometer are plotted versus the wavelength.
1
Chapter 1. Introduction
Figure 1.2: Bifacial solar cell with down- and up-converter.
malization losses can be reduced significantly in a solar cell when more than
one electron-hole pair is generated per incident photon. A maximum EQE of
nearly 40 % for a silicon solar cell can be theoretically achieved [5] by using a
down-converter with a high band gap solar cell. With an up-converting layer
it is possible to use sub band gap light that cannot be absorbed by the solar
cell to generate high energy photons that can be absorbed. Trupke, Green, and
W¨
urfel found the upper EQE limit of up-conversion solar cell systems can be
enhanced to 47.6 % for single junction cells [6]. Gibart et al. [7] reported in 1996
the first application of an up-converter on a bifacial gallium arsenide solar cell.
Seven years later Shalav et al. [8] demonstrated the first up-converter on a sili-
con solar cell. A schematic system containing, down- and up-converter with a
bifacial solar cell is shown in figure 1.2.
A brief introduction to the background on down- and up-conversion, rare
earth ions, fluorozirconate based glasses and thermal treatment is given in
chapter 2 along with methods like photoluminescence and x-ray diffraction.
In chapter 3 the results of Sm-doped materials for down-converting processes
are shown. Part one deals with BaCl2and BaBr2single crystals doped with
divalent Sm. These investigations are the basis for the glass ceramics in part
two, where BaCl2and BaBr2nanoparticles are formed in the glasses to enhance
their optical properties. The results of the Nd-doped single crystals (BaCl2)
and glass ceramics are presented in chapter 4 section one. In the second sec-
tion results on erbium doping are presented. Conclusions are presented in in
chapter 5.
Parts of the present work have already been published; a publication list is
added at the end of this thesis in the appendix.
2
2 Basics - Physical Background
2.1 Photon Conversion
2.1.1 Down-Conversion and Quantum Cutting
The process of down-conversion or down-shifting describes the conversion of
one high-energy photon (e.g. UV-photon) into one photon with a lower energy
(e.g. a visible photon) - normally called photoluminescence (PL). By absorbing
a photon an optically active center (e.g. an ion) goes into an excited state. By
emitting one photon with the same or less energy compared to the exciting
one the ground state can be reached. Quantum cutting occurs if one high-
energy photon is converted into two or more photons with lower energies. In
1974 Sommerdijk et al. [11] and Piper et al. [12] observed after a high energy
excitation a subsequent emission of two photons (photo cascade emission) in
YF3:Pr3+and NaYF4:Pr3+. A luminescent quantum efficiency of nearly 140 %
was observed for this process. Quantum cutting can be realized by a single
ion or by a combination of centers due to energy transfers. The first case is
based on one optically active center with three energy levels. Excited to state
E2, the ion subsequently emits two photons to relax stepwise into the ground
state (see figure 2.1 (a)). The second possibility is to obtain quantum cutting
uses two ions. The excited ion (I) can relax from state E2into E1by energy
transfer to the second ion (II), shown in figure 2.1 (b). After the energy transfer
both ions can emit a photon to reach the ground state. The energy resonance
condition had to be fulfilled - meaning that the energy difference for the energy
transfer transitions in both ions have to be equal. A well known example for
this type of quantum cutting is an Eu-Gd-system, which is shown in figure 2.2
[9, 10]. In the third concept, shown in figure 2.1 (c), quantum cutting occurs by
the use of three optically active centers. Ion (I) is excited by a photon. Through
a co-operative sensitization the energy can be transferred simultaneously to
two nearby centers (II + III). After the energy is transferred these centers both
emit a photon.
3
Chapter 2. Basics - Physical Background
(a) (b) (c)
Figure 2.1: Concepts of quantum cutting. Relaxation and excitation by energy trans-
fers were marked as dashed arrows. (a) One optically active center (ion) subsequently
emits two photons after being excited into the energy level E2. In (b) quantum cutting
with two ions is shown. After an energy transfer both ions emit a photon. Using three
ions, the excited ion (I) can relax into the ground state, E0, by transferring the energy
to two nearby ions - shown in (c).
Figure 2.2: Energy level diagram for a Gd3+-Eu3+system, showing the possibility
for visible quantum cutting by two-step energy transfer from Gd3+to Eu3+upon
excitation in the 6GJlevels of Gd3+[9, 10]
4
Chapter 2. Basics - Physical Background
2.1.2 Up-Conversion
The process of frequency-up-conversion describes the conversion of two or
more low-energy photons (e.g. IR-photons) into one or more photons with a
higher energy (e.g. a visible photon). The sum of the energies of the absorbed
photons must be greater or equal to the energies of the emitted photons, i.e.
∑
n
Eemitted,n=∑
n
Eabsorbed,n+Eloss. (2.1)
The up-conversion-processes can occur via three mechanisms. These mecha-
nisms are excited-state-absorption, direct two-photon-absorption, and energy-
transfer-up-conversion [13].
Excited-State-Absorption: ESA
Excited-state-absorption is a single-ion process. Initially the ion is in its ground-
state (E0). A photon with an energy E2is absorbed. After the first absorption
and a possible radiationless relaxation into a metastable state (dashed arrow,
figure 2.3, another photon with an energy which is equal to the energy differ-
ence of the E1and E3levels is absorbed and leads to a second transition. The
ion gets from this (metastable) state E1into a higher energy level E3. This is
shown in figure 2.3(a). The subsequent relaxation to the ground state can be
observed as up-converted fluorescence. The energy of the final state is almost
equal to the sum of the two-photon energies.
Direct Two-Photon-Absorption: TPA
The direct two-photon-absorption is also a single-ion process. Two photons
containing in the sum the energy of the excited state E1are simultaneously
absorbed upon exciting the ion into a virtual intermediate excited state. This
virtual state is indicated by the dashed line in figure 2.3(b).
Energy-Transfer-Up-Conversion: ETU
Two kinds of energy transfers are considered here. First, there is the cross re-
laxation between ions in their excited states. In this case, the energy absorbed
by the first ion (I) is transferred to the second ion (II), which is already in an
excited state, and reaches by this way the final state (see figure 2.3). In the
second case – cooperative sensitization – the energy, accumulated by two excited
ions, is transferred by the relaxation of these two ions to a third ion. The time
dependence of the anti-Stokes fluorescence, which may be calculated using
rate equations, does not only depend on the final-state relaxations, but also on
the relaxation rates of the intermediate states and the energy transfer proba-
bility. Auzel et al. [14] have shown that excited-state absorption and energy
5
Chapter 2. Basics - Physical Background
(a) (b)
o
o
(c) (d)
Figure 2.3: (a) Scheme of the ESA-process. Stepwise absorption of photons excite the
ion into a higher energy level E3. (b) Illustration of the TPA-up-conversion-process.
The excited state is reached via a virtual state by simultaneous absorption. (c) shows
the cross-relaxation at the ETU-process. While one ion (I) is relaxing into the ground
state another ion (II) is excited in an energetically higher level. The cooperative sen-
sitization (ETU) is shown in (d). A third ion (III) is excited by the relaxation of two
other ions (I+II). Relaxation and excitation by energy transfers were marked as dashed
arrows.
6
Chapter 2. Basics - Physical Background
mechanism efficiency
ETU cross relaxation ∼10 −3
Excited State Absorption (ESA) ∼10 −5
ETU cooperative sensitization ∼10 −6
Two-Photon Absorption (TPA) ∼10−13
Table 2.1: Comparison of the different up-conversion mechanisms regarding their
efficiencies [14].
transfer up-conversion are usually much more probable than the direct-two-
photon absorption by about 10 orders of magnitude. The different efficiencies
are listed in table 2.1.
Up-Conversion Rate Equations
Pollnau [15] et al. analyzed up-conversion rate equations for m-level systems.
For the calculations the following assumptions were made: (a) almost constant
ground-state population density, (b) a continuous wave pumped system by
ground-state absorption, (c) only ETU and ESA up-conversion mechanisms
and (d) a decay rate of Ai=τ−1
i. The rate equations for a three energy level
system (m=2) are for ETU
dN1/dt =ρpσ0N0−2W1N2
1−A1N1(2.2)
dN2/dt =W1N2
1−A2N2(2.3)
and for ESA
dN1/dt =ρpσ0N0−ρpσ1N1−A1N1(2.4)
dN2/dt =ρpσ1N1−A2N2, (2.5)
where σjis the absorption cross section from state jand W1the up-conversion
parameter. The pump parameter ρpcan be described by: ρp=λp/(hcπw2
p)Pin,
with the pump wavelength λp, the incident power Pin, Planck’s constant h, the
speed of light in vacuum c, and the pump radius wp. With the assumption that
the glasses used have low maximum phonon frequencies - decay to the next
lowest levels are negligible and luminescence directly to the ground state is
predominant - in this case the equations can be solved. Three limiting situa-
tions can be distinguished. The power dependence of the population density
for an m-energy level system is given in table 2.2. For systems with a small
up-conversion rate the power dependence is given by Nm∼Pm, whereas it is
limited to Nm∼P1
in for systems with the up-conversion being the predomi-
nant process. For systems in which both processes are present the exponent of
the power dependency has to be between these two limits, one and m.
7
Chapter 2. Basics - Physical Background
Figure 2.4: Schematically shown power dependency of different levels.
In figure 2.4 the power dependency is shown schematically for different
systems. The black curve has a slope of m=1 and displays the limit of
a dominant up-conversion decay. The red line is the other limit for the 2
photon up-conversion, representing the system with a nearly negligible up-
conversion process compared to the PL. When we now look at three pho-
ton up-conversion, the limits expand to a slope of three. In this case the ex-
cited states relax mainly due to normal PL. Therefore, systems with a low
slope in the double logarithmic plot or low exponents for the power depen-
dency, respectively, are preferable. These systems show an enhancement in
the emission intensity or population density, respectively, for the same excita-
tion power when compared to systems which are PL dominated.
From the short circuit current Isc of a solar cell we can also calculate the
external quantum efficiency (EQE) of the up-converting-layer on a solar cell
system. The EQE is given by the ratio between generated electron hole pairs
influence of up-conversion power from
up-conversion mechanism dependence level
(1) low (a) ETU & (b) ESA Ni∼Pi
in i=1 . . . m
(2) high
(a) ETU Ni∼P1/2
in i=1 . . . m−1
Ni∼P1
in i=m
(b) ESA Ni∼Pi
in i=1 . . . m
Ni∼P1
in i=m
Table 2.2: Power dependence of the population of up-conversion levels.
8
Chapter 2. Basics - Physical Background
Isc(λ)/eand incoming photons Pin(λ)/hν[16]:
EQE =Isc(λ)
e
hν
Pin(λ)(2.6)
where Isc(λ)is the short circuit current of the solar cell and Pin(λ)the power
of the incident light. λand νare the wavelength and frequency of the incident
light, respectively. his Planck’s constant, cthe speed of light in vacuum, and e
the charge of an electron. Since Isc(λ)∝Pm
in we obtain from equation 2.6:
EQE ∝Pm
in
Pin
=Pm−1
in . (2.7)
2.2 Glasses and Glass Ceramics
2.2.1 Phonons
A phonon is a quantized vibration mode occurring in a crystal lattice. An-
other description for a phonon is quantized thermal energy. Phonons play
an important role for excited states. An ion excited to an energetically higher
state has two options to relax into a lower level (e.g. ground state). First, it
can emit a photon, carrying the energy difference between the two states, to
get into the lower level. The second possibility to get to a lower level is a non-
radiative process. Hereby, the energy difference between the excited states is
transferred to one or more phonons. This relaxation is normally described as
a leakage process.
To reach luminescence from an ion with high radiative emission efficiencies
the non-radiative relaxation caused by multiphonon relaxation (MPR) has to
be minimized. For this case large energy-gaps (in the region of some 1000 cm−1
for some rare earth elements) between the excited states and the next lowest
levels are useful. Additionally minimizing the highest-phonon-energy of the
hosts ¯hωmax and νmax, respectively, leads to an increase of the number of re-
quired phonons for the MPR to bridge this energy gab. Figure 2.5 shows a
trivalent neodymium energy level diagram as an example. For four differ-
ent glasses - borate, silicate, germanate, and zirconium-fluoride - the number
of phonons with maximum phonon frequency which is required to reach the
next lowest level is shown. The more phonons are needed to bridge the gap,
the higher the efficiency of the radiative depopulation; this expands the de-
tectable lifetime of the exited state. The lifetime τaof an excited state, a, is
represented by formula 2.8, where Tab and Wab are the the radiative and non-
radiative transition probabilities from level ato level b, and the summation is
over all ground states b.
1
τa
=∑
b
(Tab +Wab)(2.8)
9
Chapter 2. Basics - Physical Background
¯hωmax Cα
Host glass Notation (cm−1) (s−1) (10−3cm)
Borate B2O31400 2.9 ·1012 3.8
Phosphate P2O51200 5.4 ·1012 4.7
Silicate SiO21100 1.4 ·1012 4.7
Germanate GeO2900 3.4 ·1010 4.9
Tellurite TeO2700 6.3 ·1010 4.7
ZrF4-based ZrF4500 1.9 ·1010 5.8
ZnCl2-based ZnCl2300 5.0 ·10 74.1
Table 2.3: Host-dependent parameters of multiphonon relaxation in glasses [17, 19].
Non-radiative processes include, in addition to the MPR, relaxation by direct
ion-ion energy transfer. For low dopant concentrations the energy transfer is
not the predominant mechanism so that it is negligible and the MPR rate can
reasonably be described with an exponential law. Equation (2.9) is often called
the energy-gap law [17].
Wab =C·exp(−α∆E)(2.9)
The parameters Candαare the positive host-dependent constants and depend
on the host crystal and strength of ion-lattice coupling. With rare exceptions
these constants are independent of the specific RE ion or electronic states [18].
Some examples for these parameters for different glasses are given in table
2.3. In figure 2.6 the MPR rate for different glasses and rare earth elements are
given with respect to the energy gap. Figure 2.7 shows the reflexion spectra
of different glasses. The measured maximum phonon frequency is in good
agreement with the literature [19, 21].
2.2.2 ZBLAN glasses
In 1974 Poulain et al. [22] discovered the heavy metal fluoride glasses (HMFG)
based on zirconium (Zr) and/or hafnium (Hf) fluoride. The most important
feature is high transparency over a large spectral range from the near UV
(∼0.2 µm) to the mid-IR (∼7.0 µm) [23]. In figure 2.8 three glasses were
compared in respect to their transparency in the IR spectral range. Fluoro-
zirconate (FZ), fluorochlorozirconate (FCZ), and fluorobromozirconate (FBZ)
glasses and glass ceramics are based on the well-known ZBLAN composition
[21] and belong to the HMFG. These glasses and glass ceramics are character-
ized by a low maximum phonon energy (<580 cm−1). This leads to increased
fluorescence efficiencies from rare-earth ions due to minimized non-radiative
decay processes. Energy levels that are separated from the next lowest lev-
els by more than a few times the maximum phonon energies decay mainly
by radiation. These glasses also possess a low refractive index, low material
10
Chapter 2. Basics - Physical Background
Figure 2.5: Advantage of low phonon glasses using the example of trivalent
neodymium. The number of phonons with maximum phonon frequency which are
required to depopulate the 4F3/2to the next lower energy level is shown for different
glasses. From left to right: borate (4), silicate (5), germanate (6), zirconiumfluoride
(11) based glasses.
Figure 2.6: Multiphonon relaxation rates for rare-earth ions in different host materials
plotted as a function of the energy gap to the next lower level [17]. For demonstration
some data points [18–20] of YAlO2(yttrium orthoaluminate), SiO2-, and ZrF4-based
glasses are shown.
11
Chapter 2. Basics - Physical Background
Figure 2.7: Reflection spectra of silicate (black curve), ZrF4 (red curve), and borate
(blue curve) glasses. The maximum phonon frequency of the samples can be deter-
mined from these measurements. A FTIR IFS66 spectrometer was used.
Figure 2.8: Transparency of a normal SiO2glass (black curve), borate glass (blue) and
an as-made FCZ glass ceramic (red curve). All samples have a thickness of 1 mm. The
measurements were performed with a Bruker Hyperion spectrometer.
12
Chapter 2. Basics - Physical Background
Material FZ FCZ FBZ
ZrF453.0 -x 53.0 -x 53.0 -x
BaF220.0 10.0 10.0
BaCl2- 10.0 -
BaBr2- - 10.0
LaF33.5 3.5 3.5
AlF3 3.0 3.0 3.0
NaF 20.0 - -
NaCl2- 20.0 (-x) -
NaBr2- - 20.0 (-x)
InF 0.5 0.5 0.5
Rare Earth - +x - +x - +x
KCl - - (+x) - (+x)
Table 2.4: Composition of ZBLAN glasses. The left column shows the basic version
of doped glasses. The compositions for FCZ and FBZ glass ceramics are listed in
the middle and right column, respectively. For some dopings (e.g. NdF3) a charge
compensation is required - this is realized by the (x)-components. The compositions
are in mol%.
dispersion, low linear scattering loss, and good chemical durability [24–26].
All these attributes are useful and/or essential for a fibre material. Thus, rare-
earth doped HMFG are interesting to the telecommunication and data transfer
industries. Also for digital radiography, FZ glasses have been doped with
europium and chlorine ions resulting in very efficient glass-ceramic radiation
detectors [27, 28] where the RE ion, europium, is incorporated into the glass as
well as in the barium chloride nanocrystals formed in the glass upon anneal-
ing. FZ-based glass ceramics offer a broad range of applications; the function-
ality can be modified not only by appropriate thermal processing but also by
appropriate rare-earth doping.
The glasses are prepared as followed: The base composition of chemicals
were additionally doped with RE ions. In table 2.4 the chemicals used and
their compositions in mol-% are given for FZ, FCZ, and FBZ glasses. The in-
gredients should be high purity chemicals. In order to minimize impurities
the preparation process has to be done in an inert atmosphere. The constituent
chemicals were pulverized and mixed in a mortar. They were melted in a
glassy carbon crucible at 745 ◦C for 30 minutes, cooled down to 650 ◦C and
then poured into a brass mold that was at a temperature of 200 ◦C, i.e. be-
low the glass transition temperature of 262 ◦C for a FZ-based glass [21], before
being slowly cooled to room temperature.
13
Chapter 2. Basics - Physical Background
Figure 2.9: FCZ glass ceramics. All samples were annealed for 20 minutes. From left
to right: as-made, 260, 270, 280, 285, and 290 ◦C.
2.2.3 Thermal Processing and Nano-Particles
In section 2.2.1 it was described that phonons and their maximum frequency
affect the observable lifetime of the excited states. However, the phonon fre-
quency is not only given by the matrix but also by the size of the crystallites;
in nanocrystals, only low-phonon frequencies are found [29].
Through subsequent annealing in the vicinity of the glass transition temper-
ature nanoparticles are formed in the glass. Therefore the FZ glasses have to
be doped beforehand with chlorine for FCZ or with bromide for FBZ glasses.
The glass transition temperature is at about 260 ◦C and depends on the com-
position and doping of the glasses. The glass transition temperature can be
determined with differential scanning calorimetry (DSC). Figure 2.9 shows a
series of FCZ samples. The annealing time for all samples was 20 minutes, and
the temperature was varied from 260 to 290 ◦C. With higher annealing temper-
ature the nanocrystals in the glass ceramics became larger. When the particle
size reaches the region of a few hundreds nanometers, visible scattering effects
can be observed. To determine the size of the particles XRD spectra or TEM
images can be used. The subsequent annealing steps after glass processing (to
form the nanoparticles) as well as the glass production were performed in an
inert nitrogen atmosphere.
2.3 Rare Earth Ions
Lanthanides and actinides are generally known as rare earth (RE) elements.
The potential application of these elements are manifold. The REs, in common
with the actinides, have the most complicated fluorescence spectra of any el-
ements [30] (see figure 2.10). This is caused by an incompletely filled 4f shell.
At first, the origin of the sharp peaks seen in these spectra puzzled the scien-
tific community. Bethe, Kramers and Bequerel suggested that the lines could
be attributed to transitions within the 4f configuration. In 1937 van Vleck [41]
showed that the parity forbidden 4f →4f transitions become partly allowed as
electric dipole transitions by the admixture of configurations of opposite par-
ity. Many low lying energy levels are produced due to the transition within
the 4f shell. Additionally due to an effect called lanthanide contraction the
14
Chapter 2. Basics - Physical Background
4f shell is surrounded by the completely filled 5s2and 5p6orbitals, shield-
ing the 4f shell nearly completely from its surrounding. Atomic lattice vibra-
tions (phonons) or other influences from the host are minimized because of
this energetic shielding. This is a reason for the similar chemical and phys-
ical attributes of the RE elements. The position of the energy levels and the
transitions itselfs depend only barely on the host material the RE element is
incorporated in. Upon the fact of forbidden 4f →4f transitions and the ener-
getic shielding, minimizing the of interactions between the RE ion and the host
material, the excited states have relatively long lifetimes; up to milliseconds.
A selection of trivalent RE element energy levels is given in figure 2.10, it is the
so called ’Dieke diagram’.
Different applications like lasers use the up-conversion process. The fol-
lowing RE elements are known for these applications: Tm, Er, Ho, Nd, and
Pr; all in their trivalent state. Figure 2.11 shows the energy level diagram of
these elements. Up-conversion routes are not shown for clarity. A grey back-
ground was added to the excited states of the energy level diagrams. The well-
known trivalent RE ions used for down-conversion and/or quantum cutting
are shown with their energy level diagram in figure 2.12. These are Tb, Eu, Ho,
Sm, and Gd. A grey background for the excited states was added, too.
2.4 Analyzing Methods
2.4.1 Photoluminescence
For fluorescence and fluorescence excitation spectra a single-beam spectro-
meter in which a 0.22 m (Spex 1681) and a 0.55 m (Horiba Jobin Yvon iHR550)
monochromator were available for excitation and emission. For excitation and
absorption measurements two lamps, a halogen (300 - 800 nm) and a xenon
lamp (250 - 500 nm), were used. For the up-conversion process more powerful
excitations sources were needed. Here continuous wave (CW) infrared laser
diodes operating at 796 nm (Toptica Photonics #LD-0795-0150-2), at ∼811 nm
(OEC GmbH 1010-C), and at 1542 nm (Anritsu GB5A016) were used. The flu-
orescence was detected in the visible spectral range with a cooled photomul-
tiplier (Hamamatsu R943-02) and in the infrared spectral range with a cooled
germanium detector (Edinburgh Instruments). Silicon, InGaAs, and PbS photo
diodes were available for measurements. All detectors were used in lock-in-
technique (SRS-SR830). The spectra have not been corrected for spectral sensi-
tivity of the experimental setup.
For the fluorescence lifetime measurements, the excitation sources were mod-
ulated by a square wave signal from a generator (Rhode & Schwarz AFS) or
by a mechanical chopper, depending on the expected lifetimes. By switching
16
Chapter 2. Basics - Physical Background
Figure 2.11: Trivalent RE elements known for up-conversion processes. The excited
states have a grey background. Closely spaced levels are depicted as bands.
Figure 2.12: Rare earth elements in its trivalent state known for down-conversion pro-
cesses. The excited states have a grey background. Closely spaced levels are depicted
as bands.
17
Chapter 2. Basics - Physical Background
Figure 2.13: Bragg scattering in a crystal. The x-rays (aiand ai+1) are scattered and
reflected by the periodic structure. The scattered x-ray beams build a “reflected” in-
terference pattern.
the excitation source on and off only the fluorescence decay was observed.
The fluorescence signal was detected with the photomultiplier amplified and
recorded with a digital oscilloscope (Tektronix TDS 1012B).
For power dependence measurements of the fluorescence and up-converted
fluorescence intensities versus the excitation power a power meter with a sili-
con optical sensor (Thorlabs S121B) was used. The different excitation powers
were achieved by adjusting the laser current or by setting neutral density fil-
ters into the optical axis.
2.4.2 X-Ray Diffraction
X-ray diffraction (XRD) is a method to characterize the structure of crystals.
It can also be used to analyze the size of nano-crystals or particles. Interfer-
ence of incident x-rays with the periodic structure of crystals is measured. The
condition for constructive interference is given by the “Bragg’s law”:
2·dhkl ·sinϑB=n·λ, (2.10)
with dhkl the lattice distance between the crystal layers, ϑBthe angle of in-
cidence, nthe order of the diffraction, and λthe wavelength of the incident
wave. The index hkl is defined by the Miller indices and describes a series of
crystallographic planes.
The nano-crystallites in the FCZ glass ceramics have no preferred orienta-
tion. These give essentially powder-diffraction conditions. For a powder, x-
ray diffraction occurs for orientation that fulfills Bragg’s law. This produces the
Debye Scherrer rings in a plane behind the sample, which are symmetric about
the origin, ϑ=0. By detecting the x-ray intensity on a line of this plane, an
angular dependent XRD spectrum can be observed.
18
Chapter 2. Basics - Physical Background
Figure 2.14: The broadening of XRD reflection peaks: the ideal XRD peak (black), with
broadening due instrumental resolution (red). The green line is additionally broaden
due to instrumental and crystal size effects. The blue line describes the reflection with
all named effects before combined with strain and stress.
Line Width and Crystal Size Determination
A Bragg reflection from a monochromatic x-ray tube diffracted for an infi-
nite, perfect crystal is a δ-function, but a crystal with imperfections produces
a Gaussian like reflection peak. These imperfections are based on crystals that
are not free from strain or stress. Also faulting and lattice defects or non-
infinite crystals e.g. small crystal sizes produce a broadening effect. Additional
to the broadening caused by the parameters of the diffracting crystals there is
an external broadening, the so called instrumental broadening. The instru-
mental broadening depends on resolution and the natural linewidth. Wave-
length dispersion of the incident and reflected beam also affects the linewidth.
Slit width, sample size or incorrect x-ray focusing causes a broadening of the
detected diffraction peak. Because of these effects the ideal delta peak of the
measurement signal is widened as shown in figure 2.14.
The broadening effect based on the crystal size can be used to determine the
size of nano-crystals formed in glasses i.e. glass ceramics. The first investiga-
tion of particle size broadening was due to Scherrer [42]. To analyze the size
of the crystals the instrumental broadening has to be subtracted from the x-
ray diffraction pattern. To determine the instrumental broadening a reflection
peak from a sample in which the particle size is nearly infinite has to be taken,
consequently obtaining a signal with only the instrumental broadening. With
this result the instrumental broadening can be corrected for all measurements
taken under the same conditions.
19
Chapter 2. Basics - Physical Background
Figure 2.15: In (a) the grain or particle size is schematic shown, (b) and (c) show the
crystal and crystallite size, respectively.
With the known instrumental broadening and the assumption of strain- and
stress-free crystals a determination of crystal sizes is possible. Therefore, we
estimate the particle sizes by using the Scherrer formula [42]
d=K·λ
(∆−0.085◦)·cos(ϑ)(2.11)
with dthe particle size, Kthe Scherrer constant, λthe wavelength of the inci-
dent x-ray beam, 2ϑthe peak position of the reflection, and ∆the full width
at half maximum of the reflection (FWHM). The Scherrer constant depends on
the crystal structure and on the reflection indices i.e. the reflection crystallo-
graphic plane [43]. However the constant varies only slightly for the different
structures and planes between 0.9 and 1.0. For first estimations of crystal sizes
the Scherrer constant Kcan be set to one. A complete derivation of the Scherrer
formula can be found in [44].
Unfortunately, there is enormous confusion in literature concerning the defi-
nition of size of particles, crystals, and crystallites. Going from large to small, it
starts with the particle or grain. It consists of one or more crystals. All crystals
in this particle may be separated by large angle differences as well as amor-
phous or crystalline intersections (see figure 2.15(a)). It is worth mentioning
that it is not really possible to determine the particle size by X-ray diffraction
measurements; the crystals consists of one or more crystallites with size equal
to or smaller than the size of a particle (fig. 2.15(b)). As well as the particle
the crystal size itself cannot be detected or measured. The size of a crystallite
is typically less than or equal to the crystal size (fig. 2.15(c)). The size of the
crystals can be determined - in the absence of crystal domains - by analyzing
x-ray diffraction patterns.
20
3 Down-Conversion based on Sm
By a process, in which high energy (UV) photons were effectively converted
or cut into one or more low energy photons with an energy higher than the
band gap energy, the efficiency of a solar cell can be increased significantly. In
the quantum cutting process one high energy photon is cut into two or more
photons in the visible or near-infrared that can be absorbed by a solar cell.
Trupke et al. [6] calculated the theoretical limit of a single junction solar cell
which uses additional quantum cutting processes to an over-all external quan-
tum efficiency of 39.63 %. The down-conversion process requires an optically
active center. Rare earth ions are known for a discrete energy levels that can
be used for the conversion process. When embedded in a low phonon glass
matrix the loss mechanisms can be reduced. A good choice due to their low
phonon frequencies of around 500 cm−1are glasses based on the well known
ZBLAN composition [21]. From prior investigations it is known that glass ce-
ramics with embedded nanocrystals containing rare earths show an enhanced
fluorescence efficiency.
3.1 Simulations
Trupke et al. [5] made several theoretical assumptions in their calculations and
vary the band gap to reach the the theoretical maximum external quantum
efficiency. To determine more realistic limits, further calculations were per-
formed; the details appear in appendix B. Figure 3.1 shows the results of the
calculations as a contour plot as a function of the absorption bang edge of the
down-converting layer and the internal quantum efficiency (IQE), given in nm
and %, respectively. The absorption of the down-conversion layer was set to
one for this calculation. All points were normalized to the initial system with-
out an down-converting layer. The maximum enhancement, for a absorption
band edge at 818 nm and 100 % IQE, is at 1.198, which means an enhancement
of 19.8 % in comparison to the system without a down-converter. In fact of
the down-conversion layer in front of the solar cell, the ratio can be less than
1 in fact of low internal quantum efficiencies. These areas are colored in blue
to green; see the color scale in the figure. A dashed line at 500 nm marks the
“position” where the next calculation (see figure 3.2 take place. With a fixed ab-
sorption edge at 500 nm, the internal quantum efficiency and absorption (both
given in %) were varied for the calculations. Figure 3.2 shows normalized
21
Chapter 3. Down-Conversion based on Sm
Figure 3.1: Calculated ratio between solar cell with and without a down-converting
layer in dependency of the absorption edge of the layer and the internal quantum
efficiency. The absorption is set to 100 %. A maximum for for 818 nm as the absorption
band edge and 100 % internal quantum efficiency of 1.198 was calculated; 0.282 for the
minimum. The color stands for the calculated ratio (see color scale in the lower left
corner).
Figure 3.2: Calculated ratio between solar cell with and without a down-converting
layer in dependency of the absorption of the layer and the internal quantum efficiency.
The absorption band edge is set to 500 nm. A maximum for 100 % absorption and
internal quantum efficiency of 1.100 was calculated; 0.943 for the minimum. The color
stands for the calculated ratio (see color scale in the lower left corner).
22
Chapter 3. Down-Conversion based on Sm
Figure 3.3: Energy level diagram of trivalent (black) and divalent (blue) samarium.
The Sm2+energy levels are a combination of the 4f6and the 4f55d states [45, 46]. The
4f55d states of Sm3+are in the far VUV (with energies of about 70,000 cm−1) [30].
All excited states are highlighted with a grey background. Closely spaced levels are
depicted as bands.
results for the solar cell without a down-converting layer. A minimum and
maximum of 0.943 and 1.100, respectively, were calculated. In both cases the
absorption of the down-converting layer is 100 %, whereas the internal quan-
tum efficiency is zero for the minimum and 100 % for the maximum value. For
all absorption values the plot shows a horizontal line with a value of one at an
IQE of 36.36 %, i.e. that a silicon solar cell with a down-converter of 36.36%
IQE on top has the same overall efficiency as the silicon solar cell itself.
3.2 Dopant: Samarium
3.2.1 Motivation
For down-conversion applications an optically active center is needed. Good
RE candidates for efficient down-conversion, both quantum cutting and nor-
mal PL, are given in chapter 2 in figure 2.12. They all have many absorption
lines and absorption bands in the UV and blue spectral range. Some elements
even have energy levels that absorb in the lower energetic spectral range. In-
tense emission lines and bands in the green, red or near-infrared spectral range
are required for a down-converter. Samarium is listed in this figure; in its triva-
lent state. By comparing the trivalent energy level of Sm with the divalent one
(see figure 3.3) it can be seen that Sm2+offers some advantages. The normal
4f6energy levels are superimposed by the broad energy levels of the 4f55d.
These energy levels mainly increase the absorption in the green, blue and UV
spectral range.
23
Chapter 3. Down-Conversion based on Sm
Figure 3.4: Normalized fluorescence spectra of 0.1 mol% Sm2+-doped BaCl2and
BaBr2single crystals. The emission was recorded under continuous laser excitation
with the 476 nm line of an Ar+laser. The lines were attributed to 5D0→7FJand
4f55d →4f6transitions [31]. The measurements were carried out at RT.
3.2.2 Barium Chloride and Barium Bromide
Single crystals of BaCl2and BaBr2were grown in the crystal growth laboratory
at the University of Paderborn using the Bridgman method. Before the crystal
growth process the BaCl2and BaBr2powders were dried in a vacuum with
subsequent melting in a SiCl4and SiBr4atmosphere, respectively to reduce
possible oxygen contamination of the starting materials. The powders were
melted in a quartz ampoule under the SiCl4and SiBr4atmosphere. 1000 molar
ppm of SmCl2and SmBr2, respectively were added as dopant to the powder.
The single crystals were slowly cooled through the cubic-orthorhombic phase
transition near 920 ◦C for BaCl2[47] and 800 ◦C for BaBr2[48], respectively.
Fluorescence and X-Ray Excited Luminescence
Fig. 3.4 shows the fluorescence spectra of BaCl2(top) and BaBr2(bottom) with
0.1 mol% Sm2+doped single crystals in the visible spectral range between 500
and 830 nm. The samples were excited with an Ar+laser at 487 nm. The ob-
served intense sharp emissions at 687, 702, and 728 nm combined with the
24
Chapter 3. Down-Conversion based on Sm
Figure 3.5: X-ray excited fluorescence spectra of 0.1 mol% Sm2+-doped BaCl2(top)
and BaBr2(bottom) single crystals. Both spectra were normalized to the emission
at 728 and 687 nm, respectively. The emission lines were attributed to 5D0→7FJ
with J ={0, 1, 2, 3, 4}and 4f55d →4f6transitions of Sm2+[31]. The Sm-doped BaCl2
crystals show additional Sm3+emissions [49]. The measurements were carried out at
RT.
broad band emission ranging from around 600 up to 800 nm are typical for
Sm2+. The sharp emissions can be assigned to 5D0→7FJtransitions with
J={0, 1, 2, 3, 4}. The broad emission band is not from a 4f →4f transition but
it can be assigned to a 4f55d →4f6transition. All emissions associated with
the 4f →4f transitions of Sm2+are noticeably split by the crystal field [46].
Both spectra were normalized to their most intense emission at 687 nm.
For a closer look all emissions of the doped crystals the samples were excited
with x-rays. In figure 3.5 the normalized x-ray excited luminescence (XL) spec-
tra of 0.1 mol% Sm2+doped BaCl2and BaBr2are shown. Comparing the X-ray
excited spectra with the normal fluorescence spectra shows no evident differ-
ences. The intense sharp emissions can be attributed to the 5D0→7FJtransi-
tions with J ={0, 1, 2, 3, 4}as depicted in the figure. The broad band emission
of Sm2+can be seen in both XL spectra. Only one main exception can be ob-
served comparing the XL with the normal fluorescence spectra. There were
three emissions on the shoulder of the 4f55d →4f6emission for the Sm-doped
25
Chapter 3. Down-Conversion based on Sm
Figure 3.6: Afterglow spectra of 0.1 mol% Sm2+-doped BaCl2(black curve) and BaBr2
single crystals (red curve). Both spectra were normalized to the most intense emission
at 687 and 730 nm. The lines can be attributed to 5D0→7FJand 4f55d →4f6transitions
of Sm2+[31]. The measurements were carried out at RT. The lower graph shows the
intensity of the 687 nm emission (BaCl2:Sm) during and after x-ray irradiation.
BaCl2. These three emissions can be attributed to transitions of a trivalent Sm
ion. The energetically highest emissions of the three at 560 nm results from a
4G5/2→6H5/2transition. The other two transitions although arise from the
same excited state but relax to different ground states; to the 6H7/2and 6H9/2
for the 595 and the 640 nm, respectively. No further emissions can be observed
- co-doping due to impurities can be excluded. The Sm3+emissions are only
in the XL spectra observable, which could be an evidence for a charge trans-
formation upon x-ray irradiation.
The Sm-doped crystals show an afterglow after they were excited with x-
rays. The afterglow spectra of BaCl2and BaBr2doped with divalent Sm are
shown in figure 3.6 (a). In contrast to a XL spectrum an afterglow spectrum
is recorded after and not during the x-ray excitation. The spectra show the
typical Sm2+emission. In the near infrared sharp emissions arising from the
5D0→7FJtransitions with J ={0, 1, 2, 3, 4}can be observed. The broad band
26
Chapter 3. Down-Conversion based on Sm
Figure 3.7: Decay of the Sm2+fluorescence in BaCl2and BaBr2. The Sm2+doping
concentration is 0.1 mol%. The emission was detected at 687 nm, which corresponds
to the 5D0→7F0transition of Sm2+. The fluorescence lifetimes of the 5D0→7FJand
the 4f55d →4f6transitions are identical within experimental error. An Ar+laser was
used for the excitation at 476 nm.
emission can be attributed to the 4f55d →4f6transition. The afterglow spectra
were recorded after a radiation time of 15 minutes. The spectra were recorded
30 minutes after the excitation. The afterglow decay time is relatively slow
- measurements performed 45 minutes after the sample was x-radiated com-
pared to measurements after 60 minutes show a loss of only 3 to 5 % in its
intensity. Therefore further investigations were performed on this effect. The
measured intensity of the 687 nm emission in dependency of the time during
and after the x-ray excitation is plotted in figure 3.6 (b). The start and end of
the x-ray excitation can be seen in figure 3.6, marked as dashed blue lines. The
afterglow signal is 5 to 6 orders of magnitude stronger than the measured dark
signal and has a relatively long decay time.
Lifetime Measurements
Lifetime measurements of the 5D0level have been performed at a wavelength
of 687 nm (see figure 3.7) which corresponds to the 5D0→7F0transition of
Sm2+. These measurements yielded different lifetimes for different host mate-
rials: The measured lifetime of the Sm-doped BaCl2sample is (1.65 ±0.02) ms,
the lifetime for BaBr2as host material for the Sm2+ions is (0.5 ±0.01) ms. The
measured lifetimes correspond very well with the literature [31].
27
Chapter 3. Down-Conversion based on Sm
3.2.3 Zirconium Fluoride Glass and Glass Ceramics
Preliminary investigations of Sm-doped BaCl2and BaBr2single crystals were
done in the last section. Now Sm-doped glass ceramics which were co-doped
with chlorine or bromine ions are investigated. In theses glasses were during
a subsequent annealing process BaCl2or BaBr2nanoparticles formed. Samar-
ium can enter the glass matrix either in its divalent form and/or as a trivalent
ion. The different energy levels were shown in the last section (see figure 3.3).
The composition of the samarium doped glasses are given in table 3.1.
Some problems occurred during the production process of the Sm-doped FZ
glasses with the charge level of the dopant. The first glasses became immedi-
ately ceramic on pouring. This was solved by decreasing the temperature of
the mold from 200 ◦C down to 150 ◦C, which enabled the glasses to remain
transparent. FBZ 95 and FBZ 104 have a reddish color which originates from
Sm2+; the other glasses stay colorless. However, fluorescence measurements
show Sm3+. An attempt to reduce Sm3+was made by adding LiH (2 wt%)
or NH4HF2(2 wt%), then remelted and pouring the mixture. FBZ 114 was
made with new SmBr2material, but also stayed colorless. In [50] a procedure
to reduce the Sm3+within an H2-N2atmosphere was described, but was not
successful in this case.
Fluorescence
The fluorescence spectrum of FBZ glass ceramics containing Sm2+ions (Fig.
3.8 (a), solid curve) shows relatively sharp lines of Sm2+at 690, 700, 730, 765,
and 815 nm which originate from 5D0→7FJtransitions with J ={0, 1, 2, 3, 4}.
In addition to the narrow line emissions, a broadband emission peaking at
∼700 nm can be observed; the broadband emission can be attributed to a
ZrF4BaF2BaX2LaF3AlF3NaF NaX3YF3InF3SmX2
FBZ 93 52 20 - 1.5 3 5 15 1.5 1 1
FBZ 95 52 5 15 1.5 3 5 15 1.5 1 1
FBZ 104 52 20 - 1.5 3 5 15 1.5 1 1
FCZ 108 52 10 10 3.5 3 - 20 - 0.5 1
FZ 109 52 20 - 3.5 3 20 - - 0.5 1*
FBZ 114 52 20 - 1.5 3 5 15 1.5 1 1
FCZ 115 52 10 10 1.5 3 - 20 1.5 1 1*
Table 3.1: Composition of the Sm-doped glasses in mole percent. The name of the
samples is a composition of the abbreviation fluorobromo- (FBZ) or fluorochloro-
zirconate (FCZ) given by the compounds of the FZ glasses and an inventory number.
For both glasses marked with an asterisk (*) SmF2was used. Only the red marked
glasses contain divalent Sm as dopant.
28
Chapter 3. Down-Conversion based on Sm
Figure 3.8: Normalized fluorescence spectra of FBZ glasses containing Sm2+(solid
curves) or Sm3+(dashed curves). (a) Emission spectra excited at 590 nm (Sm2+) and
at 476 nm (Sm3+), (b) excitation spectra detected at 730 nm (Sm2+) and at 640 nm
(Sm3+). All spectra were recorded at room temperature.
4f55d →4f6transition [31]. The Sm2+excitation spectrum is shown in fig-
ure 3.8 (b), solid curve. The spectrum was detected for the 730 nm emission
arising from the 5D0→7F2transition. It shows a broad band peaking at
583 nm with a full width at half maximum of around 80 nm. These excitation
bands can be attributed to the closely spaced 6H energy levels of the 4f55d
state. The two smaller peaks on the shoulder of the broad excitation band in
the blue spectral range result from the transition of the ground state 7F0into
the excited states 5L7(481 nm) and 5L8(460 nm). Both are 4f →4f transitions
of Sm2+.
The dashed curve in figure 3.8 (a) shows the Sm3+emission bands in FBZ
glasses. These emissions are caused by transitions starting from the same ex-
cited state 4G5/2but relaxing to different ground state levels: 6H5/2(560 nm),
6H7/2(595 nm), 6H9/2(640 nm), and 6H11/2(705 nm), respectively. The results
29
Chapter 3. Down-Conversion based on Sm
Figure 3.9: (a) Emission spectra of FBZ glasses containing both Sm2+and Sm3+, as-
made and annealed for 20 min at 270◦C. (b) Sm2+/ Sm3+fluorescence intensity ratio
vs. annealing temperature normalized to as-made FBZ. The line is a guide to the eye.
agree with previous findings in Sm3+-doped FZ fibers [51]. The correspond-
ing excitation spectrum is depicted in figure 3.8 (b), dashed curve. The spec-
trum was recorded for the 640 nm emission depending on the 4G5/2→6H9/2
transition. In the case of Sm3+the 4f →4f55d transitions are all in the far
VUV (energies about 70.000 cm−1[30] or about 8.7 eV). So all detectable exci-
tations depend on the 4f →4f transition. The observed excitation bands can
be attributed to the transitions starting at the ground state 6H5/2to the 4G3/2
(559 nm), 4F3/2(527 nm) energy levels in the green spectral range. Due to over-
lapping energy levels like the 4I11/2and 4I13/2or 6P5/2,4M19/2,4L13/2,6P3/2,
and 4F7/2with their maximum at 478 nm and 402 nm, respectively, broad ex-
citation bands were formed in the blue [52]. Between these bands lies another
excitation at 440 nm corresponding to the 6H5/2→4I15/2. In combination the
energy levels 4F9/2,4D3/2, and 6P5/2produces a wider excitation band in the
UV at around 368 nm.
30
Chapter 3. Down-Conversion based on Sm
Figure 3.10: Emission spectra of FCZ 115 glass samples containing only Sm3+. The
measured spectrum of the as-made glass is shown. For the samples annealed for
20 min at 250 ◦C, 260 ◦C, 270 ◦C, and 280 ◦C the differences in the signal compared to
the as-made are plotted.
Fig. 3.9 (a) shows the fluorescence spectra of the FBZ 104 glass samples.
When this emission spectra is compared to figure 3.8 (a) a combination of the
Sm2+and Sm3+fluorescence spectra can be seen. This FBZ glass contains
Sm2+and Sm3+ions. Upon annealing the Sm2+fluorescence intensity is in-
creased by a factor of 4-5 with respect to the 270◦C annealed sample (solid
curve); the Sm3+intensity does not change upon annealing within the exper-
imental error. The Sm2+/Sm3+ratio versus annealing temperature is plotted
in figure 3.9 (b). For an annealing temperature of 270 ◦C for 20 minutes a
maximum increase for the Sm2+fluorescence intensity in comparison with the
Sm3+intensity can be achieved. The Sm2+/Sm3+ratio was normalized to the
value of the as-made sample.
Figure 3.10 shows the emission spectra of the as-made glass of the FCZ 115
glass series. The four observed emissions can be assigned to a Sm3+doping, all
starting from the 4G5/2energy level. No evidence of Sm2+ions can be found.
Upon annealing the shape and intensity of the spectra change. The difference
between the intensities of the emission spectra of the annealed sample and the
as-made glass was calculated for all annealing temperatures and is also shown
this figure. The emission corresponding to the 4G5/2→6H7/2shifts from
31
Chapter 3. Down-Conversion based on Sm
Figure 3.11: Emission spectra of FCZ 114 glass samples containing only Sm3+. The
measured spectrum of the as-made glass is shown. For the samples annealed for
20 min at 250 ◦C, 260 ◦C, 270 ◦C, and 280 ◦C the differences in the signal compared to
the as-made are plotted.
595.6 nm for the as-made glass to 595 nm for the glass ceramics annealed at 250,
260, and 270 ◦C, observable in the difference spectra as a minimum-maximum
structure at around 595 nm. Additionally a decrease of the emission intensity
around 640 nm (4G5/2→6H9/2) is observable. In the spectral range between
620 and 780 nm where Sm2+emits a broad band (4f55d →4f6) superimposed
by sharp emissions (5D0→7FJwith J ={0, 1, 2, 3, 4}) no increase in the
emission intensity can be found. In figure 3.11 the spectra of a different Sm-
doped FCZ glass is shown. Only Sm3+emissions are observable. A broad
background in the range between 500 and 750 nm increases during the thermal
treatment process. Once again, no emission bands or peaks of Sm2+appear in
these spectra.
X-Ray Diffraction
The XRD data of FBZ glass ceramics containing Sm2+ions are comprised of
very broad peaks at about 26◦and 47◦, typical for glasses close to the ZBLAN
formulation for copper Kα1radiation [53]. In addition, relatively sharp peaks
are superimposed on the broad glass background; these peaks grow upon an-
nealing for 20 min at temperatures between 260 and 290◦C. Fig. 3.12 (a) shows
the XRD data for the as-made, the 270◦C, and the 290◦C sample. The sharp
peaks can be identified as reflections from hexagonal BaBr2; the bar graph
32
Chapter 3. Down-Conversion based on Sm
Figure 3.12: (a) XRD data for Sm2+-doped FBZ glasses as-made and annealed for
20 min at 270 and 290 ◦C. The line pattern of hexagonal phase BaBr2(PDF #45-1314) is
shown for comparison. (b) XRD peak used for the particle analysis of the 285 ◦C sam-
ple and the corresponding Lorentzian fitting curve. (c) Particle size versus annealing
temperature. The line is a guide to the eye.
33
Chapter 3. Down-Conversion based on Sm
in figure 3.12 (a) represents the corresponding powder diffraction data; pow-
der diffraction file (PDF #45-1314) [54]. There is no phase transition from the
hexagonal to orthorhombic phase BaBr2as was found in Eu-doped FBZ glass
ceramics [55].
The XRD peaks are wider than the instrumental resolution of 0.085◦, sug-
gesting size-broadening effects. The particle size can be determined by using
the Scherrer formula [42]
d=K·λ
(∆−0.085◦)·cos(ϑ)(3.1)
with dthe particle size, Kthe Scherrer constant (which is set to one in our
analysis), λthe wavelength of the incident x-ray beam, 2ϑthe peak position of
the reflection, and ∆the full width at half maximum of the reflection. Stress or
strain effects which may also lead to additional line broadening are not con-
sidered. The Scherrer formula is applied to the (111) reflection of hexagonal
phase BaBr2at around 28◦. For fitting the line profile a Lorentzian was used.
Figure 3.12 (b)) gives an example for the fitted function to the XRD pattern of
the 285 ◦C annealed sample.
The particle sizes of the as-made sample and after thermal processing at 260
and 270◦C are between 15 and 30 nm; for annealing temperatures at 280◦C and
above the particles grow from 50 nm to about 70 nm (Fig. 3.12 (c). Despite the
poor signal-to-noise ratio of the XRD reflections, the Scherrer analysis provides
a first estimate of the particle size; there is a trend to bigger particles for higher
annealing temperatures. The particle growth proceeds from the precipitation
of small hexagonal nanocrystals through Ostwald ripening: Small particles
will re-dissolve, and the larger ones will grow at the expense of the smaller
ones.
Fluorescence Lifetime
Fig. 3.13 (a) and (b) show the fluorescence decay curves for Sm2+and Sm3+in
as-made FBZ and FBZ glass ceramics annealed for 20 min at different temper-
atures. The Sm2+decay was recorded for the 5D0→7F0transition (690 nm)
and the Sm3+decay for the 4G5/2→6H7/2transition (595 nm). The fluores-
cence lifetimes of the narrow line (5D0→7FJ) and the broad band (4f55d →
4f6) emissions of Sm2+are identical within experimental error. Fig. 3.13 shows
that the Sm2+fluorescence lifetime in FBZ glass ceramics increases upon an-
nealing. The lifetime of Sm3+, however, does not change upon thermal pro-
cessing; it is (2.2±0.2) ms for the as-made and all annealed samples.
34
Chapter 3. Down-Conversion based on Sm
Figure 3.13: Decay of the Sm fluorescence in FBZ glasses. (a) The Sm2+emission
was detected at 690 nm (5D0→7F0). (b) The Sm3+emission was detected at 595 nm
(4G5/2→6H7/2). (c) Decay time of the Sm2+fluorescence in FBZ glass ceramics versus
annealing temperature. The line is a guide to the eye. All measurements were carried
out at room temperature; the fluorescence was excited at 590 nm for Sm2+and at
476 nm for Sm3+.
35
Chapter 3. Down-Conversion based on Sm
3.3 Discussion
Fluorescence, x-ray excited fluorescence and afterglow spectra show that Sm2+
can be incorporated into BaCl2and BaBr2single crystals and elucidates therein
its characteristic optical properties. In particular, a broad band emission su-
perimposed by narrow emissions when excited in the blue-UV spectral range
is found. Both the fluorescence efficiency and lifetime of Sm2+in FBZ glass
ceramics can be significantly increased upon appropriate annealing. The in-
vestigation of samples which contain Sm2+and Sm3+ions showed that the
annealing affects the fluorescence properties of Sm2+while those of Sm3+re-
main unchanged.
Most photons were emitted due to the broad band emission of Sm2+arising
from the 4f55d →4f6transition (figure 3.4). By comparing the emission spectra
of the Sm-doped BaCl2to the BaBr2crystals, the BaCl2doped crystals show
a less intense broad band emission and narrow emissions with smaller line
width. Some of the Sm-doped ZBLAN glasses show an Sm3+.
Lifetime measurement were performed for Sm2+in BaBr2, BaCl2and FBZ
glass ceramics, whereas the Sm3+fluorescence decay was only measured in
fluorozirconate based glasses. The fluorescent decay lifetime of the 687 nm
emission (5D0→7F0) differs in the single crystals; the doped BaCl2shows
lifetimes that are a factor of 3 longer compared to the BaBr2crystals. This is
evidence for less phonons or a smaller maximum phonon frequency in BaCl2.
Consequently, the Sm2+emissions in BaCl2crystals show narrower emission
linewidths. The lifetime measurements of Sm doping in the Sm2+and Sm3+
doped glass shows a lifetime for the trivalent state that is a factor of 4.7 longer
than the Sm2+lifetime. The different annealed glass ceramics show an inter-
esting phenomena that the Sm2+lifetime increases, but the Sm3+lifetime does
not change within experimental error upon annealing; this indicates that some
of the Sm2+present in the glass matrix enters the BaBr2nanocrystals, which
are formed during thermal processing. Sm2+probably substitutes for barium
on a regular lattice site, while the incorporation of Sm3+requires charge com-
pensation. The Sm2+lifetime therefore has contributions from Sm2+in the
glass and Sm2+in the nanocrystals. Since the fluorescence lifetimes increase
upon annealing the phonon energies of the BaBr2nanocrystals are smaller than
those of the FZ base glass, i.e. the non-radiative losses are reduced, the overall
measured lifetime is longer and thus the fluorescence is more efficient.
It was shown that upon annealing the BaBr2nano-particles in the FBZ glass
ceramic grow. The particle diameters, after thermal processing at 260, 270,
and 280 ◦C, are between 15 and 60 nm; for annealing temperatures at 280 ◦C
and above, the particles grow rapidly and their sizes become >100 nm (figure
3.12(b)). Due to the larger particle volume, more Sm2+-ions are incorporated
into the nanocrystalls, leading to a fluorescence emission enhancement (figure
36
Chapter 3. Down-Conversion based on Sm
3.9(a) and (b)).
Fluorescence intensity versus annealing temperature plot (see fig. 3.13(b))
does not follow the trend of the fluorescence lifetime vs. annealing temper-
ature (figure 3.12(c)) or the particle size vs. annealing temperature (figure
3.12(b)): For annealing temperatures at 280◦C and above the fluorescence is
slightly reduced with respect to the value of the 270◦C sample. It is assumed
that for higher annealing temperatures light scattering effects play an increas-
ingly important role. The BaBr2particles now have a size of greater than
50 nm, which may cause more light scattering and thus slightly reduced fluo-
rescence intensities. Additionally, due to the growing particles, higher phonon
frequencies can exist within the particles, which leads to higher losses through
non-radiative decays.
It is expected that the fluorescence efficiency, as well as the lifetime, depends
significantly on the quantity of Sm2+ions incorporated into the nanocrystals
and how large the BaBr2nanocrystals are. The nominal Sm-doping level for
the glass is 1 mol% but the number of Sm2+ions present in the nanocrystals is
probably much smaller than that. Future work will be focused on optimizing
the Sm2+concentration in the nanocrystals.
However, it was not possible to produce another glass containing only Sm2+
after the FBZ 104. Also it was not possible to reduce the trivalent samarium to
divalent samarium in the glass ceramics. An attempt was made to increase the
intensity of the Sm2+by a subsequent annealing step, as shwon in figure3.9.
No evidence was found for a Sm2+in any of the glasses (see figures 3.10, 3.11,
and A.2).
37
Chapter 3. Down-Conversion based on Sm
38
4 Up-Conversion based on Nd and Er
The efficiency of a solar cell can be increased significantly with a process, in
which infrared photons are converted to visible photons. Second harmonic
generation or two-photon absorption up-conversion fulfils this condition. In
the up-conversion process two or more IR photons are converted to a visible
photon that can be absorbed by a solar cell. Trupke et al. [5] calculated the theo-
retical limit of a single junction solar cell which uses additional up-conversion
processes to an over-all external quantum efficiency of 47.6 % which lies above
the Shockley-Queisser limit of 30.9 % [56]. The up-conversion process an opti-
cally active center. From chapter 2 it is known that intermediate excited states
with long lifetimes are needed. Rare earth ions are known for discrete en-
ergy levels and relatively long excited state lifetimes. When embedded in a
low phonon glass matrix non-radiative losses can be reduced. A good choice
due to their low phonon frequencies of ∼500 cm−1are glasses based on the
well known ZBLAN composition [21]. From prior investigations it is known
that glass ceramics with nanocrystals embedded in rare earth ions show an
enhanced fluorescence efficiency [27, 28].
4.1 Simulations
Trupke et al. [5] made several theoretical assumptions in their calculations
and vary the band gap to reach the the theoretical maximum external quan-
tum efficiency. To determine more realistic limits, further calculations were
performed; the details appear in appendix B. Figure 4.1 shows the results of
the calculations as a contour plot as a function of the absorption of the up-
converting layer and the internal quantum efficiency, both given in %. All
points were normalized to the initial system without an up-converting layer.
The maximum, for 100 % absorption and internal quantum efficiency, is at
1.264, which means an enhancement of 26.4 % in comparison to the system
without an up-converting layer. The diagram does not show any blue areas,
i.e. the calculated overall efficiency never drops below 1. An upconverter-
solar cell system provides always better EQE values than a silicon solar cell
alone. The minor enhancement compared to the calculations of Trupke is af-
fected by the difference of the AM1.5 and the blackbody spectrum (figure 1.1);
atmospheric absorption bands are centered at 1130, 1400, and 1875 nm.
39
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.1: Calculated ratio between solar cell with and without an up-converting
layer in dependency on absorption of the layer and internal quantum efficiency. For
100 % absorption and internal quantum efficiency a maximum of 1.264 was calculated;
1.0 for the minimum. The color stands for the calculated ratio (see color scale in the
lower left corner).
4.2 Dopant: Neodymium
4.2.1 Motivation
For the first system neodymium was chosen as a dopant. Nd3+is already well
known for its up-conversion application in up-conversion lasers. It offers a
large amount of energy levels that can be used for the up-conversion process
and it possess lifetimes of several hundreds µ-seconds.
The aim is to produce a model system with Nd-doped nanocrystals embed-
ded in low phonon glass ceramics that show intense up-conversion when ex-
cited with infrared photons. For a better understanding of the optical behavior
of the embedded nanoparticles it is essential to investigate the rare earth dop-
ing in the corresponding bulk material. Due to the fact that BaCl2nanoparti-
cles are formed upon subsequent annealing of the glass matrix, investigations
on Nd-doped BaCl2single crystals are performed. After a basic set of mea-
surements on the Nd-doped BaCl2, the optical and structural properties of
Nd-doped glass ceramics with different annealing temperatures as a function
of their Nd doping level are investigated.
40
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.2: Normalized fluorescence spectra of (a) 0.01 mol% and (b) 1 mol% Nd3+-
doped BaCl2. The fluorescence was recorded under CW infrared laser excitation at
796 nm. Parts of the fluorescence spectra (dotted curves) are expended as indicated.
4.2.2 Barium Chloride
Fluorescence and Excitation
Fig. 4.2 shows the fluorescence spectra of 0.01 (top) and 1 mol% (bottom)
Nd3+-doped BaCl2single crystals in the infrared spectral range between 800
and 1500 nm. The samples were excited with a 796 nm LD. The most intense
emissions observed at 890, 1065, and 1340 nm are typical for Nd3+and can be
assigned to 4F3/2→4I9/2,4I11/2, and 4I13/2transitions, respectively. Blowing
up the spectrum by a factor of 5 the 1 mol% sample clearly shows additional
emission bands at approximately 960 and 1190 nm; these line groups are barely
observable in the 0.01 mol% doped sample. Comparing these emissions with
the energy level diagram of Nd3+it can be seen that they are probably caused
by 4F5/2to 4I11/2and 4F5/2to 4I13/2transitions, respectively. All infrared Nd3+
emissions are more or less split by the crystal field [57]. The splitting effect is
stronger in the 1 mol% Nd3+doped crystal.
41
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.3: Normalized excitation spectra of (a) 0.01 mol% and (b) 1 mol% Nd3+-
doped BaCl2. The emission was detected at 880 nm. All transitions start from the
4I9/2ground state. Parts of the excitation spectra (dotted curves) are expanded as
indicated.
The corresponding excitation spectra for both samples are shown in fig-
ure 4.3. The spectra were recorded at 880 nm for the 4F3/2→4I9/2transitions.
They depend clearly on the Nd3+concentration. The relative intensity ratio of
the excitation bands of the 1 mol% sample differs significantly from that of the
0.01 mol% sample: The spectrum of the 1 mol% doped sample shows numer-
ous excitation bands over the whole visible and ultraviolet spectral range. The
excitation spectrum of the 0.01 mol% sample is in comparison relatively simple
with one dominant band at 590 nm which can be assigned to 4I9/2→(4G5/2,
2G7/2) transitions. Even in the expended spectrum none of the UV excitation
bands seen for the 1 mol% doping level appear. The splitting effect is much
stronger in the 1 mol% doped single crystal. Looking at the peaks around
515 nm, a more detailed structure can be seen in the 1 mol% Nd3+-doped sam-
ple. Different sublevels of the (4G9/2,2K13/2)→4I9/2can be ascribed to these.
Note, that for wavelengths longer than approximately 750 nm, scattered exci-
tation light enters the spectra, whereas the 806 nm excitation (4I9/2→4F5/2)
is superimposed on this.
42
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.4: Normalized up-converted fluorescence spectra of (a) 0.01 mol% and (b)
1 mol% Nd3+-doped BaCl2under CW infrared laser excitation at 796 nm. Parts of the
up-converted fluorescence spectra (dotted curves) are expanded as indicated.
Converted Fluorescence
Up-conversion in the visible range can be observed at room temperature under
excitation from a CW infrared laser diode emitting at 796 nm. The excitation
is in resonance with the 4I9/2→4F5/2transition of the Nd3+ion. Figure 4.4
shows the up-converted fluorescence spectra for 0.01 and 1 mol% Nd-doped
BaCl2in the 340 −750 nm spectral range. Scattered excitation light enters
the spectrum at about 750 nm. After the 4F5/2level is populated through the
excitation by a 796 nm photon, the ion relaxes to the 4F3/2level. This level has
a relatively long life time due to its large energetic gap of more than 0.7 eV to
the next lowest level 4I15/2. The longer lifetime induces a higher probability for
the ion to absorb another 796 nm photon before relaxing (through an emitted
photon) to the ground state. Beside this ESA process an ETU process is also a
possibility to reach levels with higher energies than the 4F5/2level. From these
higher excited states many levels can be populated by non-radiative processes.
Figure 4.5 shows the Nd3+energy levels and possible population routes to
the higher states. The most intense emission bands are located in the green
(530 nm), yellow (590 nm), and red (660 nm) region. They arise from the 4G
multiplets [58]. The 4G multiplets, comprised of the 4G9/2,4G7/2, and 4G5/2
43
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.6: Fluorescence spectra of (a) 0.01 mol% and (b) 1 mol% Nd3+-doped BaCl2
under CW infrared laser excitation at 796 nm (dashed line). The spectra are normal-
ized to the 4F3/2→4I9/2transition at 890 nm, the up-converted fluorescence parts
(λ<800 nm) are expended by factor of 100. The spectra are corrected for the spectral
sensitivity of the detector.
states, produce overlapping emission bands in the visible spectral range, and
are therefore the strongest in the spectrum. The 4G multiplets are energeti-
cally accessible with two 796 nm photons. The weak bands in the blue spectral
range originate from the higher 2P1/2excited state. The two bands at 635 and
690 nm found in the 1 mol% Nd-doped sample are tentatively assigned to
transitions from the excited states 2H11/2and 4F9/2, respectively, to the 4I9/2
ground state. The up-converted fluorescence bands in the UV spectral range
at approximately 360, 385, and 415 nm are only accessible with three photons
(Fig. 4.5). The highest observed energy band at 360 nm arises from a 4D3/2to
4I9/2transition; the 385 nm emission is from the same excited state, but depop-
ulates to the 4I11/2ground state. The 415 nm emission can be attributed to a
transition from the 4D3/2excited state to the 4I13/2ground state.
In figure 4.6 a direct comparison between the normal fluorescence and the
up-converted fluorescence of the Nd-doped crystals is given. The one pho-
ton fluorescence efficiency at 890 nm arising from the 4F3/2level is by a factor
of around 200 better than the efficiency of the strongest up-converted fluores-
45
Chapter 4. Up-Conversion based on Nd and Er
cence at 590 nm in both samples. A direct comparison of the efficiencies of the
emission and up-converted emission to the excitation intensity was not possi-
ble. The surfaces and sizes of the different BaCl2samples are not comparable;
they are fragments of the grown single crystals. In addition to the high hygro-
scopic nature of the crystals an identical preparation of the samples was not
possible. Scattering effects at the surface, different absorption volumes, and
varying radiation patterns do not allow a normalization of the excitation. For
PL and up-conversion spectra in a direct comparison they were recorded with
the less sensitive silicon photo diode, rather than a photomultiplier. Therefore
the signal-to-noise-ratio and the resolution of the spectra are not as good as
the photomultiplier detected up-conversion spectra. Nevertheless, the main
2-photon up-converted emissions in the visible spectral range are clearly iden-
tifiable.
Lifetime Measurements
For up-conversion processes, the lifetime of the intermediate energy levels is
of particular importance. Thus, lifetime measurements of the 4F3/2level have
been performed at a wavelength of 880 nm (Fig. 4.7 a) which corresponds to
the 4F3/2→4I9/2transition. These measurements yielded different lifetimes
for different Nd3+-doping level: The measured lifetime of the 0.01 mol% sam-
ple is (280 ±1) µs, the lifetime for the 1 mol% doping level is (49 ±1) µs.
This result was expected since energy transfer processes between neighboring
Nd3+ions lead to a shorter lifetime for the sample with the higher Nd3+con-
centration.
To complete the picture on energy level lifetime also lifetime measurements
on the two-photon up-converted fluorescence were performed. The Nd3+-
doped BaCl2single crystals were excited with a CW laser diode operating at
796 nm and the up-converted fluorescence decay was detected at 590 nm (4.7b)
corresponding to transitions starting at the 4G-multiplet and depopulated to
the ground state. As expected and already observed for the infrared fluo-
rescence lifetime, the measurements yielded different lifetimes for different
Nd3+-doping level: The lifetime of the two-photon up-converted fluorescence
at 590 nm is (19 ±0.5) µs and (9.5 ±0.5) µs for the 0.01 and 1 mol% doped
sample, respectively. Additionally, comparing the fluorescence with the up-
converted fluorescence lifetimes it can be seen, that the higher energy levels
(up-conversion) radiate with a shorter lifetime as the energy levels where the
normal photoluminescence starts.
46
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.7: The upper measurements show the fluorescence decay of Nd3+in BaCl2
doped with (a) 0.01 mol% and (b) 1 mol% NdCl3at 880 nm. The lower diagram shows
the up-converted fluorescence decay of BaCl2doped with (a) 0.01 mol% and (b) 1
mol% NdCl3. The excitation was carried out with a pulsed infrared laser diode at
796 nm.
Power Dependence
Figure 4.8 shows the dependence of the emission intensity on the 796 nm exci-
tation power for the infrared fluorescence band at 880 nm corresponding to the
4F3/2→4I9/2) transition and the most intense up-converted fluorescence band
at 590 nm, which arises from transitions of the 4G multiplets to the ground
states. The power dependence was measured by inserting different optical fil-
ters (e.g. neutral density filters) in the pump beam. Fluorescence spectra were
recorded over several orders of magnitude of excitation power, e.g. from ap-
proximately 130 mW (maximal laser diode output power) down to a few µW.
For the emission at 880 nm, the slope is 0.996 and 0.992 for a Nd3+-doping level
of 0.01 and 1 mol%, respectively, e.g. an almost perfect linear dependence on
the excitation power. Saturation effects or a decreasing slope for higher excita-
tion powers are not observed. Up-conversion spectra of the 590 nm emission
47
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.8: Power dependence (on double-logarithmic scale) of the fluorescence and
up-converted fluorescence intensity in (a) 0.01 mol% and (b) 1 mol% Nd3+-doped
BaCl2, recorded at the wavelengths indicated under CW laser diode excitation at
796 nm.
were recorded for excitation powers in the range from 130 mW down to 7 mW.
For lower excitation powers the signal to noise ratio was too low for an exact
analysis. For the up-converted fluorescence band, the slope is 1.85 and 1.92
for 0.01 and 1 mol%, respectively. Neither up-conversion power dependencies
show saturation effects or a decrease of the fitted slope.
48
Chapter 4. Up-Conversion based on Nd and Er
4.2.3 Fluorozirconate Glasses and Glass Ceramics
In the last section 4.2.2 the up-converted luminescence of BaCl2single crystals
doped with NdCl3was discussed. A problem during these measurements was
the hydroscopicty of BaCl2. Crystals left for a day or longer in normal air, tar-
nish and become white, lose their physical stability, and crumble on contact.
The easiest way to solve this problem is to embed BaCl2in a protective glass
matrix. The fluorozirconate glasses, especially the ZBLAN glasses, are an ideal
host material for barium chloride crystallites. In addition to the low phonon
frequencies the ZBLAN glasses offer optimal transparency over a wide spec-
tral range. Emission and excitation bands of embedded Nd3+ions are neither
influenced nor absorbed by the glass. Another advantage of these glasses lies
in their production process conditions. They are not limited in their size, vol-
ume, and form, like single crystals, where the production price grows rapidly
with the diameter of the crystal. In table 4.1 the compositions of the three
glasses synthesized with chlorine doping are listed.
Differential Scanning Calorimetry
In further investigations glasses with different doping levels were made. All
glasses are based on the well known ZBLAN composition [21]. One glass was
doped with 1 mol% of NdCl3doping. It is comprised of 52·ZrF4-20·BaF2-
20·NaF-3.5·LaF3-3·AlF3-0.5·InF3-1·NdF3(values in mol %) and has no addi-
tional chlorine doping. The other three glasses have additional chlorine dop-
ing and are comprised of (53 −x)·ZrF4-10·BaF2-10·BaCl2-(20 −x)·NaCl-x·KCl-
3.5·LaF3-3·AlF3-0.5·InF3-x·NdF3, where x=0.5, 1, 5 (values in mol%). The
differential scanning calorimetry (DSC) measurements of three fluorozirconate
based glasses are shown in figure 4.9. Here a Nd-doped FZ (a), an undoped
FCZ (b), and a Nd-doped FCZ (c) glass are investigated. The DSC data for the
1 mol% Nd-doped FZ glass (figure 4.9 (a)) shows a glass transition at ∼262 ◦C
in good agreement with that observed in a pure FZ glass [21, 60] for a heat-
ing rate of 10 K/min. No crystallization peaks can be found upon doping
the fluorozirconate glass with Nd only; however, crystallization is initiated if
the glass is additionally doped with chlorine ions. Doping the FZ base glass
ZrF4BaF2BaCl2LaF3AlF3NaCl3InF3KCl NdCl3
ZBLAN 112 52.5 10.0 10.0 3.5 3.0 19.5 0.5 0.5 0.5
ZBLAN 111 52.0 10.0 10.0 3.5 3.0 19.0 0.5 1.0 1.0
ZBLAN 113 48.0 10.0 10.0 3.5 3.0 15.0 0.5 5.0 5.0
Table 4.1: Composition of the Nd- and K-co-doped FCZ glasses in mole percent. The
name of the samples is a composition of the abbreviation ZBLAN given by the com-
pounds of the FZ glasses ([21]) and an inventory number. The glasses are sorted by
their Nd doping concentration.
49
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.9: DSC measurements for FZ and FCZ glasses. From top to bottom are the
measured curves of (a) Nd-doped FZ, (b) undoped FCZ and (c) Nd-doped FCZ glass.
Figure 4.10: Picture of the sample illuminated from the back. The rows are from top to
bottom: 1 mol% Nd-doped FZ, follwing the three Nd-doped FCZ glasses with 0.5, 1,
and 5 mol% neodymium doping. The columns are the annealing temperatures; from
left to right: as-made, 240, 250, 260, 270, 280, and 290 ◦C for 20 minutes.
50
Chapter 4. Up-Conversion based on Nd and Er
additionally with chlorine (figure 4.9 (b)) shifts the glass transition tempera-
ture to ∼211 ◦C which is significantly less than that for the undoped FZ base
glass. The exothermic peak at about 250 ◦C is assigned to the crystallization of
hexagonal BaCl2. The two peaks at ∼310 ◦C and ∼365 ◦C are also crystalliza-
tion peaks; the latter one is also observed in pure FZ glass where the main glass
crystallization starts at ∼350 ◦C [21]. For the as-made 5 mol% Nd-doped FCZ
glass (4.9 (c)) the glass transition temperature is around 216 ◦C. Interestingly,
the peak for the hexagonal phase BaCl2crystallization is now at ∼280 ◦C. The
crystallization peak at 310 ◦C is shifted by 10 ◦C to 320 ◦C whereas the glass
crystallization peak is now at ∼335 ◦C. Further DSC measurements were per-
formed on the 5 mol% doped sample. They are added in appendix C.
Figure 4.10 shows the four synthesized glasses. The samples were illumi-
nated from the back when the picture was taken. The glasses from top to bot-
tom are: 1 mol% Nd-doped FZ, followed by the three Nd-doped FCZ glasses
with 0.5, 1, and 5 mol% neodymium. Due to subsequent annealing of the
sample at temperatures in the range of the crystallization temperature of the
samples in three of the four glasses nano-crystals were formed in the glass ma-
trix. Light scattering effects can be observed and became stronger with higher
annealing temperatures. The columns are 20 minutes annealing temperatures
for 20 minutes of as-made, 240, 250, 260, 270, 280, and 290 ◦C, from left to
right. The first glass without additional chlorine doping, did not form any
nano-particles. The characteristic BaCl2crystallization temperature shifted to
lower temperatures due to additional chlorine doping or to higher tempera-
tures with increasing Nd-doping; this can be observed in the picture.
X-Ray Diffraction
Figure 4.11 shows the x-ray diffraction (XRD) data for the as-made and an-
nealed (for 20 minutes) 5 mol% Nd-doped FCZ glasses; temperatures are as
indicated. The XRD data of the as-made sample consists of very broad peaks at
about 26◦and 47◦, typical for glasses close to the ZBLAN formulation [55] for
copper Kαradiation. For an annealing temperature of 240 ◦C (data not shown)
no significant change in the XRD pattern can be observed. However, sharper
peaks arise upon increasing the annealing temperature from 250 ◦C to 290 ◦C.
These peaks can be identified as reflections from hexagonal BaCl2(space group
P62m(189), a=0.8066 nm, c=0.4623 nm, bar graph (PDF #45-1313) in fig-
ure 4.11) [54]. There is no phase transition from hexagonal to orthorhombic
phase BaCl2as was found in other RE-doped FCZ glass ceramics [61, 62]. For
the 290◦C annealed sample, however, there are peaks from phases which have
not yet been identified; they are marked with asterisks in figure 4.11.
The XRD peaks are wider than the instrumental broadening of 0.085◦. The
FWHM is becoming narrower with higher annealing temperatures, suggesting
51
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.11: XRD data for the 5 mol% Nd-doped FCZ glass ceramic, as-made and
annealed, for 20 min at 250 ◦C, 270 ◦C, and 290 ◦C. The curves are vertically displaced
for clarity. The line pattern of hexagonal phase BaCl2(PDF #45-1313) is shown for
comparison. The asterisks mark additional peaks from unidentified phases.
Figure 4.12: Particle size (full squares) and integrated intensity (open squares) of the
(201) reflection vs. annealing temperature for hexagonal BaCl2. The lines are a guide
to the eye.
52
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.13: Optical density of a Nd3+-doped FCZ glass sample. The labeled transi-
tions start from the 4I9/2ground state and end on the levels indicated. The sample
thickness was 1 mm. The inset shows an expanded version of the absorption band in
the IR spectral range that will be used for excite the up-conversion.
size-broadening effects. The particle sizes can be estimated by using the Scher-
rer formula (see equation 2.11 in chapter 2) [42]. Any stress or strain effects,
which may also lead to additional line broadening are not considered. The
Scherrer formula is applied to the (201) reflection of hexagonal phase BaCl2at
about 32◦. For fitting the line profile a Lorentzian function was used. The par-
ticle diameters, after thermal processing at 250, 260, and 270 ◦C, are between
20 and 60 nm. The particles grow rapidly and their sizes become >100 nm
(figure 4.12, full squares) for annealing temperatures at 280 ◦C and above. The
XRD linewidth of the (201) reflection for 280 ◦C and 290 ◦C annealing is too
close to the instrumental resolution to allow a precise estimate of the particle
size (see error bars in figure 4.12, full squares). In addition, figure 4.12, open
squares, shows that for annealing above 270 ◦C the hexagonal phase BaCl2
nano-crystals start to dissolve, i.e. the hexagonal phase BaCl2volume fraction
in the glass decreases.
Excitation and Fluorescence
Figure 4.13 shows the optical density of the 5 mol% Nd3+-doped FZ sample.
The material has strong Nd3+absorption at 578, 742, and 797 nm and very
weak absorption at 427, 625, and 679 nm. Additional small absorption bands
can be found at 469 and 427 nm. At other wavelengths the absorption strengths
are intermediate and can observed at 867 nm and two bands are observable at
53
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.14: Optical density of four Nd3+-doped FCZ glass samples. The sample
thicknesses were 1 mm each. The measurements for the as-made glass (black) and
these annealed at 250 ◦C (red), 270 ◦C (green), and 290 ◦C (blue) are shown.
∼521 and 353 nm. The absorption bands are in good agreement with literature
[30, 58, 59] and the energy level diagram of trivalent neodymium (see figure
4.5). The energy levels to which the trivalent Nd is excited due to the ab-
sorptions [30] are indicated in the diagram. In addition to the observed Nd3+
absorption bands, the material shows some background absorption in the UV
spectral range below 300 nm. The inset in the upper right area shows the ab-
sorption bands in the NIR spectral range in more detail. The up-conversion
experiments were carried using the absorption bands at ∼797 nm. The CW
laser diode used has an excitation wavelength of 796 nm and is in resonance
with the transition from the 4F5/2ground state to the 4F3/2excited state. The
relatively high optical density is caused by the sample thickness being ∼1 mm
and cords in the glass. Figure 4.14 shows the optical density of the as-made
and three annealed glasses. The black curve is from the as-made glass already
shown in figure 4.13. The colored curves are the measurements of the annealed
glass ceramics: the red one is annealed at 250 ◦C, the green one at 270 ◦C, and
the blue one at 290 ◦C, respectively; the higher annealing temperature corre-
sponds to higher optical density. The increase in the optical density is stronger
for higher photon energies, e.g. for the UV and blue spectral range. This is
affected by scattering effects of the nanoparticles; growing with the annealing
temperature.
54
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.15: Fluorescence spectra for 5 mol% Nd-doped FCZ glass ceramics. From
bottom to top: as-made and annealed at 250, 270, and 290 ◦C for 20 minutes. The
spectra were blown up by 25 and 50 (dashed curves) for the low intensity emissions
at 950 and 1165 nm, respectively.
Figure 4.15 shows the infrared fluorescence spectrum of different 5 mol%
Nd-doped FCZ glass samples: as-made, annealed at 250, 270, and 290 ◦C for
20 minutes. The excitation was carried out in resonance with the 4I9/2→4F5/2
transition of Nd3+. All samples show three main emission bands peaking at
885, 1060, and 1340 nm which can be assigned to transitions from the 4F3/2ex-
cited state to the 4I9/2,4I11/2, and 4I13/2ground state of Nd3+, respectively. The
expanded IR emission spectra of the samples (see figure 4.15 dashed curves,
expanded by a factor of 25 and 50) clearly show additional emission bands be-
tween the three main emissions at approximately 950 and 1165 nm. These are
probably caused by 4F5/2to 4I11/2and 4F5/2to 4I13/2transitions, respectively.
The experimental resolution for the fluorescence measurements was 0.2 nm.
The corresponding excitation spectra (see figure 4.16), were recorded (a) at
1060 nm for the 4F3/2→4I11/2transition, and (b) for the 4F3/2→4I9/2transi-
tion at 880 nm. The relative intensity ratio as well as the shape of the excitation
bands are nearly identic for the spectra detected at 1060 and at 880 nm with
one exception; an additional excitation peak at 703 nm in the spectrum de-
tected at 1060 nm. The energy of this emission is equal to the energy difference
of the 4I11/2→2H11/2transition. Both spectra of the Nd3+-doped samples
show numerous excitation bands over the whole visible and in the ultraviolet
spectral range. Note, that for wavelengths longer than approximately 820 nm,
scattered excitation light starts to enter the spectra detected at 880 nm.
55
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.16: Excitation spectra of the Nd3+-doped FCZ glass. The emission was de-
tected at (a) 880 nm and (b) at 1060 nm, respectively. The dashed curves show the
measurements in the range below 500 nm and are expanded a factor of ten. Both
spectra were normalized to their most intense emission.
In figure 4.17 the measurements of the excitation spectra of the as-made
glass and the annealed glass ceramics in the spectral range from around 400 to
820 nm are shown. For the sake of clarity the measurements of the annealed
samples are not shown as excitation spectra, but the difference between the
measurements of the as-made glass and the annealed samples are displayed.
The dashed lines mark the most intense emission lines. At these position the
difference curves show a double peak structure. This means that the excita-
tion peaks have a broadening effect and a larger FWHM due to the thermal
treatment whereas the maximum intensity of the excitation peak only grows
slightly. Whereas annealed samples with annealing temperatures of 250 ◦C
and above show about the same in intensity, shape and spectral position the
glass ceramic annealed at 240 ◦C shows no significant difference compared to
the as-made sample. The enhancement of the FWHM of the 579 nm excitation
peak compared between the as-made and annealed samples versus the anneal-
ing temperature is shown in table 4.2. The values are are given to an accuracy
of 0.1 and the enhancement factor has a calculated error of ∼ ±0.5 %.
56
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.17: Excitation spectra of the 5 mol% Nd3+-doped FCZ glass and annealed
glass ceramics. The emission was detected at 880 nm. The spectrum of the as-made
and the differences between this and that of the annealed samples are shown. The
dashed lines mark the intense emission peaks.
annealing temperature FWHM enhancement
(◦C) (nm) (%)
as-made 22.1 -
240 22.9 3.6
250 26.5 19.9
260 26.6 20.5
270 26.5 20.1
280 26.4 19.2
290 26.4 19.3
Table 4.2: Enhancement factor of the FWHM of the 579 nm excitation listed after
annealing temperature. The FWHM was approximated to one position after decimal
point for this table. The enhancement factor has an error of ±0.5 %.
57
Chapter 4. Up-Conversion based on Nd and Er
Converted Fluorescence
Up-converted fluorescence in the visible and in the ultraviolet range was ob-
served under excitation at 796 nm in resonance with the 4I9/2→4F5/2tran-
sition of Nd3+. The recorded spectra for the as-made and annealed samples
at 250, 270, and 290 ◦C are shown in figure 4.18. The spectra are shown in
the range of 340 - 750 nm and were normalized to their most intense emission
at 590 nm. The spectral behavior at ∼750 nm is not shown, because scat-
tered excitation light enters the spectrum. The most intense emission bands in
the visible region are located in the green (530 nm), yellow (590 nm), and red
(660 nm). They arise from the 4G multiplets which, comprised of the 4G9/2,
4G7/2, and 4G5/2states, produce overlapping emission bands, and are there-
fore the strongest in the spectrum [58]. The 4G multiplets are energetically
accessible with two 796 nm photons. The two up-converted bands in the ul-
traviolet region at about 360 and 385 nm correspond to transitions from the
4D3/2excited state to the 4I9/2and 4I11/2ground state, respectively [58]. The
4D3/2level is only accessible with three 796 nm photons. The up-converted flu-
orescence spectra change their shape and relative intensities upon annealing.
For annealing temperatures of 250 ◦C and above the UV emissions become
stronger; they reach a maximum for the 270 ◦C annealed sample and then de-
crease slightly. All up-converted emissions show a splitting after annealing.
Crystal Field Splitting
For a more detailed examination of this effect, a comparison for the three emis-
sions in the green and blue spectral range between the as-made sample and the
sample annealed at 270 ◦C is given in figure 4.19. The intensities and offsets
of the curves are plotted as in figure 4.18. The spectra were recorded with the
same spectral resolution. The left curves shows the emission band at 530 nm
in the green belonging to the 4G7/2→4I9/2transition. The two emissions in
the blue range are the 2P1/2→4I9/2(435 nm) and 2P3/2→4I11/2(413 nm)
transition. The as-made sample shows emission bands without an explicit
substructure. Upon annealing the up-converted emission bands splits. With
higher annealing temperatures the splitting of the emission becomes stronger.
The strongest observable splitting shows the 270 ◦C annealed sample. For the
2 samples annealed at higher temperatures, 280 and 290 ◦C, the splitting de-
creases instead of becoming stronger. Additionally, a change in the intensities
of the up-converted emission can be observed for the emissions in the blue
and ultraviolet spectral range. These transition can only be reached by subse-
quently absorbing three 796 nm photons.
58
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.18: Normalized up-converted fluorescence spectra of 5 mol% Nd3+-doped
ZBLAN glass ceramics under CW infrared laser excitation at 796 nm. The spectra for
the as-made and three annealed glasses are shown. For clarity the curves were plotted
with an offset.
Figure 4.19: Up-converted emission spectra of the 530 nm (4G7/2→4I9/2) emission,
435 nm (2P1/2→4I9/2), and the 413 nm emission (2P3/2→4I11/2) are shown for the
as-made and the 270 ◦C annealed sample.
59
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.20: Fluorescence spectra for 5 mol% Nd-doped FCZ glass ceramics. From
bottom to top: as-made and annealed at 250, 270, and 290 ◦C for 20 minutes. The
excitation was carried out with a LD at 811 nm.
Figure 4.21: Comparison of the IR emission spectra of Nd-doped FCZ glass ceramics.
The spectra for the as-made and 270 ◦C annealed sample are shown. The excitation
for the left spectra was a LD emitting at 796 nm, on the right at 811 nm.
60
Chapter 4. Up-Conversion based on Nd and Er
Additional measurements of the IR emission spectra were performed. An-
other CW laser diode with λ=811 nm was used to excite the 5 mol% Nd-
doped samples this time. In figure 4.20 the IR emission spectra in the range
between 830 and 1500 nm are shown for the as-made and three annealed sam-
ples. The observed emissions can be attributed to the transitions between the
energy levels of the trivalent Nd. For the annealed sample a splitting of the
emission bands is observed. The intense emission at 1050 nm shows a split-
ting into a substructure of three noticeable peaks superimposed on the main
emission from the as-made glass. For an annealing temperatures of 270 ◦C
the splitting is a maximum; for temperatures above 270 ◦C the splitting re-
duces again. The relatively broad emission band between 850 and 900 nm
(4F3/2→4I9/2) shows a similar behavior; it splits into 5 smaller peaks. The
lowest emission band at ∼1350 nm does not shows such a splitting effect.
Figure 4.21 shows the most intense IR emission of trivalent Nd for the as-
made glass (bottom) and the 270 ◦C annealed glass ceramics (top) excited at
796 nm (left) and 811 nm (right). The spectrum of the as-made sample excited
at 796 nm is identical in shape and spectral position within the experimental
error to the spectrum excited at 811 nm. The 270 ◦C annealed sample shows
different spectra upon the two excitations. The spectrum excited at 796 nm
is the same as the spectra of the as-made sample. However, the spectrum of
the 270 ◦C annealed sample excited at 811 nm shows a strong splitting in the
emission band. The emission splitting is similar to those seen in the Nd-doped
BaCl2single crystals. From this it can be assumed that the 270 ◦C annealed
sample contains Nd ions both in the glass matrix and in the particles.
For deeper understanding of the crystal field splitting further temperature
depended measurements were made. The IR emission spectra of an annealed
sample of the 5 mol% Nd-doped FCZ glass ceramics were measured at tem-
peratures between 20 and 290 K. Figure 4.22 (a) shows four of the measured
emission spectra in the range between 860 and 920 nm. These measurements
had to be performed with a PMT instead of a germanium detector due to the
experimental setup for low temperature measurements. The shape of the ob-
served emission changes due to the different spectral responses of the detec-
tors. In (b) the emission spectra were plotted versus the temperature in the
range between 20 and 290 K and 855 to 920 nm. All spectra were normalized
to their most intense emission at 901 nm. It can be see, that the emission spec-
trum at 20 K is relatively simple and shows three strong peaks. A shoulder
on the lower wavelength side can be observed. At higher temperatures the
three clear emission peaks spectrum broaden with a few minor peaks super-
imposed.
61
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.22: (a) shows four IR emissions at around 880 nm according to the
4F3/2→4I9/2transition for temperatures from 20 to 290 K. The spectra were mea-
sured at the temperatures indicated. (b) is a contour plot containing all temperature
depending IR emission measurements.
62
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.23: Normalized fluorescence decay of Nd3+in the 270 ◦C annealed FCZ glass
ceramic in a logarithmic plot. The fluorescence was detected for normal fluorescence
(880 nm) and up-converted fluorescence (590 and 387 nm) as indicated and was ex-
cited at 796 nm with an infrared laser diode in resonance with the 4I9/2→4F5/2
transition of trivalent Nd. The measurements were fitted (red lines) with linear func-
tion.
Lifetime Measurements
Figure 4.23 shows the normalized fluorescence decay of Nd3+in the 270 ◦C
annealed FCZ sample. The Nd3+decay was recorded for the infrared fluores-
cence at 880 nm, the yellow, 2-photon up-converted fluorescence at 590 nm,
and the ultraviolet, 3-photon up-converted fluorescence at 387 nm. The life-
time of the 880 nm infrared fluorescence is (200 ±4) µs, that of the 2-photon
up-converted fluorescence (116 ±5) µs, and that of the 3-photon up-converted
fluorescence (68 ±6) µs. As was already stated in section 2.1.2, of particu-
lar importance to the up-conversion efficiency is the intermediate energy level
lifetime, specifically for the 2-photon up-conversion processes the energy level
lifetime of the 4F3/2level is important. Figure 4.24 shows that the lifetime of
the 4F3/2to 4I9/2(880 nm) and that of the 4G7/2→4I9/2(593 nm) transition
depends significantly on the annealing temperature. The longest lifetime of
the intermediate energy level 4F3/2was found to be for the 260 ◦C sample with
(204 ±4) µs. By annealing the samples at higher temperatures the lifetimes of
the intermediate level decrease again. The up-converted fluorescence lifetime
of the 4G7/2energy level is dependent on the annealing temperature and ob-
tains the longest lifetime of (123 ±5) µs for the sample which was annealed at
280 ◦C for 20 minutes. It seems that the energy levels are all slightly different
effected by the changing crystal field upon the annealing process. This could
led to the different annealing temperatures for the maximum fluorescence and
up-converted fluorescence lifetimes.
63
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.24: Fluorescence (880 nm, black squares) and up-converted fluorescence
(590 nm, open circles) decay time of Nd3+doped samples is shown vs. annealing
temperature.
Power Dependence
Figure 4.25 shows the dependence of the emission intensity on the 796 nm exci-
tation power for the infrared fluorescence band at 880 nm corresponding to the
4F3/2→4I9/2transition and the most intense up-converted fluorescence band
at 590 nm, which arises from transitions of the 4G multiplets to the ground
states. The three-photon-up-conversion emission at 385 nm (4D3/2→4I11/2) is
also plotted. The power dependence was measured by inserting different opti-
cal filters (e.g. neutral density filters) in the pump beam. Fluorescence spectra
were recorded over several orders of magnitude of excitation power, e.g. from
approximately 130 mW (maximal laser diode output power) down to a few
µW. For the emission at 880 nm, the slope is 1.00 with a confidence level of
almost one, e.g. an almost perfect linear dependence on the excitation power.
Up-conversion spectra of the 590 nm emission were recorded for excitation
powers in the range from 130 mW down to less than 2 mW. For lower excita-
tion powers the signal to noise ratio was too low for an exact analysis. For the
two photon up-converted fluorescence emission, the slope is 1.87. The three
photon up-conversion was measured in an excitation range between 130 mW
and 17 mW. At this level the signal to noise ratio becomes too low for fur-
ther measurements. The slope can be determined as 2.76. Both up-conversion
power dependencies show nor saturation effects neither a decrease of the fitted
slope.
64
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.25: This figure shows the intensities of the PL, two- and three-photon-up-
conversion versus excitation power. The intensities of the emissions are normalized
to their intensity at 130 mW. The 880 nm PL are plotted as squares, the two photon-
up-conversion at 590 nm as circles and the three-photon-up-conversion at 385 nm as
triangles. The slope of the fits is indicated as well as the confidence level R.
Nd Doping Level Dependencies
After the optical and structural characterization of the 5 mol% Nd-doped FCZ
glass ceramics the same measurements were performed on the 0.5 and 1 mol%
doped glasses and glass ceramics. These sample were annealed after synthe-
sis at the same temperatures as the 5 mol% doped sample. A picture of the
samples is shown in figure 4.10, which depicts differences in the scattering
effect when illuminated from the back. Due to the lower crystallization tem-
perature of the BaCl2nanocrystalites for lower Nd doping concentrations (see
figure 4.9) different crystallite sizes for the same annealing temperatures are
expected. Prasad et al. [29] showed that the optical behavior and properties of
optical active nanoparticles are critically effected by their size.
In figure 4.26 the emission intensity of the different annealed samples is
plotted for the three Nd-doped FCZ glass ceramics. The intensity is given
by the color; going from blue to red for higher intensities. The spectra were
normalized to the integrated up-converted emission intensities to obtain bet-
ter comparison. The spectral fraction of the normalized spectra are shown in
figure 4.27. Here the percentage for the different spectral ranges to the whole
up-converted emission spectrum are shown as a function of the annealing tem-
peratures of the glass ceramic samples. The ranges are chosen with respect to
the main emissions and are shown in table 4.3. From the three diagrams of
figure 4.26 the same up-converted emissions of trivalent Nd can be observed.
65
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.26: The spectral behavior of the normalized up-conversion intensity is plot-
ted versus the annealing temperature. The three figures are from top to bottom: 0.5,
1.0, and 5.0 mol% Nd-doped FCZ glasses and glass ceramics. The intensities are given
by the color; from blue to red for higher intensities.
66
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.27: These figures show the percentage for the different spectral emission
ranges to the whole integrated normalized up-converted emission spectra. The three
diagrams are from top to bottom: 0.5, 1.0, and 5.0 mol% Nd-doped samples. The color
indicates the chosen spectral range (see legend above and table 4.3).
67
Chapter 4. Up-Conversion based on Nd and Er
number range emission(s)
start (nm) end (nm) (nm)
1 339 370 359
2 370 400 385
3 400 425 413
4 425 443 436
5 436 477 450, 461
6 477 551 530
7 551 620 590
8 620 700 663
Table 4.3: List of chosen spectral ranges for up-converted emissions of Nd-doped FCZ
glasses ceramics.
The blue and UV spectral emissions are relatively more intense the higher the
neodymium doping. The two upper plots of the 0.5 and 1 mol% doped sam-
ples show conspicuous features when annealed at temperatures in the range
of 240 to 260 ◦C. The intensities of the 530 nm emission rise up to nearly 50 %
and 40 % of the integrated intensity for the 0.5 and the 1.0 mol% doped sam-
ple, respectively. In both samples is the orange-red spectral range the least
intense. The emissions in the blue and UV spectral range contribute only a
small part to the integrated intensity. The sum is almost constant at ∼20 %
for the lowest doped samples and did not change significantly for higher an-
nealing temperatures. The 1 mol% doped sample shows a slightly intensity
which goes up when annealing up to 250 ◦C. After this maximum point the
intensity decreases and ends up at ∼25 %. The spectra of the 5 mol% doped
sample have already been shown in figure 4.18 where the spectrum was nor-
malized to the strongest emission for that figure. Here it can be seen that the
up-converted emissions in the blue and ultraviolet get stronger in comparison
to the emissions in the green and red spectral range as annealing temperature
increases. They reach a maximum of ∼50 % for annealing temperatures of
∼260 and 270 ◦C, and decrease slightly for temperatures up to 290 ◦C. Even
the sample annealed at 290 ◦C shows a blue-UV emission of ∼40 %.
In figure 4.28 the infrared emission spectra of the different doped glasses
is shown. The spectra of the as-made glasses were plotted with an offset to
clarify the figure. All glasses show the typical trivalent Nd emissions. The
three most intense emissions, all starting from the 4F3/2energy level, can be
observed in all glasses with nearly the same intensities. Although the spec-
tral positions of the emissions are identical within the experimental error. All
emissions of the three spectra have the same shape. Only the emission corre-
sponding to the 4F3/2→4I13/2transition of the 1 mol% doped samples differs
68
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.28: This figure shows the measured IR emission spectra of the three as-made
glasses. From top to bottom the spectra of 0.5, 1, and 5 mol% Nd-doped FCZ glass
ceramics are shown.
slightly in its shape from the others. The IR emission spectra of different tem-
perature annealed glass ceramics are shown for the 5 mol% doped sample in
figure 4.2. There is no significant change in their intensities, shapes, or emit-
ting wavelengths. The same result is found for the emission spectra of the
annealed samples of the 0.5 and 1 mol% doped glass ceramics. Therefore the
spectra are not shown in this chapter but were added to the appendix. They
are shown in figure A.3 on page 93.
The excitation spectra of the three (0.5, 1.0, and 5.0 mol%) Nd-doped glasses
are shown in figure 4.29. Only the measured spectra from the as-made glasses
are depicted here for comparison. The excitation spectra of the annealed sam-
ple can be found in the appendix (figure A.4, 0.5 and 1.0 mol% doped Nd3+)
and in figure 4.17 (5.0 mol% doped FCZ). The spectra are normalized to the
most intense line at 579 nm corresponding to the 4I9/2→4G5/2transition. No
significant difference between the three as-made glasses can be observed. The
excitation intensities as well as their spectral position and shape are nearly
identical. Only the FWHM of some excitation peaks variy with the doping
level; the higher the doping level the broader the FWHM of the peaks. In addi-
tion the excitation at around 741 nm increases from 0.67 up to 0.78 (normalized
intensity) when raising the doping level from 0.5 up to 5.0 mol% Nd3+.
69
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.29: Measured excitation spectra of the three as-made glasses. From top to
bottom the spectra of the following FCZ glasses are shown: 0.5, 1, and 5 mol% Nd-
doped.
4.2.4 Discussion
Infrared fluorescence and up-converted fluorescence spectroscopy (figure 4.2
and 4.4) show that Nd3+-ions can be incorporated into BaCl2single crystals
and elucidates therein its characteristic optical properties, in particular, an up-
conversion effect upon excitation with a laser diode operating at 796 nm. It
is also possible to form hexagonal BaCl2nanoparticles (figure 4.11) in a Nd-
doped ZBLAN glass matrix. Additionally, it was shown that Nd3+ions enter
these particles and leads to enhanced two and three photon up-conversion
emissions in the visible and UV spectral range (see figure 4.30).
It was found that the glasses without additional chlorine doping show no
crystallization peak (figure 4.9). Therefore the chlorine doping is essential to
initiate crystallization of the particles in the glass. Additionally, the glass tran-
sition temperature as well as the crystallization peak shifts to higher tempera-
tures for higher doping levels. Both effects are confirmed by a back illuminated
picture of samples, shown in figure 4.10. Samples without chlorine show no
scattering effects. The scattering effect gets stronger for less doped samples
compared to samples with the same annealing temperature. Upon thermal
processing, nanoparticles are formed in the glass; these can be identified as
hexagonal phase BaCl2. There is no phase transition observable as was found
in Eu-doped [61, 62] or Ce-doped [63] FCZ glass ceramics. From analysis of
measurements the particle size grows from a few tens up to about 150 nm and
70
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.30: FCZ glass ceramic sample with its intense yellow-white up-converted
fluorescence. The excitation is carried out with an IR laser diode operating at 796 nm.
their volume fraction (maximum at an annealing temperature of 270 ◦C) can
be determined. The particle size determination by XRD has large errors for
sizes of 100 nm and more. Transmission electron microscopy measurements
should be a possibility to get more information on the particle size and size
distribution. The appearance of a crystalline phase in the XRD patterns (see
figure 4.11) after thermal processing can be correlated to crystallization peaks
in the DSC data (plotted in figure 4.9).
The optical density increases (figure 4.14) due to the scattering effects from
the larger particles precipitate for higher annealing temperatures. The infrared
emission spectra show the characteristic Nd3+emissions for all samples. The
emissions show no significant change in their shape or intensity upon the an-
nealing. The IR spectra of the glasses and glass ceramics are comparable with
the spectra from the BaCl2single crystals (figure 4.2), whereas the single crys-
tals show a crystal field splitting. In the excitation spectra (figure 4.17) the sam-
ples annealed at temperatures of 250 ◦C and above show a relative increase in
the excitation bands in the UV and visible spectral range. The FWHM of some
excitation bands becomes wider (table 4.2); an increase of 20.5 % was observed
for the at 260 ◦C annealed and with 5 mol% Nd-doped sample. This effect
could be useful for an up-conversion excitation with normal sunlight due to
the possibility for absorbing a higher number of photons. The excitation spec-
tra of the lower doped glasses (figure A.4) also show these spectral broadening
effect.
The up-conversion spectra (figure 4.4 and 4.18) suggest that the most impor-
tant excited states for up-conversion fluorescence are 4D3/2,2P1/2, and the 4G
multiplets. The 796 nm photons are strongly absorbed and excite ions into the
71
Chapter 4. Up-Conversion based on Nd and Er
4F5/2level from which they relax non-radiatively to the relatively long lived
4F3/2level. The 4G multiplets can be populated by excited state absorption
(ESA) and/or energy transfer up-conversion (ETU). In the ESA process a sec-
ond 796 nm photon excites the ion from the 4F3/2to 2D5/2, for which there is
a good energy match with 796 nm photons; this is followed by non-radiative
relaxation into the 2P1/2level and into the various 4G states. In the ETU pro-
cess, each ion is separately excited into the 4F3/2state. From that level de-
excitation of one ion can result in excitation of the other. The Nd level scheme
provides a number of such possibilities (figure 4.5). Close energy matching
is obtained with excitation from 4F3/2to 4G9/2accompanied by de-excitation
to 4I11/2and also with excitation to 4G7/2and de-excitation to 4I13/2. Excita-
tion to the 4D3/2state clearly requires the energy of three 796 nm photons.
However, it is difficult to identify the excitation mechanism without more de-
tailed information on the energy levels beyond those depicted in the energy
level diagram. An ESA process could occur in these higher levels from one
of the 4G levels, which are known to be significantly populated by 2-photon
excitation followed by non-radiative relaxation. Alternatively an energy trans-
fer mechanism could operate with an ion in a 4G5/2state interacting with a
neighbor in a 4F3/2state, resulting in excitation of one ion to 4D5/2accompa-
nied by de-excitation to 4I9/2. The 4D5/2state would rapidly relax to populate
4D3/2, giving rise to the observed emissions. The spectra of the up-converted
emissions (see figure 4.18) suggest likewise that Nd3+is incorporated in the
nanoparticles during the annealing process. The emission intensities in the
UV and blue spectral range increase upon annealing compared to the other
up-converted emissions. The 5 mol% doped glass ceramics reaches a maxi-
mum at around 260 ◦C, whereas it is reached at lower temperatures for the
lower doped glass series. Another effect of the annealing process is a splitting
of the up-converted emission (figure 4.19) which is affected by the local crys-
tal field [57]. The maximum splitting was found for the sample annealed at
270 ◦C. This sample had shown the highest volume fraction. The largest of the
splitting effect is again at lower temperatures for lower doped glass ceramics;
at 240 and 250 ◦C for the 0.5 and 1.0 mol% doped samples, respectively. Mea-
surements with another laser diode operating at λem =811 nm for excitation
lead to Nd3+infrared emission spectra (see figure 4.20) that show also such
a crystal field splitting effect depending on the annealing temperature. These
spectra show various similarities with the measured infrared emission spec-
tra of the Nd-doped BaCl2single crystals; compare figure 4.20 with figure 4.2.
From this it can also be assumed that upon annealing some of the Nd3+ions
enter the nanocrystals leading to a splitting of the infrared and up-converted
fluorescence spectra.
72
Chapter 4. Up-Conversion based on Nd and Er
As already stated in the section 2.1.2, of particular importance to the up-
conversion efficiency is the intermediate energy level lifetime, specifically for
two-photon up-conversion processes the energy level lifetime of the 4F3/2level
is important. Lifetime measurements have been performed (see figure 4.7 and
figure 4.24) for fluorescence (880 nm) and up-converted fluorescence (593 nm).
For both cases the lower doped single crystal has a longer lifetime. It is as-
sumed that quenching effects decrease the fluorescence lifetime due to non-
radiative energy transfer between neighboring Nd3+ions. The measurements
on the glass ceramics show a significant dependency on the annealing temper-
ature. The longest lifetime of around 204 µs for this level was found for the
sample annealed at 260 ◦C; leading to an enhancement of 15.1 % when com-
pared to the lifetime (177 µs) of the as-made glass sample. The energy levels
are affected slightly differently by the changing crystal field upon annealing.
Due to this effect the maximum fluorescence and up-converted fluorescence
lifetime were not measured for the same sample, but for glass ceramics an-
nealed at 260 and 280 ◦C, respectively.
In principle, for low excitation powers a quadratic dependence of the two-
photon up-converted fluorescence at 590 nm and a cubic dependence of the
three-photon up-converted fluorescence on the excitation power is expected
[15]. The power dependence of the Nd-doped single crystals shown in figure
4.8 yields, however, for the two-photon up-conversion exponents of 1.85 and
1.92 for 0.01 and 1 mol%, respectively. The power dependence of the two and
three photon up-conversion in the glass ceramics shown in figure 4.25 yields
exponents of 1.87 and 2.76, respectively whereas the normal PL shows a nearly
perfect linear decay. This experimentally observed decrease is determined by
the competition between linear decay and up-conversion processes for the de-
pletion of the intermediate excited states. The shorter fluorescence lifetime for
the higher doped BaCl2single crystal (figure 4.7), i.e. the faster depletion of the
4F3/2state leads to a larger slope in the up-converted intensity vs. excitation
power dependence. The intensity of up-converted fluorescence that is excited
by the sequential absorption of mphotons has a dependence on absorbed ex-
citation power Pin, which may range from Pm
in in the limit of infinitely small
up-conversion rates down to P1
in for the upper state and less than P1
in for the
intermediate states in the limit of infinitely large up-conversion rates (see table
2.2 and [15]). The fact that the deviation from m=2 is small in the case of 0.01
and 1 mol% Nd3+-doped BaCl2indicates that these systems are far away from
the point where saturation effects will play a role. It is the same for the glass
ceramics, with the fitted slopes of 1.85 (two photon up-conversion m=2) and
2.76 (three photon up-conversion m=3).
73
Chapter 4. Up-Conversion based on Nd and Er
4.3 Dopant: Erbium
4.3.1 Motivation
In the last section Nd-doped glasses and glass ceramics were investigated
for up-converting processes. These up-converting glass ceramics were pro-
duced as a model system for an up-converting back layer for bifacial solar
cells. However, the Nd-doped glasses are not applicable as up-converters for
silicon solar cells due to their excitation wavelength being around 800 nm.
Since this light can be absorbed by the solar cell itself, the dopant was changed
from neodymium to erbium as the energy levels of Er3+have an absorption
band at ∼1540 nm. Therefore erbium as dopant is much more suitable for
up-converting backlayers for silicon solar cells. Additionally, excitation with
800 nm photons is also possible.
4.3.2 Fluorozirconate Glasses
As mentioned in the last section the ZBLAN glass system is an ideal host for
optically active rare earth ions. In addition to the low phonon frequencies the
ZBLAN glasses offer optimal transparency over a wide spectral range. Emis-
sion and excitation bands of embedded Er3+ions are neither influenced nor
absorbed by the glass. The composition of the erbium doped glasses are given
in table 4.4.
Differential Scanning Calorimetry
All glasses are based on the well known ZBLAN composition [21]. For the
2 and 5 mol% doped samples, the ErF3-doping was done at the expense of
ZrF4. These glasses are comprised of (53 −x)·ZrF4-20·BaF2-20·NaCl-3.5·LaF3-
3·AlF3-0.5·InF3-x·ErF3, where x=2 or 5 (values in mol%). For 9.1 and 13 mol%
doping, the adding of ErF3is compensated by reducing all other components
proportionally. The DSC data of these fluorozirconate based glasses are shown
in figure 4.31. The DSC data for the 2% Er-doped FZ glass has a glass transition
sample ZrF4BaF2LaF3AlF3NaF InF3ErF3
ZBLAN JJ33 51.0 20.0 3.5 3.0 20.0 0.5 2.0
ZBLAN JJ34 48.0 20.0 3.5 3.0 20.0 0.5 5.0
ZBLAN JJ54 48.2 18.2 3.2 2.7 18.2 0.5 9.1
ZBLAN JJ57 46.1 17.4 3.0 2.6 17.4 0.4 13.0
Table 4.4: Composition of ErF3FZ glasses in mole percent. The name of the samples is
a composition of the abbreviation ZBLAN given by the compounds of the FZ glasses
([21]) and an inventory number.
74
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.31: DSC data of Er-doped FZ glasses. From bottom to top: The samples
contain 2, 5, 9.1, and 13 mol% ErF3.
temperature (Tg) of (267.5 ±0.3) ◦C which is slightly higher than that observed
for pure FZ glass [21, 60]. No crystallization peaks can be found upon doping
the fluorozirconate glass with Er. However, crystallization is only initiated if
the glass is additionally doped with chlorine ions. Increasing the ErF3in the
FZ base glass the glass transition temperature shifts up to ∼290 ◦C for the 13%
doped glass which is significantly higher than that for the lower Er-doped FZ
glass. An exothermic peak can be observed at temperatures of 380 to 400 ◦C.
This also appears in the from pure FZ glass spectrum where the main glass
crystallization starts at ∼350 ◦C [21]. In table 4.5 the glass transition temper-
ature is given for the ErF3-doped glasses. The Tgfrom literature was added
to the table for the undoped FZ glass. The measurements show higher glass
transition temperatures for higher ErF3doping levels. For doping concentra-
tions of 9 mol% ErF3and more the glass transition temperature saturates at
∼290 ◦C.
sample doping level (%) Tg(◦C)
ZBLAN [21, 60] 0 262.0
ZBLAN JJ33 2.0 267.5
ZBLAN JJ34 5.0 274.8
ZBLAN JJ54 9.1 289.6
ZBLAN JJ57 13.0 289.5
Table 4.5: The table shows the glass transition temperatures of the different ErF3-
doped FZ glasses. The data for the undoped FZ glass was taken from literature [21,
60].
75
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.32: Optical density of the 5 mol% Er3+-doped FZ glass sample. The labeled
transitions start from the 4I15/2ground state level and end on the levels indicated. The
sample thickness was 2.2 mm. The inset shows a closer look on the lowest absorption
band in the IR spectral range.
Excitation
Figure 4.32 shows the optical density of the 5 mol% Er3+-doped FZ sample.
The material has strong Er3+absorption at 254, 377, and 520 nm and very
weak absorption at 229, 274, and 800 nm. At other wavelengths the absorption
strengths are intermediate. The absorption bands are in good agreement with
the literature [64] and the energy level diagram of trivalent erbium (see figure
4.33). The energy levels of the trivalent Er affecting the absorption ([30]) are
indicated in the graph. In addition to the observed Er3+absorption bands, the
material shows some background absorption in the UV spectral range below
300 nm. The inset in the upper right shows the lowest absorption band in more
detail. The up-conversion experiments were carried out by an excitation into
this. The CW laser diode is operating at a wavelength of 1540 nm and thus is
in good resonance with the transition from the 4I15/2ground state to the 4I13/2
excited state. The relatively high optical density is due to the sample thickness
of about 2.2 mm.
76
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.33: Energy level diagram of trivalent Er in FZ glasses: Up-conversion emis-
sions (solid arrows) and possible up-conversion routes (dashed arrows) for excitation
at 1540 (left) and at about 800 nm (right) are indicated.
77
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.34: Up-converted fluorescence spectrum of the 5 mol% Er3+-doped FZ glass
sample. Excitation is carried out with a CW laser diode operating at 1540 nm. The
range from 900 down to 400 nm is blown up by a factor of 4 (dashed curve).
Figure 4.35: The up-converted emission (red) and photoluminescence (black) of Er3+-
doped FCZ glass ceramics are shown. The upper curve was excited at 1540 nm, the
lower one at 796 nm. All spectra were normalized to their most intense emission.
Note: This is no comparison between the up-conversion efficiencies in dependency of
the excitation.
78
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.36: Up-converted fluorescence spectra of Er3+-doped FZ glass samples. Ex-
citation is carried out with a CW laser diode operating at 1540 nm (17.2 mW). The
range from 900 down to 400 nm is blown up by a factor of 4, the dashed line marks
this border at 900 nm.
Converted Fluorescence
Upon 1540 nm excitation with the light of a CW laser diode the Er3+-doped
FZ glasses show intense up-conversion. The corresponding spectrum for the
5 mol% doped sample in the spectral range between 400 to 1100 nm can be seen
in figure 4.34. Besides the 980 nm emission in the near infrared spectral range,
the most intense bands in the visible are located in the green at about 530 and
550 nm. Another emission is located in the red spectral range at 660 nm. The
NIR emission at 980 nm can be attributed to a transition from the 4I11/2excited
state to the 4I15/2ground state. The weaker emission band between 800 and
830 nm is a combination of two emissions at 820 and 810 nm corresponding to
the 2H9/2→4I9/2and 4I9/2→4I15/2transition, respectively. The 4I9/2and
4I11/2levels are accessible with two and four 1540 nm photons, respectively.
The emission at 850 nm is caused by a transition from 4S3/2to the 4I13/2energy
level [65]. As shown in [64] the electric dipole transition rate of the 850 nm
band is almost three times higher than the one of the 810 nm emission result-
ing in a higher fluorescence intensity of the 850 nm band. The visible emissions
can be attributed to transitions from the 2H11/2(530 nm), 4S3/2(550 nm), and
4F9/2(660 nm) excited states to the 4I15/2ground state. These levels are en-
ergetically accessible with three 1540 nm photons. Four-photon up-converted
fluorescence can be observed at 415 nm, which arises from a transition of the
2H9/2state to the ground state. In figure 4.33 possible excitation and emission
79
Chapter 4. Up-Conversion based on Nd and Er
routes are shown. All transitions to the 4I15/2ground state are split by about
0.03 eV which is caused by the crystal field splitting of the ground state.
Figure 4.35 shows the emission spectra of Er3+doped FCZ glass ceramics.
The upper spectrum shows up-conversion when excited with 1540 nm pho-
tons. The lower curves were excited at around 796 nm with a different laser
diode. All spectra were normalized to their most intense emission. No shift in
the position of the emissions can be observed but the intensity ratio changes
depending on the excitation wavelength. The weak emission at 850 nm in the
up-conversion spectra of the sample excited at 1540 nm cannot be observed in
the emission spectra excited at 800 nm.
The up-converted emission spectra of the whole series of different doping
concentrations is shown in figure 4.36. All measured curves are normalized to
the highest up-conversion intensity at 980 nm. The range from 400 to 900 nm
is blown up by a factor of 4, the border at 900 nm is marked with a dashed line.
In Nd-doped FCZ samples the main up-conversion emissions were observed
in the visible and in the ultraviolet range. Upon increasing the Nd doping level
the intensity ratio between the VIS and the UV lines changes in favor of the UV
emissions. Fig. 4.36 shows that this trend cannot be observed for the Er-doped
FCZ samples: Here, an increased Er doping level did not lead to changes in
the intensity ratio between the NIR emission at 980nm and the visible emis-
sions (550 and 660 nm). In contrast, the emission band in the UV (415 nm)
and the less intense emission shoulder (around 820 nm) increase with higher
doping levels. A closer look at the energy level diagram (figure 4.33) gives
the explanation for this similar behavior. Both emission correspond to tran-
sitions starting from the energetically high lying 2H9/2energy level (3.05 eV).
The 820 nm up-conversion is also a four-photon up-converted emission.
Lifetime Measurements
Lifetime measurements for the two- and three-photon up-converted fluores-
cence have also been performed. Er3+-doped FZ glasses were excited with a
CW laser diode operating at 1540 nm and the up-converted fluorescence de-
cay was detected at 980 nm, 850 nm, 660 nm, and 550 nm. These emissions
correspond to transitions starting from the 4I11/2(980 nm), 4F9/2(660 nm), and
4S3/2(550 nm), depopulating to the ground state 4I15/2. The 850 nm emis-
sion belongs to the 4S3/2→4I13/2transition and therefore relaxes from an
excited state to another excited state. The up-converted fluorescence decay
curves are shown in figure 4.37. For the two-photon up-converted emission
at 980 nm a fluorescence decay lifetime of (7.9 ±0.1) ms is determined for the
5 mol% doped sample. All emission starting from an energy level that can
only be reached by three 1548 nm photons, 4F9/2(660 nm), 4S3/2(550 nm), and
2H11/2(530 nm), seems to have nearly the same lifetime. Due to the lower
signal to noise ratio an exact analysis is not possible. The best fit for these
80
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.37: Fluorescence decay of Er3+doped ZBLAN glasses. The fluorescence
lifetime of the 980 nm emission (black dots), 850 (blue), 660 (red) and 550 nm (green)
have been measured. The excitation was carried out with a infrared laser diode at
1540 nm. The signal was modulated with a chopper.
Figure 4.38: Fluorescence lifetimes of the Er3+-doped FZ glasses. The data was ex-
tracted from lifetime measurements as shown in figure 4.37.
81
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.39: Digital photo of the experimental setup. Excitation is carried out from
the top with a CW laser diode operating at 1540 nm (not shown). The light is incident
perpendicular to the solar cell and the glass sample.
measurements is a lifetime of 3.6 ms with a relatively high error of ±0.5 ms
for the measurement of the emission with the lowest intensities (850 nm). The
errors for the 660 and 550 nm emission lifetimes are ±0.4 ms and ±0.3 ms, re-
spectively. The fluorescence decay measurements of the other glasses are not
shown. The fitted lifetimes are plotted as a function of the doping level in fig-
ure 4.38. The lifetimes of the 660 and 550 nm emission are identical within the
experimental error. All emissions show the same trend, the higher the doping
concentration the shorter the lifetime of the emission. The ratio of the decay
times (τ980 nm/τ660 nm,550 nm) is 2.1 ±0.2 for all measurements.
Power Dependence
To measure the power dependence of the Er3+-doped FZ glass the samples
were placed on top of a commercial mono-crystalline 2 cm x 2 cm solar cell
without any additional optical coupling (see figure 4.39). The power depen-
dence shown in figure 4.40 was measured by recording the short circuit current
for different laser diode output power. A short circuit current was observed
from 17.2 mW (maximum laser diode output power) down to a few hun-
dred µW. Below this limit the solar cell shows only the dark current induced
short circuit current. A direct illumination of the solar cell with a 1540 nm laser
diode yields a negligible dark current induced signal. On a double logarithmic
scale a saturation of the up-conversion intensity, i.e. the short circuit current
can be observed in contrast to the power dependent measurements of the Nd-
doped glasses. The point of saturation depends critically on the ratio between
excitation power, intermediate energy level lifetime, and the relative contri-
bution of excited state absorption (ESA) and energy transfer up-conversion
(ETU) processes to the overall up-converted fluorescence. For the 5 mol% Er-
82
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.40: Power dependence of the short circuit current (on a double-logarithmic
scale) of the up-converted fluorescence intensities in the Er3+-doped FZ glasses,
recorded under CW laser diode excitation at 1540 nm. The slopes are fitted for the
lower and higher excitation power range.
Figure 4.41: Power dependence of the external quantum efficiency (on a double-
logarithmic scale) of the Er3+-doped FZ glasses. The samples were excited with a CW
laser diode lasing at 1540 nm. The slopes are fitted to two excitation power ranges.
83
Chapter 4. Up-Conversion based on Nd and Er
ErF3lower excitation power higher excitation power
(mol %) m(m−1)cal (m−1)meas m(m−1)cal (m−1)meas
2.0 1.82 0.82 0.75 1.50 0.50 0.51
5.0 1.63 0.63 0.61 1.34 0.34 0.32
9.1 1.22 0.22 0.23 0.94 -0.06 -0.06
13.0 1.13 0.13 0.13 1.06 0.06 0.06
Table 4.6: The fitted slopes mof the short circuit current vs. excitation power curves
are shown for two excitation power ranges. The calculated (m−1)cal and the fitted
data from EQE vs excitation power measurements (m−1)meas are also listed.
doped sample, the slope of the short circuit current versus excitation power
curve is m=1.6 and 1.3 in the range from 0.5 to 3 mW and from 3 to 17.2 mW,
respectively. This experimentally observed decrease is determined by the com-
petition between linear decay and up-conversion processes for the depletion
of the intermediate excited states as shown in section 2.1.2. The intensity of
up-converted fluorescence that is excited by the sequential absorption of m
photons has a dependence on absorbed excitation power Pin, which may range
from Pm
in in the limit of infinitely small up-conversion rates down to P1
in for the
upper state of infinitely large up-conversion rates [15] (see table 2.2).
From the short circuit current Isc we can also calculate the EQE of this simple
and unoptimized system shown in figure 4.39.The EQE is given by equation
2.7:
EQE ∝Isc
Pin
∝Pm
in
Pin
=Pm−1
in .
From this equation the slope of the EQE versus excitation power curve in
the double logarithmic plot can be calculated with m−1. For the 5 mol% ErF3
doped sample the slope can be determined as m−1=1.6 −1=0.6 in the
range from 0.5 to 3 mW and m−1=1.3 −1=0.3 for the second range from 3
to 17.2 mW, respectively. This is in good agreement with the experimental data
(see figures 4.40 and 4.41) in which the EQE vs. excitation power curves can
be fitted with (m−1)meas of 0.61 and 0.32, respectively. For the setup shown
in figure 4.39 a maximum short circuit current of more than 300 µA for the
5 mol% doped sample was measured. With an excitation power of 17.2 mW for
the laser diode operating at λ= 1540 nm we can obtain a maximum quantum
efficiency of almost 1.5 %. All short circuit current and EQE versus excitation
power curves of the Er-doped glasses with different doping concentrations
were fitted and the results can be seen in table 4.6.
The Er3+absorption band used for this experiment is in the spectral range
from 1480 to 1580 nm. In this range the sun has an energy current density of
2.35 mW/cm2; taken from the AM1.5 spectrum [66]. The exciting laser beam
is focused in the experiment with a lens down to a spot of only a few tens of
84
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.42: Power dependence of the external quantum efficiency (on a double-
logarithmic scale) of the Er3+-doped FZ glasses. The lines are second order poly-
nomial fits.
micrometers in diameter; measurements of the spot size with a CCD camera
determine 70 µm as the smallest possible spot diameter. Therefore the excita-
tion power of 17.2 mW corresponds to an energy current density of 450 W/cm2
in this case. Therefore the excitation power of 17.2 mW leads to an energy cur-
rent density of 190,000 suns or less; depending on the real spot size used in the
experiment. For slightly larger spot sizes e.g. a diameter of 100 µm, the en-
ergy current density decreases to less than the half, leading to a decrease of the
number of the suns. Figure 4.42 shows again the EQE versus excitation power.
This time the measured curves are fitted with polynomial to obtain better pre-
dictability of their behavior in the range of low excitation power. Additionally,
the calculated number of required suns are displayed as a second axis. The
real size of the excitation spot is not known for these measurements, therefore
the number of suns was calculated with the smallest possible spot size.
The spectral behavior of the Er3+emission, figure 4.43, shows that rela-
tive intensity of the up-converted fluorescence bands depends significantly on
the excitation power. Upon increasing the excitation power from 0.7 mW to
17.2 mW the relative intensity ratio between the two-photon up-conversion
band at 980 nm and the three-photon up-conversion bands at 530, 550, and
660 nm changes in favor of the three-photon up-conversion bands. All spectra
were normalized to the 980 nm emission intensity.
85
Chapter 4. Up-Conversion based on Nd and Er
Figure 4.43: Spectral behavior of the Er3+up-converted emission intensity with re-
spect to the excitation power. All spectra of the 5 mol% doped sample were normal-
ized to the 980 nm emission intensity.
4.3.3 Discussion
Optical measurements (figure 4.32 and 4.34) show that trivalent Er can be in-
corporated into FZ-based glasses. For higher doping concentrations a shift of
the glass transition temperature to higher temperatures is observed (figure 4.5)
which indicates an influence of the Er doping in the glass on the glass transi-
tion temperature. Upon excitation with infrared laser diodes, for both 1540 nm
and 796 nm excitation, the doped glasses show an intense up-converted emis-
sion (figure 4.36). Due to two-, three-, and four-photon up-conversion pro-
cesses the Er3+ions are excited to different energy levels leading to emissions
over a broad spectral range from the NIR to the UV. Upon increasing the Er
doping concentration the up-converted emission, when normalizing to the
two-photon up-conversion emission (980 nm), does not change shape or spec-
tral position. The intensities of the emissions in the visible region vary slightly
but do not follow a clear trend, whereas the small UV emission becomes more
intense.
An important parameter for the up-conversion efficiency is the intermediate
energy level lifetime. For the two energy levels, from which the 980 nm two-
photon up-conversion and the 660 and 550 nm three-photon up-conversion
initiate, a decrease of the lifetime upon higher erbium doping concentrations
is observed. It is assumed that this is influenced by quenching effects.
86
Chapter 4. Up-Conversion based on Nd and Er
Up-conversion spectra (figure 4.43) show that the relative intensity ratio
between the 2-photon up-conversion band and the 3-photon up-conversion
bands change in in favor of the 3-photon up-conversion upon increasing the
excitation power (from 0.7 to 17.2 mW). This is a clear indicator for saturation
of the energy levels involved in the two photon up-conversion. This means
that for lower excitation powers higher up-conversion rates are achievable in
comparison to the normal fluorescence.
First results of a simple experimental setup (see figure 4.39) with a 5 mol%
Er-doped up-converter on a silicon solar cell show that an EQE of 1.5 % can be
achieved. Estimations show that for a worst case scenario the excitation power
of the laser diode was almost 290,000 suns. For lower excitation power the EQE
of the sample decreases rapidly. In contrast to this the higher doped (9.1 and
13 mol%) samples show a maximum EQE of only 1.18 and 0.75 %, respectively,
but their great advantage is a more “stable” behavior upon a decrease of the
excitation power (figure 4.42). Therefore, they have a calculated EQE of 0.6 and
0.45 %, respectively, for an excitation of 90 µW which equates to incident light
of 1,000 suns. This value can be achieved by concentrator cells. State-of-the-art
concentrator systems have already reached factors of 1,000. Measurements of
these up-converting samples with a concentrator system in combination with
a solar light simulator could prove this predictions.
87
Chapter 4. Up-Conversion based on Nd and Er
88
5 Conclusion
Sm-doped BaCl2and BaBr2single crystals have been grown for the applica-
tion of down-conversion. Typical emission and excitation bands of Sm2+are
observable. The corresponding fluorescence lifetimes are relatively long about
1.65 ms in BaCl2and shorter by a factor of three for the BaBr2single crystals.
Both the fluorescence efficiency and lifetime of Sm2+in FBZ glass ceramics
can be significantly increased upon appropriate annealing. The investigation
of samples which contain Sm2+and Sm3+ions show that annealing affects the
fluorescence properties of Sm2+while leaving those of Sm3+unchanged. The
fact that the Sm2+lifetime is increased, but that the Sm3+lifetime does not
change within experimental error indicates that a part of the Sm2+present in
the glass matrix enters the BaBr2nanocrystals which are formed during ther-
mal processing. Since the fluorescence lifetimes are increased upon annealing
the phonon energies of the BaBr2nanocrystals are likely to be smaller than
those of the FZ base glass, i.e. the non-radiative losses are reduced, the overall
measured lifetime is longer and thus the fluorescence is more efficient. How-
ever, most of the glasses produced contain Sm3+instead of Sm2+or a combina-
tion of both. It was not possible to control the Sm charge state in the FZ glasses.
For up-conversion applications Nd doped materials were investigated. It
has been shown that Nd can be incorporated into BaCl2single crystals. Therein
it shows the typical Nd3+emissions and excitations as described in the litera-
ture or calculated from the energy level diagram. An additional substructure
is observable in the infrared emission which is caused by crystal field splitting.
The doped single crystals show up-converted fluorescence in the UV and VIS
spectral range under 800 nm laser diode excitation.
The hygroscopicity of BaCl2is a problem; crystals left for a day or longer in
normal air tarnish and become white, lose their physical stability, and crum-
ble on contact. The easiest way to solve this problem is to embed BaCl2in
a protective glass matrix. Therefore, fluorozirconate based glasses contain-
ing neodymium and chlorine ions were produced. On subsequent annealing,
nano-crystallites were formed in the glasses. The particles can be identified as
hexagonal phase BaCl2, which grow in size with increasing annealing temper-
ature. Optical measurements show typical Nd3+transitions and emissions as
well as an intense up-conversion in the visible spectral range upon excitation
at ∼800 nm. The spectral positions of these emissions are nearly identical in
the glass ceramics but slightly shifted to shorter wavelengths with respect to
89
Chapter 5. Conclusion
the single crystals. Due to the slightly different excitations of Nd-doped single
crystals (806 nm) and FZ-glasses (798 nm), it is possible with two different exci-
tation sources (811 and 796 nm) to observe both specific Nd IR emission spectra
- with and without crystal field splitting - for the annealed glass ceramics. This
is a direct evidence of the incorporation of the Nd3+ions into the nanocrystals
during the annealing process. From the split up-conversion emissions (under
796 nm excitation) it is assumed that the up-conversion is mainly effected by
the Nd3+ions in the nanocrystals. Additionally, the up-conversion emission
intensities in the UV and blue spectral range increase upon annealing. Fluo-
rescence lifetime measurements of the 4F3/2energy level, which is important
for the efficiency for the two-photon up-conversion process, show that the life-
time has a significant dependent of the annealing temperature. A lifetime 22 %
longer than that in the as-made glass was achieved for an annealed glass ce-
ramic. The power dependency of the the up-converted emissions shows no
saturation effects.
Nd-doped glass ceramics are useful as a model system, but are not ap-
plicable for silicon solar cells due to the excitation of the up-conversion at
∼800 nm. However, erbium-doped FZ glasses are more appropriate systems
for up-conversion-based silicon solar cells due to their excitation at 1540 nm.
The investigated erbium doped fluorozirconate glasses show absorption and
up-conversion spectra which can be attributed to the erbium dopant. The
strong up-converted fluorescence is emitted in the NIR and in the visible spec-
tral range upon excitation in the IR at around 1540 nm. All up-converted emis-
sions lie energetically above the band gap of Si and can be absorbed by a silicon
solar cell, whereas the excitation is below. In a simple experimental setup, an
EQE of almost 1.5 % can be achieved for one of the doped FZ glasses. From
power-dependent measurements, an EQE of 0.4 % or more is assumed for an
excitation from 1,000 times concentrated sunlight.
The next step will be to increase the erbium doping level and to find the opti-
mal sample thickness. Analogous to the results obtained for Nd3+-doped FZ
glasses, additional chlorine doping should initiate the crystallization of BaCl2
nanocrystals therein upon thermal processing. This will hopefully lead to a
significant increase in the Er3+up-conversion efficiency. Therefore the Er3+
doped FZ glasses or FZ glass ceramics have huge potential as an up-converting
layer for photovoltaic applications.
90
A Additional Measurements and
Graphs
In these section additional graphs are shown. They contain in contrast to the
in the chapters 3 and 4 shown figures more measured data. They are shown to
complete the comparison of the different glasses and glass ceramics.
Figure A.1: Emission spectra of all ZBLAN glasses doped with Sm. The measured
spectra of the as-made glasses are shown. The upper graph shows the glasses con-
taining Sm2+or Sm2+,3+. The lower graph shows the samples with Sm3+.
91
Appendix A. Additional Measurements and Graphs
Figure A.2: Emission spectra of FZ 109 glass samples containing Sm3+. The measured
spectrum of the as-made glass is shown. For the samples annealed for 20 min at
250 ◦C, 260 ◦C, 270 ◦C, and 280 ◦C the differences in the signal compared to the as-
made glass are plotted.
The first two graphs contain measurements of Sm-doped glasses of the down-
conversion chapter. Figure A.1 shows the different produced glasses doped
with Sm. Sm2+and/or Sm3+emission spectra were measured. In figure A.2
the differences of the annealed samples of FZ109 are shown in dependency of
the annealing temperature.
The both figures A.3 and A.4 shows the measured infrared emission and
excitation spectra of the lower (0.5 and 1.0 mol%) Nd-doped glass and glass
ceramics.
92
Appendix A. Additional Measurements and Graphs
Figure A.3: This figure shows the measured IR emission spectra of annealed glass
ceramics doped with 0.5 mol% (upper graph) and 1 mol% (lower graph) Nd3+. From
bottom to top the spectra of as-made glass and these annealed for 20 minutes at 250 ◦C,
270 ◦C, and 290 ◦C are shown.
93
Appendix A. Additional Measurements and Graphs
Figure A.4: This figure shows the measured excitation emission spectra of annealed
glass ceramics doped with 0.5 mol% (upper graph) and 1 mol% (lower) trivalent Nd.
From bottom to top the spectra of as-made glass and these annealed for 20 minutes at
250 ◦C, 270 ◦C, and 290 ◦C are shown.
94
B Conversion Calculations
Equations
In the calculation model the current of the solar cell (Isc) is given by a convo-
lution:
Isc =[S⊗Dtr ⊕Dabs ⊗Deff ⊗Dem]⊗EQE (B.1)
Isc =[S⊗Utr ⊕Uabs ⊗Ueff ⊗Uem]⊗EQE (B.2)
where Sis the solar spectrum, Deff is the internal quantum efficiency of the
down-converter. Dtr is the transmission through the down-converter calcu-
lated from the given absorption Dabs, and Dem is the emission of the down-
converter. The same coefficients for the up-converter are labeled with Uand
the corresponding indexes. Note that every part of equations B.1 and B.2 is
dependent on the wavelength λ.
To normalize the results of these 2-dimensional parameter spaces, the gener-
ated electrons of the normal solar cell system without any additional down- or
up-converting layer had to be calculated first. This can be done by setting the
absorption of the converting layer to zero.
Assumptions
Trupke et al. [5, 6] have performed calculations on the theoretically reachable
EQE limit of an undefined single junction solar cell. To get a better overview
of the potential of down- and up-converters, more realistic calculations were
performed. Instead of an undefined material with a band gap energy varying
parameter ([5, 6]), a measured EQE of a silicon detector/solar cell (see figure
B.1) was used. In contrast to a 6,000 K blackbody spectrum, an AM1.5 spec-
trum [66] was taken. Both spectra are shown in figure B.2. Comparing the
two regions of interest for conversion processes, it can be seen that the 6000 K
blackbody spectrum has in the range from 1000 to 2000 nm 748 ·1018 more
photons than the AM1.5 spectrum, which corresponds to an artificial enhance-
ment of convertible photons of 51.5 %. However, the greatest mismatch can
be observed in UV range between 100 and 350 nm. Here the AM1.5 spec-
trum shows no significant photon flux due to the atmospheric absorbtion. In
this region, the blackbody spectrum offers more than 800 % of the real photon
flux. Figure B.3 schematically shows the calculation process for down- and
up-converter.
95
Appendix B. Conversion Calculations
Figure B.3: Down- and up-conversion process simulated with AM1.5 spectrum and
real EQE. The absorption range and hight of the converting layers are shown as red
(down) and blue (up) colored areas. The converted emission is painted in green and
emits at ∼1000 nm. The absorption band edge or the absorption high for the down-
and up-converter calculations, respectively, was varied, as indicated by the arrows.
97
Appendix B. Conversion Calculations
The optical parameters of the converting layers were chosen as follows:
Down-Conversion: The absorption of the down-converter is defined as:
absorption(λ) = (1 280 nm ≤λ≤b
0 otherwise )
with the wavelength parameter b=280 . . . 900 nm. The down-converter emits
the converted photons at 1000 nm (gauss profile) with an internal quantum ef-
ficiency (IQE) of 0 . . . 100 %, where an IQE of 100 % means that all incoming
photons were converted and emitted.
Up-Conversion: The absorption of the up-converter is defined as:
absorption(λ) = (a1220 nm ≤λ≤2000 nm
0 otherwise )
with the absorption parameter a=0 . . . 100 %. The up-converter emits the
converted photons with an IQE of 0 . . . 100 %, where an IQE of 100 % means
that from 2 incoming photons all were converted and 1 up-converted photons
were emitted (two-photon up-conversion). The up-converted photons were
emitted at 1000 nm using a Gaussian. Figure B.3 schematically shows the pro-
cess for down- and up-conversion.
In addition to the calculations shown in chapter 3 and chapter 4 a third system
containing a down-converter and a quantum cutting layer was calculated. The
result is shown in figure B.4.
Quantum-Cutting: The absorption of the quantum-cutting (QC) and down-
converting (DC) layer is defined as:
absorption(λ) =
a280 nm ≤λ≤500 nm for DC
a500 nm <λ≤980 nm for QC
0 otherwise
with the absorption parameter a=0 . . . 100 %. The quantum-cutter and down-
converter emit the converted photons with an IQE of 0 . . . 100 %, whereas an
IQE of 100 % means that all incoming photons were converted; therefore the
quantum-cutter emits two converted photons for one absorbed. The converted
photons were emitted at 1000 nm using a Gaussian.
98
Appendix B. Conversion Calculations
Figure B.4: Calculated ratio between solar cell with and without a down-converting
and quantum-cutting layer dependent on the absorption of the layers and the inter-
nal quantum efficiencies. A maximum for 100 % absorption and internal quantum
efficiency of 1.353 was calculated; 0.167 for the minimum. The color stands for the
calculated ratio (see color scale in the lower left corner).
99
Appendix B. Conversion Calculations
100
C DSC Measurements and Analysis
Measurements were performed with a NETZSCH DSC 204 F1 and analyzed
in combination with the software package NETZSCH PROTEUS V4.8.5.
The glass transition temperature Tgis correlated with the melting temper-
ature Tmby the “two-thirds rule” Tg/Tm=2/3 [67]. The onset, the math-
ematical beginning of the first crystallization peak, is called Txand belongs
to BaCl2for the investigated glass. The temperature of the maximum of the
crystallization peak is Tp.
∆T=Tx−Tg(C.1)
Hr=Tx−Tg/(Tm−Tx)(C.2)
S=Tx−TgTp−Tx/Tg(C.3)
Equation C.2 describes the devitrification tendency of the glass. Hrdefined
by Hruby [68] informs about the nucleation rate of the crystals embedded in
Tx−Tgand the magnitude of growth-rate between Tmand Tx[60]. The
parameter Sintroduced by Saad and Poulain also includes the width of the
devitrification peak [69]. If the Tp−Txfactor is high, the growth rate de-
creases. Moreover DSC measurements with different heating rates (α) make
it possible to investigate the apparent activation energy Eaof crystallization
and the crystallization kinetics. The activation energy can be determined by
equation C.4 [70] or equation C.5 [71].
dln T2
p/α
d(1/Tp)=Ea
R(C.4)
d(ln α)
d(1/Tp)=−Ea
R(C.5)
According to the Avrami model ([72–74]), under isothermal conditions and
assuming a constant crystallization temperature, the degree phase transition
or crystallization xis given by C.6, where Ktis a temperature dependent rate
constant depending on the nucleation and growth rate, and nAv is the Avrami
exponent reflecting the process dimensionalities. Using the Avrami equation
in the double logarithmic form (equation C.7), the parameters nAv and Kt
can be determined as the slope and intersection respectively of straight line
fits. Jeziorny [75] modified the Avrami equation to account for non-isothermal
101
Appendix C. DSC Measurements and Analysis
Figure C.1: Plot of ln T2
p/α(full circles) and lnα(open circles) vs. (1/Tp) for the
5 mol% Nd-doped glass. The calculated energies Eαare indicated.
crystallization. Therefore he takes into consideration the influence of heating
or cooling rate on the parameter K. The corrected parameter Kcis given by
equation C.8.
1−x=exp(−Kt·tnAv )(C.6)
log [−ln (1−x)] =nin ·log(t) + log (Kt)(C.7)
log (Kc)=log (Kt)
α(C.8)
Results
Table C.1 shows important peak values and thermal stability parameters for all
heating rates. The activation energy Eais obtained from the linear fits of the
slopes by plotting ln T2
p/αor ln αversus (1/Tp)(figure C.1). The results are
indicated in the graph and additionally listed in table C.2. The dependence
of the relative crystallization xon temperature for different heating rates is
shown in figure C.2. To determine the Avrami exponent (nAv) and corrected
constant (Kc) the measured data is plotted in a double logarithmic graph and
determined by linear fits and equations C.7 and C.8. The mean values nAv
listed in table C.2 is nAv ≈3 which corresponds for randomly dispersed nuclei
in the system with an instantaneous nucleation process in which every nucleus
develops into a spherulite [76].
102
Appendix C. DSC Measurements and Analysis
Figure C.2: The relative crystallization xof the glass sample at different heating rates:
from 3 K/min to 30 K/min.
Figure C.3: Avrami plots of the FCZ glass sample for different heating rates.
103
Appendix C. DSC Measurements and Analysis
heating rate TgTxTpTmTg/Tm∆T HrS t0.5 nav log(Kt)Kc
(K / min) (◦C) (K) (◦C) (K) (◦C) (K) (◦C) (K) (K) (K) (s)
3.0 220.6 493.8 261.2 534.4 268.6 541.8 412.5 685.7 0.720 40.6 0.268 0.686 214.9 1.97 -1.219 0.392
5.0 218.7 491.9 266.1 539.3 274.0 547.2 412.2 685.4 0.718 47.4 0.324 0.762 133.8 3.03 -1.725 0.452
7.5 219.6 492.8 270.2 543.4 278.6 551.8 412.7 685.9 0.718 50.6 0.355 0.863 92.1 3.75 -1.842 0.568
10.0 220.4 493.6 272.9 546.1 281.6 554.8 412.5 685.7 0.720 52.5 0.376 0.926 71.3 3.31 -1.078 0.780
15.0 221.9 495.1 277.5 550.7 286.8 569.0 412.1 685.3 0.722 55.6 0.413 1.045 49.3 3.64 -0.682 0.900
20.0 222.6 495.8 280.6 553.8 290.4 563.6 411.7 684.9 0.724 58.0 0.442 1.147 38.2 3.50 -0.322 0.964
25.0 224.7 497.9 283.4 556.6 293.7 566.9 411.7 684.9 0.727 58.7 0.457 1.215 31.6 3.31 0.077 1.007
30.0 226.4 499.6 285.8 559.0 296.4 569.6 411.7 684.9 0.729 59.4 0.471 1.261 27.2 3.45 0.245 1.019
Table C.1: Thermal stability parameters of the 5 mol% Nd-doped FCZ glasses. Whereas Tg, Tx, Tp, and Tmare given in ◦C and K.
Eα(Chen) Eα(Ozawa) nav
212 ±4 kJ/mol 195 ±5 kJ/mol 3.25
Table C.2: Activation energy Eaand averaged Avrami exponent of the 5 mol% Nd-doped FCZ glass.
104
D Bibliography
[1] International Energy Agency (IEA). Renewables in Global Energy Supply -
an IEA Fact Sheer (2007).
[2] Federal Ministry of Education and Research (BMBF). The High-Tech Strat-
egy on Climate Protection (2008).
[3] E. Becquerel. R. C. Acad. Sci. 9, 561 (1839).
[4] M. A. Green, K. Emery, Y. Hishikawa, and W. Warta. Prog. Photovolt: Res.
Appl. 16, 435–440 (2008).
[5] T. Trupke, M. A. Green, and P. W¨
urfel. J. Appl. Phys. 92(3), 1668 (2002).
[6] T. Trupke, M. A. Green, and P. W¨
urfel. J. Appl. Phys. 92(7), 4117 (2002).
[7] P. Gibart, F. Auzel, J.-C. Guillaume, and K. Zahraman. Jpn. J. Appl. Phys.
35(8), 4401–4402 (1996).
[8] A. Shalav, B. S. Richards, T. Trupke, K. W. Kr¨
amer, and H. U. G¨
udel. Appl.
Phys. Lett. 86, 013505 (2005).
[9] R. T. Wegh, H. Donker, K. D. Oskam, and A. Meijerink. Science 283, 663
(1999).
[10] R. T. Wegh, H. Donker, K. D. Oskam, and A. Meijerink. J. Lumin. 82(2), 93
(1999).
[11] J. L. Sommerdijk, A. Bril, and A. W. de Jager. J. Lumin. 8(4), 341–343 (1974).
[12] W. W. Piper, J. A. DeLuca, and F. S. Ham. J. Lumin. 8(4), 334–348 (1974).
[13] S. Guy, M. F. Joubert, B. Jacquier, and M. Bouazaoui. Phys. Rev. B 47(17),
11001 (1993).
[14] F. Auzel. The Luminescence of Inorganic Solids. Plenum Press, New York,
(1979).
[15] M. Pollnau, D. R. Gamelin, S. R. L¨
uthi, H. U. G¨
udel, and M. P. Hehlen.
Phys. Rev. B 61(5), 3337–3346 (2000).
[16] H. G. Wagemann and H. Eschrich. Photovoltaik. Vieweg+Teubner Verlag,
(2007).
105
D Bibliography
[17] M. Shojiya, M. Takahashi, R. Kanno, Y. Kawamoto, and K. Kadono. J.
Appl. Phys. 82(12), 6259 (1997).
[18] M. J. Weber. Phys. Rev. B 8(8), 54 (1973).
[19] C. B. Layne, W. H. Lowdermilk, and M. J. Weber. Phys. Rev. B 16(1), 10
(1977).
[20] M. D. Shinn, W. A. Sibley, M. G. Drexhage, and R. N. Brown. Phys. Rev. B
27(11), 6635 (1983).
[21] I. D. Aggarwal and G. Lu. Fluoride Glass Fibre Optics. Academic Press,
INC., (1991).
[22] Mi. Poulain, Ma. Poulain, J. Lucas, and P. Brun. Mater. Res. Bull. 10(4), 243
(1975).
[23] J. M. Parker. Annu. Rev. Mater. Sci. 19, 21 (1989).
[24] B. Bendow, M. G. Drexhage, and G. Lipson. J. Appl. Phys. 52(3), 1460
(1981).
[25] B. Bendow, R. N. Brown, M. G. Drexhage, T. J. Loretz, and R. L. Kirk. Appl.
Opt. 20(21), 3688 (1981).
[26] C. J. Simmons, H. Sutter, J. H. Simmons, and D. C. Tran. Mater. Res. Bull.
17(9), 1203 (1982).
[27] J. A. Johnson, S. Schweizer, B. Henke, G. Chen, J. Woodford, P. J. Newman,
and D. R. MacFarlane. J. Appl. Phys. 100, 034701 (2006).
[28] J. A. Johnson, S. Schweizer, and A. R. Lubinsky. J. Am. Ceram. Soc. 90(3),
693–698 (2007).
[29] P. N. Prasad. Nanophotonics. John Wiley & Son, Inc., New Jersey, (2004).
[30] G. H. Dieke and H. M. Crosswhite. Appl. Opt. 2(7), 675–686 (1963).
[31] H. V. Lauer Jr. and F. K. Fong. The Journal of Chemical Physics 65(8), 3108
October (1976).
[32] W. T. Carnall, H. Crosswhite, H. M. Crosswhite, and J. G. Conway. J.
Chem. Phys. 64(9), 3582 (1976).
[33] J. A. Koningstein. Phys. Rev. 136(3A), A717 (1964).
[34] C. G¨
orller-Walrand, E. Huygen, K. Binnemans, and L. Fluyt. J. Phys. B:
Condens. Matter 6(38), 7797 (1994).
[35] H. M. Crosswhite and R. L. Schwiesow. J. Chem. Phys. 50(11), 5032 (1969).
106
D Bibliography
[36] J. B. Gruber, B. Zandi, and L. Markle. J. Appl. Phys. 83, 1009 (1998).
[37] G. Blasse and B. C. Grabmaier. Luminescent Materials. Springer Verlag,
Berlin Heidelberg, (1994).
[38] B. M. Walsh, G. W. Grew, and N. P. Barnes. J. Phys.: Condens. Matter 17,
7643 (2005).
[39] P. A. Tanner and C. S. K. Mak. Phys. Rev. B 66, 165203 (2002).
[40] C. Tiseanu, A. Lupei, and V. Lupei. J. Phys.: Condens. Matter 7(44), 8477
(1995).
[41] J. H. van Vleck. J. Phys. Chem. 41(1), 67 (1937).
[42] P. Scherrer. Nachr. Ges. Wiss. G¨ottingen 2, 98–100 (1918).
[43] J. I. Langford and A. J. C. Wilson. J. Appl. Cryst. 11, 102–113 (1978).
[44] B. E. Warren. X-Ray Diffraction. Dover Publications, (1990).
[45] G. S. Ofelt. J. Chem. Phys. 38, 2171 May (1963).
[46] G. Grenet, M. Kibler, A. Gros, J. C. Souillat, and J. C. Gˆ
acon. Phys. Rev. B
22(11), 5052 (1980).
[47] E. B. Brackett, T. E. Brackett, and R. L. Sass. J. Phys. Chem. 67(10), 2132
(1963).
[48] E. Monberg and Y. Ebisuzaki. J. Cryst. Growth 21, 307–309 (1974).
[49] P. Mikhail, K. Ramseyer, G. Frei, F. Budde, and J. Hulliger. Optical Com-
munications 188, 111 (2001).
[50] K. Kojima, A. Kubo, M. Yamashita, N. Wada, T. Tsuneoka, and Y. Komat-
subara. J. Lumin. 87-89, 697 (2000).
[51] A. Saissy, N. Azami, J. Jones, and G. Maze. Appl. Opt. 36(24), 5931–5933
(1997).
[52] J. B. Gruber, B. Zandi, and M. F. Reid. Phys. Rev. B 60(23), 15643 (1999).
[53] A. Edgar, S. Schweizer, S. Assmann, J. M. Spaeth, P. J. Newman, and D. R.
MacFarlane. J. Non-Cryst. Solids 284, 237 (2001).
[54] G. Liu and H. A. Eick. J. of the Less Common Metals 129, 47 (1989).
[55] A. Edgar, M. Secu, G. V. M. Williams, S. Schweizer, and J.-M. Spaeth. J.
Phys.: Condens. Matter 13, 6259–6269 (2001).
[56] W. Shockley and H. J. Queisser. J. Appl. Phys. 32(3), 510–519 (1961).
107
D Bibliography
[57] J. Lucas, M. Chanthanasinh, M. Poulain, M. Poulian, P. Brun, and M. J.
Weber. J. Non-Cryst. Solids 27(2), 273–283 (1978).
[58] A. T. Stanley, E. A. Harris, T. M. Searle, and J. M. Parker. J. Non-Cryst.
Solids 161, 235–240 (1993).
[59] K. Binnemans, D. Verboven, C. G¨
orller-Walrand, J. Lucas, N. Duhamel-
Henry, and J. L. Adam. J. Non-Cryst. Solids 204(2), 178–187 (1996).
[60] M. D. Bar´
o, A. Otero, S. Suri˜
nach, A. Jha, S. Jordery, M. Poulain, A. Soufi-
ane, D. W. Hewak, E. R. Taylor, and D. N. Payne. Mater. Sci. Eng., A
179-180(1), 303 (1994).
[61] S. Schweizer, L. W. Hobbs, M. Secu, J.-M. Spaeth, A. Edgar, and G. V. M.
Williams. Appl. Phys. Lett. 83(3), 449 July (2003).
[62] S. Schweizer, L. W. Hobbs, M. Secu, J.-M. Spaeth, A. Edgar, G. V. M.
Williams, and J. Hamlin. J. Appl. Phys. 97, 083522 (2005).
[63] J. Selling, S. Schweizer, Spaeth J.-M., G. Corradi, Edgar A., and G. V. M.
Williams. phys. stat. sol (c) 2(1), 592 (2005).
[64] S. R. Bullock, B. R. Reddy, P. Venkateswarlu, S. K. Nash-Stevenson, and
J. C. Fajardo. Optical and Quantum Electronics 29, 83–92 (1997).
[65] T Catunda, L . A. O. Nunes, A. Florez, Y. Messaddeq, and M. A. Aegerter.
Phys. Rev. B 53(10), 6065–6070 (1996).
[66] P. W¨
urfel. Physics of Solar Cells - From Principles to New Concepts. WILEY-
VCH Verlag GmbH & Co. KGaA, Weinheim, (2005).
[67] W. Kauzmann. Chem. Rev. 43 (2), 219–256 (1948).
[68] A. Hruby. Czech. J. Phys. B 22, 1187 (1972).
[69] M. Saad and M. Poulain. Mater. Sci. Forum 19-20, 11 (1987).
[70] M.S. Chen. J. Non-Cryst. Solids 27, 257 (1978).
[71] T. Ozawa. Polymer 12, Issue 3, 150–158 (1971).
[72] M. Avrami. J. Chem. Phys. 7(12), 1103 (1939).
[73] M. Avrami. J. Chem. Phys. 8, 212 (1940).
[74] M. Avrami. J. Chem. Phys. 9(2), 177 (1941).
[75] A. Jeziorny. Polym. J. 19(10), 1142–1144 (1978).
[76] Z. Zhang, C. Xiao, and Z. Dong. Thermochim. Acta 466(1-2), 22–28 (2007).
108
E List of Publications
(1)B. Ahrens, B. Henke, P. T. Miclea, J. A. Johnson, S. Schweizer
Enhanced up-converted fluorescence in fluorozirconate based glass ceramics for
high efficiency solar cells
Proceedings of SPIE: Photonics for Solar Energy Systems II, 7002, 700206
(2008)
(2)B. Ahrens, C. Eisenschmidt, J. A. Johnson, P. T. Miclea, S. Schweizer
Structural and optical investigations of Nd-doped fluorozirconate-based glass
ceramics for enhanced upconverted fluorescence
Appl. Phys. Lett., 92, 061905 (2008)
(3)S. Schweizer, P. T. Miclea, B. Henke, B. Ahrens
Up- and down-conversion in fluorozirconate based glass ceramics for high efficiency
solar cells
Proc. of 23rd European Photovoltaic Solar Energy Conference (EU PVSEC),
Feria Valencia, Valencia, Spain (2008)
(4)B. Ahrens, J. Selling, C. Eisenschmidt, A. Engel, S. Schweizer
Sm-activated barium halide nanocrystals in fluorozirconate glasses
J. Phys.: Condens. Matter, 20, 295227 (2008)
(5)B. Ahrens, P. L¨
oper, J. C. Goldschmidt, S. Glunz, B. Henke, P. T. Miclea,
S. Schweizer
Neodymium-doped fluorochlorozirconate glasses as an upconversion model sys-
tem for high efficiency solar cells
Phys. Status Solidi A, 205, 2822-2830 (2008)
(6)B. Ahrens, P. T. Miclea, S. Schweizer
Upconverted fluorescence in Nd3+-doped barium chloride single crystals
J. Phys.: Condens. Matter, 21, 125501 (2009)
109
Appendix E. List of Publications
(7)B. Henke, B. Ahrens, J. A. Johnson, P. T. Miclea, S. Schweizer
Erbium-doped Fluorozirconate Glasses For High Efficiency Solar Cells
Prog. Photovoltaics Res. Appl., (submitted)
(8)B. Henke, B. Ahrens, P. T. Miclea, C. Eisenschmidt, J. A. Johnson, and S.
Schweizer
Erbium- and chlorine-doped fluorozirconate-based glasses for up-converted flu-
orescence
J. Non-Cryst. Solids (accepted)
In Preparation
(1)B. Henke, J. A. Johnson, B. Ahrens, P. T. Miclea, S. Schweizer
Saturation effects in the up-conversion efficiency of Er-doped fluorozirconate
glasses
J. Phys.: Condens. Matter
(2)B. Ahrens, B. Henke, J. A. Johnson, P. T. Miclea, S. Schweizer
Upconverted fluorescence in Er-doped fluorozirconate based glass ceramics for
high efficiency solar cells
Proc. of SPIE
(3)S. Schweizer, B. Henke, J. A. Johnson, B. Ahrens, P. T. Miclea
Progress on erbium-doped fluorozirconate glass ceramics for upconversion-based
solar cells
Proc. of 24th EU PVSEC
110
F Declaration
Hereby I declare that this submission is my own work and to the best of my
knowledge it contains no materials previously published or written by another
person, nor material which to a substantial extent has been accepted for the
award of any other degree or diploma. Any contribution made to the research
by others, with whom I have worked at University of Paderborn or elsewhere,
is explicitly acknowledged in the thesis.
I also declare that the intellectual content of this thesis is the product of my
own work, except to the extent that assistance from others in the project’s de-
sign and conception or in style, presentation and linguistic expression is ac-
knowledged.
Date, Location Bernd Ahrens
111
Appendix F. Declaration
112
G Acknowledgement
Das Deckblatt der Arbeit tr¨
agt zwar nur einen Namen dennoch haben viele
Menschen zum Gelingen dieser Arbeit beigetragen und mich auf verschieden-
ste Art und Weise unterst¨
utzt.
Ihnen allen sei mein Dank ausgesprochen.
Als erstes m¨
ochte ich mich bei Herrn PD DR. STEFAN SCHWEIZER f¨
ur die
Aufnahme in seine Arbeitsgruppe bedanken. Ebenso f¨
ur die M¨
oglichkeit der
Promotion, die immer freundliche, ermutigende, fordernde und f¨
ordernde Be-
treuung.
Herrn PROF. DR. J ¨
ORG LINDNER danke ich f¨
ur die ¨
Ubernahme des Koreferats
sowie sein Interesse an dieser Arbeit.
Dank m¨
ochte ich DR. BASTIAN HENKE und DR. JULIA SELLING aussprechen,
gute Kollegen und Freunde, die in jeder Lebenslage ein offenes Ohr haben,
sei es wissenschaftlich oder privat. Auch f¨
ur ihre gute Laune und den damit
verbundenen Spaß bei und nach der Arbeit. J ¨
ORG HALLMANN sei im gleichen
Atemzug erw¨
ahnt und f¨
ur Selbiges gedankt. steht hinten an, weil k¨
urzer da
I would like to thank DR. JACQUELINE A. JOHNSON for reading my thesis
and correcting the worst misunderstandings between my and the official En-
glish caused by grammatical bloomers.
DR. PAUL MICLEA danke ich f¨
ur zahlreichen und hilfreichen Diskussionen
bez¨
uglich meiner teilweise kreativen Auslegung der optischen Spektroskopie
und deren Auswertung.
Auch MARCEL DYRBA und CHRISTIAN PASSLICK sei gedankt f¨
ur unz¨
ahlige
Gespr¨
ache und Diskussionen.
Kollektiv geht mein Dank auch an alle (aktuellen und ehemaligen) Mitarbeiter
des A4 FLURES (Uni Paderborn) sowie der AG WEHRSPOHN (Uni Halle/Saale)
f¨
ur das immer gute Arbeitsklima, die kurzweiligen Mittagspausen, interessan-
ten Gespr¨
ache, neue Experimente, die Kuchenrunden und den Rest.
113
Appendix G. Acknowledgement
Den Mitarbeitern der Kristallzucht, Hr. NIGGEMEIER und Hr. WINTERBERG,
als auch Herrn PAULI sei mein Dank.
Bedanken m¨
ochte ich mich auch bei meinen ehemaligen Kommilitonen, mit
denen ich w¨
ahrend des Studiums Vorlesungen geh¨
ort und zusammen gelernt
habe: MARINA PANFILOVA, ELENA TSCHUMAK, STEPHAN BLANKENBURG
und STEFAN WIPPERMANN.
Ebenso bedanke ich mich bei allen, die nicht namentlich Erw¨
ahnung fanden,
aber zum Gelingen dieser Arbeit, physikalisch oder privat, beigetragen haben.
Insbesondere danke ich meinen FREUNDEN, GESCHWISTERN und ELTERN,
die mich immer nach besten Kr¨
aften unterst¨
utzt haben, mich aber auch nicht
zweifeln ließen, dass eine Welt außerhalb des Studiums und der Arbeit ex-
istiert.
Support
This work was supported by the FhG Internal Programs under Grant No.
Attract 692 034. Funding by the German Science Foundation (DFG project
PAK88 - NanoSun) and the Federal Ministry of Education and Research (BMBF)
within NanoVolt is gratefully acknowledged.
114
