Degradation mechanism of AlInGaP light
emitting diodes during PMMA encapsulation
and operation
Dem Department Physik
der Universität Paderborn
zur Erlangung des akademischen Grades eines
Doktors der Naturwissenschaften
genehmigte
Dissertation
von
Stephan Preuß
Paderborn, November 2007
Contents
1 Introduction 1
2 Fundamentals of semiconductor devices 5
2.1 Semiconductors 5
2.2 Band structure of p-n junctions 8
2.2.1 Doping of semiconductors 8
2.2.2 p-n junction 9
2.2.3 Double heterostructures and quantum wells 11
2.3 Radiative recombination 14
2.4 Modelling of the emission spectra 15
2.5 Light extraction from a LED 18
2.6 Non-radiative recombination 21
3 Micro electroluminescence set-up 23
3.1 Principle of the set-up 23
3.2 Optics 25
3.2.1 Step size at the chip surface 26
3.2.2 Space resolution 26
3.3 The positioning system 27
3.4 The Ulbricht sphere 27
4 Micro electroluminescence - results and discussion 29
4.1 Description of the samples 29
4.1.1 The HWFR-B 410 LED: basic properties 30
4.2 Sample preparation 34
4.3 Experimental details for micro electroluminescence
measurement 34
4.4 Temperature analysis 35
4.5 Experiment 36
4.5.1 Glue hardening 36
4.5.2 Spectral measurements along the z-axis 37
i
4.6 Discussion of the results 40
4.6.1 Glue hardening 40
4.6.2 Micro electroluminescence along the z-axis 41
4.6.3 Modelling A1 and A2 versus z 41
4.6.4 Current dependence 45
4.6.5 Measurements along the x-axis 47
5 Aging of LEDs 49
5.1 Degradation under high forward current-chip only 49
5.2 Positive aging 49
5.3 Long-term aging 55
6 Injection molding of LED clusters 61
6.1 General set-up and temperature measurements 61
6.1.1 The module (molded part) 63
6.1.2 Optics 65
6.1.3 Temperature measurement 66
6.1.4 Temperature simulation 71
6.2 Influence of injection molding on the LED performance 74
6.2.1 Results and discussion 74
6.3 Radiation pattern of the PMMA optic 80
7 Diffusion model for LED degradation 83
7.1 The dependence of electroluminescence on trap
concentration 83
7.2 Identifying magnesium as the p-type doping material 86
7.3 Magnesium diffusion into the active layer 89
7.4 Data analysis 92
7.4.1 The activation and diffusivity of magnesium
in AlInGaP 95
8 Summary 99
9 Appendix 101
Bibliography 107
ii
List of Figures
2.1 Bandstructure of direct semiconductor InP. 6
2.2 Bandstructure of indirect semiconductor GaP. 7
2.3 Band gap and wavelength of alloy AlInGaP versus lattice constant.
For large In-content the direct bandgap dominates. The red line indicates
the spectral range of AlInGaP LEDs lattice matched on GaAs with a direct
bandgap. 8
2.4 Band structure of the p-n junction under zero bias 10
2.5 Band diagram of a forward-biased double heterostructure. The p-type
confinement layer consists of a lightly doped layer close to the active
region and a higher doped layer further away from the active layer. 12
2.6 Square-well potential of a quantum well structure. The well is formed by
decreasing the mole fraction x 13
2.7 Density of state for a quantum well structure (2D) and for bulk material (3D). 13
2.8 The emission spectrum of a QW LED, red curve, is the product of the density
of state and the distribution of carriers. 16
2.9 Emission spectrum of an AlInGaP LED. The black line indicates the
measurement, the blue line the model function 17
2.10 A captured photon in an LED structure. The active layer is dark red, the
escape cones are hatched. Only photons inside the escape cones can escape
the device. A captured photon with an angle >βcrit is shown as light red rays. 19
2.11 Auger-recombination. The energy released by a electron hole recombination
is absorbed by another electron. 21
3.1 Electroluminescence set-up, the emission from the LED chip is coupled
into a monochromator. A diode array detects the diffracted light and the
data are sent to a PC. 23
3.2 Sketch of the micro-electroluminescence set-up. The light from the LED is
magnified by a factor of 500 and projected onto a screen. One point of the
image with a diameter of 1mm is in coupled to an optical fibre and guided
to the monochromator. Data is stored in a PC. 24
3.3 Optical set-up. The LED is outside the focal length of lens 1. Lens 1 generates
a magnified real intermediate image. Lens 2 magnifies this intermediate image. 25
3.4 Schematic set-up of an Ulbricht sphere. The coating has a high reflectivity and
the reflection is diffuse. Thus the light flux is uniform in the sphere. 28
4.1 Layer structure of the LED chips. “EP” denotes the edge planes (red hatched). 30
iii
4.2 SEM image of the LED chip (side view). The optically active MQW region
is inside the groove and marked with an arrow. 30
4.3 Current-voltage characteristic of the HWFR-B410 LED at room temperature. 31
4.4 Reverse current versus reverse voltage of the HWFR-B410 LED. 31
4.5 Luminous flux Φ versus forward current. The maximum luminous flux at
T=20°C is measured at a forward current of I=70mA. 32
4.6 Junction temperature versus forward current. 32
4.7 Calibration of the monochromator with two laser lines at λ1=632.8nm
and λ2=639.3nm. 35
4.8 Forward current-voltage behaviour after different thermal treatments. 36
4.9 Electroluminescence along the z-axis in the n-type region. The black curve
is the emission from the top surface. The active layer is at z=0, negative
z points toward the n-contact. 38
4.10 Electroluminescence along the z-axis in the p-type region. The black curve is
the emission from the top surface. The active layer is at Z=0, positive z points
toward the p-contact. 38
4.11 Emission at different drive currents at z=-1µm, just below the active layer. 39
4.12 Emission in the GaP substrate at z=-7µm at different drive currents. 39
4.13 Emission at z=-1µm, below the active layer and at different x-values. 40
4.14 Intensity A1 of maximum E1 and intensity A2 of maximum E2 as
function of distance z from the active layer. 42
4.15 Observation cone inside the LED chip. The cone (hatched) divides the active
layer (red). 43
4.16 Peak energy E1 versus forward current at z=-1µm. 46
4.17 Peak energy of the emission versus carrier temperature. 47
4.18 Increase of the carrier temperature as function of the LED forward current. 47
4.19 Peak energy E1 as function of the x-coordinate. A-red shift is observed when
moving the point of measurement from the centre to the edge of the LED. 48
5.1 Sample 1, luminous flux versus time and different measurement currents.
Burn-in behaviour and positive degradation with increasing luminous
flux for is observed. 50 mAImeas 5≥
5.2 Luminous flux versus forward current of sample 1. Idegra=70mA,after t=65h. 51
5.3 Sample 2, normalised luminous flux versus aging time at Idegra=120mA. The
increase of luminous flux is 54% after t=16h. 51
5.4 Equivalent circuit of a real diode with its serial and parallel resistance. 52
5.5 I-V characteristics of sample 2 before annealing and after 14h and 40h of
annealing. The diode was treated by a forward current of Ideagra=120mA.
A reduction of the parallel shunt resistance is clearly observable. 53
5.6 Intensity versus degradation time of sample 3 at Imeas=40mA. 53
iv
5.7 Intensity versus degradation time of PMMA encapsulated LEDs. Blue curve
for glued/bonded sample 5, red curve for soldered sample 4 at Imeas=40mA. 54
5.8 Current voltage characteristics of a LED at various degradation times.
The serial resistance increases with the degradation time. 55
5.9 Serial resistance of sample LZ-UL 2 versus aging time. The black
curve is a logarithmic fit of the experimental data. 56
5.10 Luminous flux versus aging time measured at different currents.
After t=120h the luminous flux starts to decrease. 57
5.11 Emission spectra before degradation for various forward currents.
The intensity increases by increasing the forward current thus shifting the
Peak to lower energies. 58
5.12 Emission after 54h of aging at Idegra=120mA. 58
5.13 Maximum intensity (black dots) and energy (red dots) versus
degradation time for different drive currents. 59
6.1 Drawing of the tool, set-up for “bubble in” the LED chips at low pressure. 62
6.2 Drawing of the tool, set-up for optic production at high pressure. 62
6.3 Top view into the cavity of the injection moulding tool. The slide bars are
wide open to load the cavity with the frame. The investigations were
performed on an Arburg A380 injection molding machine. 63
6.4 Copper web for die/wire bonding. Asymmetry of the electrode guarantees
that the LED chips are in the focal point of the parabolic optic. 64
6.5 Copper web used for die/die bonding. Two symmetric noses with a gap of
d=100µm. The chip is placed with its active layer over the gap. 64
6.6 Frame with die/wire bonded AS AlInGaP LED chip (X-ray image). 64
6.7 Die/die bonded TS LED chip. 65
6.8 Total reflection optic, principle. 65
6.9 LED module in two different viewing angles. 66
6.10 Optic module. 66
6.11 Band gap versus temperature of GaP. 67
6.12 Forward voltage versus temperature at I=1mA for AlInGaP LEDs. 68
6.13 Forward voltage versus temperature of three different LEDs degraded
at 0min, 3min and 5min at T=205°C. 68
6.14 LED and cable feed through arrangement in the cavity of the injection
molding tool. 69
6.15 Temperature versus time of the LED chip in the cavity of the injection
molding tool during the encapsulation process. The arrow marks the
temperature increase at the glass transition. 70
6.16 Time delay of the PT100 temperature sensor in comparison to the LED chip
sensor. 71
v
6.17 Temperature distribution in the LED module just after filling the cavity of the
tool with liquid PMMA. 72
6.18 Simulated temperature versus time at the virtual temperature sensor which is
near the real LED device. 73
6.19 Experimental set-up to investigate the polarisation of the emitted light after
the encapsulation process. 74
6.20 Current-voltage characteristics before (dashed line) and after (full line) the
encapsulation process. 75
6.21 Luminous flux as function of the drive current. Squares before encapsulation,
triangles after the injection moulding process 75
6.22 Emission spectra of a typical LED before and after the encapsulation
process at different drive currents. The FWHM increases after the
encapsulation process. 77
6.23 A) depicts the decrease of the emission energy E2 versus the drive current.
B) shows the peak broadening after the encapsulation process. 78
6.24 Spectra at different polarization angles relative to the axis of symmetry of the
module. 78
6.25 Light intensity of an encapsulated LED versus angle ϕ of the polarisation filter. 79
6.26 Radiation pattern of a LED chip without encapsulation. 80
6.27 Radiation pattern of a LED encapsulated in the parabolic optic. 81
7.1 Intensity versus sputter time for phosphor, magnesium and gallium. 87
7.2 Mass-spectra of Mg24, Mg25 and Mg26. Red line describes the intensity
ratio of the Mg isotopes in nature. 88
7.3 Schematic coordinate system for the diffusion of Mg through the active layer
during thermal degradation. 90
7.4 The calculated relative intensity r(t) versus degradation time for θ = 79.5h
and γ = 0.2. 92
7.5 Relative intensity versus degradation time. Red dots for Idegra=120mA.
The green curve is a fit using θ=79.5h and γ=0.2 . Blue dots for
I
degra=150mA, the black curve is a fit using θ= 7.5h and γ=0.33. 93
7.6 Relative intensity versus degradation time at Idegra=40mA and Imeas=40mA. 94
7.7 Diffusion time =
θ
versus reciprocal junction temperature. The arrow Dwd4/
2
marks T=46°C. 95
7.8 Lifetime prediction at a forward current of Idegra=20mA. Arrow 1: Time to
failure (TTF) at r(t)=0.9 is 10 . Arrow 2: TTF at r(t)=0.5 is 1. 96 h
5h
6
105. ⋅
A.1 Calibration curve to focus inside the LED chip. FP is the distance between the
surface of the LED chip and the focal point inside the LED. ∆d is the distance
difference between lens 1 and lens 2. 101
B.1 Observation cone in the LED chip. 103
C.1 Sample arrangement in the SIMS unit. Red square marks the LED chip. 105
vi
List of abbreviations
AL Active Layer
AS Absorbing Substrate
CF Confinement Layers
CL Cladding Layer
DBR Distributed Bragg Reflector
DH Double Heterostructure
EL Electroluminescence
EP Edged Plane
FEM Finite Element Method
FWHM Full Width at Half Maximum
LED Light Emitting Diode
LCL Lower Confinement Layer
µ-EL Micro Electro Luminescence
MQW Multi Quantum Well
PMMA Polymethylmethacrylate
QW Quantum Well
SEM Scanning Electron Microscopy
SIMS Secondary Ion Mass Spectroscopy
ToF Time of Flight
TS Transparent Substrate
UCL Upper Confinement Layer
vii
viii
List of symbols
A junction area (m2)
A0 intensity of spectra model function
An numerical aperture
α half aperture angle
α, β
Varshni parameters
α, β
capture rates
Cx diffusion material concentration (cm2)
cv specific heat capacity (J/gK)
D diffusion coefficient (cm2s-1)
d diameter (m)
E Energy (eV)
e elementary charge (C)
Eg(T) energy gap (eV)
Eg 0 Energy gap at 0 K (eV)
ε0 vacuum dielectric permittivity (AsV-1m-1)
εr relative dielectric permittivity
Φ luminous flux (lm)
h Plancks constant (Js)
ηt electroluminescence efficiency
ηB internal quantum efficiency
ηinj current injection efficiency
ηoc outcoupling efficiency
I current (A)
Idegra degradation current (A)
Imeas measurement current (A)
Is saturation current (A)
Ie electron diffusion current (A)
Ih hole diffusion current (A)
ϕ polarization angle (°)
ix
k Boltzmann constant (JK-1)
Lx diffusion length (m)
L light output
LQW width of the quantum well (nm)
λ wavelength (nm)
m magnification
Nn,d density of non radiative traps (cm2)
n, p carrier concentration (cm-3)
n refraction index
ν
frequency (Hz)
P power (W)
q charge (C)
θ diffusion time (s)
R resistance (Ω)
Rs serial resistance (Ω)
Rdiff differential resistance (Ω)
r relative efficiency
σnr capture cross section (m2)
t time (s)
T temperature (K)
Tj junction temperature (K)
τx carrier lifetime (s)
V voltage (V)
vth thermal electron velocity (m/s)
Wn width of the depletion zone in the n-type region (m)
Wp width of the depletion zone in the p-type region (m)
WD width of the depletion zone (m)
wD diffusion distance (m)
x
1 Introduction
Fast progress in the development of new semiconductor materials and an enhanced
development of established material systems during recent years have led to new high
brightness light emitting diodes (LEDs) [1], [2]. Today, millions of those LEDs are
manufactured weekly for a manifold applications. Transformation from light sources with
rather low light output used as indicators in many applications to devices able to illuminate
rooms is under progress. But not only for ambient lighting, LEDs are promising light sources
with a wide field of applications, also car manufacturers have discovered these new solid-state
light sources [3]. Efficiency improvement of about 50% for monochromatic light, point light
sources for new designs and fast switching times are some benefits unachievable with
conventional light bulbs.
Both, the increase of internal quantum efficiency as well as the improvement of light out-
coupling leads, from a physical point of view, to an increase in efficiency. Improved light
extraction from the LED devices is reached by changing the geometry of the semiconductor
material in a way that more photons can pass the interface between the semiconductor and the
encapsulation material. This means, less photons are totally reflected at the interface and
absorbed in the substrate. In [4] the light out-coupling efficiency from a cubic-shaped device
is denoted with ηext=25%. By changing the geometry to a truncated cone with a backside
metal mirror the out-coupling efficiency increases to ηext=60%. The limited reflectivity of
metal mirrors, which is in the range of R=90%, can be enhanced by using distributed Bragg
reflectors (DBRs). In addition to their high reflectivity which can exceed R=99.9% [5], DBRs
can be epitaxially overgrown by optically active layers like quantum wells. Further
enhancement of light out-coupling can be achieved by surface structuring. In [6] the
extraction efficiency enhancement of a LED is denoted by a factor of 4.4 using a photonic
crystal.
The improvement of internal quantum efficiency is reached by the invention of quantum wells
and multi quantum well LEDs [7]. These structures confine both types of carriers in a narrow
range, which leads to a large overlap of the carrier wave functions. This increases the
probability of recombination and the creation of photons.
1 Introduction 2
To avoid electron diffusion from the n-type to the p-type contact, diffusion blocking layers
are inserted. Due to their large band gap they stop electron diffusion after the multi quantum
well (MQW) region.
In order to extend the range of the LED applications further, it is useful to incorporate the
semiconductor LED chip directly into total reflection optics. Direct encapsulation by an
injection molding tool is a useful way to control the optical geometry to produce the required
radiation pattern and also to protect the semiconductor against environmental influences.
In collaboration with the Hella automotive lighting manufacturing company, we developed
and tested the influence of the injection moulding encapsulation process on AlInGaP LEDs.
We used LEDs manufactured at Lumileds Lighting with an emission wavelength of
λ
=630nm
and a power input of about P=50mW. Directly dispensing the chips onto a web and using a
copper frame, we combined four LED chips to form a module of 32mm length, 9mm height
and 8mm thickness. The copper frame integrates the electrical connections and acts as a heat
sink to conduct the heat out of the semiconductor chip to the environment. Die bonding glues
the anode of each LED chip to the frame, and wire bonding with 25µm diameter gold wires
connects the cathode to the frame. An Arburg 380 injection moulding machine encapsulates
the frame with polymethylmethacrylate (PMMA) instead of standard resin and forms the four
parabolic-shaped optics with the semiconductor light sources inside. The PMMA gives much
more flexibility than standard optical resin concerning the shapes achievable, as well as
accelerates faster processing because of its quick hardening characteristic.
A disadvantage of injection molding encapsulation is the thermal treatment of the LED chips
during this encapsulation process. One known reason for degradation in semiconductor
materials is temperature-induced diffusion of impurities and lattice defects. For this reason,
specific knowledge of the thermal load of semiconductor materials during an injection
molding process is required. We developed a method to investigate the thermal load of the
LED chips during the encapsulation process. We also show that thermal degradation during
encapsulation is negligible and has no effect on the lifetime of the LED. The mechanical
treatment during the encapsulation process is also investigated and different chip contact
methods are tested with reference to their application possibilities.
Accelerated degradation experiments on PMMA encapsulated LEDs are carried out. Optical
and electrical parameters during aging are measured. The results are analyzed by means of a
diffusion model and a lifetime prediction for nominal power rating is calculated. The lifetime
Introduction 3
of PMMA-encapsulated LEDs is predicted to t=105h using a 5% reduction in the luminous
flux as failure criteria.
In chapter 2 we describe the fundamentals of semiconductor devices. The light generation and
extraction from the semiconductor is described in detail. The micro electroluminescence set-
up is described in chapter 3. We discuss space resolved measurements of the LED chip in
chapter 4. Chapter 5 describes the LED aging under high forward current. Positive as well as
negative aging effects are observed and discussed in detail. The injection molding of LED
clusters is presented in chapter 6. We also compare the characteristics of the non-encapsulated
LEDs to the characteristics of the PMMA-encapsulated LEDs. A diffusion model to describe
the decrease of luminous efficiency versus degradation time is presented in chapter 7. We
determine from accelerated degradation experiments the lifetime of the LEDs under the
recommended maximum forward current. Chapter 8 gives a summary of the results.
1 Introduction 4
2 Fundamentals of semiconductor devices
This chapter discusses the basic physics of semiconductors with the focal point on LEDs used
in this thesis. Starting with the band structure in solids the functional principle of pn-junctions
and multi quantum well LEDs (MQW LEDs) made of AlInGaP are described in detail.
Furthermore, the light generation by radiative recombination and the non-radiative
recombination is illustrated, followed by an analysis of the emission spectrum. Subsequently
the theoretical spectrum is compared to a real spectrum and a modelling function is
introduced. Light-out coupling from the solid to the environment is described and it is shown
that only a small amount of generated photons can escape to the environment.
2.1 Semiconductors
Due to the Pauli principle, the energy splitting of quantized atomic energy levels leads to
quasi-continuous energy bands in crystalline solids. Essentially all the transport and optical
properties of a semiconductor are determined by these energy bands. Energy levels have
regions with energy values called “not allowed”. This means that an electron can exist in
allowed regions which are separated by forbidden band gaps. At T=0K we have the situation
that an allowed band is completely filled with electrons while the next allowed band,
separated in energy by a gap Eg, is completely empty. At T>0K the highest occupied band is
partially filled. When an allowed band is completely filled with electrons, the electrons in the
band cannot conduct any current. Therefore in this case the material has an infinite resistance
and is called an insulator. The material in which a band is only half full with electrons has a
low resistance and is called a conductor. The band which is normally filled with electrons at
T=0K in semiconductors is called the valence band, while the upper unfilled band is called the
conduction band. At finite temperatures, the occupation of electrons and holes is described by
the Fermi distribution function. As the temperature rises, the Fermi distribution function
smears and some electrons are emitted from the valence band into the conduction band.
6 2 Fundamentals of semiconductor devices
Now the electrons in the conduction band and holes in the valence band can carry current.
The bandedge properties are of particular interest since they dominate the electrical and
optical properties. Generally one distinguishes between direct and indirect semiconductors. In
direct materials the minimum of the conduction band is at the same point in the reciprocal
space as the maximum of the valence band. A direct bandgap ensures excellent optical
properties, momentum conversation needs a “vertical k”. Figure 2.1 depicts the bandstructure
of indiumphosphide (InP). The arrow depicts the transition from the minimum in the
conduction band to the maximum in the valence band.
Figure 2.1: Bandstructure of the direct semiconductor InP [8]. The arrow depicts the
transition from the conduction band to the valence band.
In a direct semiconductor the maximum of the valence band is below the minimum of the
conduction band at the gamma point (Γ). The smallest difference between the conduction
band and the valence band is called the energy gap and for InP
[9].
eVE InP
G)0006.04236.1( ±=
The most popular indirect semiconductor is silicon (Si). Due to the indirect bandgap Si has
very poor optical properties and cannot be used to form LEDs. Also hole transport properties
are quite poor as the hole masses are large. However, for electronic devices, silicon is the
material of choice, because of its highly reliable processing technology.
An interesting material with respect to optical properties is galliumphosphide (GaP). Material
compositions of GaP, indiumphosphide (InP) and aluminiumphosphide (AlP) are commonly
used to produce LEDs in the red and amber spectral range. The bandstructure of GaP is
shown in Figure 2.2, it can be seen that GaP is an indirect semiconductor material. The
2.1 Semiconductors 7
minimum of the conduction band is at the X-Point, the maximum of the valence band at the
Γ-Point. The energy gap is [9]. The arrow depicts the transition
from the minimum in the conduction band to the maximum in the valence band. To excite an
electron from the valence band to the conduction band the energy of E
eVEGaP
XG )006.0344.2(
)( ±=
Γ−
G is required.
Figure 2.2: Bandstructure of the indirect semiconductor GaP [8]. The arrow depicts the
transition from the conduction band to the valence band.
Additionally, the momentum p from a third particle is required due to the non-vertical k. The
momentum of a photon is too small, but a phonon can contribute the needed momentum.
Therefore a three-particle process is required to excite an electron from Γ15 to X1. The inverse
process, a recombination of an electron at X1 with a hole at Γ15, also needs the momentum of
a third particle. These processes are very rare and the internal quantum efficiency would be
very low.
The bandstructure of AlP is similar to GaP. AlP like GaP is an indirect semiconductor. The
energy Gap is [9].
eVE AlP
G)02.051.2( ±=
The materials GaP, AlP and InP can be alloyed in any composition. Starting from GaP,
several Ga-atoms can be replaced by Al-atoms or In-atoms, respectively. The result is the
alloy AlInGaP. The properties of this quaternary material system, for example bandgap and
lattice constant, are in the limits of the pure materials. Therefore, by varying the composition
of the materials, the bandgap and the lattice constant can be tuned. The correlation of the
bandgap and the lattice constant is shown in Fig. 2.3. For large InP-content the direct bandgap
dominates the alloy. For small InP-content the bandgap of the alloy is indirect. The
composition of (Alx Ga1-x)0.5In0.5P is lattice-matched to GaAs. Therefore GaAs is a perfect
2 Fundamentals of semiconductor devices 8
substrate for AlInGaP LEDs. The red line indicates the spectral range of AlInGaP LEDs
lattice-matched on GaAs which covers the purple to green field. This spectral range is equal
to a variation in the band gap between 1.89eV and 2.33eV.
Wavelength [nm]
Band gap [eV]
Lattice constant [Å]
Figure 2.3: Band gap and wavelength of the alloy AlInGaP versus lattice constant. For large
In-content the direct bandgap dominates. The red line indicates the spectral
range of AlInGaP LEDs lattic- matched on GaAs with a direct bandgap [7].
Because of the small band gap of GaAs, Eg=1.45eV [10], it is non transparent for the emission
range of AlInGaP LEDs. This leads to light absorption in the substrate and less efficient
devices. These disadvantages could be overcome by removing the substrate after the
production process.
2.2 Band structure of p-n junctions
2.2.1. Doping of semiconductors
The density of free carriers that can carry current in pure semiconductors is very low. To
increase the free carrier concentration, impurities known as dopants are introduced. The
dopants are chosen from the periodic table so that they either have an extra electron in their
outer shell in comparison to the host semiconductor, or have one electron less. Absent
electrons are called holes and act as a positive quasi-particle.
2.2 Band structure of pn-junctions 9
The resulting dopant is called a donor or acceptor. In chapter 7.4 it is shown that the p-type
dopant in the LEDs used during this investigation is magnesium (Mg). The n-type doping is
not investigated, but a common material for this application in AlInGaP LEDs is tellurium
(Te) [11]. Donors create an energy level in the energy gap near the conduction band edge.
Acceptors also create an energy level in the energy gap, but near the valence band edge. In the
p-doped area the holes are the majority carriers, the electrons are the minority carriers. In the
n-doped area the notation is vice versa.
Non-doped semiconductors are called intrinsic and the Fermi level is in the middle of the
band gap. By n-type doping the Fermi level moves towards the conduction band, p-type
doping lowers the Fermi-level to the valence band.
Holes and electrons can recombine and photons with the energy difference EPh are created and
radiated. The energy of a photon is given by:
λ
ν
⋅
== n
hc
hEPh (2.1)
with
gPh EE ≥
where h is the Planck’s constant,
ν
is the frequency,
λ
is the wavelength of the photon, c is the
speed of light, and n is the refraction index.
The inverse process is also possible, a photon can be absorbed by the semiconductor and a
electron hole pair can be generated. Typical applications are solar cells.
2.2.2. pn-junction
The pn-junction is one of the most important structures in solid state electronics. To fabricate
a pn-junction a p-type layer is formed by epitaxial growth. The dopant species is changed and
a n-type layer is created directly on the p-type layer. We assume in our analysis that the
junction is abrupt. In the absence of any applied bias no current flows and the Fermi level is
uniform throughout the structure. The material at both ends of the structure is neutral. In the
p-type region, far away from the junction, the density of acceptors exactly balances the
density of holes. In the n-type region, far away from the junction, the density of immobile
donors exactly balances the free electron density. In the depletion region (WD) the bands are
bent and a field exists which has swept out the mobile carriers leaving behind negatively
charged acceptors in the p-region and positively charged donors in the n-region. In the
depletion zone, which extends a distance Wp in the p-region and a distance Wn in the n-
region, an electric field exists. Any electron or hole is swept away by this field. Thus a drift
2 Fundamentals of semiconductor devices 10
current exists which counterbalances the diffusion current which arises because of the
difference in electron and hole densities across the junction. Figure 2.4 depicts the band
structure of such an unbiased pn-junction. For similar space charge density in the p- and n-
regions of the depletion zone we get applying Poisson’s law:
ρ
εε
0
1
r
V=∆− (2.2)
a quadratical behaviour of the bands, because:
qVE
=
(2.3)
where
V is the potential difference, εr is the relative permittivity, ε0 is the electric field constant, q is
the charge and
ρ
is the space carrier density.
Figure 2.4: Band structure of the p-n junction under zero bias. [7].
In the present of an applied bias the balance between drift and diffusion current will no longer
exist and a current flows. One distinguishes between a forward bias and a reversed bias. In
forward bias Vf, the p-side is at a positive potential with respect to the n-side. The potential
barrier eVD is reduced by the applied voltage Vf which leads to a reduced electrical field at the
junction. Therefore the drift current is also reduced and a diffusion current flows through the
device. The current increases exponentially with the applied forward bias. If the applied
voltage is equal to the potential barrier eVD this situation is called “flat band condition” and
the applied voltage is called the threshold voltage.
If the bias is applied in the reverse direction, the barrier eVD increases with increasing Vr and
the electric field at the junction increases as well. Thus, in the equilibrium, the increased drift
current must be compensated by an increased diffusion current. Therefore the depletion zone
2.2 Band structure of pn-junctions 11
expands in the n- and p-type region. A low reverse current flows through the device which is
independent from the applied reverse voltage.
The current flow through a pn-diode has a strongly nonlinear and rectifying behaviour. The
current saturates at a value I0 if a reverse bias is applied. Since the value is quiet small, the
diode is essentially non-conducting. If a positive bias is applied the current increases
exponentially and the diode becomes strongly conducting. The current density at an abrupt
pn-junction is given by [12]:
−
−
−=+= 1
)(
exp
00
Tk
IRVe
n
L
neD
L
peD
jjj
B
S
Diff
n
pn
p
np
npDiff (2.4)
where jp is the hole current density, jn is the electron current density, Dp is the diffusion
coefficient of the holes, pn0 is the hole concentration in thermodynamical equilibrium in the n-
region, Lp the diffusion length of the holes, Dn is the diffusion coefficient of the electrons, np0
is the electron concentration in thermodynamical equilibrium in the p-region, Ln the diffusion
length of the electrons, and T the temperature. Rs describes the serial resistance of the device
and nDiff is the quality factor which theoretically is one.
2.2.3 Double heterostructures and quantum wells
A double heterostructure (DH) consists of the active region in which recombination occurs
and two confinement layers cladding the active region. The bandstructure of a double
heterostructure LED is shown in Fig. 2.5. The two cladding or confinement layers have a
larger bandgap than the active region. If the bandgap difference between the active and the
confinement region is ∆, then the band discontinuity occurring in the conduction and
valence band follows the relation:
g
E
(2.5)
VCg
act
g
clad
gEEEEE ∆+∆=∆=−
Both band discontinuities, and
C
E∆V
E
∆
, should be much larger than kT in order to avoid
carrier escape from the active region into the confinement regions.
In the case of a homojunction, carriers diffuse to the adjoining side of the junction under
forward bias conditions. Minority carriers are distributed over the electron and hole diffusion
lengths. In III-V semiconductors, diffusion lengths can be 10µm or even larger. The wide
distribution of carriers can be avoided by employing double heterostructures. Carriers are
2 Fundamentals of semiconductor devices 12
confined to the active region as long as the barrier heights are much higher than the thermal
energy kT. Additionally, reabsorption of the generated photon is avoided due to the lower
bandgap of the active layer. Nowadays all high efficiency LEDs contain double
heterostructure designs.
Figure 2.5: Band diagram of a forward-biased double heterostructure. The p-type confinement layer
consists of a lightly doped layer close to the active region and a higher doped layer
further away from the active layer [7].
When the thickness of the active layer of a DH is reduced to the order of the de Broglie carrier
wavelength two dimensional quantization occurs. Quantum well active regions provide
additional carrier confinement to the narrow well regions which can further improve the
internal quantum efficiency. On the other hand, if a quantum well active region is used, the
barriers between the wells will impede the flow of carriers between adjacent wells. Thus the
barriers in a multi-quantum well (MQW) region need to be sufficiently transparent (low or
thin barriers) in order to allow efficient carrier transport between the wells and to avoid the
inhomogeneous distribution of carriers within the active region. In [13] it is described that
QW have discrete energy levels which strongly depend on the barrier depth and thickness.
Therefore the emission energy of the device can be tuned by changing the geometry of the
QW. The wave functions of a captured electron and hole in the QW have a large overlap
which leads to a probable recombination.
Figure 2.6 shows the band diagram of a quantum well for the (AlxGa1-x)0.5In0.5P material
system, the QW width is denoted by LQW. The energy eigenvalues are designated by E1 and E2
for electrons and Eh1 and Eh2 for holes.
All LEDs used in these investigations are MQW LEDs. We assume a well thickness and a
barrier thickness of about 5nm. In [14] the well width of an AlInGaP LED is 3nm.
2.2 Band structure of pn-junctions 13
(AlxGa1-x)0.5Ino.5P
Eg, ba
r
Eh2
Eh1
E2
E1
LQW
E
g
, QW
∆
Ec
∆
Ev
Ev
Ec
Figure 2.6: Square-well potential of a quantum well structure. The well is formed by
decreasing the mole fraction x.
Interband recombination transition (
∆
n=0) occurs from a bound state in the conduction band
e.g. E1 to a bound sate in the valence band e.g. E2. The energy of transition is given by:
11, hQWgt EEEE
+
+
=
(2.6)
Thus the recombination can proceed between two well-defined energy levels in contrast to a
bulk semiconductor.
Figure 2.7 depicts the density of states for quantum wells (2D) as well as for bulk material
(3D). The step-like density of state is characteristic for a quantum well structure [16]. The
half parabola originates from the conduction band edge.
Densit
y
of state
g(
E
)
2D
3D
Eg, QW E1 E2
Energy
E
Figure 2.7: Density of state for a quantum well structure (2D) and for bulk material (3D).
2 Fundamentals of semiconductor devices 14
2.3 Radiative recombination
Electrons and holes in semiconductors recombine either radiatively, i. e. accompanied by the
emission of a photon, or non-radiatively (see chapter 2.6). In LEDs, the former is clearly the
preferred process.
Considering a free electron in the conduction band, the probability of an electron recombining
with a hole is proportional to the hole concentration. The number of recombination events
will also be proportional to the concentration of electrons. Thus the recombination rate is
proportional to the product of electron and hole concentrations. Using a proportionality
constant, the recombination rate per unit time per unit volume can be written as the well
known bimolecular rate equation [7]
Bnp
dt
dp
dt
dn
R=−=−= (2.7)
where R is the recombination rate, n and p are the electron and hole concentrations and B is
the proportional constant.
Further we discuss the recombination dynamics as a function of time. The equilibrium and
excess electron and hole concentrations are n0, p0,
∆
n and
∆
p, respectively. Since electrons
and holes are generated and annihilated (by recombination) in pairs, the steady-state electron
and hole excess concentrations are equal,
)()( tptn
∆
=
∆ (2.8)
Using the bimolecular rate equation (2.7), the recombination rate is given by:
[
]
[
]
)()( 00 tpptnnBR
∆
+
∆
+
=
(2.9)
Quantum wells provide a means of confining the free carriers to a narrow quantum well
region by using the two barrier regions cladding the quantum well. The well thickness is LQW,
the conduction band and the valence band wells have a carrier concentration of n2D/LQW and
p2D/LQW, respectively. Using these values as the 3D carrier concentrations, the recombination
rate can be inferred from Eq. (2.9), and is given by:
2.3 Radiative recombination 15
QW
D
QW
D
L
p
L
n
BR
22
= (2.10)
One of the essential advantages of quantum well LEDs is illustrated by this equation. A
decrease of the QW thickness allows to attain high 3D carrier concentrations in the well
(carriers per cm3). The carrier lifetime is then reduced and the radiative efficiency is increased
[7]. For small QW thicknesses, the wave function no longer scales with the physical well
width. LQW must be replaced by the carrier distribution width, which for sufficiently small
well thicknesses is larger than LQW, since the wave function will extend into the barriers.
2.4 Modelling of the emission spectra
The emission spectra of QW LEDs is given by the product of the density of state and the
distribution of carriers in the allowed bands. The distribution of carriers is given by the
Boltzmann distribution, i.e.
(2.11)
)/(
)( kTE
BeEf −
=
The density of state in a QW is given by:
∑
−∝
n
n
EEHEg )()( (2.12)
where H(E-En) is the Heaviside function. It takes the value of zero when E is less than En and
1 when E is equal to or greater than En.
Figure 2.8 depicts the emission spectrum of a QW LED (red curve). The low energy regime is
confined by the 2D density of states. The high energy emission is given by the Boltzmann
distribution [15].
2 Fundamentals of semiconductor devices 16
)/exp( kTE
−
∝
Boltzmann distribution
2D
3D
g(E)f(E)
Eg, QW E
E1
Energy
Figure 2.8: The emission spectrum of a QW LED, red curve, is the product of the density of
state and the distribution of carriers.
The real emission spectrum of a LED is quite different. Figure 2.9 depicts the emission of a
non-encapsulated LED chip with a peak energy of E=1.93eV in logarithmic scale. An
exponential slope to the maximum emission is significant. Then the emission decreases
exponentially with a different slope. The exponential increase of the measured emission
spectra is called Urbach edge. According to [17] the Urbach edge yields an exponential
increase of the intensity.
Energy levels in QWs strongly depend on the QW thickness. In the MWQ region of a LED
the thickness of the individual QWs may vary due to fluctuations in the production process
[18]. Therefore single QWs may emit at a different energy level. Thus the emission broadens.
These effects lead to an exponential increase of the emission [19]. Also the exponential
decrease of the emission in the high energy regime is effected by QW inhomogeneous
broadening. Therefore the carrier temperature could not be calculated directly by Eq. (2.11).
Details of temperature measurements are given in chapter 4.4.
2.4 Modelling of the emission spectra 17
1.84 1.88 1.92 1.96 2.00 2.04
Intensity [a. u.]
Energy [eV]
Urbach edge
Figure 2.9: Emission spectrum of an AlInGaP LED. The black line indicates the
measurement, the blue line the model function.
To analyse the emission with respect to peak energy and exponential decay of the emission a
fit function is required. The peak energy and the exponential decay can be calculated by the fit
function. The following function describes the emission:
)
(2.13)
()()( 2EkEAhEg =
with
)(
)( DEa
eEh −
=
1
1
)( )(
+
=−dEab
e
Ek
ab
b
Ed
−
−= 12
1
ln
0
2
1
0
12
2
12
1
−
−
=
b
b
b
AA
b
where A0 is the intensity at the maximum and a is the exponent of the emission increase.
2 Fundamentals of semiconductor devices 18
The parameter b influences the exponent of the emission decay. The parameter d must be
tuned so that E0 describes the maximum of the emission. For E<<E0 Eq. (2.13) becomes:
(2.14)
)(
)( dEa
eEg −
=
and for E>>E0
(2.15)
))(2(
)( dEaba
eEg −−−
=
with
abaC 2
−
=
(2.15a)
The peak energy of the emission E0, the maximum intensity A0 and with Eq. (2.15a) the
exponent of the emission decay can be calculated. In Fig. 2.9 the model function is depicted
in blue, the experimental data in black. The model function matches the LED emission
perfectly.
2.5 Light extraction from a LED
Light generated inside a semiconductor cannot escape from the semiconductor, if it is totally
reflected internally at the semiconductor-air interface. If the angle of incidence of a light ray
is close to normal incidence, light can escape from the device. However, total internal
reflection occurs for light rays with oblique and grazing-angle incidence. Total internal
reflection reduces the external efficiency significantly, in particular for LEDs consisting of
high-refractive-index materials. This is the case for the investigated AlInGaP. The critical
angle for total internal reflection is given by:
1
2
sin n
n
crit =
β
(2.16)
where n are the refraction indexes. For GaP the refraction index is nGaP=3.02 for photons with
an energy of E=1.9eV [20]. Thus at the GaP-air interface the critical angle of total reflection
2.5 Light extraction from a LED 19
is . The light in the device is generated by spontaneous emission and radiates
uniformly in all directions. This is equal to a point-like light source in the active layer of the
semiconductor. Only a fraction of light is emitted inside the critical angle ,
allowing escape to the ambient region (air). All other photons are reflected at the interface.
Once they are reflected in the LED they are captured in the device until they are absorbed by
the substrate or reabsorbed. Figure 2.10 depicts this behaviour (side view), for a photon with a
arbitrarily chosen angle
°=
−3.19
airGaP
crit
β
airGaP
crit
−
<
ββ
crit
β
β
>
exp . The active layer is dark red, at the emission point the
escape cones with the critical angles crit
β
are depicted. Light rays inside this escape cone can
escape the device. The red arrows are the ray of one arbitrarily chosen photon with an angle
>βcrit. As can be seen the photons are captured in the LED and cannot escape. Normally those
photons are absorbed in the substrate or reabsorbed in a QW.
Figure 2.10: A captured photon in a LED structure. The active layer is dark red, the escape
cones are hatched. Only photons inside the escape cones can escape the device.
A captured photon with an angle >
β
crit is shown as light-red rays.
Additionally, the external efficiency depends on the reflectivity of the interface. Neglecting
absorption the reflectivity is given by :
2
11
11
+
−
=nn
nn
R (2.17)
Thus the reflectivity at the GaP-air interface is calculated to:
%2.25
=
−airGaP
R (2.18)
2 Fundamentals of semiconductor devices 20
With respect to a fully metallised n-side and a partially metallised p-side and without respect
to photon recycling the escape efficiency is calculated to:
airGaP
ext
−
η
(
)
(
airGaP
airGaP
crit
airGaP
ext R
)
r
r
−
−
−−
−⋅
=1
4
cos125.4
2
2
π
βπ
η
(2.19)
%5.9≈
Using
2
)1(
4
1+
=− n
n
R (2.20)
and
()
(
)
airGaP
crit
airGaP
crit
−− −−≈−
ββ
sin11
2
1
cos1
2
1 (2.21)
2
4
1
n
≈
The escape efficiency can be written as a function of the refraction index and is given by
2
)1(
5.4
+
=
−
nn
airGaP
ext
η
(2.22)
for a homogenous photon distribution.
The escape efficiency can be increased by changing the geometry of the LED chips. Examples
are given in [7]. Additionally, the device can be embedded in a material with a refraction
index between that of GaP and that of air, i.e. resin. Thus the critical angle of reflection is
increased and more light can escape the LED [21].
2.6 Non-radiative recombination 21
2.6 Non-radiative recombination
In non-radiative recombination processes, the electron energy is converted into vibrational
energy of lattice atoms, i.e. phonons. Thus the electron energy is converted into heat. Non-
radiative recombination reduces the efficiency of a LED.
Defects in crystal structures are the most common cause for non-radiative recombination.
These defects include foreign atoms, native defects and dislocation. In compound
semiconductors, native defects include interstitials, vacancies and antisite defects [7]. It is
quite common for such defects to form one or several energy levels within the energy gap.
These defects are efficient recombination centres, in particular if the energy level is close to
the middle of the energy gap which reduces the band-band recombination.
Another non-radiative recombination mechanism is the Auger-recombination. In this process,
energy becomes available by an electron hole recombination. The energy is immediately
absorbed by another electron which is excited high in the conduction band, or by a hole
excited deep in the valence band. The highly excited carriers will lose energy by multiple
phonon emission until they are close to the band edge. Figure 2.11 depicts the Auger-
recombination.
Energy
Reciprocal space k
E
Hot electron
Eg
Conduction band
Valence band
Figure 2.11: Auger-recombination. The energy released by an electron hole recombination is
absorbed by another electron.
2 Fundamentals of semiconductor devices 22
3 Micro electroluminescence set-up
This chapter describes the construction of an electroluminescence measurement set-up with
high spatial resolution. The set-up has been used to accomplish spatially resolved
measurements of the emission wavelength and the emission intensity on the lateral LED
planes. Measurements perpendicular and parallel to the growth direction are investigated.
Additionally, the set-up of the Ulbricht sphere is explained in detail at the end of this chapter.
3.1 Principle of the set-up
Electroluminescence (EL) is the luminescence of a solid caused by an applied electrical
current. Luminescence can also be excited by photons (photoluminescence) or by an electron
beam (cathode luminescence). EL is the method of choice to investigate the characteristics of
completely structured LEDs. A schematic set-up for electroluminescence measurements is
shown in Fig. 3.1. The light emission is measured with respect to wavelength (energy) and
intensity
+
Monochromator Detector PC
Figure 3.1: Electroluminescence set-up, the emission from the LED chip is coupled into a
monochromator. A diode array detects the diffracted light and the data is sent
to a PC.
The monochromator 500M from Spex Industries has a focal length of f=50cm and is directly
driven by a PC. The grating has a blaze wavelength of
λ
blaze=750nm and 600 grids per mm.
3 Micro electroluminescence set-up 24
The detector is a silicon diode array from EG&G, model 1421B. The 1024 diodes are placed
on a length of 40mm. A spectral range of 80nm is imaged on the detector array. An optical
multi channel analyser OMA 2000 from EG&G reads out the array and sends the data to a
PC. The spectral resolution is given by
n
a
λ
∆
= (3.1)
where n=1024 is the number of diodes of the array. Thus the resolution is given by:
nm
nm
a08.0
1024
80 ≈= (3.2)
As control software for the monochromator and detector system WinSpec is used. This
software allows the implementation of macros. Therefore new functions can be added to the
system.
Investigating local effects in a scale much smaller than the edge length of the device requires
a micro electro luminescence set-up (µ-EL). Structural investigations in a range of a few µm
are possible with this method.
Optical fiber
Optic Screen
PC Detector Monochromator
+
Figure 3.2: Sketch of the micro-electroluminescence set-up. The light from the LED is
magnified by a factor of 500 and projected onto a screen. One point of the image
with a diameter of 1mm is coupled into an optical fibre and guided to the
monochromator. Data is stored in a PC.
3.2 Optics 25
To expand the EL-set-up (Fig 3.1) for µ-EL measurements, a further device is inserted into
the light path which magnifies the LED Chip by a factor of 500. The image is projected onto a
screen. A point from this image with a diameter of 1mm is coupled into an optical fibre and
guided to the monochromator. The position of in-coupling can be moved by stepping motors
over the hole image range of the LED. Thus every point of the emitting surface can be
investigated. The schematic set-up is depicted in Fig. 3.2.
3.2 Optics
To obtain a magnification factor of , two convex lenses can be used. The set-up is
shown in Fig. 3.3. Lens 1 generates a magnified real intermediate image. Lens 2 magnifies
this intermediate image.
500≥m
Figure 3.3: Optical set-up. The LED is outside the focal length of lens 1. Lens 1 generates a
magnified real intermediate image. Lens 2 magnifies this intermediate image.
Lens 2 can be adjusted with respect to lens 1, thus exact focusing is possible. Instead of
simple lenses, objectives from Ernst Leitz Wetzlar GmbH are used. The objective near the
LED has a nominal magnification of 10 and a numerical aperture of 0.22. The second
objective has a nominal magnification of 20 and a numerical aperture of 0.45.
3 Micro electroluminescence set-up 26
3.2.1 Step size at the chip surface
The minimum measurement step size at the chip surface depends on the magnification factor
of the optics and the diameter of the in-coupling point. As mentioned before, the
magnification factor should be . It can be estimated experimentally by comparing the
image size with the object (LED chip) size. The described set-up achieves a magnification of
. The size of the in-coupled light spot is equivalent to the diameter d=1mm of
the optical fibre. The minimum step size O
500≥m
15610 ±=m
A at the chip surface is given by:
64.1≈= m
d
OAµm (3.3)
3.2.2 Space resolution
The space resolution d of an optical device is the minimum distance of two points required to
percipience them as two separate points. Is the distance between two points below this limit
one observes them as one point. Edge diffraction occurs in every optical system, thus two
points can be separated if the main maximum of the first point will be at the first minimum of
the second point. So the spatial resolution is given by:
N
A
d2
λ
= (3.4)
where
λ
is the wavelength of the light and AN is the numerical aperture of the optical device. It
is defined as:
α
sinnAN
=
(3.5)
where n is the refraction index and
α
is the half aperture angle of the objective.
The numerical aperture of the µ-EL set-up is AN=0.28. The emission of the investigated LEDs
is )20640( ±=
λ
nm. Thus the spatial resolution of the µ-EL set-up is given by:
)04.014.1(
±
=
dµm (3.6)
3.3 The position system 27
By comparing Eq. (3.3) with Eq. (3.6) we see that the step size of the set-up and the space
resolution are in the same order. Structures which are imaged by the system are also in the
resolution limit of the system.
3.3 The positioning system
The optical fibre is driven by two stepping motors. One motor moves the fibre along the x-
axis and the other along the z-axis. Both are perpendicular to the optical axis. The minimum
step size for both axes is l=1.25µm. The software to control the positioning system has been
implemented into the WinSpec software. Thus a synchronisation of luminescence
measurement and fibre movement is obtained. Fully automation to measure large areas of the
LED chip space resolved is achieved by this set up.
The measured data are directly stored with respect to the coordinates of the emission area,
intensity and energy.
3.4 The Ulbricht sphere
An Ulbricht sphere is an optical device for measuring the optical flux from a laser diode,
light-emitting diode or bulb. It is a hollow sphere with a diffusely reflecting internal surface,
in our case bariumsulphate. The sample is mounted in the centre of the sphere, the detector is
attached to the shell of the sphere. A light barrier is used to prevent direct illumination of the
detector by the light source. The arrangement causes many diffuse reflections of the emitted
light before it reaches the detector. So the light flux becomes uniform at the detector and
nearly independent of the spatial distribution of the light emitted by the light source. Basically
the detected optical power depends solely on the total emitted intensity. Figure 3.4 depicts a
schematic drawing of our Ulbricht sphere.
The spectral response of the selenium detector is shown in Ref. [22].
3 Micro electroluminescence set-up 28
Figure 3.4: Schematic set-up of an Ulbricht sphere. The coating has a high reflectivity and
the reflection is diffuse. Thus the light flux is uniform in the sphere.
Ideally the coating on the inner side of the sphere has a very high reflectivity over the required
wavelength range, and the reflection is diffuse. Good optical efficiency is obtained, if the
sphere is much larger than the light source and the detector.
4 Micro electroluminescence - results and discussion
This chapter discusses the results obtained with the µ-EL set-up described in chapter 3. The
samples are described in detail, and the emission of the lateral planes and surfaces of the LED
is analysed.
4.1 Description of the samples
The investigated samples are commercially available AlInGaP-MQW LEDs type HWFR-
B410 by Lumileds Lighting. The device structure is depicted in Fig. 4.1, the quadratic surface
has an edge length of l=232µm. The lateral length is h=280µm. The p-type contact is a
circular metal layer with a diameter of d=100µm. The contact material is aluminium followed
by a gold – zinc layer which guarantees a perfect ohmic contact to the highly p-type doped
GaP window layer. In chapter 2.5 the advantages of window layers are described. Below the
GaP window layer is the p-type AlInGaP-UCL (upper confinement layer). The optically
active multi quantum well region and the n-type AlInGaP –LCL (lower confinement layer)
are aligned. A n-type GaP substrate with a gold contact completes the LED [14]. The
metallised areas of the LED are identified as the surfaces of the LED. In the following all the
other surfaces are called edge planes and they are denoted “EP” in Fig. 4.1. The orientation in
space is also depicted in Fig. 4.1, the z-axis is the direction of the layer deposition.
The emission wavelength of the LED chip at T=20°C and at I=5mA is )5630( ±=
λ
nm
which is equivalent to an emission energy of 96.1
≈
EeV.
4 Micro electroluminescence - results and discussion 30
EP MQW
Figure 4.1: Layer structure of the LED chips. Figure 4.2: SEM image of the LED chip
“EP” denotes the edge planes (side view). The optically active MQW
(red hatched). region is inside the groove and marked
with an arrow.
A scanning electron microscopy (SEM) image of the LED chip is depicted in Fig. 4.2. A
groove 50µm below the circular p-type contact is indicated by an arrow. All epitaxial layers,
the UCL, LCL and the MQW region are in this grove. This epitaxial structure has originally
been grown on GaAs. The GaAs substrate is removed and is replaced by GaP which is
transparent to red light.
More details of the production process and the material composition are not known. Also the
width of the QWs and barriers are not known, but can be estimated to be 5nm.
4.1.1 The HWFR-B 410 LED: basic properties
In this chapter the basic electrical and optical properties of the HWFR-B 410 LEDs used for
our investigations are described. Figure 4.3 depicts the I-V characteristics of the LED chip
measured in a range between –2.5V and 2.5V.
4.1 Description of the samples 31
-5
0
5
10
15
20
25
-3 -2 -1 0 1 2 3
Current [mA]
Voltage [V]
dU/dI=18.2Ω
Us=2.12V
RT
Figure 4.3: Current-voltage characteristic of the HWFR-B410 LED at room temperature.
The threshold voltage is determined to Vth=2.12V and the differential resistance in forward
direction is RDiff=18.2Ω.
10-11
10-10
10-9
-3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5
Current [A]
Voltage [V]
RT
Figure 4.4: Reverse current versus reverse voltage of the HWFR-B410 LED.
To investigate the proper reverse current a picoamperemeter Keithley 6485 is used. The
results are shown in Fig. 4.4. The reverse current increases from A at about 0V
to A at V=-3V.
11
105.3 −
⋅=I
10
108.7 −
⋅=I
4 Micro electroluminescence - results and discussion 32
0
2
4
6
8
10
12
14
16
0 204060801001
Luminous
Current [mA]
I=70mA,
Φ
=15.3
Φ=10.2
T=20°C
flux [a. u.]
Φ=5.2
20
Figure 4.5: Luminous flux
Φ
versus forward current. The maximum luminous flux at T=20°C
is measured at a forward current of I=70mA.
Figure 4.5 depicts the total luminous flux Φ, which was measured in the Ulbicht sphere,
versus forward current. The luminous flux increases linearly up to I=40mA, then the luminous
flux starts to saturate. At an ambient temperature of T=20°C and at I=70mA the maximum
luminous flux is reached. The maximum current recommended by the manufacturer is
I=20mA.
0
20
40
60
80
100
120
140
160
0 102030405060708
Temperature [°C]
Current [mA]
0
Figure 4.6: Junction temperature versus forward current.
Figure 4.6 shows the junction temperature versus forward current up to I=70mA. This result is
obtained in a two-step process. The LED was heated in a cryostat applying a constant forward
4.1 Description of the samples 33
current of I=1mA. The emitted light was coupled into a monochromator and the peak
wavelength shift versus temperature was measured. A linear relationship between the peak
wavelength decay versus temperature was found. In a second experiment the peak wavelength
shift versus increasing forward current of the same sample was measured. Thus a relationship
between the LED temperature and the forward current is created by combining both results.
The junction temperature for 1≤
I
mA is approximately room temperature. The dependence
between the temperature and the dissipated power Pd, converted to heat is:
d
PT
∝
(4.1)
According to [15] Pd is given by:
P(= (4.1.a)
sfffd RIVI 2
)1
ηη
+−
where
η
is the external quantum efficiency, Vf is the forward voltage, If is the forward current
and Rs is the serial resistance. The changes in Vf caused by current is much smaller than Vf
(), and
η
and R
ff VV <<∆ s
are constant. Thus Eq. (4.1) and Eq. (4.1.a) indicate that the
junction temperature depends linearly and quadratically on the forward current.
The experimental data have been fitted by
(4.2)
2
01.02.14.291)( IIITj⋅+⋅+=
For a forward current of I=70mA the junction temperature is determined to T=152°C.
Heating the LED by forward current we observe a red shift of the emission due to the
decreasing energy gap with increasing temperature. The blue shift of the emission which is
caused by band filling with carriers is superposed by the red shift. According to [7] the blue
shift is negligible in the AlInGaP system.
4 Micro electroluminescence - results and discussion 34
4.2 Sample preparation
For mechanical stability, electrical and thermal conductivity reasons the samples are glued
with their n-side onto a copper board. The glue, XCA 80210 from Emerson & Cuming,
hardens within 2 min at T=100°C. In chapter 4.6.1 we show that this hardening process has no
effect on the performance of the LED. The p-side of the LEDs are contacted by a tip probe.
For practical reasons five LEDs are deposited in a row on the above mentioned copper board
and labelled E01 to E05. The board with the five LEDs is installed in the µ-EL set-up.
4.3 Experimental details of micro-electroluminescence
measurements
It has been shown in Chapter 3.1 that the detector of the µ-EL set-up is a diode array which
can detect a specified spectral range during one measurement. Therefore the monochromator
has been adjusted to a centre wavelength of
λ
=640nm. By means of the grating with 600 grids
per mm light with a wavelength of
λ
=640nm is projected onto the centre of the detector array
and a full range of 80nm is projected to the diode array. This means a spectrum from
λ
1=600nm to
λ
2=680nm can be measured instantly without moving the grating in the
monochromator. This range covers the full area of interest during the investigations and a fast
measurement is guaranteed.
The calibration of the monochromator has been done with two lasers with wavelengths of
λ
1=632.8nm and
λ
2=639.3nm, respectively. The spectra of the calibration measurements is
shown in Fig. 4.7. The emission of laser 1 has been measured to nm. The
emission of laser 2 has been measured to nm. The differences
∆λ
)08.084.632(
'
1±=
λ
)08.0
±
32.639(
'
2=
λ
1
=0.04nm and
∆λ
2
=0.02nm are within the resolution limit (see Eq. (3.2)), of the
monochromator-detector system.
4.3 Experimental details of micro-electroluminescence measurements 35
620 625 630 635 640 645 650 655 660
632,84
639,32
Intensity
Wavelenght [nm]
Laser 1: λ = 632,8 nm
Laser 2: λ = 639,3 nm
Figure 4.7: Calibration of the monochromator with two laser lines at
λ
1=632.8nm and
λ
2=639.3nm.
Before the surfaces or edge planes of the LED chip can be investigated the magnified image
must be focused on the screen. Focusing the image is easy if one of the surfaces of the chip is
investigated. One can directly use the image edges of the device as reference. Focusing to the
active layer is more difficult. First one has to focus to the top (p-contact) surface and then
moving the focal point 50µm (distance from the top surface to the active layer) inside the chip
by adjusting the second lens. A detailed description is given in appendix A. To find out the
magnification the dimensions of the mapped image can be measured and compared with the
dimensions of the LED chip. After measuring the background count rate at Imeas=0mA,
adjusting the monochromator slit and integration time, the measuring routine can be started.
4.4 Temperature analysis
The carrier temperature in the device can be calculated by analysing the exponential decay of
the emission. This exponential decay c of the emission can be extracted from Eq. (2.15a). The
Fermi distribution of the electrons in the conduction band describes this exponential decay
and is given by:
()
−
=
Tk
EE
EF
B
F
exp
1 (4.3)
4 Micro electroluminescence - results and discussion 36
If we equate the exponential decay c with the Fermi function one obtains:
() (
G
B
EE
Tk
dEc −=− 1
)
(4.4)
with we can derive a relationship between the exponential decay of the emission and
the carrier temperature.
G
Ed ≈
ck
T
B
e
1
= (4.5)
4.5 Experimental details
4.5.1 Glue hardening
First the influence of the glue-hardening process must be investigated. The glue hardens
within 2 minutes at T=100°C. To investigate a change in the LED performance under these
conditions, LEDs are thermally treated at T=200°C and for different lengths of time. Before
starting the experiment the I-V characteristic is measured. The next measurements are
performed after 2 min at T=200°C, after 3 min at T=200°C up to 5min at T=200°C. Each time
the LED chip is cooled down and the I-V characteristic is measured. The results are shown in
Fig. 4.8, the current is depicted in a logarithmic scale.
1.21.41.61.82.02.22.4
10-10
10-9
10-8
10-7
10-6
10-5
10-4
10-3
10-2
10-1
Current [A]
Voltage [V]
untreated
2 min at 200°C
3 min at 200°C
4 min at 200°C
5 min at 200°C
Sample G H 1
Figure 4.8: Forward current-voltage behaviour after different thermal treatments.
4.5 Experimental details 37
4.5.2 Spectral measurements along the z-axis
Space resolved measurements from the EP and from the top surface have been made. The
spectral range for all measurements is nmnm 674606
≤
≤
λ
which is equivalent to
. eVEeV 045.284.1 ≤≤
The emission from the EP with respect to z has been investigated. Selected spectra are
depicted in Fig. 4.9 and Fig. 4.10. The measurements are made with sample E04 at
x=131.8µm and at Imeas=10mA. The emission from the n-side is depicted in Fig. 4.9 and from
the p-side in Fig. 4.10. The top surface emission is shown in both figures as black curve. Due
to the fact that the location of the active layer is not exactly known, the emission with the
highest intensity is chosen as reference and set to z=0. Positive z-values are directing toward
the p-region. In the low energy region the increase of the intensity is exponential. The
emission also decreases exponentially in the high energy region. The energies E1 and E2 are
the two emission maxima, and their intensities depend on z. For large distances from z=0 the
intensity of the total emission decreases whereas E2 remains constant and the intensity of E1
decreases. Also, the emission for large distances is equivalent to the top surface emission.
Additionally the influence of the drive current on the emission spectrum is investigated. Space
resolved measurements of sample E04 along the z-axis at different drive currents and constant
x are investigated. The forward current is varied from I=1mA to I=20mA in steps of 1mA.
The characteristic of the emission is similar to the characteristic described in the previous
paragraph. Thus only the emission at z=-1µm and at z=-7µm (in the GaP substrate) are shown.
Figure 4.11 depicts the emission at z=-1µm (just below the active layer). By increasing the
forward current Imeas the intensity increases. The intensity maximum at E1 dominates the
emission. E1 also decreases with increasing forward current.
Figure 4.12 depicts the emission at z=-7µm, in the GaP substrate. The emission E2 dominates
the spectral behaviour and E2 decreases with increasing forward current.
4 Micro electroluminescence - results and discussion 38
Figure 4.9: Electroluminescence along the z-axis in the n-type region. The black curve is the
emission from the top surface. The active layer is at z=0, negative z points
toward the n-contact.
1.84 1.88 1.92 1.96 2.00 2.04
Intensity [a. u.]
Energy [eV]
z = 0,2 µm
z = -1,5 µm
z = -2,6 µm
z = -3,7 µm
z = -4,8 µm
z = -13,8 µm
Surface
E1E2
Sample E04
Figure 4.10: Electroluminescence along the z-axis in the p-type region. The black curve is the
emission from the top surface. The active layer is at z=0, positive z points
toward the p-contact.
1.84 1.88 1.92 1.96 2.00 2.04
Intensity [a. u.]
Energy [eV]
z = 0,2 µm
z = 1,9 µm
z = 4,1 µm
z = 9,7 µm
z = 20,4 µm
Surface
E1E2
Sample E04
4.5 Experimental details 39
1.84 1.88 1.92 1.96 2.00 2.04
I = 1 mA
I = 2 mA
I = 3 mA
I = 6 mA
I = 12 mA
I = 20 mA
Intensity [a. u.]
Energy [eV]
E1
E2
Sample E04
Figure 4.11: Emission at different drive currents at z=-1µm, just below the active layer.
1.84 1.88 1.92 1.96 2.00 2.04
Intensity [a. u.]
Energy [eV]
I = 6 mA
I = 12 mA
I = 20 mA
E1
E2
Sample E04
Figure 4.12: Emission in the GaP substrate at z=-7µm at different drive currents.
4 Micro electroluminescence - results and discussion 40
Figure 4.13: Emission at z=-1µm, below the active layer and at different x-values.
1.84 1.88 1.92 1.96 2.00 2.04
Intensity [a. u.]
Energy [eV]
x = 0 µm
x = 50,2 µm
x = 83,7 µm
E1
E2
Sample E04
Figure 4.13 depicts the dependence of the emission spectra for different x-values. The
forward current is Imeas=10mA and z=-1µm, x=0 is the edge of the LED chip. The intensity
decreases at the edge of the LED and E1 increases.
4.6 Discussion of the results
4.6.1 Glue hardening
This paragraph discusses the influence of glue hardening on the LED performance. Figure 4.8
shows the I-V characteristics after different thermal loads as described in chapter 4.5.1. For
voltages V >1.2V an exponential increase is observed up to V=1.8V. Then the serial resistance
dominates the characteristic and the device acts like an ohmic contact. No significant change
in the I-V characteristics can be observed during this experiment.
Between V=1.2V and V=1.8V the serial resistance is slightly different, but no correlation with
the degradation time can be observed. The reason for this behaviour is a variation in the
contact resistance. Before every measurement the sample must be contacted by a tip probe.
4.6 Discussion of the results 41
Different pressures of the tip causes different contact resistances. Therefore, we conclude, that
the performance of the LED is not affected by the gluing process.
4.6.2 Micro EL along the z-axis
Figure 4.9 shows that the emission spectrum is a superposition of two lines with maxima at E1
and E2. To fit the emission spectra Eq. (2.13) has to be applied for each line. Thus the fit
function is modified to:
() ()()
()()
[]
(
)
(
)
()()
[]
2
2
22
2
1
11
1exp
exp
1exp
exp
+−
+−
+
++−
+−
=EEab
eEEa
N
A
eEEab
eEEa
N
A
Eg (4.6)
where
2
1
12
2
12
1
−
−
=
b
b
b
N
b
and ab
b
e
−
=12
1
ln
A1 is the intensity at E1 and A2 is the intensity at E2.
Due to the strong correlation of the parameters in Eq. (4.6) both terms use the same a and b
parameters. The values of the energies E1 and E2 are extracted from an emission spectrum
with pronounced energy peaks and set to constant values. Thus one obtains the intensities A1
and A2 in dependence of the z-position. Figure 4.14 depicts the intensity A1 of the maximum
E1 and the intensity A2 of maximum E2 as a function z.
4 Micro electroluminescence - results and discussion 42
Figure 4.14: Intensity A1 of maximum E1 and intensity A2 of maximum E2 as function
-12 -8 -4 0 4 8 12 16 20
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
A1
A2
Intensity [a. u.]
z [µm]
Sample E04
n-side p-side
of distance z from the active layer.
The origin of the z-axis is the position of the active layer which corresponds to the intensity
maximum A1. Positive values of z point to the p-contact. The intensity is plotted over a space
interval mzm
µ
µ
2114
+
≤≤− . For z<-5µm the intensity A1 increases slightly (5)
with increasing z. For z>-5µm the intensity A
13
10 −−
⋅m
µ
1 strongly increases up to z=0. Then the intensity
decreases strongly up to z=4µm. After this point the slope flattens.
The behaviour of intensity A2 at E2 is quite different. It is nearly constant throughout the
whole range of the measurement. From z=-4µm to z=-1µm the intensity decreases linearly and
then increases linearly at z=2µm to the same level of intensity.
4.6.3 Modelling A1 and A2 versus z
The area under the emission spectrum of an LED is understood as the emitted energy per solid
angle. Since with increasing intensity the shape of the emission spectrum does not change, the
peak intensity and the area under the spectrum are proportional. Therefore one can investigate
the behaviour of the peak energy along the z-axis instead of the integral of the emission
spectrum. The active layer (AL) inside the LED chip is a volume V with the length xa, the
4.6 Discussion of the results 43
width ya and the depth za. The coordinates x, y, z are similar to the coordinate system
introduced in chapter 4.1. Due to the flatness of the active layer, za is much smaller than xa
and ya, thus xa and ya>>za. Because of the light generation by spontaneous recombination,
every part volume of the active layer emits the same energy per solid angle. The detector
focuses on the active layer with a surface of πr2 . Thus the detector is measuring the volume
π
r2za(1+tan
β
NA) which has the geometry of a truncated cone. Angle
β
NA is the angle of the
observation cone with respect to the z-axis. Light rays with an angle larger then
β
NA are not
detected. The numerical aperture A of the set-up is defining
β
NA, (A=nairsin
α
NA).
Active layer
Detection cone
Figure 4.15: Observation cone inside the LED chip. The cone (hatched) divides the active
layer (red).
The light is refracted at the semiconductor-air interface and
β
NA is given by the law of
Snellius:
NAGaPNAair nnA
β
α
sinsin == (4.7)
where nair is the refraction index of air and nGaP is the refraction index of the semiconductor
material. The light intensity which is measured by the detector is proportional to the
intersection of the active layer and the observation cone. During the measurement the
observation cone is moved through the active volume (where the light is generated) along the
z-axis. Therefore the detected light is generated in a volume which has the geometry of conic
sections, (see Fig. 4.15). The observation cone is the hatched area inside the LED chip. The
active layer inside the LED chip is red cross hatched. By moving the cone (in z direction)
over the active layer the intersection increases or decreases and more or less light is detected.
4 Micro electroluminescence - results and discussion 44
The intersection can be approximated by triangles and then the intensity as a function of z is
given by:
()
22
2
22 ln
zRR
z
R
z
zRzIx−+
−−∝ (4.8)
where R is the radius of the cone base area. The derivation of Eq. (4.8) is shown in appendix
B. Thus the behaviour of A1 near the active layer is described by Eq (4.8). We conclude that
the emission E1 is homogeneously generated in the active layer of the LED chip. The intensity
A1 is a function of the intersection between the active layer and the detection cone inside the
LED chip. Therefore the intensity A1 is a function of the z-position of the in-coupling fibre.
However, surface roughness and scattering effects affect out-coupling of light at large
distances from the active layer. Thus the intensity cannot be described by Eq (4.8) for large
distances.
The intensity A2 at E2 remains constant over the measurement range. The small decrease of
intensity around the active layer is neglected at this point and discussed later. The reason for
the emission around E2 must be different to the emission around E1 due to their fundamental
differences. The light of A2 can be generated in the active layer, but must then escape with a
precipitous angle to the surface normal. Then the light is uniformly distributed in the whole
chip. Surface roughness and scattering allow some light to escape causing a homogenous light
distribution at the surface as a consequence. Self absorption at the active layer can cause such
a precipitous angle with respect to the surface normal, but this effect seems to be too weak.
Alternatively the light could have been generated in the GaP substrate and the GaP window
layer. This would directly lead to homogenous light distribution. Due to the larger bandgap of
GaP in comparison to the AlInGaP alloy, the emission of E2 is at higher energies. This is
shown in Fig. 4.10.
In both cases the reduction of A2 near the active layer is explained by the self absorption in the
active layer. The bandgap in the active layer is smaller than in the GaP substrate and the
window layer. Therefore the self absorption is larger in the active layer. The probability of
self absorption increases with the distance the photons cover in the active region. Thus the
value of A2 decreases over the active layer.
Both models cannot explain the energy value of E2. There cannot be two transitions energies
in the active layer. Also E2 is too small to be caused by impurities in GaP. The bandgap of
4.6 Discussion of the results 45
GaP is eV [8] and E2.2=
GaP
G
E2<1.97eV. Therefore the lattice impurity must have an
activation energy of EA>230meV. Also they must be homogeneously distributed and the
concentration of these impurities must be rather high. Doping materials are in concentrations
which are sufficient to generate light with this intensity. However, the activation energy of the
doping materials is rather low, Mg has an activation energy in GaP of EA=53.5meV [24]. The
emission E2 could be generated in the confinement layers (CFs). The energy gap could be in
the range of E2. But a homogenous light distribution cannot be explained by a light generation
in the CFs.
4.6.4 Current dependence of the peak energies
In chapter 4.5 it is shown that the energies E1 and E2 shift to lower energies by increasing the
forward current. Figure 4.16 depicts the shift of E1 (AL) versus the drive current in a range of
1mA<Imeas<20mA. The characteristic is linear, plotted as black curve in the diagram and can
be fitted by:
A
eV
dI
dE 38.0
1−= (4.9)
This red shift is a temperature effect. Every real diode has a serial resistance Rs where a part
of the applied electrical power is converted to heat. By increasing the electrical power the
produced heat also increases and the temperature at the device rises. The lattice constant of
the semiconductor increases by heating up. The bandgap energy EG decreases by increasing
lattice constant and the emission shifts to lower energies. Therefore AlInGaP LEDs shift to
the red with increasing forward current. This characteristic can be quite different in other
semiconductor materials. For example see [7] for blue emitting GaN LEDs.
Also the increased electron-phonon interaction leads to a band-gap shift to lower energies.
The band-gap shift due to electron-phonon interaction in AlInGaP is not known. But in [23]
the band-gap shift in GaAs/AlAs superlattices is measured to 198
−
=
dTdE µeVK-1.
4 Micro electroluminescence - results and discussion 46
0 2 4 6 8 101214161820
1.918
1.919
1.920
1.921
1.922
1.923
1.924
1.925
1.926
Energy E1 [eV]
Current [mA]
Figure 4.16: Peak energy E1 versus forward current at z=-1µm.
The carrier temperature inside the LED chip can be calculated by the exponential decay of the
emission (see chapter 4.4). A temperature calibration can be made by assuming the carrier
temperature to be at room temperature at I=2mA. Increasing forward current causes an
increasing carrier temperature and leads to a decrease of the E1 peak energy. Figure 4.17
depicts the decrease of peak energy versus carrier temperature. The temperature range is
and a linear behaviour is observed which can be described by: CTC °≤≤° 6020
K
meV
dT
dE 16.0
1−= (4.10)
Combining Eq (4.9) and Eq (4.10) one obtains the relation between carrier temperature and
forward current. The result is shown in Fig. 4.18. A linear characteristic is obtained which can
be described by:
mA
K
dI
dE 6.1
1= (4.11)
A temperature dependence of E2 could not be observed because of the low intensity of this
emission.
4.6 Discussion of the results 47
20 25 30 35 40 45 50 55 60
1.918
1.919
1.920
1.921
1.922
1.923
1.924
1.925
1.926
Energy E1 [eV]
Carrier temperature [°C]
Figure 4.17: Peak energy of the emission versus carrier temperature.
0 2 4 6 8 10 12 14 16 18 20 22
20
25
30
35
40
45
50
55
60
Carrier temperature [°C]
Current [mA]
Figure 4.18: Increase of the carrier temperature as function of the LED forward current.
4.6.5 Measurements along the x-axis
The emission shift as a function of the x-coordinate was also measured to investigate the
current homogeneity as a function of x. In chapter 4.5 we have demonstrated that the emission
shifts towards lower energies by moving the point of measurement towards the centre of the
LED. Figure 4.19 depicts the change of E1 from the edge of the LED chip (x=0µm) to the
centre of the LED chip (x=90µm). We assume an inhomogeneous current distribution inside
the LED chip due to the circular p-side contact. The p-side contact covers only a part of the
LED surface and the n-side contact covers the hole back side of the LED chip. Due to the
4 Micro electroluminescence - results and discussion 48
distance between the active layer and the p-contact, the current distribution in the active layer
is not homogeneous. In the centre a higher current density is expected. Thus we assume a
higher temperature in the centre of the LED than at the edges. This leads to a red shift of the
emission as described in chapter 4.6.5
0 20406080
1.917
1.918
1.919
1.920
1.921
1.922
1.923
1.924
1.925
1.926
Energy E1 [eV]
x [µm]
Figure 4.19: Peak energy E1 as function of the x-coordinate. A red-shift is observed when
moving the point of measurement from the centre to the edge of the LED.
The edge plane emission of the LED chip strongly depends on the z- position. Near the active
layer the emission is dominated by E1. The energy E1 depends on the device temperature and
therefore also on the current through the device. The energy of E1 shifts to lower values by
increasing the drive current. Temperature analysis can be done calculating the high energy
decay of the emission. The intensity of E1 is a function of z. At z=0 - at the active layer - the
intensity is at maximum. The emission E1 also depends on the x-coordinate. At x=0 at the
edge of the LED the intensity of E1 is lower and the energy value of E1 is higher. A higher
current density and thus a higher temperature in the centre of the LED causes this
characteristic.
Energy E2 dominates the emission for larger distances from the active layer. Also E2 is a
function of the drive current and shifts to lower energies by increasing drive current. The
intensity of E2 is almost independent of z. Only near z=0 the intensity decays.
5 Effects of LED aging
5.1 Degradation under high forward current
In this chapter we describe the change of the LED properties aged under high forward current.
In paragraph 5.2 we focus on the positive aging effects. The characteristics of five different
samples treated with two different currents is investigated. Paragraph 5.3 describes the
characteristics of two LED chips treated at a forward current of Idegra=120mA. The treatment
took as long as 704h and the focus of the investigation is on the negative degradation effects.
5.2 Positive aging
During the degradation experiments an increase of luminous flux is observed which is defined
as “positive aging”. This positive aging during the first few hours of aging time leads to an
increase of luminous efficiency and therefore to a higher light output. In addition, an
improved current-voltage characteristic in forward direction below the threshold voltage is
observed. Also a decreased leakage current is measured. The samples listed below are
investigated and the results are discussed in detail.
Sample 1: Resin encapsulated LED Idegra=70mA, #33_C1
Sample 2: Chip only Idegra=120mA, LZ-EL 1
Sample 3: Chip only Idegra=120mA, LZ-UL 2
Sample 4: PMMA encapsulated LED at Idegra=120mA, soldered chip, H G 1
Sample 5: PMMA encapsulated LED at Idegra=120mA, glued chip, H G 2
5 Effects of LED aging 50
0.1
1
10
110
Log luminous flux [a. u]
Time [h]
I=70mA, T=151°C
I=40mA, T=85.1°C
I=20mA, T=44.7°C
I=10mA,T=30.1°C
I=5mA, T=24.6°C
sample 1
100
Figure 5.1: Sample 1, luminous flux versus time and different measurement currents. Burn-in
behaviour and positive degradation with increasing luminous flux for
is observed. mAImeas 5≥
The LED #33_C1 is treated with a forward current of I=70mA which causes a temperature of
T=151°C at the junction of the device. From time to time the degradation is interrupted and
the device is cooled down to room temperature. After cooling down the device is driven by
forward currents of Imeas=5mA, 10mA, 20mA, 40mA and 70mA and the total luminous flux is
measured in an Ulbricht sphere. After the measurements the degradation is continued at
Idegra=70mA. The experiment is stopped after t=65h. Figure 5.1 shows the results. The
luminous flux versus aging time in a logarithmic scale is plotted. Clearly an exponentially
increasing luminous flux is observed for measuring currents . In Fig. 5.2 the
normalized luminous flux versus the measurement current is depicted.
mAImeas 5≥
For a measurement current of Imeas=5mA the light output is increased by a factor of two after
an annealing time of t=65h at Idegra=70mA. The efficiency improvement decreases by
increasing measurement current. At Imeas=20mA the enhancement is only 30% and then seems
to saturate. At I=40mA the improvement of light output is only 25% which is the same as at
Imeas=70mA within the experimental error.
5.2 Positive aging 51
100
120
140
160
180
200
220
0 102030405060708
Luminous flux at t=65h/t=0
Forward current [mA]
sample 1
t=65h
0
Figure 5.2: Luminous flux versus forward current of sample 1. Idegra=70mA,after t=65h.
By increasing the degradation current the annealing process accelerates. To investigate this
characteristic and to show the influence of the housing on the LED, two non-encapsulated
LED chips are degraded under a forward current of Idegra=120mA for several hours. Also the
I-V characteristic is measured and analysed. Figure 5.3 depicts the measured luminous flux
versus aging time. The measurement current is Imeas=10mA.
0.9
1
1.1
1.2
1.3
1.4
1.5
1.6
0.1 1 10 100
Luminous flux [a. u.]
Time [h]
Sample 2
Sample 1
Figure 5.3: Sample 2, normalised luminous flux versus aging time at Idegra=120mA. The
increase of luminous flux is 54% after t=16h.
5 Effects of LED aging 52
An increase of the luminous flux of 54% is observed after an annealing time of t=16h (Fig.
5.3). In comparison to sample 1, the increase of the intensity is accelerated. We find an
increase of the luminous flux of 50% for sample 1 at Imeas=10mA. This is the same value as
for sample 2 within the experimental error.
To prove the annealing effect and the reduction of non-radiative recombination centres during
the first hours of degradation, I-V curves in forward direction are investigated. Most diodes
have unwanted parasitic parallel resistance. This bypass can be caused by defects in the
crystal structure where non-radiative recombination occurs. At high voltages these defects are
saturated and the current flow is dominated by radiative recombination in the depletion layer.
Figure 5.4 depicts schematically the equivalent circuit diagram of a real diode with its parallel
and serial parasitic resistances.
Rse
r
Rpar
LED
Figure 5.4: Equivalent circuit of a real diode with its serial and parallel resistance.
Therefore the I-V characteristic in the low voltage regime must be investigated. Reducing the
defects in the crystal structure leads to an increase of the parallel resistance which increases
the total resistance. Thus the current flow through the device is decreased [7]. This decrease in
current flow is clearly shown in the I-V characteristic at sample 2. Figure 5.5 depicts the result
of this measurement before annealing (red curve) and after 14hours of annealing time (blue
curve) and after 40hours (green curve). The reduction of current flow at V=1.5V is 19% and
marked with arrows. This result indicates a clear improvement of the electrical properties of
the device by annealing.
The increased light output and the decrease of the parallel shunt resistance leads to the
conclusion that the crystal structure is improving during the first hours of the degradation
process.
5.2 Positive aging 53
4 10-10
6 10-10
8 10-10
1 10-9
1.2 10-9
1.4 10-9
1.6 10-9
1.8 10-9
2 10-9
00.511.52
0h
14h
40h
Current [A]
Voltage [V]
sample 2
Figure 5.5: I-V characteristics of sample 2 before annealing and after 14h and 40h of
annealing. The diode was treated by a forward current of Ideagra=120mA. A
reduction of the parallel shunt resistance is clearly observable.
1.55
1.60
1.65
1.70
1.75
1.80
0.1 1 10 100
Intensity [a. u.]
Time [h]
sample 3
Figure 5.6: Intensity versus degradation time of sample 3 at Imeas=40mA.
The burn-in behaviour of sample 3, a non-encapsulated LED Chip, has also been investigated.
The result is depicted in Fig. 5.6. In contrast to sample 2 the intensity is only increased by
10% at Imeas=40mA. This is the lowest value observed during these investigations. The reason
for the large fluctuations in the positive degradation is unknown.
5 Effects of LED aging 54
One reason could be a different lattice defect density due to different production parameters.
Also the thermal treatment during gluing or soldering influences the annealing process.
Figure 5.7: Intensity versus degradation time of PMMA encapsulated LEDs. Blue curve for
glued/bonded sample 5, red curve for soldered sample 4 at Imeas=40mA.
1
2
3
4
5
6
7
0 5 10 15 20 25 30
Intensity [a. u.]
Time [h]
glued, sample 5
soldered, sample 4
Figure 5.7 depicts the intensity versus degradation time of sample 4 and 5. The blue curve is
obtained from a sample which is glued with the n-contact to the frame. The p-contact is
bonded. Details of the experimental set-up are described in chapter 6.1.1. The degradation
current is Idegra=120mA, the measurement current is Imeas=40mA. An increase of the luminous
flux is clearly observed during the first hours of degradation time. The intensity then
saturates, the total increase of light output is about 21%.
The characteristic of the soldered LED is different and given by the red curve in Fig. 5.7. The
degradation and measurement parameters are the same as with sample 5. The total luminous
flux is a factor 3.3 lower in comparison to the bonded sample. Also no increase of light output
during the first 25h of treatment is observed.
We conclude that treating the LED with a high forward current has a positive effect on the
light output. This positive aging is based on the improvement of the crystal structure [25],
[26] and has been demonstrated with LED chips as well as PMMA encapsulated LEDs. Also
an improvement of the electrical characteristic has been demonstrated, the parallel shunt
resistance has been reduced.
Soldered LEDs did not show positive aging during our investigations. We believe that the
positive aging is effected by the thermal treatment during soldering.
5.3 Long term aging 55
5.3 Long term aging
Two LED chips are treated under a forward current of Idegra=120mA. From the first LED the
total luminous flux at Imeas1=5mA, Imeas2=10mA, Imeas3=20mA, Imeas4=40mA and Imeas5=70mA
is measured. Also the I-V characteristic versus degradation time is measured. This sample is
called LZ-UL 2. The second sample is called LZ-EL 2 and the surface emission spectra at
similar currents are measured during the degradation. Also the I-V characteristic is
investigated. Both LEDs are non-encapsulated chips. The p-contact is realised by a tip probe.
The tip remains on the contact during the degradation experiment. This avoids variations in
the serial resistance. The n-contact is glued to a copper frame by using glue XCA 80210
described in chapter 4.2. This glue grants a perfect electrical and thermal conductivity from
the semiconductor to the copper frame. The sample LZ-UL 2 is aged over a period of t=704h
and sample LZ-UL 2 over a period of t=230h. First we investigate the I-V characteristics of
these samples. Figure 5.8 depicts the forward current versus forward voltage characteristics
over the degradation time of sample LZ-UL 2. The serial resistance increases clearly over the
degradation time.
Figure 5.8: Current voltage characteristics of a LED at various degradation times.
1.8 2.0 2.2 2.4 2.6 2.8 3.0
0
5
10
15
20
25
30
35
40
Current [mA]
Forward voltage [V]
t = 0h
t = 1h
t = 6h
t = 124h
t = 354h
LZ-UL 2
The serial resistance increases with the degradation time.
5 Effects of LED aging 56
This characteristic is shown in Fig. 5.9 where the time is plotted in a logarithmic scale. The
data are fitted by:
()
=
0
0ln t
t
RtRS (5.1)
The parameters are calculated to R0=0.27Ω and th
33
01038.1 −
⋅=
Figure 5.9: Serial resistance of sample LZ-UL 2 versus aging time. The black line is a
0.1 1 10 100 1000
20.0
20.5
21.0
21.5
22.0
22.5
23.0
Measurement
Fit
RS [Ω]
Time [h]
LZ-UL 2
logarithmic fit of the experimental data.
Also the total luminous flux of sample LZ-UL 2 is measured in an Ulbricht sphere. The
measurement currents are Imeas1=5mA, Imeas2=10mA, Imeas3=20mA, Imeas4=40mA and
Imeas5=70mA. The results are depicted in Fig. 5.10. The luminous flux is given in arbitrary
units versus the degradation time. During the first few hours an increase of the luminous flux
is observed as described in the previous paragraph. After t=120h the luminous flux starts to
decrease and remains constant after t=380h. The increase of efficiency at the beginning of the
treatment is an effect of thermal annealing. This effect of structural improvement by heating
the semiconductor is well known and described for example in [25] and [26]. The effect
seems to be exhausted after t=120h and degradation starts to reduce the efficiency. Details
and modelling of the degradation will be given in chapter 7.
5.3 Long term aging 57
Figure 5.10: Luminous flux versus aging time measured at different currents. After t=120h
1 10 100
0.1
1
10
log luminous flux [a. u.]
Time [h]
I1 = 5 mA I3 = 20 mA I5 = 70 mA
I2 = 10 mA I4 = 40 mA
LZ-UL 2
the luminous flux starts to decrease.
Improvement in efficiency is observed for all measurement currents, but is more significant
for low forward currents.
Figure 5.11 depicts the emission spectra of the non-degraded LED at different measurement
currents. The intensity is plotted in a logarithmic scale versus the energy. By increasing the
forward current the intensity increases and the peak energies E1 and E2 decrease. At
Imeas=70mA the peaks E1 and E2 cannot be distinguished any longer. The peaks E1 and E2
result in a peak energy E0. Figure 5.11 shows the emission after t=54h of aging at
Idegra=120mA. By comparing the Imeas=70mA measurements from Fig. 5.11 and Fig. 5.12 one
sees that the intensity is increased and the peak emission E0 is shifted to higher energies. By
fitting the emission with Eq. (2.13) the intensity and the energy of Emax can be calculated. For
low currents Emax is equal to E2. For high currents Emax is equal to E0.
5 Effects of LED aging 58
1.84 1.88 1.92 1.96 2.00 2.04
Intensity [a. u.]
Energy [eV]
I = 5 mA
I = 10 mA
I = 20 mA
I = 40 mA
I = 70 mA
LZ-EL 2
Figure 5.11: Emission spectra before degradation for various forward currents. The
intensity increases by increasing the forward current thus shifting the peak to
lower energies.
Figure 5.12: Emission after 54h of aging at Idegra=120mA.
1.84 1.88 1.92 1.96 2.00 2.04
Intensity [a. u.]
Energy [eV]
I = 5 mA
I = 10 mA
I = 20 mA
I = 40 mA
I = 70 mA
LZ-EL 2
Figure 5.13 depicts the increase of intensity as well as the increase of the peak energy for
different forward currents versus aging time. The intensity is plotted in black, the peak energy
in red.
5.3 Long term aging 59
Figure 5.13: Maximum intensity (black dots) and energy (red dots) versus degradation
0.1 1 10 100
I1 = 05 mA I4 = 40 mA
I3 = 20 mA I5 = 70 mA
Intensity [a. u.]
Time [h]
LZ-EL 2
1.935
1.940
1.945
1.950
1.955
1.960
1.965
1.970
1.975
Emax [eV]
time for different drive currents.
In this chapter we have described the change of the LED chip properties aged under high
forward current. An increase of the serial resistance versus the degradation time is observed,
which is attributed to degradation of the ohmic contact on top of the p-layer [27].
The luminous flux versus degradation time has been measured in an Ulbricht sphere. During
the first hours of degradation an increase of luminous efficiency is observed which is
attributed to an improvement of the crystal structure. The atomic motion is increased by
higher device temperatures thus a self arrangement improves the lattice purity and reduces the
dislocation density. The increase of the peak energy is also an annealing effect. By reducing
the defect density the efficiency of the device is increased. Thus the temperature of the LED
chip decreases during the annealing period. A decreasing device temperature leads to a blue
shift in the emission. The maximal energy shift is measured to
∆
E=10meV.
The decrease of luminous intensity can be explained by the generation of non-radiative
recombination centres in the active layer.
5 Effects of LED aging 60
6 Injection molding of LED clusters
6.1 General set-up and temperature measurements
A standard method to encapsulate LEDs is casting them in resin [28]. The LED chip is
mounted in a reflector cup on a lead frame which in general is formed out of the cathode lead.
The p-contact is normally bonded with a gold wire to the anode lead. The entire device is
encased in epoxy which also serves as a lens. The resin hardens due to a chemical process.
This process has some advantages. It is easy to handle and the temperature load during the
hardening process is negligible. The development of non-shrinking epoxy resins avoids
pressure pile-up at the devices. Limitations are the very high thermal resistance Rth>200 K/W,
the limited temperature stability of epoxy resin TEpoxy<120°C and therefore a limited input
power of PIn<0.1W. Also there are limitations in forming the geometry of the housing. With
respect to lightweight and miniaturization, especially in the automotive industry, external
optical components to tailor the light beam should be avoided. Using external optical devices
leads to an adjustment process where the optical device has to be adjusted against the light
source. Also every optical unit has an interface which leads to an intensity reduction due to
reflection. This means a direct encapsulation process which combines the housing and the
optic element in one unit is necessary. Those processes are impossible with resin as an
encapsulation material.
By contrast direct encapsulation of LEDs by an injection molding tool is a useful way to
control the optical geometry to produce the required radiation pattern, whilst also protecting
the semiconductor against environmental influences. To investigate the injection molding
encapsulation process, an injection molding tool has been developed. It meets the following
requirements: The frame which is described later in this chapter should be easily insertable in
the tool. Is has been taken into account that in a future step of automation the frame should be
replaced by an endless band of copper. The possibility to pre-encapsulate the LED chips with
a small sphere at low pressure before creating the optic must be provided. To meet these
requirements a split mold seems to be the best approach. The cavity insert is removable to
6 Injection molding of LED clusters 62
pre-encapsulate the LED chips in a first step for protection reasons. In a second step the optic
is formed at high pressure with a different mold insert. All surfaces in the cavity of the tool
are polished to obtain perfect optical surfaces. Figure 6.1 is a drawing of the set-up of the
injection molding tool. Marked in red is the runner which is divided by four and leads to the
individual LED chips. This is the approach to encapsulate each chip under low pressure with a
small sphere. Further down it is shown that this pre-encapsulation process has no effect on the
failure rate. Therefore it has only been performed in the early steps of the investigation.
Figure 6.2 depicts the tool with the cavity which forms the parabolic shaped optics. The sprue
is realised by a single channel located at one site allowing to fill the cavity at high pressure.
High pressure, one should say high backpressure, is necessary to avoid shrinking of the
module. Shrinking would lead to a deformation of the polymer and a less effective optic.
Figure 6.1: Drawing of the tool, set-up for “bubble in” the LED chips at low pressure.
Source: Hella
Figure 6.2: Drawing of the tool, set-up for optic production at high pressure. Source: Hella
6.1 General set-up and temperature measurements 63
For indirect encapsulation both steps (bubble in and optic creation) must be carried out. For
direct encapsulation only the second step (Fig.6.2), is necessary. Our experiments have clearly
shown that the two-step process has no advantages compared to direct encapsulation.
Therefore after the first experiments, all investigated samples are encapsulated using the one-
step method. Figure 6.3 is a view directly into the cavity of the tool. The slide bars form the
parabolic optics, the frame is loaded from the top and a polished stamp close the cavity.
During the closing process the slide bars shut under a certain angle and fix the copper frame.
It acts as a gasket for the liquid polymer in this situation. This means there is a mold parting
surface just between the two slide bars. The slide bars and the stamp is water-tempered to
guarantee stable conditions.
Figure 6.3: Top view into the cavity of the injection molding tool. The slide bars are wide
open to load the cavity with the frame. The investigations were performed on an
Arburg A380 injection molding machine.
6.1.1 The Module (molded part)
The module consists of two parts. A copper web was chosen as a basis, it has been metalised
with nickel and gold. The nickel layer acts as diffusion barrier and has a thickness of d=10
µm. Copper has a rather high diffusivity in semiconductor materials, therefore the nickel layer
is necessary. To improve the ohmic contact between the web and the LED, a gold layer with a
thickness of d=100nm has been deposited on the nickel layer. Both layers have been
deposited in a galvanic bath. The LED chips are directly glued onto the gold.
Figure 6.4 shows the geometry of the web chosen for die/wire bonding. Four LEDs are placed
on every frame.
6 Injection molding of LED clusters 64
1mm
Figure 6.4: Copper web for die/wire bonding. Asymmetry of the electrode guarantees that the
LED chips are in the focal point of the parabolic optic.
The LED chips are glued with their anode on the “nose”. This asymmetric set-up is necessary
to place the LED chips in the focal point of the parabolic optic.
Figure 6.5 shows the geometry of the web chosen for die/die bonding.
1mm
Figure 6.5: Copper web used for die/die bonding. Two symmetric noses with a gap of
d=100µm. The chip is placed with its active layer over the gap.
For die/die bonding both sides of the chip are metalised. Then the chip is turned by 90° and
placed over the gap of the symmetric noses. The active layer has to be over the gap to avoid a
short circuit of the device.
Detailed description of the two different kinds of bonds which have been tested:
(a) Die/wire bonding (Fig. 6.6): The LED chip is glued with its anode to the frame (die bond)
and the cathode is bonded with a gold wire, diameter d=25µm, to the frame (wire bond).
200
µm
Figure 6.6: Frame with die/wire bonded AS AlInGaP LED chip (X-ray image).
6.1 General set-up and temperature measurements 65
(b) Die/die bonding (Fig. 6.7): Both sides of the chip (anode and cathode) are metallised. The
chip is turned by 90° in comparison to (a). Both sides of the chip are soldered in an oven.
In Fig. 6.7 a vertical cut through the die/die bonded transparent substrate (TS) chip shows the
principle of die/die bonding. Clearly visible is the LED chip and the two soldering points. The
chip is well positioned and has good contact to the soldering material.
Figure 6.7: Die/die bonded TS LED chip.
200
µm
The yellow strip is the active layer which is located near the p-contact (left side). Typically
for TS chips is the red-shining substrate. With this bonding method no light could escape
through the contact region.
6.1.2 Optics:
The system consists of four contiguous, parabolic shaped total reflection optics. The focal
length is f=1,1mm. Figure 6.8 shows the principle of total reflection in one parabolic optic.
Figure 6.8: Total reflection optic (principle).
The light beam is tailored to a narrow beam with a spread angle of
α
=10°. Detailed radiation
patterns are given in chapter 6.3. The material is PMMA 7N which is amorphous. The
refraction index is n=1.49, the glass transition temperature is T=106°C [29]. This material is
commonly used for optical parts such as lenses, but also for car tail-lights. Therefore it is a
proven material for this application.
6 Injection molding of LED clusters 66
In Fig. 6.9 the complete module in two different viewing angles is illustrated.
Figure 6.9: LED module in two different viewing angles.
Figure 6.10 shows the final LED module consisting of the frame and the four parabola-shaped
optics with the semiconductor light sources inside. The length of the device is l=32mm, the
height h=9mm, and the depth d=8mm.
4mm
Figure 6.10: Optic module.
6.1.3 Temperature measurements
To investigate the thermal load of the semiconductor during the encapsulation process the
temperature of the polymer melt must be measured at the LED location. Therefore a
temperature sensor which can replace the LED is necessary. Standard temperature sensors for
injection molding are not suitable for this application. They are bulky and compared to the
cavity need a rather big feed through which could not be realized in the tool. They have a
relatively high heat capacity and therefore a time constant which exceeds the time constant of
semiconductor temperature sensors. Semiconductor temperature sensors are widely used for
different applications and temperature ranges. This could be an alternative due to the small
sensor size, the small heat capacity and the easy availability. In the following we show that
the LED chip itself could be used as a temperature sensor and that it fulfills all the
requirements. The LED itself can be used as a temperature sensor due to the fact that the
forward voltage of a pn-junction is a function of temperature at a constant current. According
to [7], the temperature depending voltage can be expressed by:
6.1 General set-up and temperature measurements 67
e
TE
I
I
e
kT
TV g
S
)(
ln)( += (6.1)
where V(T) is the forward voltage, k the Boltzmann constant, T the temperature, e the
elementary charge, I the operation current, Is the saturation current and Eg(T) the energy gap.
The temperature dependence of the Fermi level is described by the first term on the right-hand
side. This level moves towards the middle of the gap with increasing temperature. The energy
gap Eg(T) is the distance between the conduction band and the valence band. Hence,
increasing temperature results in a smaller gap and therefore leads to a lower forward voltage.
According to [30], the energy gap can be expressed by Eq. (6.2). The parameters
α
and
β
depend on the particular semiconductor material.
β
α
+
−= T
T
ETE gg
2
0,
)( (6.2)
where Eg,0 is the energy gap at 0K and
α
and
β
are the Varshni parameters [28]
Figure 6.11 illustrates the energy gap EG of GaP calculated with Eq. (6.2).
2.10
2.15
2.20
2.25
2.30
2.35
0 100 200 300 400 500 600 700
E gap[eV]
Temperature [K]
Figure 6.11: Band gap versus temperature of GaP.
Polymer melts are typically processed in a range between 400K to 650K.
6 Injection molding of LED clusters 68
1.40
1.45
1.50
1.55
1.60
1.65
1.70
1.75
1.80
0501001502002
Forward voltage [V]
Temperature [°C]
50
KmV
dT
dU I/8.1
|−=
Figure 6.12: Forward voltage versus temperature at I=1mA for AlInGaP LEDs.
Figure 6.12 shows the forward voltage versus temperature characteristic of the AlInGaP
LEDs used in the current investigations, the drive current is I=1mA. The slope of the linear fit
in Fig.6.12 is dV/dT=-1.8 mV/K.
To ensure the accuracy of the measurement, the influence of thermal load on the calibration
curve has been investigated. For this reason, five samples (Temp. 1-5) have been annealed in
an oil bath at 205°C for different durations (1 to 5 minutes). The subsequent measurements of
the temperature sensitivity depicted in Fig. 6.13, do not show much difference between the
samples which emphasizes the suitability of LED chips for temperature measurements.
1.40
1.45
1.50
1.55
1.60
1.65
1.70
1.75
1.80
0 50 100 150 200 250
0min
3min
5min
Forward voltage [V]
Temperature [°C]
Samples: Temp 1, 3 and 5
Figure 6.13: Forward voltage versus temperature of three different LEDs degraded at 0min,
3min and 5min at T=205°C.
6.1 General set-up and temperature measurements 69
To summarize, the LED chip temperature sensor is based on the voltage variation inside a
semiconductor caused by the temperature change during direct encapsulation. Due to the low
heat capacity of the device, the melt temperature is almost unaffected and a fast response is
achieved. This method enables the user to determine the thermal load of directly encapsulated
electrical devices and thereby to estimate the thermal degradation of semiconductors.
Furthermore, it could also be possible to measure the temperature distribution within the
molded part using multiple probes.
Figure 6.14 illustrates that the “LED temperature sensor” is placed into the second chamber.
To ensure the electrical supply of the diode, a plug connection has been mounted into the first
cavity on the right side. Beside the mentioned LED chips conventional PT 100 resistors were
used to investigate the temperature variation in the tool.
Temperature Sensor/
LED Chip
Plug / Cable
feed through
Injection
orifice
LED moulds
Figure 6.14: LED and cable feed through arrangement in the cavity of the injection molding
tool.
Experimental Results:
Using the set-up described above, in-situ temperature measurements during the injection
molding process have been performed.
The probes were supplied with constant current and the voltage variation caused by the
temperature changes during the injection and cooling process has been recorded. By means of
the calibration shown in Fig 6.12, the fluctuation of voltage is transformed into a temperature
curve.
In Fig. 6.15 a typical temperature measurement of the LED chip during the PMMA
encapsulation process is shown.
6 Injection molding of LED clusters 70
20
40
60
80
100
120
140
160
0 10203040
Temperature [°C]
Time [s]
50
Figure 6.15: Temperature versus time of the LED chip in the cavity of the injection molding
tool during the encapsulation process. The arrow marks the temperature
increase at the glass transition.
The water-cooled injection molding tool is kept at T=50°C. This results in a starting
temperature of T=40°C as the probe is mounted in the open, moving half of the mold. At
t=9.1s the tool is closed and the temperature in the cavity increases due to the tool
temperature. At t=11.8s the polymer melt reaches the probe and the temperature rises to a
peak of 153°C. Then the temperature decreases, the melt solidifies and the sample is ejected
at t=47s. All PMMA samples containing the LED probe show a small peak (arrow) at a
temperature of about T=80°C. PMMA has a strong decay of its heat capacity during the
cooling process at the glass transition temperature. Thus, intrinsic energy is transformed into
heat during the glass transition process. This characteristic leads to a temperature increase
which has been observed during our investigations. Assuming a decrease in the heat capacity
of , the temperature increases by
∆
T=42°C. Due to the inhomogenous glass
transition in the module we observe a constant temperature instead of a temperature increase.
The results of the PT 100 and the LED probe are compared in Fig. 6.16. The different starting
temperatures are a result of their mounting position: While the LED is glued onto the copper
frame, the PT 100 was mounted by a small rod to the cavity of the molding tool. The diagram
shows a considerable time delay for the PT 100 compared to the LED probe.
KgJCp⋅=∆ /2.0
6.1 General set-up and temperature measurements 71
20
40
60
80
100
120
140
160
180
0 1020304050
Temperature [°C]
Time [s]
60
Tmax PT100=164 °C
Tmax LED=153 °C
Figure 6.16: Time delay of the PT100 temperature sensor in comparison to the LED chip
sensor.
This can be explained by the low specific heat capacity of the LED chip, which is only
cv=0,31J/gK. Nevertheless, the maximum temperature can also be estimated by using the
PT100: Extrapolating the slope of the curve with a linear function and determining the point
of intersection with a vertical line at t=11.5s results in a temperature of T=164°C, while the
maximum temperature measured by the LED is T=153°C.
6.1.4 Simulation of the temperature distribution in the injection molding tool
To verify the temperature measurements presented above, a simulation of the temperature
distribution based on the finite element method (FEM) has been carried out. Therefore the
geometry of the molded part is implemented in the Moldex 3D software, as well as all other
parameters from the Arburg 380 injection molding machine. Figure 6.17 depicts the
temperature distribution in the module just after filling the cavity of the injection molding
tool. The temperature scale in false colour ranges from T=32°C (blue) to T=250°C (red). The
core of the module reaches T=250°C except the upper left corner, where the maximum
temperature is approximately T=200°C. This is an indication, that the thermal load of the
LED chips could by reduced by modifying the position of the injection orifice. The surfaces
of the module are immediately cooled down to the temperature of the injection molding tool
due to its large heat capacity. It is interesting that the LED chip in the right optic is treated
with a higher thermal load as the other three. The temperature gradient from the shell to the
core is higher in this region. Therefore hotter liquid polymer reaches the LED chip. The LED
6 Injection molding of LED clusters 72
chip which behaves (as described above) as a temperature sensor is marked with a black
square. A virtual sensor is placed in the simulation as close as possible to the real position of
the temperature sensor.
virtual sensor real sensor
Figure 6.17: Temperature distribution in the LED module just after filling the cavity of the
tool with liquid PMMA.
Only at the position of the knots of the finite element net such sensors could be positioned.
Therefore the virtual sensor, which is marked with a blue circle, is directly above the real
sensor position. From the simulation the temperature of the LED chip can be determined to
T=150-175°C. This result is in perfect agreement with our measurement depicted in Fig. 6.16.
To estimate the duration of the thermal load, a time-dependent temperature measurement at
the point of the virtual sensor is simulated. The result is shown in Fig. 6.18. The temperature
starts at T=50°C which is the cooling temperature of the molding tool. At t=0s the liquid
polymer reaches the sensor and the temperature rises immediately to T=240°C. The
temperature decreases fast and after t=2s it is below T=100°C. After t=9s the temperature has
dropped to T= 60°C and at this point the module is removed from the tool. The higher peak
temperature of the simulation is due to the fact described above that the virtual temperature
sensor is not at exactly the same position as the real one. The shift of the virtual sensor to the
inner core leads to an increased temperature.
Summarizing the experimental and theoretical results, the thermal load is defined by a peak
temperature of about T=170°C and a subsequent fast decrease of the temperature. Ten seconds
6.1 General set-up and temperature measurements 73
after the peak, the temperature has dropped to T=75°C (experiment) or below T=60°C
(simulation). The thermal load is in the range of a standard gluing process used to glue the
LED chip to the board. In Ref. [14] the maximal thermal load for the LED chip is given by a
maximal temperature of T and a duration of tCC °±°= 5235 ss 5.05.2
±
=
. Our result shows
that the thermal load during the encapsulation process is within these limits.
Thus, in comparison to a soldering process the thermal load during this encapsulation process
is negligible.
Figure 6.18: Simulated temperature versus time at the virtual temperature sensor which is
50
100
150
200
250
02468
Temperature [°C]
Time [s]
10
near the real LED device.
Also, the temperature simulation shows a small temperature increase at t=0.8s which is
attributed to the glass transition.
6 Injection molding of LED clusters 74
6.2 Influence of injection molding on the LED
performance
In this chapter the influence of the injection molding process on the LED chips is
investigated. Before encapsulation the I-V characteristics and the luminous flux at different
drive currents are measured. Additionally the emission spectra at I1=5mA, I2=10mA,
I3=20mA and I4=40mA are recorded. After the encapsulation process the samples are
measured again and the results are compared. The polarization of the emitted light is also
measured after the encapsulation process in order to find polarization effects caused by
strained PMMA. The electroluminescence set-up described in chapter 4 is used as a detector.
A polarization filter is adapted into the light path and the polarisation angle is measured in
relation to the axis of symmetry of the module. Figure 6.19 depicts the set-up schematically.
Figure 6.19: Experimental set-up to investigate the polarisation of the emitted light after the
encapsulation process.
6.2.1 Results and discussion
The current-voltage characteristic of a typical sample is depicted in Fig. 6.20. The current is
plotted in an linear scale. At low voltages an exponential increase of the current with
increasing voltage is observed. At V=1.9V the device starts to behave like an ohmic contact.
Up to V=1.9V no difference before and after encapsulation is observed. After the
encapsulation process the current is increased in the regime of larger forward voltages. This is
clearly depicted in Fig. 6.20 A. For V>2.1V the characteristic is linear and a decreased serial
resistant after the encapsulation process is observe. The serial resistant after encapsulation
6.2 Influence of injection molding on the LED performance 75
decreases by R=8.5Ω in comparison to the chip. The decrease of the serial resistance is caused
by a higher device temperature during operation, after encapsulation. Before encapsulation,
air flows around the chip and it is cooled by convection. After encapsulation cooling by
convection is impeded. Using the temperature calibration in Fig. 6.12 we find an increase of
the junction temperature of the encapsulated chip of about 15°C.
1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0
0
5
10
15
20
25
30
35
40
3.19=
dI
dU
Ω= 8.27
d
I
dU
Current [mA]
Forwad Voltage [V]
before encapsulation
after encapsulation
Sample: Dit 19
Figure 6.20: Current-voltage characteristics before (dashed line) and after (full line) the
encapsulation process.
Additionally, the luminous flux for different drive currents has been measured in an Ulbricht
sphere. Results are shown in Fig. 6.21. The luminous flux is plotted versus the drive current in
a range from 0mA<I<45mA.
0 5 10 15 20 25 30 35 40 45
before encapsulation
after encapsulation
Luminous flux [a. u.]
Current [mA]
Sample: Dit 19
Figure 6.21: Luminous flux as function of the drive current. Squares before encapsulation,
triangles after the injection molding process.
6 Injection molding of LED clusters 76
The total luminous flux at Imeas=45mA is 1.8 times higher compared to the non-encapsulated
chip. As shown in chapter 2.5 the external quantum efficiency can be increased by
encapsulation. If the refractive index of the encapsulation material is between that of air and
the semiconductor material (GaP), the angle of total reflection increases and losses at the
surfaces can decrease. The PMMA used in these investigations has an index of refraction of
nPMMA=1.49. Using Eq. (2.17) the reflectivity at the surface is calculated to:
%5.11
≈
−PMMAGaP
R
%9.3
≈
−AirGaP
R
The transmission from the semiconductor through the PMMA to air is then:
%1.85)1)(1(
≈
−
−
=−−−− AirPMMAPMMAGaPAirPMMAGaP RRT
The absorption in the material is neglected.
Thus the external quantum efficiency of the system is 10.7%
AirPMMAGaP
ext
−−
η
To calculate the theoretical improvement of the external quantum efficiency, the ratio of
external quantum efficiency from GaP to air to the external quantum efficiency from GaP-
PMMA-air must be calculated:
14.1≈==
Φ
Φ
−
−−
−
−−
AirGaP
AirPMMAGaP
AirGaP
ext
AirPMMAGaP
ext
before
after
T
T
η
η
(6.3)
Experimentally this factor has been determined to 1.8. In the calculation above the geometry
of the PMMA optic is neglected. Therefore a lower theoretical value can be expected. Before
encapsulation the angle of total reflection at the surface GaP-air is . After
encapsulation the total reflection is:
°=
−3.19
AirGaP
G
β
°=
−6.29
PMMAGaP
G
β
°=
−8.41
AirPMMA
G
β
If the geometry is not changed, the angle of total reflection for the system GaP-PMMA-air
will remain the same as for the GaP-air system. But in our investigations the geometry does
6.2 Influence of injection molding on the LED performance 77
not remain constant. Therefore the experimentally obtained value is greater than the
theoretical one.
Figure 6.22 depicts the emission spectra before and after the encapsulation process for two
different drive currents. The intensities are normalized for easy comparability. Also included
is the FWHM for Imeas=40mA. As explained in chapter 4.6.6 the peak energy of the emission
shifts to the red spectral range by increasing the forward current. A peak energy shift due to
the encapsulation process could not be observed, but the FWHM increases from
∆
E=58 meV
before to
∆
E=67 meV after the encapsulation process.
Figure 6.22: Emission spectra of a typical LED before and after the encapsulation process
1.90 1.92 1.94 1.96 1.98 2.00 2.02 2.04
Intensity [a. u.]
Energy [eV]
Ivor = 5 mA
Ivor = 40 mA
Inach = 5 mA
Inach = 40 mA
Sample: Dit 19
at different drive currents. The FWHM increases after the encapsulation
process.
Figure 6.23 B depicts the FWHM as a function of the forward current. The FWHM of the
LED emission increases with higher forward current. Additionally the FWHM at a constant
current is larger after the encapsulation process. This is also a temperature-related behaviour.
Because of the absence of convection the encapsulated LED chip has a higher temperature. As
described in [7] the FWHM of the emission spectra broadens with increasing temperature.
This is due to a temperature-related distribution of carriers in the conduction and valence
band.
6 Injection molding of LED clusters 78
0 5 10 15 20 25 30 35 40 45
50
55
60
65
70
75
80
before encapsulation
after encapsulation
FWHM [meV]
Current [mA]
B
0 5 10 15 20 25 30 35 40 45
1.974
1.976
1.978
1.980
1.982
1.984
1.986
1.988
before encapsulation
after encapsulation
Energy E2 [eV]
Current [mA]
A
Figure 6.23: A) depicts the decrease of the emission energy E2 versus the drive current.
B) shows the peak broadening after the encapsulation process.
Figure 6.23 A depicts the decrease of the emission energy E2 versus the drive current. As
expected, the emission energy decreases with higher forward current because of the device
heating. Not fully understood is the on average
∆
E=1.425meV higher emission energy after
the encapsulation process. In chapter 3 it is shown that the maximum difference from the
emission along the z-axis to the emission from the top surface is
∆
E=37meV. The parabolic
optic directs the edge-emitted light parallel to the surface emitted light (chapter 6.1.2).
Therefore the change of
∆
E=1.425meV in the peak emission could be caused be a different
light composition. Composition in this case means the ratio from “edge” light to “surface”
light launched into the monochromator is influenced by the optic.
1.88 1.92 1.96 2.00 2.04
Intensity [a. u.]
Energy [eV]
ϕ = 90°
ϕ = 70°
ϕ = 50°
ϕ = 30°
ϕ = 0°
Sample: Dit 19
Figure 6.24: Spectra at different polarization angles relative to the axis of symmetry of the
module.
6.2 Influence of injection molding on the LED performance 79
After encapsulation the polarisation angle
ϕ
of the emission is measured. The maximum
intensity of the emission spectra depicted in Fig. 6.24 is plotted versus the polarisation angle
ϕ of the polarisation filter. Figure 6.25 shows the results. The data are fitted by Eq. (6.4)
000 )cos()( IAI
+
−=
ϕ
ϕ
ϕ
(6.4)
The intensity I(
ϕ)
only varies about
∆
A0=0.16%. A slight polarisation with an orientation of
ϕ
0=18.9° is observable.
-10 0 10 20 30 40 50 60 70 80 90 100
0.88
0.90
0.92
0.94
0.96
0.98
1.00
Measurement
Fit
Intensity [%]
ϕ [°]
Figure 6.25: Light intensity of an encapsulated LED versus angle
ϕ
of the polarisation filter.
According to [31] strained PMMA has a strong polarisation effect to light. Thus we believe
that the slight polarisation effect depicted in Fig. 6.25 is an indication for a strained PMMA
encapsulation.
6 Injection molding of LED clusters 80
6.3 Radiation pattern of the PMMA optic
To demonstrate the focusing effect of the parabolic PMMA optic the radiation pattern of the
chip and of the complete module is measured with a far field goniometer. Figure 6.26 shows
the radiation pattern of the non-encapsulated LED chip as supplied by the manufacturer. The
light source is mounted in the centre of the coordinate system. The active layer is
perpendicular to the 90° direction, the n-contact is not transparent and points to the 270°
direction. Therefore no light emission could be observed at angles between 180° and 360°.
The goniometer measures the light intensity per solid angle and creates a vector where the
direction of the vector corresponds to the direction of the light emission. The length of the
vector corresponds to the intensity. As we clearly see, the main part of the light is emitted at
angles between 0° and 180°. The light intensity emitted in the 90° direction, is about half the
maximum intensity, noticeable are the two bulky edges which correspond to the edges of the
square-shaped semiconductor. This means that in a 3D view there are four of these bulky
edges which correspond to the four edges of the LED chip. The total luminous flux at an
electrical current of I=20mA is
Φ
=0.83lm. The emission spectrum is typically for edge-
emitting diodes.
0
0.05
0.1
0.15
0.2
0
30
60
90
120
210
240
270
300
330
Intensity [a. u.]
Figure 6.26: Radiation pattern of a LED chip without encapsulation.
6.3 Radiation pattern of the PMMA optic 81
Figure 6.27 shows the radiation pattern of the LED with the parabolic-shaped total reflection
optic. The emission is focused straight forward in a narrow beam with a beam spread of
α
=10°. Due to the focusing effect of the parabolic optic, the intensity in the 180° direction
clearly increased. A total luminous flux of
Φ
=1.11lm has been determined.
0
5
10
15
20
0
30
60
90
120
210
240
270
300
330
Intensity [a.u.]
10°
Figure 6.27: Radiation pattern of a LED encapsulated in the parabolic optic.
6 Injection molding of LED clusters 82
7 Diffusion model for LED degradation
In this chapter we analyze the aging characteristics of PMMA encapsulated LEDs. The
reduction of the electroluminescence efficiency
η
t during aging time t by keeping the junction
at temperature Tj is measured. The junction temperature is caused by a forward current Idegra.
The reduction of the efficiency is attributed to the formation of non-radiative recombination
centres in the active layer. We show that this non-radiative recombination centres are Mg
atoms which diffuse from the p-type layer into the active layer.
We assume that the Mg diffusion can only occur when the repelling p-n junction potential is
reduced by the forward voltage. In the case of device heating by forward current this is taken
for granted.
7.1 The dependence of electroluminescence on trap
concentration
In this paragraph we calculate a relationship between the concentration of non-radiative
trapping centres Nn,d (in the active layer of the LED) and the electroluminescence efficiency
r(t).
An increase of the non-radiative trap concentration during degradation decreases the lifetime
of the electrons. This reduces the probability that a given electron will recombine radiatively.
It has been shown in [32] that the external electroluminescent efficiency
η
of a non-degraded
device can be expressed as:
oceinjB IRf
η
η
η
η
)(= (7.1)
where
η
B is the internal quantum efficiency,
η
inj is the current injection efficiency defined as
the ratio of the electron and hole diffusion current density (Ie+Ih) to the total current density It,
R is the ratio of Ie to (Ie+Ih), f(Ie) accounts for the decay in quantum efficiency due to
7 Diffusion model for LED degradation 84
saturation of the radiative recombination mechanism, and
η
oc is the external optical out-
coupling efficiency of the device to the environment. Hence
η
can be written as:
octeeB IIIf
η
η
η
)/)((=. (7.2)
According to the kinetics of photoluminescence [33], [34]
η
B is related to the minority-carrier
lifetime
τ
e and to the lifetime for electrons captured by nonradiative traps
τ
n:
β
α
τ
βτ
α
τ
η
e
n
n
B=
+
= (7.3)
which leads to
βττ
111 +=
ne
(7.4)
In Eq. (7.3) and Eq. (7.4)
α
and
β
are capture rates which are unaffected by degradation. In an
undegraded diode . After a period of degradation t, the density of nonradiative traps,
N
0
ee
ττ
=
n,d, increases. Thus the electron decay rate is given by [35]:
)(
1
)(
1
,
0tNv
tdnnrth
e
e
σ
τ
τ
+= (7.5)
where vth is the thermal velocity of electrons and
σ
th is the capture cross section of the non-
radiative traps.
Another factor which contributes to
η
is the saturation function f(te) which accounts for the
decay in quantum efficiency due to saturation of the radiative recombination process. The
theory of electroluminescence provides the following approximation for this function [31]:
1
00 expexp1ln)(
−
+= kT
qV
n
n
kT
qV
n
n
tf
tt
e (7.6)
7.1 The dependence of electroluminescence on trap concentration 85
where V is the forward bias voltage applied to the junction, n0 is the electron concentration in
a p-type crystal in absence of a bias voltage, nt is the excess electron density required for the
radiative recombination.
For an abrupt p-n junction Ie is [32]
=kT
qV
L
nqD
I
e
e
eexp
0 (7.7)
where De is the electron diffusivity and Le=(De
τ
e)1/2 is the electron diffusion length.
Substituting Eqs. (7.3), (7.6) and (7.7) into Eq. (7.2), we obtain
[]
oc
t
ekT
qV
n
n
tBt
ητη
+= exp1ln)()( 0
2/1 (7.8)
where
t
te
I
nDq
B
β
α
2
1
)(
=
whereas the electroluminescent efficiency of the undegraded device is given by
oc
t
ekT
qV
n
n
B
ητη
+=
0
0
2
1
0
0exp1ln)( (7.9)
assuming that
η
oc does not degrade and affect the measurement.
Using Eq. (7.8) and Eq. (7.9) and defining r(t)=
η
(t)/
η
0 we get
Qtr
e
e⋅
=2
1
0
)(
τ
τ
(7.10)
where
1
0
00 exp1lnexp1ln
−
+
+= kt
qV
n
n
kT
qV
n
n
Q
tt
(7.11)
We assume that the voltage at the p-n junction is not affected by degradation [33]:
0
VV ≈
Thus we assume that Q=1.
7 Diffusion model for LED degradation 86
The relative intensity r(t) is then given by:
2
1
0
)(
=
e
e
tr
τ
τ
(7.12)
A combination of Eq. (7.5) and Eq (7.12) gives an expression for Nn,d(t):
−= 1
)(
11
)( 2
,tr
C
tN dn (7.13)
where C.
nrthe v
στ
0
=
For simplicity reasons C is assumed to be not affected by degradation. Thus C is absorbed in
the amplitude of the diffusion equation which is derived later.
Equation (7.13) provides a relationship between the concentration of non-radiative trapping
centres which are generated during degradation and the relative intensity r(t). Since LED
degradation is a function of aging time t, one must obtain an expression for the increase in
Nn,d in terms of a reasonable time-dependent process. Diffusion processes are assumed to be
responsible for the observed time dependence of degradation. The connection between
diffusion and degradation of solid-state devices has been described earlier. In [36] it is
suggested that the degradation of GaAs tunneling diodes may be explained by the unimpeded
interstitial diffusion of the p-type dopant zinc (Zn) into the junction. Similar investigations
have been performed with Cu-contaminated GaP red LEDs [42].
7.2 Identifying magnesium as the p-doping material
In this paragraph we show that the p-type doping material is magnesium. Therefore, depth-
resolved time of flight (ToF) SIMS measurements are performed. The experimental set-up is
described in appendix C.
To guarantee a planar abrasion, the p-type contact of the LED has been removed by HCL. The
surface of the LED chip has been sputtered with cesium ions at an angle of 25° to the surface
normal. The Cs+ ions where accelerated by V=3KV, the ion current was I = 20nA and
7.2 Identifying magnesium as the p-doping material 87
the sputtered surface was A=125x125 µm2. The secondary ions have been extracted with
V=25KV. Figure 7.1 shows the results obtained under these conditions.
Figure 7.1: Intensity versus sputter time for phosphor, magnesium and gallium.
01234
100
P
Intensity
Time [h]
P
Ga69
Mg26
Ga
Mg
The intensity versus the sputter time which is equivalent to the sputter depth is plotted for the
elements and isotopes P, Ga69, Mg26. For Phosphor a constant count rate during the sputter
time is observed. The same characteristic is observed for Ga. Gallium and phosphor are the
main substances of the LEDs. Due to the fact that above and under the active layer a GaP
window is placed, the constant and high signal in mass spectra is perspicuous. The
characteristic of the Mg count rate is quite different. Near the surface the Mg content is quite
high. The Mg content decreases in the first 30 minutes of sputtering which is equivalent to a
sputter depth of d=5.9µm. Mg is diffused into the first microns of the semiconductor for
optimizing the ohmic contact on the p-side. Mg behaves as an acceptor in GaP and a high
conductivity is reached. This reduces the serial resistance and a device heat-up is avoided.
After a sputter depth of 5.9µm Mg is still observed. This proves that Mg is used as an
acceptor in this device. After a sputter time of t=4.5h which is equivalent to a sputter depth of
d=53µm the magnesium content starts to increase. At this point the ion beam has reached the
boundary of the LED. The ion beam starts to sputter the glue.
The ion beam hits the sample at an angle of α=25° to the surface normal. So the maximum
sputter depth is given by:
α
tan
max
⋅
=
e
Ld (7.14)
7 Diffusion model for LED degradation 88
where dmax is the maximum of depth, Le is the edge length of the crater, and α is the angle
between the surface normal and the ion beam.
With Eq. (7.14) dmax is calculated to 58µm which corresponds to our observed maximum
depth of 53µm. During the sputtering process mass-spectra of the magnesium isotopes Mg24,
Mg25 and Mg26 have been measured. Figure 7.2 shows the mass-spectra of Mg24, Mg25 and
Mg26.
Figure 7.2: Mass-spectra of Mg24, Mg25 and Mg26. Red line describes the intensity ratio of
the Mg isotopes in nature.
23 24 25 26 27
100
101
102
103
104
105
106
Intensity [a. U.]
Mass [amu]
The red line describes the intensity ratio of the Mg isotopes in nature. For Mg25 we measured
exactly the expected intensity, for Mg24 the measured intensity is larger. This can be
explained as mass interference with C2.
The depth resolution strongly depends on the sputtering depth. Near the surface, layers with a
thickness of d=5nm can be resolved. Due to the increasing surface roughness with increasing
sputter depth the resolution decreases. SEM measurements have shown that the depletion
zone of the device is in a depth of d=50µm.
Due to the decreased resolution in a depth of d=50µm it is impossible to detect the
magnesium diffusion from the p-type layer into the depletion zone. It has been shown that
magnesium is the p-type doping material of the AlInGaP LEDs.
7.3 Magnesium diffusion into the active layer 89
7.3 Magnesium diffusion into the active layer
In [11] Mg atoms diffusing into the active layer are identified as efficient non-radiative
recombination centres which reduce the internal quantum efficiency. We will now combine
the Mg diffusion into the active layer with the increase of non-radiative trapping centres. The
diffused quantity of Mg in the active layer, FMg, at time can be expressed as the time
integral of the Mg flux [35]:
t
()
' (7.15)
0
dt
z
C
DtF
d
wz
Mg
t
Mg
=
∂
∂
−= ∫
where D is the diffusivity, CMg is the magnesium concentration, and wd is the location of the
plane where the accumulation occurs. To affect the radiative recombination and decrease the
relative intensity, this plane must be in the MQW region.
We assume that the trap density Nn,d is proportional to the diffused quantity of Mg atoms FMg.
Thus we take:
Mgdn KFN
=
, (7.16)
where K is a constant of proportionality which also includes the ratio of the junction area to
the effective volume of the accumulation.
To evaluate FMg an analytic expression for Cis necessary. Before the degradation the
Mg content in the p-type layer is C on average. It drops down to zero at and beyond the
edge of the active layer. Figure 7.3 shows the concentration gradient in the active layer. The
origin of the coordinate system is at the edge of the p-side of the active layer. Then the initial
conditions are:
)',( tz
Mg
0
Mg
for z<0
0
)0,( MgMg CzC =
0
for z
)0,( =zCMg d
w
≤
≤
0
The boundary conditions are:
7 Diffusion model for LED degradation 90
0
)',0( MgMg CtC = for all 't
for 0t.
0)',( =twC dMg '>
Thus, we assume a Mg concentration of 0)',(
=
twC dMg at the border of the active layer at the
edge of the n-type side [37].
Relative Mg concentration CMg / CMg
0
0
0
Mg
C
p-type
layer
MQW region
T=423°C
t‘=1000h
t‘=100h
t‘=10h
20 40 60 80 100
z [nm]
Figure 7.3: Schematic coordinate system for the diffusion of Mg through the active layer
during thermal degradation.
The solution of the diffusion equation (Fick’s second law) for the initial and boundary
conditions can be deduced from Ref. [38, Eq. (4.12)]. The concentration of Mg at z and t is
given by the infinite series
()
−⋅−
−
=∑
∞
=d
m
d
d
d
MgMg w
Dtm
m
wzm
w
zw
CtzC '
exp
)/sin(
2
)',(
22
1
0
π
π
π
(7.17)
The resulting concentration profile is drawn in Fig. 7.3 for the diffusion times t’=10h, t’=100h
and t’=1000h. The junction temperature was T=423°C and the diffusivity
(see chapter 7.4.1). We assume 10 quantum wells and barriers with a
width of w=5nm. Thus the width of the MQW region was assumed to be w
1213
101.1 −−
⋅= scmD
d=100nm.
Unretarded diffusion as in Eq. (7.17) can only take place if the device is under a relatively
high forward bias voltage. If the device is heated up by a forward current, the retarding effect
of the inner E-field is negligible.
7.3 Magnesium diffusion into the active layer 91
Substituting Eq. (7.17) into Eq. (7.15)
−+=
∂
∂
−∑
∞
=
=1
22
0'
exp)cos(
2
mddd
Mg
wz
Mg
w
Dtm
m
w
D
w
D
C
z
C
D
d
π
π
and after integration over time and taking the limits
−−
−
+= ∑
∞
=1
22
22
0
4
exp1
)1(2
4m
m
dMgMg
tm
m
t
wCF
θ
π
π
θ
(7.18)
with the diffusion time:
Dwd4/
2
=
θ
. (7.18a)
Using Eq.(7.14), Eq. (7.16) and Eq. (7.18) r(t) can be expressed as:
[]
2
1
)/(1)( −
+=
θγϕ
ttr (7.19)
where
∑
∞
=
−−
−
+=
1
22
22 4
exp1
)1(2
4m
mtm
m
tt
θ
π
π
θθ
ϕ
(7.20)
and
(7.21)
dMgnrthedX wKCvwCKC 000
στγ
==
γ
is expected to be a constant due to the fact that C is already assumed as constant.
The proportionality factor K is not affected by degradation nor is w
nrthe v
στ
0
=
d. The Mg content is
assumed to be constant that means not reduced by degradation. Only small variations in C
between different samples are expected.
0
Mg
Figure 7.4 depicts the calculated relative intensity r(t) versus degradation time. The
parameters
θ
and
γ
have been chosen to
θ
= 79.5h and
γ
= 0.2. After a degradation time of
t=5000h the intensity is reduced to r = 0.5.
7 Diffusion model for LED degradation 92
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
0.1 1 10 100 1000 10000
Rel. intensity r(t)
Time [h]
q=79.5h
g=0.2
Figure 7.4: The calculated relative intensity r(t) versus degradation time for
θ
= 79.5h and
γ
= 0.2.
7.4 Data analysis
Assuming that Mg diffusion is responsible for the degradation, we can extract the diffusivity
of Mg by Eq (7.18a). Therefore the experimental data of LED degradation r(t) will be fitted
by Eq. (7.19). and Eq. (7.20).
The first experiment (sample H G 4) was made under a degradation current of Idegra=150mA
(Tj=423°C ). The luminous flux has been measured at Imeas=40mA. Figure 7.5 shows the
relative intensity r(t) versus the degradation time. The blue dots indicate the experimental
data. A fast decrease of intensity is clearly observable. After t=263h the intensity has
decreased by 50%. Using Eq. (7.19) and Eq. (7.20) the experimental data have been fitted by
a successive approximation of the parameters
θ
and
γ
. For practical reasons the infinite series
in Eq. (7.20) has been calculated up to m=40. For m>40 no significant change in the result is
observable.
The black curve is calculated using the parameters
θ
=7.5h and
γ
=0.33. A perfect agreement of
experimental and calculated data is obtained. Degradation experiments were also performed
7.4 Data analysis 93
under a forward current of Idegra=120mA (T=306°C). The relative intensity has been measured
at Imeas=40mA and the experimental data are plotted as red dots in Fig. 7.5.
0.2
0.4
0.6
0.8
1
1.2
0.1 1 10 100 1000 10000
Rel. intensity
Time [h]
H G 4, Tj=306°C
H G 2, Tj=423°C
Figure 7.5: Relative intensity versus degradation time. Red dots for Idegra=120mA. The green
curve is a fit using
θ
=79.5h and
γ
=0.2 . Blue dots for Idegra=150mA, the black
curve is a fit using
θ
= 7.5h and
γ
=0.33.
After an aging time of t=360h the intensity decreases to 90% of its starting value and after
t=5150h the intensity decreases by 50%. In this case the parameters
θ
and
γ
have been set to
θ
=79.5h and
γ
=0.2 and the green curve depicts the calculation. The experimental and
calculated data are in a perfect agreement with one exception. In the time range between
t=25h and t=260h an increase of intensity is observed. This so called “positive aging” is also
observed in [39] and has been noticed regularly in our experiments. This increase of intensity
cannot be explained by the model presented earlier in this paragraph.
In a further experiment the device has been treated with Idegra=40mA (Tj=85°C). The duration
of the aging process exceeded t=8000h. Long-term experiments allow to predict an accurate
time to failure. This is due to the fact that the stress current is only twice as high as the
recommended maximum drive current. Figure 7.6 shows the relative intensity r(t) versus
aging time. The red dots indicate the experimental data. During the first 1000 hours the
intensity increases approximately by 5%. After t=1350h the intensity has decayed to r(t)=1
(100%) and continues to drop to 93% after 8800h of aging time. The curve is calculated using
the parameters
θ
=15000h and
γ
=0.7. The fit matches the experimentally observed decay at
t=1000h, however, for t<1000h an increase of the intensity is observed which is not
reproducible by the diffusion model.
7 Diffusion model for LED degradation 94
0.92
0.94
0.96
0.98
1.00
1.02
1.04
1.06
1 10 100 1000 10000
Rel. intensity
Time [h]
H G 3, Tj=85°C
Figure 7.6: Relative intensity versus degradation time at Idegra=40mA and Imeas=40mA.
The data for the degraded LEDs are listed in Table 1. The following key features are found:
a)
θ
decreases rapidly with increasing Tj as anticipated and the diffusivity of Mg
() increases with increasing T
θ
4/
2
d
wD =j.
b) The parameter
γ
is essentially constant, independent of Tj. This is in agreement with the
definition of
γ
in Eq. (7.21). Small changes from diode to diode in the initial Mg doping level
(
)
0
Mg
C and electron lifetime
(
)
0
e
τ
can affect
γ
.
Sample Idegra Tj θ γ
H G 3 40mA 85°C 15000h 0.7
H G 2 120mA 306°C 79.5h 0.2
H G 4 150mA 423°C 7.5 0.33
Table 7.1: Values of
θ
and
γ
together with degradation data.
7.4. Data analysis 95
7.4.1 The activation and diffusivity of magnesium in AlInGaP
In this paragraph we calculate the diffusion coefficient and the activation energy of Mg in
AlInGaP. For the recommended forward current of I=20mA the diffusion time
θ
will be
calculated. As a next step the time to failure for I=20mA will be determined.
The temperature dependence of the diffusion coefficient in solids is usually given by
∆−
=kT
E
DD a
exp
0 (7.23)
where is the activation energy and D
a
E∆0 is the diffusion coefficient at infinite temperatures.
This equation is only valid in the case of free diffusion, unaffected by retarding electric fields.
Therefore the applicability of Eq. (7.23) to the diffusion values listed in table 7.1 have been
tested.
Figure 7.7 shows the Arrhenius plot of log(
θ
) versus 1000/Tj. Dots are experimental data from
diodes degraded at Tj=423°C, Tj=306°C, and Tj=85°C. Clearly, the function relationship is
linear, in agreement with Eq. (7.23). A linear fit indicated by the blue line yields
()
T5349exp005.0 ⋅=
θ
.
Figure 7.7: Diffusion time versus reciprocal junction temperature. The arrow
Dwd4/
2
=
θ
100
101
102
103
104
105
106
11.522.533
θ [h]
1000/T[K-1]
.5
activation energy Ea=0.46eV
T=46°C
marks T=46°C.
7 Diffusion model for LED degradation 96
The activation energy is eVeVEa04.046.0
±
=∆ . In the literature no values for the activation
energy of Mg in AlInGaP have been found. But in [35] an activation energy ∆E=0.70eV of an
unknown impurity in GaP has been determined. This value is of the same order of magnitude
as the activation energy we calculated.
A very well studied material is GaAs. In [40] the activation energy of Mg in GaAs has been
found to Ea=1.5eV. This value is three times higher than the value we calculated for the
activation of Mg in AlInGaP. One reason for the large difference is the different host crystal.
According to [37] we assume a carrier concentration of of Mg in
AlInGaP. Thus the width of the depletion zone has been calculated to W
317
1014.2 −
⋅= cmp
1210 −
scm
D=0.11µm. With
respect to ∆ a diffusivity of Mg in AlInGaP at
infinite temperatures has been calculated. At a junction temperature T
eVeVEa04.046.0 ±=
1213 −− scm
.5=D
0102 −
⋅=D
1212 −
scm
1−
j=423°C a diffusivity
has been calculated. The diffusion coefficient for Mg in AlInGaP has
not been found in the literature. Therefore we compared our results with the diffusivity of Mg
in GaAs. In [40] a diffusivity of at T=900°C for Mg in GaAs is
determined. A diffusivity of at T=785°C for Mg in GaAs is presented
in [41].
101.1 ⋅=D
102.1 −
⋅=D
215
1037 −
⋅scm
To obtain a lifetime characteristic at the drive current of Idegra=20mA the linear fit was
approximated to Tj=46°C. The resulting diffusion time is
θ
= 90473h. As
γ
was to be assumed
to be a constant, it is set to
γ = 0.7
(table 7.1). With these parameters the relative intensity
degradation r(t) for a forward current of Idegra= 20mA is calculated. Figure 7.8 shows the
result.
Figure 7.8: Lifetime prediction at a forward current of Idegra=20mA. Arrow 1: Time to failure
0.2
0.4
0.6
0.8
1.0
1.2
100101102103104105106107
Rel. intensity r(t)
Time [h]
1
2
(TTF) at r(t)=0.9 is 10 . Arrow 2: TTF at r(t)=0.5 is 1.
h
5h
6
105. ⋅
7.4 Data analysis 97
The degradation is clearly weak. For the first 104h of degradation time no intensity decrease is
expected. After 105h of aging time the intensity is reduced to 90% (marked by arrow 1).
Therefore, if the time to failure criteria is a loss of 10% of intensity, the diode lifetime will be
105h. At r(t)=0.5 which is equivalent to a reduction of 50% of luminous intensity the time to
failure is 1(marked by arrow 2).
h
6
105. ⋅
In this chapter we have shown that the reduction of intensity can be described by a diffusion
of the p-type doping material (Mg) into the MQW region. Using this model we have
calculated the reduction of intensity for different junction currents and degradation times. A
good match between experimental and theoretical data was achieved.
The activation energy of magnesium in AlInGaP has been determined to
. A diffusivity of magnesium in AlInGaP at
infinite temperature was found. The reduction of relative intensity r(t) versus degradation
time for a drive current of I
eVeVEa04.046.0 ±=∆ 1210
0102 −−
⋅= scmD
degra=20mA has been calculated. The lifetime is 1 defining
a decrease of 50% of relative intensity( r(t)=0.5) as failure criteria.
h
6
105. ⋅
7 Diffusion model for LED degradation 98
8 Summary
In this thesis we investigate the degradation mechanism of AlInGaP LEDs during
encapsulation and operation. A new method to encapsulate AlInGaP LEDs by means of an
injection molding tool using polymethylmethacrylat (PMMA) 7N as an encapsulation
material is analysed. The combination of semiconductor light source, optics and physical
housing in one unit with diminutive dimensions leads to a luminous source with tailored
radiation pattern. We demonstrate light beams with a spread angle of 10° obtained by total
reflection optics.
Before encapsulation a detailed electrical and optical characterization of the LED chip such as
space resolved measurements of the light emission from the edges and top surface are
performed. An injection molding tool has been created which allows the encapsulation of four
LED chips, each in a parabolic optic with a focal length of f=1.1mm. Encapsulation was done
with an Arburg A380 injection molding machine. To investigate the thermal load during the
encapsulation process, a new approach for measuring the temperature inside the injection
molding tool is presented. It is based on the temperature dependence of the electrical
conductivity of a pn-junction. During encapsulation we measure a maximum LED
temperature of T=153°C. Due to the low heat capacity of the device, the melt temperature is
almost unaffected and a very quick response is achieved. With this method we measure the
thermal load of directly encapsulated electrical devices. Furthermore, this method allows to
measure the temperature distribution within the molded part using multiple probes. To verify
the temperature measurements, a finite element simulation is carried out. The variation of the
temperature at the LED chip location is calculated and shows good agreement with the
experimental results.
After the encapsulation process the LED properties are investigated and compared with the
properties before encapsulation. We find a reduction of the serial resistance as well as an
enhanced luminous efficiency. The peak emission wavelength remains constant, but the full
width at half maximum (FWHM) of the emission increases by
∆
E=9meV. A slight
polarisation of the emitted light indicates a polarisation effect of the PMMA.
8 Summary 100
Accelerated degradation experiments using forward currents of I1=40mA, I2=120mA and
I3=150mA for estimating the lifetime of the PMMA encapsulated LEDs are performed.
The measured lifetime under I2 is t2=5150h, and under I3 is t3=263h using a reduction in the
luminous flux of 50% as failure criteria. A diffusion model is presented to explain the decay
of luminous flux versus degradation time and degradation current. We support the idea that
the reduction of quantum efficiency is caused by diffusion of the p-type dopant into the active
layer where it acts as a non-radiative recombination centre. This is supported by other
investigations [11]. SIMS measurements show that the p-type doping material is magnesium.
Using the diffusion model we determine the lifetime under the recommended drive current of
I=20mA. The resulting lifetime is t using a reduction of 50% in the luminous flux
as failure criteria.
h
6
105.1 ⋅=
Our results reveal that a direct encapsulation of LEDs by means of an injection molding tool
is possible. The effects of thermal degradation of the LED are negligible and the LED lifetime
is sufficient for most applications. Especially for application in tail lights for passenger cars
where a lifetime of t=6000h is required, this method opens a wide new field of space
economy and design potential. Ultra thin tail lights which fulfill the legal requirements for
passenger cars are possible by this encapsulation technique.
Appendix A: focusing the active layer of a LED chip
In order to obtain the EL from the active region of the LED, the following method is used.
First one has to focus on the top (p-contact) surface and then move the focal point 50µm
(distance from top surface to the active layer) inside the chip by adjusting the second lens.
Figure A.1 depicts the calibration curve for moving the focal point from the surface of the
LED chip into the semiconductor material.
0
1
2
3
4
5
6
7
8
0 0.05 0.1 0.15 0.2 0.25 0.3
∆d [mm]
Focal point inside the LED chip (FP) [mm]
Figure A.1: Calibration curve to focus inside the LED chip. FP is the distance between the
surface of the LED chip and the focal point inside the LED,
∆
d is the distance
difference between lens 1 and lens 2.
The x-coordinate of Fig. A.1 describes the distance from the LED chip surface to the point
which should be focused by the optics. The y-coordinate is the difference in distance of the
two lenses. For example the point of interest is FP=0.1mm beyond the chip surface. Thus the
distance of lens 1 to lens 2 must be decreased by
∆
d=2.202mm. The relationship is given in
Eq. (1).
(1) 02.22][][ ⋅=∆ mmFPmmd
102 Appendix A
Appendix B: capture volume
The intersection plane of the observation cone and the active layer (red) is a hyperbolic
bordered area. This border is depicted in Fig. B.1 as blue line.
active layer
Figure B.1: Observation cone in the LED chip.
The edge planes of the LED chip have the distance xa, the height of the cone is hk, the radius
of the base area is R. The radius of the top surface area of the cone is r. The spread angle of
the cone is given by
β
. Thus the hyperbola is given by:
22 ax
a
b
y−= (1)
with:
z
R
hk
=
a and z=b (2)
The surface of the hyperbolic bordered area is then given by :
∫−=
k
h
a
dxax
a
b
A22 (3)
−+
+−= 22
2
22 ln
2zRR
x
R
x
zR
hk (4)
104 Appendix B
Due to the small value of za (thickness of the active layer), the volume is given by:
(5)
AzV a⋅=
Thus the detection volume is:
−+
+−= 22
2
22 ln
2zRR
x
R
x
zR
zh
Vak (6)
with
α
tan
r
xh ak += and
β
β
tantan ak xrhR
+
=
=
Appendix C: SIMS measurement set-up
Depth resolved ToF – SIMS measurements on the AlInGaP LEDs are performed. Goal of
these measurements is to identify the p-type doping material and to investigate if a doping
material diffusion in the depletion region can be measured with this method. Due to the small
dimensions of the samples and the need for a planar surface with an edge length of 5mm the
sample is embedded into GaAs. This is illustrated in Fig. C.1. The green rectangles are the
GaAs, the red square is the LED. The p-contact of the LED is removed by HCL to get a
planar abrasion.
Cs+ beam
sample
25°
GaAs
Figure C.1: Sample arrangement in the SIMS unit. Red square marks the LED chip.
The surface is sputtered with Cesium ions under an angle of α=25° to the surface normal.
These ions are accelerated by U=3kV, the ion current is I = 20nA and the sputtered surface is
A=125x125 µm2.
106 Appendix C
Bibliography
[1] DenBaars, S.P. Katona, T. Cantu, P. Hanlon, A. Keller, S. Schmidt, M. Margalith, T.
Pattisson, M. Moe, C. Speck, J. Nakamura, GaN based high brightness LEDs and UV,
IEEE Electron Devices Meeting, 16.1.1 (2003)
[2] D. A. Steigerwald, J. C. Bhat, D. Collins, R. M. Fletcher, M. O. Holcomb, M. J.
Ludowise, P. S. Martin, S. L. Rudaz, IEEE J. Quantum Electron. 8, 310 (2002)
[3] Hella, Technische Information Licht-Scheinwerfer (2003)
[4] G. Moos, private communication
[5] P. Modak, D. Delbeke, I. Moerman, R. Baets, P. Van Daele, P. Demeester, ICMOVPE-X,
workbook of the tenth int. conf. on metalorganic vapour phase epitaxy, June 5-9, Sapporo,
Japan, 319 (2000)
[6] M. Zoorob, G. Flinn, LEDs Magazine, August, 21 (2006)
[7] E. F. Schubert, Light-emitting Diodes 2nd edition, (Cambridge University Press,
Cambridge, 2006)
[8] M. L. Cohen; T. K. Bergstresser, Phys. Rev. 141, 789 (1966)
[9] I. Vurgaftman, J. R. Meyer, L. R. Ram-Mohan, Appl. Phys. Rev. 89, 5815 (2001)
[10] J. Singh, Semiconductor Optoelectronics, (Mc Graw Hill, New York, 1997)
[11] K. Streubel, N. Linder, R. Wirth, A. Jäger, IEEE J. Quantum Electron. 8, 321 (2002) and
Workshop Erlangen (2005)
[12] N. W. Ashcroft, N. D. Mermin, Solid state physics (Thomson Learning Inc., Stamford
1976)
[13] F. Schwabl, Quantenmechanik, (Springer-Verlag Berlin Heidelberg New York, 2002)
[14] Lumileds Lighting, U.S. LLC, Technical datasheet DS31, (2004)
[15] Y. Xi, T. Gessmann, J. Xi, J. K. Kim, J. M. Shan, E. F. Schubert, A. J. Fischer, M. H.
Crawford, K. H. A. Bogart, A. Allermann, J. J. Appl. Phys. 44, 7260 (2005)
108 Bibliography
[16]D. Schwedt, Räumlich und spektral hochaufgelöste Sepktroskopie der
Resonanzfluoreszenz an Exzitonen in ungeordneten Halbleiterfilmen, (Dissertation,
Universität Roststock, 2005)
[17] P. K. Basu, Theory of optical processes in semiconductors, (Clarendon Press, Oxford,
1997)
[18] U. Jahn, J. Menniger, R. Hey, B. Jenichen, H. T. Grahn, E. Runge, Mat. Science and
Engineering: B, 42, 133 (1996)
[19] M. A. Herman, D. Bimberg, J. Christen, J. Appl. Phys. 70, R1 (1991)
[20] S. Adachi, J. Appl. Phys. 66, 6030 (1989)
[21] A. Zukauskas, M. S. Shur, R. Gaska, Indroduction to solid-state Lighting, (Wiley-
Interscience Publication, Vilnius, 2001)
[22] Fa. Hartmann & Braun, Gebrauchsanweisung, Beleuchtungsmesser EBLX 3, Frankfurt
(1971)
[23] D. Lüerßen, R. Bleher and H. Kalt, Phys. Rev. B 61 812 (2000)
[24 ] P. J. Dean, E. G. Schönherr, R. B. Zetterstrom, J. Appl. Phys., 41, 3475 (1970)
[25] C. F. Zhu, W.-K. Fong, B.-H. Leung, C.-C. Cheng, C. Surya, IEEE Trans. Electron.
Devices, 48, 1225 (2001).
[26] S. Nakamura, G. Fasol, The blue laserdiode, GaN based light emitters and lasers,
(Springer, Berlin, 1997)
[27] G. Meneghesso, S. Levada, R. Pierobon, F. Ramazzo, E. Zanoni, A. Cavallini, A.
Castaldini, G. Scamarcio, S. Du, I. Eliashevich, Proc. of IEEE Int. Electron Device
Meeting, 103, San Francisco, California, December 8-11 (2002)
[28] E. F. Schubert, Basics of LED encapsulants, (Troy, New York, 2002)
[29] Kenndaten für die Verarbeitung thermoplastischer Kunststoffe Teil 1, VDMA, (Carl
Hanser Verlag München, 1979)
[30] Y. P. Varshni, Physica, 34, 149 (1967)
[31] M. Lasch, F. Rudolph, L. Schreiner, H. Schulze, J. Appl. Polymer Science, 11, 369
(2005)
[32] J. M. Ralston, J. Appl. Phys. 44, 2635 (1973)
[33] W. Rosenzweig, W. H. Hackett, Jr., J. S. Jayson, J. Appl. Phys. 40, 4477 (1969)
[34] J. S. Jayson, R. N. Bhargava, R. W. Dixon, J. Appl. Phys. 41, 4972 (1970)
[35] A. S. Jordan, J. M. Ralston, J. Appl. Phys. 47, 4518 (1976)
Bibliography 109
[36] R. L. Longini, Solid State Electron 5, 127 (1962)
[37] P. N. Grillot, S. A. Stockman, J. W. Huang, J. Appl. Phys. 91, 4891 (2002)
[38] J. Crank, The Mathematics of Diffusion (Oxford U. P., London, 1956)
[39] O. Pursiainen, N. Linder, A. Jaeger, R. Oberschmid and K. Streubel, Appl. Phys. Lett.,
79, 2895 (2001)
[40] N. Nordell, P. Ojala, W. H. van Berloa, and G.Landgren, J. Appl. Phys. 67, 778 (1990)
[41] V. A. Labunov, O. I. Velichko, S. K. Fedoruk, J. Of Eng. Phys. and Thermophys. 67,
1067 (1994)
[42] A. A. Bergh, IEEE 8th Annual Proceedings on Reliability Physics, Las Vegas, 48 (1970)
110 Bibliography
List of samples
Note: All samples are Lumileds HWFR-B410 LEDs.
Degradation experiments:
Sample 1: Resin encapsulated LED Idegra=70mA, #33_C1
Sample 2: Chip only Idegra=120mA, name LZ-EL 1
Sample 3: Chip only Idegra=120mA, name LZ-UL 2
Sample 4: PMMA encapsulated LEDs at Idegra=120mA soldered chip, H G 1
Sample 5: PMMA encapsulated LEDs at Idegra=120mA glued chip, H G 2
Micro EL
G H 1, non-encapsulated LED chip
E04, non-encapsulated LED chip
SIMS Measuremnts
SIMS 2, non-encapsulated LED chip
Diffusion model:
H G 2, degradation current I=120mA
H G 3, degradation current I=40mA
H G 4, degradation current I=150mA
Temperature measurements
Temp 1 – Temp 5, non-encapsulated LED chips
Influence of injection moulding on LED performance
Dit 19, PMMA encapsulated LED
Publication List
S. Preuß, K. Lischka
Injection-Moulding Encapsulation of LEDs Improves Car Tail Lights
Europhotonics Magazine, February/March (2007)
S. Preuß, D. Potthoff, Th. Preuß, K. Lischka
LED encapsulation- a new approach of rear light design
Proceedings of SPIE 61980I, Strasbourg (2006)
H. Potente, Th. Preuß, S. Preuß
A New Approach for Temperature Measurement inside an Injection Molding Tool,
Antec, Paper 103176 Charlotte NC, (2006)
Acknowledgements
I would like to thank Prof. Dr. K. Lischka who gave me the opportunity to work in his group
and for the valuable discussions during this thesis.
For their support and their useful advices I would like to thank apl. Prof. Dr. D. J. As and Dr.
D. Schikora.
I would like to thank the referee apl. Prof. Dr. S. Greulich-Weber.
I am indebted to my students M. Rempe, D. Potthoff, M. Paluga and C. Mietze.
I have to express my thanks to my PhD fellows S. Potthast, Ch. Arens, J. Schörmann,
M. Panvilova, E. Tschumak and our Postdocs Dr. N. Rousseau, Dr. M. Peron Franco de
Godoy and Dr. A. Pawlis.
I am also grateful for the support of I. Zimmermann, S. Igges and B. Volmer.
The L-Lab members have to be mentioned here. Thank you for your encouragement.
I want to thank S. Strauss for the 3 years of LED investigations and the goniometer
measurements.
I have to mention O. Stahlsmeyer and Dr. J. Ante, thanks for the rides on the “Flickenteppich”
and the explorations deep down.
I would also like to thank the coffee producers of Costa Rica for keeping me alive.
Furthermore, I have to thank Dortje Rieken for her love, endless patience ( the desk in the
living room!) and encouragement.
Last but not least I would like to thank my parents for their love and financial support.
This work was financially supported by the Hella KGaA Hueck & Co.
