FinalUndergraduateThesis_AlGaN

AbstractAlGaN alloys are of interest as being efficient, reliable, and high-power sources of DUV (deep-ultraviolet) light. DUV light has a number of key applications that motivate AlGaN research, including water treatment, disinfection, UV-curing and non-line-of-sight communication. AlGaN alloys hold many advantages over conventional UV-light sources such as mercury vapor lamps which have low-efficiency, short lifetimes, and contain harmful chemicals. However, current AlGaN-based DUV light-emitters have efficiencies of less than 10%. In order to realize high efficiency AlGaN-based UV-light emitters it is necessary to optimize a number of key material properties. Loss of photons produced in the device through total internal reflection and Fresnel losses stemming from the large difference in the index of refraction between AlGaN and air greatly reduce the external quantum efficiency. The internal quantum efficiency is limited at room temperature due to the presence of crystal defects that act as nonradiative recombination pathways. Spontaneous and piezoelectric fields further reduce radiative recombination due to the separation of electron and hole wavefunctions in quantum wells due to the polar-nature of III-nitride materials along their c-axis. It is possible to restore the overlap through careful modulation of the composition of AlGaN multiple quantum well structures. Samples with ideal alloy modulation for maximum increase of the overlap integral were previously modeled using SiLENSe software. A 4nm well AlGaN heterostructure with the proposed modulation was subsequently grown. After providing necessary background, this paper analyzes preliminary characterization of the emission properties of this modulated sample and highlights the need for future testing.

Content

s1.0 Introduction..........................................................................................................................................2

2.0 Applications for LED Produced UV-light...............................................................................................3

3.0 Scientific Background............................................................................................................................4

3.1 AlGaN Structure and Properties.......................................................................................................4

3.2 Basic LED Operating Principles.........................................................................................................6

3.3 Multiple Quantum Well Structures..................................................................................................7

3.4 Quantum Efficiencies of LEDs...........................................................................................................8

3.4.1 External Quantum Efficiency.....................................................................................................8

3.4.2 Internal Quantum Efficiency......................................................................................................9

3.4.2.1 The Role of Defects on the IQE...........................................................................................9

3.4.2.2 The Role of Polarization and the QCSE on the IQE...........................................................11

4.0 Motivation for Research: A Method to Increase the Overlap Integral in MQWs..............................14

5.0 Experimental Procedure.....................................................................................................................15

1

5.1 CL Spectroscopy..............................................................................................................................16

5.2 TRPL Spectroscopy..........................................................................................................................16

6.0 Data and Analysis................................................................................................................................16

6.1 CL Spectroscopy Data and Results..................................................................................................16

6.2 TRPL Spectroscopy Data and Results..............................................................................................18

7.0 Discussion and Future Testing............................................................................................................22

8.0 References..........................................................................................................................................23

1.0 Introduction

The interest in group-III nitrides has grown exponentially since the first demonstration of high efficiency GaN blue LEDs by Dr. Nakamura over two decades ago.1-3 Since then, GaN-based LEDs in the blue and green wavelength regimes have reached efficiencies greater than traditional white light sources such as incandescent bulbs. Now, in much the same way that visible wavelength LEDs have replaced incandescent bulbs, AlGaN-based UV-LEDs are poised to replace traditional sources of UV-light such as mercury and metal halide lamps. The addition of AlN to GaN allows the creation of LEDs capable of producing UV-light with wavelengths as low as 210 nm. Today’s AlGaN-based UV-LEDs have peaked at efficiencies of only 10%.1 However, the strong demand for UV-light due to its many applications (see Figure 1) such as water purification, disinfection, and non-line-of-site communication continues to fuel research that will inevitably lead to high efficiency AlGaN-based LEDs.1,4,5

This paper will begin with a brief discussion of the needs and applications of UV-light. An overview of the scientific background needed to understand the significance of the research presented is then provided. This overview first examines the structure and properties of AlGaN and basic LED operating principles and then extends into the unique properties of quantum well structures. This is followed by the key factors that determine the quantum efficiency of AlGaN devices with a focus on the active region. After the relevant materials science and physics have been established, the motivation and direction for research in alloy-modulated AlGaN quantum well structures is introduced. Experimental procedures and data from cathodoluminescence and time-resolved photoluminescence are discussed. Finally, conclusions on the research are made and plans for future research are proposed.

2

2.0 Applications for LED Produced UV-lightThere are a large number of current and future applications of UV-light that drive the

interest in AlGaN research. Some of these applications can be seen in Figure 1 along with their relative output powers and wavelengths required. UV-light consists of wavelengths from 100-400 nm. Figure 1 shows how the wavelengths of UV-light are divided into three subgroups A, B and C in which C is the highest energy. It also shows approximate compositions of AlxGa1-xN required to produce such wavelengths. We can see from Figure 1 that AlGaN is used primarily to achieve wavelengths in the B and C range. This makes AlGaN-based LEDs and laser diodes suitable for applications such as photodetection, water and medical disinfection/sterilization, phototherapy, sensing and UV-curing to name a few.1,4-7

3

Figure 1. Applications of UV-Light.6

As one of the largest potential applications of UV-LEDs, a more detailed look into how UV-light is used for disinfection purposes is deserved. UV-light can be used in conjunction with

and/or replace chemical treatment for disinfection, sanitization, purification and sterilization of water and medical equipment.4,7 Treatment of water with chemicals such as chlorine is common practice. However, LED produced UV-light has many potential advantages over chemical disinfection. Firstly, LEDs remove the need for handling and use of potentially hazardous chemicals. LEDs are also relatively compact and easy to install and use. UV-light disinfects by disrupting the DNA of microorganisms such as bacteria and viruses as shown in Figure 2. This disruption can either kill the microorganism or prevent it from reproducing, making it harmless. Mercury lamps are commonly used for this application. However, because of UV-LEDs increasing efficiency, they are being viewed as potential replacements, as well as the fact that they can be tuned to 265nm output which is the ideal wavelength for disruption of DNA of most organisms.8

3.0 Scientific Background3.1 AlGaN Structure and Properties

AlGaN is a member of the III-V semiconductor (and more specifically nitride) family. The group-III nitride family consists of cations with a 3+ charge such as Al, Ga and In (group-III elements) and an anion, N (group V element,) with a 3- charge. The overall ratio of cations to anions is 1 to 1. The cations can be a mixture of multiple elements resulting in an alloy such as in AlxGa1-xN where x is a number between 0 and 1. The first group-III nitrides commercially developed were GaN and InGaN which produced blue and green light and thus completed the visible light color spectrum for LEDs.1-3 The output wavelength of group-III nitride LEDs can be controlled by tuning the mole fraction of InN in InxGa1-xN or AlN in AlxGa1-xN. The emission wavelength is roughly equal to the bandgap of the material (at low operating currents.) The relationship between the bandgap and lattice constant of InN, GaN and AlN are shown in Figure 3. The band gap clearly tends to increase with decreasing lattice constant. AlN has the smallest

4

Figure 2. UV-light causing adjacent thymine molecules to bond instead of creating a bridge.8

lattice constant and thus highest band gap energy at around 6 eV. GaN has a band gap of about 3.4eV, meaning that AlGaN alloys have band gap energies between 3.4-6, equivalent to photon wavelengths between 206-365 which covers the B and C UV-light ranges shown in Figure 1.

The most common crystal structure of the nitrides is the wurtzite structure depicted in Figure 4 which has unique polarization properties.10-12 The group-III nitrides in the wurtzite phase have a spontaneous polarization pointing in the [000-1] direction. This is due to the unsymmetrical arrangement of cation (Al3+, Ga3+, In3+) and anion (N3-) charges along the c-axis. The right side of Figure 4 shows how every layer of metal atoms is bonded to a layer of nitrogen atoms just beneath it, resulting in a dipole pointing from the metal atoms toward the nitrogen. This causes the faces to obtain fixed polarization sheet charges opposite in sign to the respective cation(s)/anion location at the surface, which can be visualized in Figure 4. In heterostructures such as quantum well, the discontinuity of polarization at the interfaces causes fixed charges which produce interesting quantum effects such as the quantum confined stark effect (QCSE.)10,11 These effects have important consequences on the design and efficiency of AlGaN LEDs.

5

Figure 3. Bandgaps of wurtzite and cubic phases of Group III nitrides vs their lattice parameter at 300K.9

3.2 Basic LED Operating PrinciplesFigure 5 depicts a PN-junction to illustrate how an LED produces light. In this device, a p-

type and n-type material are joined at an interface. The n-type material’s majority carriers are electrons. In the case of the group-III nitrides, n-type materials are typically created through silicon, Si, doping. The basic mechanism is that Si4+ ions will substitute onto 3+ metal cation sites. The material then compensates this net +1 charge by releasing an electron. Conversely, in p-type materials the majority carriers are holes where the typical dopant is magnesium, Mg. An Mg2+ ion sits on a 3+ metal cation site, causing an overall -1 charge at the site, requiring the release of a hole to compensate. When electrons and holes meet at the interface, they recombine. This causes the donor ions to be exposed (i.e. no shielding from electrons and

6

Figure 4. Crystal Structure of GaN/AlN (left)13 and a top view of the crystal structure (right)11 with a slab at the showing the fixed charges at the Ga-terminated and N-terminated surfaces.

Figure 5. Illustration of the radiative recombination process in a basic PN homojunction LED.14

holes) which results in a net electric field pointing from the n-type material towards the p-type. This causes holes to want to move down the potential and remain in the p-type material and electrons to remain in the n-type material. However, in order to release a photon we need electrons and holes to recombine. When electrons and holes recombine it is possible for them to release a photon equal to the bandgap of the material (at the lowest excited state.) We can cause electrons and holes to recombine by forward biasing the PN device, where we connect the p-type material to the positive electrode of a power source and the n-type to the negative. This causes an external electric field in the material opposite in sign to that of the built in internal electric field. When the external electric field is greater than the internal, the overall potential is reversed causing holes to flow towards the n-type material and vice versa.5

3.3 Multiple Quantum Well StructuresReal LEDs are much more sophisticated than a typical PN-junction device. They are

complicated heterostructures with many layers that serve different purposes.1,15-18 Figure 6 shows the layers in a state of the art AlGaN LED as well as the challenges that researchers face in the design of each layer.1 However, it is only the MQW (multiple quantum well) region shown in Figure 6 from which light is actually emitted. It is in this active MQW region where recombination occurs, analogous to the interface in the simple PN device in Figure 5.

Multiple quantum wells are made by alternately stacking thin nanometer sized layers of two materials with different bandgaps. The material with the higher bandgap is referred to as

7

Figure 6. Typical layered structure of an AlGaN-based LED along with the challenging requirements needed to optimize efficiency in each layer.1

Al.6Ga.4NAl.6Ga.4N Al.6Ga.4NAl.6Ga.4N Al.7Ga.3NAl.7Ga.3NAl.7Ga.3N

the barrier region while the material with the lower bandgap is referred to as the well region. In the case of an AlGaN LED with alternating AlxGa1-xN/ AlyGa1-yN layers, the layer with the higher Al content would be the barrier since it has a higher bandgap (reference Figure 3.) Figure 7 gives an example of relative well and barrier sizes and compositions in a typical AlGaN MQW.

3.4 Quantum Efficiencies of LEDs3.4.1 External Quantum Efficiency

There are two main ways to quantify the efficiencies of LED devices similar to that shown in Figure 6. The first is called the EQE (external quantum efficiency.) The external quantum efficiency is the result of many factors within the AlGaN device. The EQE can be thought of the amount of photons extracted from the device per electron/hole injected into the device, or as the ratio of the energy of photons emitted from the devices versus the energy of the carriers injected. It turns out that many photons produced from radiative recombination with the MQW region never exit the device. This is largely due to the large difference in the refractive index of AlGaN alloys and air. This large difference causes Fresnel losses and total internal reflection, allowing the photon to be reabsorbed by the device and converted to heat. Lenses with large indexes of refraction and wide-bandgaps with geometries that reduce the angle at which photons attempt to escape from the interface can help prevent these losses. Even with many state of the art tactics employed, one can see in Figure 10 that the highest EQE achievable at wavelengths below 350nm is about 10%. This is small compared to the EQEs of GaN/InGaN heterostructures greater than 80%.1 Thus, researchers must face every challenge shown in Figure 6 if they are to greatly increase the EQE. The EQE can defined by the equation below where ηext is the EQE and ηint is the IQE (internal quantum efficiency.)1

Eq. 1 ηext¿ Photons Extracted ¿Device ¿Carriers Injected IntoDevice

=¿ηint * Extraction

Efficiency * Injection Efficiency

3.4.2 Internal Quantum EfficiencyNearly every layer in Figure 6 directly contributes to extraction and injection efficiency;

however, only the MQW region directly contributes to the IQE. The internal quantum efficiency is a measure of how many photons are produced via radiative recombination within the MQW per injected electron. Thus, it is a measure of the ratio of radiative events to nonradiative events. The time between radiative events is τ Rad while the time between nonradiative events is τ Nonrad. In a nonradiative event, when an electron and hole recombine they release energy in the form of heat instead of light. In order to optimize the IQE, we want to minimize τ Rad

8

Figure 7. Typical compositions and structure of the MQW region of AlGaN-based LEDS.19

and maximize τ Nonrad. The IQE can be defined in terms of the radiative and nonradiative lifetimes seen in the equation below.17,18

Eq. 2 ηint ¿Photons Created∈the DeviceCarriers Injected Into Device =(1+τ Rad

τ Nonrad)−1

3.4.2.1 The Role of Defects on the IQENonradiative recombination can be further classified between Auger recombination and

recombination through defects, also known as Shockley-Read-Hall (SRH) recombination.20,21 In Auger recombination, when an electron and hole recombine they give their energy to a third carrier. This carrier then thermalizes back down to its respective valence or conduction band edge. Auger recombination increases with increased carrier concentration via increased doping or under high injection currents. Auger recombination is widely considered to be an unavoidable consequence of intrinsic material properties. SRH recombination, on the other hand, is a result of defects within the material which is an extrinsic property. Both threading dislocations22 and point defects17,18 cause act as trapping sites that introduce energy levels within the bandgap of the material. Excited electrons (hole) move down (up) from the conduction (valence) band where they are trapped. If an oppositely charged carrier then moves to the same energy level before the trapped carrier is thermally re-emitted into its respective band, they recombine. Figure 8 visualizes radiative (left) and SRH (right) recombination processes. Unsurprisingly, high concentrations of defects cause high levels of SRH recombination. For this reason, many groups are focused on growth processes in an effort to reduce the number of defects and thus improve IQE.15-18,21 If the number of defects are reduced,

9

Figure 8. Illustration of the differences between radiative recombination (left) and defect-assisted (SRH) nonradiative recombination (right).20

it is observed that the time between nonradiative events, τ Nonrad, is increased. In order to achieve IQEs over 60%, TDD (threading dislocation density) must be below 5 x 108 cm-2. This has been achieved in recent years by the development of low-TDD AlN templates on sapphire substrates, shown as the bottom two layers in Figure 6. However, even with TDDs less than 5 x 108 cm-2 DUV EQEs are still limited below 10%.17,18 In order to achieve higher IQEs (and EQEs) the concentration of point defects must also be reduced. Cation vacancy complexes (i.e. VGa-X and VAl-X) are thought to be the main NRCs (nonradiative recombination centers) in III-nitrides. These point defects are present not only within the MQW region, but also the other regions. In the n-type region, Chichibu et. al have studied the effects of Si doping.17,18 As previously discussed, Si doping releases an electron into the conduction band when the in the Si4+ ion sits on a 3+ cation site. However, when doping a material it is a concern that it will encourage the formation of native defects with a counter-charge due to the Fermi-level effect. In the case of AlGaN this would mean the formation of cation vacancies which would act as NRCs. However, after intensive study it was observed that Si doping up to 2 x 1017 cm-3 actually reduces the number of cation vacancies in AlGaN. It is thought that Si doping provides wetting conditions on the surface during growth due to the strong Si-N bonds formed which increases quantum well width uniformity. This in turns causes the amount of NRCs to decrease. However, Si doping above 2 x 1017 cm-3 begins to be dominated by the Fermi-level effect which causes NRCs to be formed.16-18,20

10

3.4.2.2 The Role of Polarization and the QCSE on the IQEReducing the amount of defects increases τ Nonrad which increases the IQE. In order to

further in increase the IQE we must also decrease τ Rad. This can be done by increasing carrier concentrations in the active region which can be accomplished through quantum confinement, which is the reason quantum wells are employed in LEDs.3,10,11

11

Figure 9. Cathodeluminescence data showing the effects of silicon doping in AlGaN MQW structures.17

Figure 10. External quantum efficiencies achieved at various wavelengths by various organizations.1

The purpose of quantum wells is to confine electrons and holes together. This confinement increases the concentration of holes and electrons in the well regions which increases radiative emission rates. An idealized case of this is shown on the right in Figure 11. There are two wavefunctions, red and blue, which correspond to the average position of an ensemble of electrons and holes in their first excited state. These wavefunctions are derived from Schrodinger’s equation. The square of these wavefunctions results in a curve similar to the original and represents the probability being at some position between 0 and x, where x is the length of the quantum well. It is known that the overlap of the square of the electron and hole wavefunctions is directly proportional to emission probability, resulting from Fermi’s golden rule. In the idealized case on the right in Figure 11 we would expect an overlap of unity. However, the right side of Figure 11 is for a structure with no internal electric field. In the case of group-III nitride heterostructures we have a situation similar to the left side of Figure 11 which is a result of the spontaneous and piezoelectric fields. This is a consequence of a phenomena known as the quantum confined stark effect (QCSE.)10,11

Within the well we have regions of high and low potential. Due to the opposite charges of holes and electrons, this causes them to move to opposite sides of the well. This is seen on the left side of Figure 11 as the shift of the electron and hole wavefunctions to opposite sides of the well. This causes the overlap of the square of the wavefunctions to be decreased and thus reduces the emission probability. Stated another way, since holes and electrons are spatially

12

LargeOverlap SmallOverlap Large τ Rad Small τ Rad

Figure 11. Conduction and valence band edges in a MQW with the first excited state electron (red) and hole (blue) wavefunctions. The right side shows perfect overlap of the wavefunctions when no external, spontaneous or piezoelectric fields are present. The left side shows reduced overlap of the wavefunctions due to the effects of spontaneous and piezoelectric fields which are present in AlGaN MQWs.23

separated at opposite sides of the well, the time it takes for a hole and electron to find each other, τ Rad, is increased.10,11

A number of methods have been proposed to deal with the separation of electrons and holes due to the QCSE in group-III nitride heterostructures. One method is to grow devices along non-polar or semipolar directions as opposed to traditional growth along the polar c-axis. Nonpolar directions would not be effected by the QCSE and would thus have their overlap expected to return to unity similar to that in the right side of Figure 11. The electric field in semipolar directions has only be a fraction of the component of c-axis grown devices, meaning the severity of the QCSE is not as strong. However, as the polarization is reduced the stability of growth in that direction tends to decrease. This means that attempting to growth in directions other than the c-axis tends to cause formation of defects which can act as NRCs.11

Another possible method is the growth of cubic phase AlGaN, which has no polarization in any direction; however this approach is also limited for two reasons. Firstly, growth is difficult due to lack of a suitable substrate under high temperature growth conditions. More importantly, the band gap of cubic phase group-III nitrides is lower than that of their wurtzite counterparts (shown in Figure 3), especially in the case of AlN.24

A third method is the use of extremely small well widths (<.2nm.) By reducing the well width, electron and hole wavefunctions are forced to overlap due to the limited space they can occupy. The solid black triangles in Figure 13 show how the overlap integral falls off steeply at well widths above 2nm in AlGaN MQW structures modeled by SiLENSe. However, such small well widths are undesirable for a number of reasons. Small well widths are difficult to achieve from a growth perspective since wells under 2nm are only several monolayers thick. Additionally, small well widths causes extremely high carrier densities even under low injection currents. This causes efficiency drooping due to nonradiative Auger recombination which prevents high power operation. Larger well widths must be used to lower the volume carrier density in order to use devices for high power applications under high injection current.10,11

13

4.0 Motivation for Research: A Method to Increase the

Overlap Integral in MQWsA method has been proposed to increase the overlap integral of AlGaN MQW structures

while maintaining relatively large well widths by modulating the alloy composition.25 Figure 12 shows the model of an alloy unmodulated sample on the left and a modulated sample on the right. The well widths and barriers widths are 4 nm and 7 nm respectively for both samples. On the bottom of the figure, the mole fraction of AlN is displayed. In the unmodulated sample, there are only two different alloy compositions, 0.7 and 0.6, in the barrier and well, respectively. In the modulated sample, it can be seen that the AlN mole fraction modulation was quadratic, the only fixed parameter being the center well composition of .6 AlN mole fraction. In the upper part of the figure, the valence and conduction bands as well as the wavefunctions for holes and electrons are shown. In the unmodulated sample, the conduction band minimum (CBM) and valence band maximum (VBM) are at different positions along the x-axis. In the modulated sample however, the location of the CBM and VBM coincide at the left side of the well. It is clear that the overlap of the wavefunctions in the modulated sample is much greater than that of the unmodulated sample. An overlap of .27 was calculated for the unmodulated sample and an overlap of .93 was calculated for the modulated sample. This gives clear evidence that alloy modulation may be used to reverse the spatial separation of electron and hole wavefunctions due to the QCSE.25

14

Figure 12. Band edges and resulting electron and hole wavefunctions modeled by SiLENSe for an unmodulated MQW structure (left) and an alloy modulated MQW structure (right.) The mole fractions of AlN used in each structure are displayed at the bottom of the figures.25

-5.4

-5.2

-5.0

-0.6

-0.4

-0.2

Pote

ntial

(eV)

I 2=0.27

-10 -5 0 5 100.40.50.60.70.8

mole

frac

tion

Position z (nm)

AlN

Squa

red

wave

fucti

on

-5.4

-5.2

-5.0

-0.6

-0.4

-0.2

Pot

entia

l (eV

) I2=0.93

-10 -5 0 5 100.4

0.5

0.6

0.7

0.8

mol

e fra

ctio

n

Position z (nm)

AlN

Squ

ared

wav

efuc

tion

-5.4

-5.2

-5.0

-0.6

-0.4

-0.2

Pot

entia

l (eV

) I2=0.93

-10 -5 0 5 100.4

0.5

0.6

0.7

0.8

mol

e fra

ctio

n

Position z (nm)

AlN

Squ

ared

wav

efuc

tion

n 層側p 層側

0 2 4 6 8 100.0

0.2

0.4

0.6

0.8

1.0

Quadratic Modulation Standard Linear modulation

Ove

rlap

inte

gral

I2

Well width (nm)

Fig. 3

LCM

No modulation

QCM

4

0 2 4 6 8 100.0

0.2

0.4

0.6

0.8

1.0

Quadratic Modulation Standard Linear modulation

Ove

rlap

inte

gral

I2

Well width (nm)

Additional models were created for well widths up to 10nm. The calculated overlap integrals as a function of well width are plotted in Figure 13. The solid red circles represent modulated samples while the solid black triangles represent the unmodulated samples. The modulated sample shows increased overlap at every well width, and the difference is more pronounced at larger well width. At well widths greater than 8nm, the overlap of the unmodulated sample falls to virtually zero, while the overlap of the modulated sample remains above .8.25

These encouraging results demand further exploration into the use of alloy modulation to improve the IQE of AlGaN MQWs. This research includes growth and testing of the modulated and unmodulated samples shown in Figure 13, as well as further extensions and improvements to the models shown.

5.0 Experimental ProcedureTwo (modulated and unmodulated) 2nm well samples were grown via MOCVD in by

members of Miyake-sensei’s group from Mei University and then sent to Chichibu-sensei’s lab for characterization. These two samples were grown to have composition profiles similar to those in Figure 12, except the well widths are 2nm. The two characterization methods that have completed thus far are CL (cathode luminesce) spectroscopy and TRPL (time-resolved photoluminescence) spectroscopy.

15

Figure 13. Overlap integrals of unmodulated (black triangles) and modulated (red circles) structures with various well widths modeled by SiLENSe.25

5.1 CL SpectroscopyThe first characterization method used on the samples was CL spectroscopy. CL

spectroscopy bombards the sample with a high energy beam of electrons which excites electrons/holes in the conduction/valence bands where they then recombine over time. The intensity (quantity) and wavelength of photons produced from this recombination process are recorded, allowing the emission spectra of the sample to be calculated. Samples were tested at incremental temperatures between 10 and 300K. The electron beam was operated at 3.5kV acceleration voltage. This procedure was performed a total of 4 times. The unmodulated sample was tested once while the modulated sample was tested three times. The repeated testing of the modulated sample was due further investigate and confirm results from the previous tests.

5.2 TRPL SpectroscopyThe second characterization method used was TRPL spectroscopy. This method uses a

laser with photon energy below that of the band gap of the material to excite carriers within the sample. The carriers then recombine over time to emit light or heat. The light is collected similarly to that of CL spectroscopy. The difference between the methods is that TRPL simultaneously measures the time it takes for photons to reach the detector after excitation in addition to their wavelength. This allows luminescence intensity maps to be created as a function of time and energy. This was performed on both samples at temperatures of 10K and 300K. The excitation source was a frequency-quadrupled mode-locked Al2O3:Ti laser with 200 fs pulses. The wavelength of the frequency-quadrupled pulse is 200nm, ensuring its energy is well below the bandgap of the samples.

6.0 Data and Analysis6.1 CL Spectroscopy Data and Results

Figure 14 shows emission spectra of the modulated and unmodulated samples. The unmodulated sample was only tested once while the modulated sample was tested three times. The curves on each graph are arranged so that the highest temperature curves are at the top and descend to the lowest temperatures at the bottom. The intensities of the curves were multiplied by factors that increase with increasing temperature so that the curves do not overlap, allowing for easier analysis.

16

The unmodulated sample has a peak wavelength of nearly 5eV at 12K and decreases gradually with increasing temperature up to 300K. This redshift was observed in all four tests. It is a result of the lattice expanding with increasing temperature. As the distance between the nuclei of the atoms increases, the potential seen by electrons decreases which causes a decrease in the band gap of the material. In the unmodulated sample we also see the presence of additional peaks that become present at higher temperatures. This is due to defects becoming activated that create deep states within the energy gap. At room temperature operation around 300K these defects are fully activated.17,18

The modulated samples were tested three times due to some peculiarities in their emission spectra. It can be seen that the width of the emission spectra of the modulated sample was larger than that of the unmodulated sample in all three tests. This is not surprising

17

Figure 14. CL spectroscopy data for an unmodulated (top-left) and a modulated sample (remaining three) from 12K at the bottom of each figure to 300K at the top.

4.5 5.0

CL

Inte

nsity

(Arb

itrar

y U

nits

)

Photon Energy (eV)

300K 275K 250K 225K 200K 175K 150K 125K 100K 75K 50K 40K 30K 25K 20K 15K 12K

Modulated Sample v2

4.5 5.0

CL

Inte

nsity

(Arb

itrar

y U

nits

)

Photon Energy (eV)

300K 275K 250K 225K 200K 175K 150K 125K 100K 75K 50K 40K 30K 25K 20K 15K 12K

Unmodulated Sample

4.5 5.0

300K 275K 250K 225K 200K 175K 150K 125K 100K 75K 50K 40K 30K 25K 20K 15K 12K

CL

Inte

nsity

(Arb

itrar

y U

nits

)

Photon Energy (eV)

Modulated Sample

4.5 5.0

CL

Inte

nsity

(Arb

itrar

y U

nits

)

Photon Energy (eV)

300K 275K 250K 225K 200K 185K 175K 165K 150K 135K 125K 110K 100K 85K 75K 50K 40K 30K 25K 20K 15K 12K

Modulated Sample v3 (large spot)

as the modulation of the sample causes the material to have varying bandgap energies within the well. In the first and second modulated samples test, it appears as though a higher energy peak appears as temperature is increased. This may be due to high energy electrons at high temperatures being able to hop into the barrier region and recombine there. This would help explain the appearance of the high energy peak since the barrier has a higher bandgap than the well region. Another likely reason is that there are inhomogeneities in the sample. Through visual inspection it is easy to see contrasts between the first and second modulated sample tests. This prompted a third test of the modulated sample, but this time under larger spot size. By increasing the spot size of the electron beam, we are able to cause radiative emission to occur from a larger portion of the sample. In this way, it is possible to test multiple areas of the sample that may not be of the same composition. When we did this, a very clean spectrum was produced with only one distinct peak. It appears as though the spectrum produced from the modulated sample lies somewhere between the first and second tests, but it not conclusive evidence the sample contains inhomogeneities.

6.2 TRPL Spectroscopy Data and ResultsFigure 15 shows an example of the image obtained by the streak camera from TRPL

testing of the modulated sample at 300K. Each purple point represents a photon. The intensity ranges from green, blue, purple and black with green being the highest. Data similar to the graphs in Figure 14 were obtained for the modulated and unmodulated samples at 10K and 300K for a total of 4 data sets. From these graphs, it is possible to extract meaningful information.

18

Figure 15. Streak camera image from TRPL testing of a modulated AlGaN Sample at 300K.

Wavelength (nm)

Time (ns)

By integrating the data in Figure 15, it is possible to obtain intensity versus energy plots similar to those in Figure 14. The data was integrated over 2 ns time intervals as well as over the entire decay time (~8 ns) for all 4 data sets. The results of this are represented in Figure 16. The differences in overall intensities are very similar when comparing between the modulated and unmodulated sample at both temperatures. Similar to what was observed in Figure 14, we also see broader emission spectrums in the modulated sample. Additionally, it appears as though the modulated samples have a small extremely long decay component evidenced by the low energy tails in the modulated data. The 1/e values plotted with their values on the secondary (right hand) y-axis are proportional to the calculated delay lifetimes.

19

4.4 4.6 4.8 5.0 5.2

Modulated 10K

Inte

nsity

(arb

. uni

ts)

Photon Energy (eV)

0ns to 2ns 2ns to 4ns 4ns to 6ns 6ns to 8ns 0ns to 8 ns

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1/e

4.4 4.6 4.8 5.0 5.2

Unmodulated 10K

Inte

nsity

(arb

. uni

ts)

Photon Energy (eV)

0ns to 2ns 2ns to 4ns 4ns to 6ns 6ns to 8ns 0ns to 8 ns

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1/e

4.4 4.6 4.8 5.0 5.2

Modulated 300K

Inte

nsity

(arb

. uni

ts)

Photon Energy (eV)

0ns to 2ns 2ns to 4ns 4ns to 6ns 6ns to 8ns 0ns to 8 ns

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1/e

4.4 4.6 4.8 5.0 5.2

Unmodulated 300K

Inte

nsity

(arb

. uni

ts)

Photon Energy (eV)

0ns to 2ns 2ns to 4ns 4ns to 6ns 6ns to 8ns 0ns to 8 ns

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1/e

Figure 15. Intensity versus wavelength plots of radiative emission at 2 ns intervals of the unmodulated (left) and modulated (right) samples at 10K (top) and 300K (bottom.)

Intensity versus time plots at various energies were also calculated by integrating over 1 nm resolutions at photon energies between 4.75-5eV at .05eV intervals. The results are shown in Figure 16. The larger the slope of the curve, the faster the decay component. The slope is related to the decay lifetime through the following equation where A1 and A2 are preexponential constants, t is time and τ 1and τ 2 are the fast and long lifetimes.

Eq. 3 I ( t )=A1exp(−tτ 1 )+A2 exp(−tτ 2 )The lifetime just after excitation, τ 1, is dependent on the radiative and nonradiative lifetimes through the equation:

Eq. 4 1

τ 1= 1

τ Rad+ 1

τ Nonrad

20

0 2 4 6 8

Unmodulated 300K

Inte

nsity

(arb

. uni

ts)

Time (ns)

5eV 4.95eV 4.9eV 4.85eV 4.8eV 4.75eV

0 2 4 6 8

Unmodulated 10K

Inte

nsity

(arb

. uni

ts)

Time (ns)

5eV 4.95eV 4.9eV 4.85eV 4.8eV 4.75eV

0 2 4 6 8

Modulated 10K

Inte

nsity

(arb

. uni

ts)

Time (ns)

5eV 4.95eV 4.9eV 4.85eV 4.8eV 4.75eV

0 2 4 6 8

Modulated 300K

Inte

nsity

(arb

. uni

ts)

Time (ns)

5eV 4.95eV 4.9eV 4.85eV 4.8eV 4.75eV

Figure 16. Intensity vs time plots at various emission energies for the unmodulated (top) and modulated (bottom) samples at 10K (left) and 300K (right.)

The results between the modulated sample and unmodulated samples are extremely similar at both temperatures. The only remarkable difference is the small intensity of the 5eV curves in the modulated sample. This is due to the modulated sample having a slightly lower energy emission peak. Because of this, there were few photons emitted (and hence collected) at higher energies for the modulated sample.

21

0 2 4 6 8

Modulated and Unmodulated at 300K

*Unmodulated 300K intensity multiplied by factor of 2

1 = .657 ns

2 = 1.961 ns

1 = .519 ns

2 = 1.365 ns

Inte

nsity

(arb

. uni

ts)

Lifetime (ns)

Modulated 300K Unmodulated 300K

0 2 4 6 8

Unmodulated Samples

1 = .657 ns

2 = 1.961 ns

1 = .645 ns

2 = 3.963 ns

Inte

nsity

(arb

. uni

ts)

Lifetime (ns)

Unmodulated 10K Unmodulated 300K

*300K intensity multiplied by factor of 2

0 2 4 6 8

1 = .642 ns

2 = 2.714 ns

Modulated Samples

1 = .519 ns

2 = 1.365 ns

Inte

nsity

(arb

. uni

ts)

Lifetime (ns)

Modulated 10K Modulated 300K

0 2 4 6 8

Modulated and Unmodulated at 10K

1 = .642 ns

2 = 2.714 ns

1 = .645 ns

2 = 3.963 ns

Inte

nsity

(arb

. uni

ts)

Lifetime (ns)

Modulated 10K Unmodulated 10K

Figure 17. Intensity vs time graphs over the entire range of emitted wavelengths for the modulated and unmodulated samples at 10K and 300K.

Intensity versus time graphs similar to those in Figure 16 were also made with integration over the entire range of energies which can be seen in Figure 17. The colors associated with the curves are associated with the same data set in each graph. By taking the log of equation 3 and fitting curves to each data set, it was possible to extrapolate τ 1 and τ 2 which are shown for on each graph in Figure 17. Unfortunately, data for τ 2 is somewhat unreliable due to the peak at 3.9 ns which was due to an error in the instrument. However, it is still possible to draw some conclusions about the slow decay component through visual inspection. In every graph, we see differences in the slower decay component. 300K slow decay lifetimes are larger than the 10K decay lifetimes as can be seen in the top two graphs. This is expected as trapping sites are activated at room temperature which cause slow decay.17,18 The initial (fast) decay lifetimes also appear to be very similar in all of the comparisons except for the modulated and unmodulated samples at 300K in the bottom right graph. Here we see the initial decay is faster in the modulated sample. Since the initial decay τ 1 is dependent on both τ Rad and τ Nonrad it is not possible to determine whether this difference is due to an increase in overlap due to modulation, or an increase in defects as a result of the attempted modulation. If overlap is increased, we expect to see a reduction in τ Rad but if the amount of NRCs were increased then τ Nonrad would be decreased.10,11,25

7.0 Discussion and Future TestingBoth CL and TRPL spectroscopy revealed the differences in luminescence properties of

the unmodulated and modulated samples to be similar. One difference included the lower peak energy emission of the modulated samples as well as overall broader emission wavelengths. This was attributed to the variation in the bandgap of the modulated sample in the well region. However, some small variations were discovered from both CL and TRPL studies.

The variations in the CL spectroscopy data after repeated testing leads to the hypothesis that there may be inhomogeneities in the sample. To further explore this possibility, spatio-time-resolved cathodoluminescence studies will be used.26 In spatio-time-resolved cathodoluminescence, the standard electron beam in an SEM is replaced with a pulsed electron beam that is created from a laser exciting a photocathode in short bursts. The pulses of electrons are directed at the sample. These electrons simultaneously excite the sample to produce luminescence as well as scatter so they can be collected to produce an image. This allows for luminescence testing of very small areas of the sample (< 10nm.) By testing various locations of the sample and comparing luminescence data, we can determine if there are inhomogeneities.

In regards to TRPL we saw very few differences in the fast decay component in three of the comparative graphs. A very likely reason for this is that the samples were tested had well widths of 2 nm. The models shown in Figure 12 the wells were 4nm instead of 2nm. Looking at Figure 13, the difference in the overlap integral of modulated and unmodulated sample is only .2. This is due to the very small confinement in the 2 nm that forces electron and hole wavefunctions to overlap despite the QCSE. This slight difference would make it difficult to observe much change in the luminesce properties between the samples. Future plans are to

22

test samples with larger well widths, starting with 4 nm, in order to observe a larger difference between the modulated and unmodulated samples.

In the bottom right graph of Figure 17 comparing the modulated and unmodulated samples at 300K, we see the initial decay of the modulated sample is faster. The dependence of τ 1 on τ Rad and τ Nonrad in equation 3 make it difficult to determine the origin of this faster decay. If the intent of the modulation was successful, the decrease in τ 1 was due to the decrease in τ Rad due to the increased overlap. It is also possible that the decrease in τ 1 was a result of increased defects as a consequence of the modulation, which would decrease τ Nonrad. In order to determine the cause of the decrease in τ 1 we must determine τ Rad and τ Nonrad. This can be accomplished by determining the IQE of the samples. Recall that the IQE in equation 2 is a function of τ Rad and τ Nonrad. We already have values of τ 1 which is also a function of τ Rad and τ Nonrad as can be seen in equation 4. Thus once we calculate the IQE, we will have two equations in which we can use to calculate our two unknowns τ Rad and τ Nonrad. The IQE can be calculated from ODPL (omnidirectional photoluminescence) spectroscopy using an integrating sphere. The integrating sphere allows simultaneous counting of the photons from the excitation source (laser) and from the sample. The ratio of photons produced from the sample to the number of photons absorbed by the sample can then be used to derive the IQE.

8.0 References(1) Kneissel, M. A Brief Review Of III-Nitride UV Emitter Technologies And Their Applications. III-

Nitride Ultraviolet Emitters 2015, 1-25.(2) Nakamura, S.; Mukai, T.; Senoh, M. High-Power Gan P-N Junction Blue-Light-Emitting

Diodes.Jpn. J. Appl. Phys. 1991, 30, L1998-L2001.(3) Nakamura, S.; Senoh, M.; Mukai, T. P-Gan/N-InGan/N-Gan Double-Heterostructure Blue-

Light-Emitting Diodes. Jpn. J. Appl. Phys. 1993, 32, L8-L11.(4) Lange, M. Ultraviolet Light-Emitting Diodes For Water Disinfection. III-Nitride Ultraviolet

Emitters 2015, 267-288.(5) Dreyer, C. Applications Of Leds For UV-Curing. III-Nitride Ultraviolet Emitters 2015, 415-432.(6) Stericase http://www.innopharm.com.my/stericase.html (accessed Jul 7, 2016).(7) Grandusky, J. Fabrication Of High Performance UVC Leds On Aluminum-Nitride

Semiconductor Substrates And Their Potential Application In Point-Of-Use Water Disinfection Systems. III-Nitride Ultraviolet Emitters 2015, 171-191.

(8) UV-Light Interacting With A DNA Molecule; 2016.(9) Bandgap Vs Lattice Constant Of Group-III Nitrides; 2016.(10) Deguchi, T.; Sekiguchi, K.; Nakamura, A.; Sota, T.; Matsuo, R.; Chichibu, S.; Nakamura, S.

Quantum-Confined Stark Effect In An AlGan/Gan/AlGan Single Quantum Well Structure. Jpn. J. Appl. Phys. 1999, 38, L914-L916.

(11) Speck, J.Chichibu, S. Nonpolar And Semipolar Group-III Nitride-Based Materials. MRS Bull. 2009,34, 304-312.

(12) Sharma, T.Towe, E. Impact Of Strain On Deep Ultraviolet Nitride Laser And Light-Emitting Diodes. J. Appl. Phys. 2011, 109, 086104.

(13) Crystal Structure And Lattice Constants Of Algan; 2016.

23

(14) File:PnJunction-LED-E.svg - Wikimedia Commons https://commons.wikimedia.org/wiki/File:PnJunction-LED-E.svg (accessed Jul 7, 2016).

(15) Richter, E. Vapor Phase Epitaxy Of Algan Base Layers On Sapphire Substrates For Nitride-Based UV-Light Emitters. III-Nitride Ultraviolet Emitters 2015, 47-70.

(16) Hirayama, H. Growth Techniques Of Aln/Algan And Development Of High-Efficiency Deep-Ultraviolet Light-Emitting Diodes. III-Nitride Ultraviolet Emitters 2015, 75-112.

(17) Chichibu, S. Impacts Of Dislocation And Point Defects On The Internal Quantum Efficiency Of The Near-Band-Edge Emission In Algan-Based DUV Light-Emitting Diodes. III-Nitride Ultraviolet Emitters 2015, 1-25.

(18) Chichibu, S.; Miyake, H.; Ishikawa, Y.; Furusawa, K.; Hiramatsu, K. Reduction In The Concentration Of Cation Vacancies By Proper Si-Doping In The Well Layers Of High Aln Mole Fraction Alxga1–Xn Multiple Quantum Wells Grown By Metalorganic Vapor Phase Epitaxy. Appl. Phys. Lett. 2015, 107, 121602.

(19) Compound Semiconductor, E-H Recombination In A Quantum Well; 2016.(20) Non-Radiative Recombination In Leds; 2016(21) Dreyer, C.; Alkauskas, A.; Lyons, J.; Speck, J.; Van de Walle, C. Gallium Vacancy Complexes

As A Cause Of Shockley-Read-Hall Recombination In III-Nitride Light Emitters. Appl. Phys. Lett.2016, 108, 141101.

(22) Speck, J. The Role Of Threading Dislocations In The Physical Properties Of Gan And Its Alloys.Materials Science Forum 2001, 353-356, 769-778.

(23) Compound Semiconductor,. Quantum Confined Stark Effect.(24) Okumura, H.; Ohta, K.; Feuillet, G.; Balakrishnan, K.; Chichibu, S.; Hamaguchi, H.; Hacke, P.;

Yoshida, S. Growth And Characterization Of Cubic Gan. Journal of Crystal Growth 1997, 178, 113-133.

(25) Kojima, K. A Design Strategy To Achieve More Than 90% Of The Overlap Integral Of Electron And Hole Wavefunctions In AlGaN Multiple Quantum Wells. NP 2016.

(26) Chichibu, S.; Hazu, K.; Ishikawa, Y.; Tashiro, M.; Ohtomo, T.; Furusawa, K.; Uedono, A.; Mita, S.; Xie, J.; Collazo, R. et al. Excitonic Emission Dynamics In Homoepitaxial Aln Films Studied Using Polarized And Spatio-Time-Resolved Cathodoluminescence Measurements. Appl. Phys. Lett. 2013, 103, 142103.

24