Study on ingan gan quantum structures and their applications in semiconductor saturable absorber mirror 2

Chapter 2 Growth and Characterizations In this work, the GaN-based quantum wells and quantum dots were all grown by an EMCORE D125 metal organic chemical vapor deposition MOCVD system..

Chapter 2

Growth and Characterizations

In this work, the GaN-based quantum wells and quantum dots were all grown by an EMCORE D125 metal organic chemical vapor deposition (MOCVD) system Section 2.1 will present the essential theories of this growth technique and highlight the major components of our MOCVD system To fabricate a broadband GaN-based SESAM, the SiO2/Si3N4 dielectric DBR was deposited by the plasma enhanced chemical vapor deposition (PECVD) system The principle of the PECVD technique will be presented in Section 2.2 Section 2.3 will be devoted to the primary characterization techniques used in this work, including photoluminescence (PL) (Section 2.3.1), spectrophotometry (Section 2.3.2), atomic force microscopy (AFM) (Section 2.3.3), scanning electron microscopy (SEM) (Section 2.3.4), transmission electron microscopy (TEM) (Section 2.3.5), and X-ray diffraction (XRD) (Section 2.3.6) The PL technique was used to investigate the emission properties of the quantum well and quantum dot samples To study the linear transmission and reflection spectra of different samples, the spectrophotometry technique was applied

and the values of surface roughness can be obtained But the accuracy of the surface profiles obtained by this technique depends on the size of the tip compared to the sizes and aspect ratios of the surface features In comparison, the SEM technique gives more trustable images than the AFM technique, but the values of surface roughness cannot be obtained by using SEM In addition, the cross-sectional SEM technique was also used to examine the quality of the interfaces and estimate the layer thickness When the feature sizes scale down to a few nanometers or tens of nanometers, the TEM technique was applied to study the nanostructures Either the plan-view or the cross-sectional crystal structures can be investigated Also, to precisely study the crystal quality, structural property and the stress development of the samples, the XRD technique was applied

2.1 Metal organic chemical vapor deposition (MOCVD)

The MOCVD growth, also called metal organic vapor phase epitaxy (MOVPE), is a non-equilibrium thin film epitaxial growth technique It relies on the vapor transport of precursors and the subsequent chemical reactions of the precursors

in a heated zone The reaction chamber of an MOCVD system can either be a quartz tube or a stainless steel chamber, which contains a heated substrate susceptor The hot susceptor has a catalytic effect on the decomposition of the gaseous products and the following growth of the materials on this hot surface

Similar to other crystal growth techniques, fundamental processes in MOCVD are commonly divided into thermodynamic and kinetic regimes

Thermodynamics determines the driving force for the overall growth process, while kinetics defines the rates at which the various processes occur Hydrodynamics and mass transport, which are intimately linked, control the material transportation rate to the interface between the growing solid and the vapor In the meantime, the chemical reaction rates at the growing interface also play a role Actually, each of these factors dominates a certain stage of the overall growth process Since the last decade, MOCVD has become the most established growth technique for III-nitride material growth and device application [Dupuis1997; Keller2003]

Figure 2.1 Simplified schematic diagram of GaN growth process

Figure 2.1 shows the simplified schematic diagram of the GaN growth

process Although the actual growth process of GaN is complicated, it basically consists of the following four steps: 1) Ga(CH3)3 diffusion through the boundary layer

to the substrate, 2) surface reactions, 3) formation of GaN, and 4) removal of the reaction products

All the GaN-based samples discussed in this thesis were grown with an EMCORE D125 vertical-geometry rotating-disk reactor (RDR) MOCVD system The schematic diagram of this system is shown in Fig 2.2 This system can be grouped into four major parts, i.e., (1) the gas handling system consisting of alkyl and hydride sources, necessary gases and all instruments used to control the gas flows and mixtures; (2) the reaction chamber, where the pyrolysis reaction and deposition occur, and the loadlock; (3) the heating and temperature controlling system; and (4) the exhaust, pumping and pressure controlling system [Ludowise1985]

other necessary gases), and the gas mixing system (or manifold) The purpose of the gas handling system is to deliver precisely metered amount of uncontaminated reactants into the growth chamber without the transients due to the changes of pressure or gas flows In this work, trimethylgallium (TMGa), trimethylindium (TMIn) (from Epichem) were used as the group-III precursors of gallium (Ga) and indium (In), respectively Highly purified ammonia (NH3) (ammonia blue from Solkatronic) was used as the group-V precursor of nitrogen (N) Highly purified H2 and highly purified

N2 (from National Oxygen) were used as the carrier gases Extra purifiers are also used for further reduction of oxygen and other contaminants in the highly purified gases of NH3, N2 and H2 to the order of ppb (parts per billion) They are Matheson Nanochem NH3 purifier, Mykrolis Aeronex N2 purifier and Johnson Mathey H2

purifier

In order to control the reaction precisely, a specially designed mass flow controllers (MFC) can accurately and reliably measure and control the molar flow rates of the gases For gas phase sources such as NH3, H2 and N2, which are stored in pressurized cylinders, the MFC can be solely used to control the molar mass flow rates precisely The maximum flow rates for NH3, H2 and N2 are 20 L, 10 L and 10 L, respectively For solid or liquid phase sources such as metalorganic precursors of TMGa and TMIn, which are stored in the stainless-steel cylinders called bubblers, the exact amount of sources is also controlled by the MFC, as well as the pressure controller and the temperature controller (thermal bath) More specifically, the partial pressure of the source vapor is actually regulated by precisely controlling the

temperature of the metal organic source bubbler, and the temperature of metalorganic

sources can be regulated by means of precisely controlling the temperature of the

thermal bath where the metalorganic source bubbler is held In addition, the pressure

of the bubbler is regulated by the pressure controller Thus, a controlled amount of

metalorganic precursor can be transported by controlling the exact amount of

abducting gas flow through the bubbler using the MFC:

bubbler vapor

vapor abduct

MO

P P

P F

where FMO is the molar flow rate per minute of the metalorganic precursor, Fabduct is

the molar flow rate of abducting gas, Pvapor is the pressure of the source vapor and

P bubbler is the pressure of the bubbler In our system, there are two TMGa cylinders

and two TMIn cylinders With the control by MFC, the maximum flow rates are 50

sccm from the TMGa#1 cylinder, 200 sccm from the TMGa#2 cylinder, 200 sccm

from the TMIn#1 cylinder, and 1000sccm from the TMIn#2 cylinder It should be

noticed that, at normal operating temperatures, the TMGa is liquid and TMIn is solid,

and the bubbler temperatures for TMGa and TMIn are -10 oC and 30 oC, respectively

The other important part of the gas handling system is the gas mixing system,

or the manifold At the manifold, the individual reactants are either combined and

switched to the growth chamber or vented to the exhaust line The manifold also

keeps the delivery pressures of the growth chamber and exhaust line in balance This

pressure balance is very important to stabilize the gas flow to obtain the atomically

abrupt interfaces in the multi-layer heterostructures In our EMCORE D125 MOCVD

system, there are three manifolds, one for the group-V precursor (NH3) and carrier

gases (H2 or/and N2), the second for the trimethylaluminum (TMAl) precursor, and the third for the other precursors The purpose of separating TMAl from other alkyl precursors is to prevent parasitic reactions of Al because Al is very reactive, and TMAl is not used in this work A push line that always injects H2 is also added to each manifold to achieve the fast delivery of the reactants through the gas tube to the growth chamber and to minimize the switching time during the growth of an abrupt interface

For the reaction chamber, our EMCORE D125 MOCVD system has a vertical-geometry RDR It is made of stainless steel and is water-cooled A loadlock maintained at high vacuum with a turbo molecular pump is connected to the reaction chamber and it can transfer wafers in/out of the reaction chamber In RDR, substrates sit on a circular horizontal disk, called a wafer carrier For our system, the wafer carrier is 5.25-inch in diameter, which allows up to three 2-inch wafers to be grown simultaneously Also, the wafer carrier is supported by a spindle which is connected to

a motor Thus, during the growth, the wafer carrier can rotate at a high speed, i.e.,

1000 revolution per minute (rpm) in our growth cases This fast rotation rate during the growth improves the gas flow hydrodynamics over the wafer carrier, leading to the improvement of the growth rate uniformity across the wafer surface

In addition, it is noticed that, for the growth of GaN-based quantum wells and quantum dots, the uniform thickness and composition are essential In the reaction chamber of our system, the top flow flange consists of several inlets, through which the precursors and carrier gases are injected downward Upon arrival in the growth

chamber, the separated alkyl and hydride source flows are distributed to different zones in the top flow flange The alkyl flow is distributed among the inner and outer zones, while the hydride and carrier gas flow is distributed among the inner, middle, and outer zones The distribution of alkyl flows to each zone is controlled with two MFCs, while the distribution of hydride and carrier gas flows is adjusted by three needle valves Hence, a customized flow distribution can be created and optimized with setting the MFCs and the individual needle valves The uniformity of growth rate and composition can therefore be controlled and optimized by setting the flow distribution

Besides the careful control of the growth rate and the composition, accurate and uniform control of the temperature is also critical For the growth of quantum wells, the small oscillation in the temperature may result in the non-uniformity of the growth rate as well as the composition The careful temperature control is even more important for the growth of quantum dots, because the temperature window for the formation of quantum dots is rather narrow Hence, in our system, a two-zone resistive heater system is used, and it consists of an inner heater and an outer heater, providing uniform temperature across the wafer carrier [Walker1995] Also, each heater has its own power supply, proportional-integral-differential (PID) temperature controller, and a thermocouple The thermocouples are installed under the heaters, which measure the corresponding heater temperature The two heaters, therefore, can

be controlled separately for better growth temperature uniformity [Gurary1994] In addition, since the thermocouples do not directly measure the temperature of the

wafer carrier, two pyrometers are also installed on the top of the reaction chamber to monitor the temperatures of the inner and outer parts of the wafer carrier separately Because the thermocouples only control the heater temperature and the pyrometers only monitor the temperature of the wafer carrier without the ability to directly control the heater temperature, a temperature calibration run is usually conducted before an actual run to ensure the better temperature control during the actual growth The temperature calibration run is exactly the same as the actual run except that no group-III precursors flow into the growth chamber Thus, with the usage of two-zone resistive heater system, the growth temperature can be controlled uniformly cross the wafer carrier up to about 1100oC

Finally, the pumping and low-pressure control system (or vacuum system) is

an important part for controlling the pressure of the reaction chamber The vacuum system basically consists of three parts, i.e., (1) a rotary pump, (2) two stage gas exhaust particle filters, and (3) a throttle valve The reaction chamber pressure is controlled by adjusting the position of the throttle valve flap with a MKS pressure controller unit The exhaust gases are pumped out of the reaction chamber by the pumps The exhaust particles are trapped by the two-stage particle filters The toxic parts in the exhaust gases are then absorbed in the toxic gas absorber unit before the exhaust gases are released into the environment

Thus, accurate thickness and composition control in the growth of GaN-based quantum wells and quantum dots can be effectively achieved through the careful control of the flow rate, temperature and the pressure in our MOCVD system

The uniform growth temperature and pressure during the MOCVD growth will also help improve the crystal quality of the GaN-based structures

2.2 Plasma enhanced chemical vapor deposition (PECVD)

The PECVD technique － i.e., film growth using gas phase precursor activated in a glow discharge environment － has been widely employed in semiconductor device fabrication for several decades It is heavily used for deposition

of nitrides and oxides The typical diagram of a standard PECVD system is shown in Fig 2.3 (a)

Figure 2.3 Diagrams of (a) stand PECVD and (b) PECVD with a plasma box

(Courtesy of Unaxis Semiconductor)

In a PECVD system, the plasma is generated by applying a radiative-frequency (RF) field to a low-pressure gas The electrons in the reactor, which gain sufficient energy from the electric field, collide with gas molecules Then

the dissociation and ionization of reactant gases occur, and the energetic species including ions and radicals are adsorbed on the surface and are able to migrate easily along the surface This explains the reason that PECVD can provide good step coverage Subsequently, the species adsorbed on the surface are subjected to the bombardment by charged species such as ions and electrons, rearranged, and reacted with other adsorbed species Hence, the thin film is finally grown [Plummer2000]

Because the RF-induced plasma supplies energy into the reactant gases, the reactions required for deposition can occur at much lower temperatures Therefore, PECVD is a low-temperature (< 400°C) process and the resulting films have one order of magnitude less stress than those grown by thermal CVD On the other hand, the use of plasma in the PECVD growth requires the reliable control of various plasma parameters such as RF power, pressure, gas flow, as well as the thermal parameters The films are usually deposited in the reaction-rate limited regime and the temperature control is especially important The advantages of PECVD also include good adhesion, good step coverage, and compatibility with small-feature sized device manufacturing

The PECVD deposition of Si3N4 and SiO2 in this work was performed using

a Nextral (NE) D200 Unaxis system This system is different from the standard PECVD system shown in Fig 2.3 (a) It applies a special reactor design – plasma box concept – where the plasma is confined in a box made of aluminum, as shown inFig 2.3 (b) The plasma box, in which the substrate holder is placed, is located in a furnace Because of this design, the accurate control of the substrate temperature is

allowed as all the surfaces in the plasma box are isothermal and equally exposed to the plasma Also, film deposition with excellent uniformities and refractive index can also be achieved because the plasma discharge is confined in a uniformly heated reactor In addition, in this system, the furnace/plasma box assembly is placed in a vacuum chamber; and a roughing pump is used for the plasma box, while a turbo and roughing pump system is used for the vacuum chamber During a process run, the two separated vacuum systems can produce up to 100 times lower pressure in the vacuum chamber than in the plasma box As a result, no contamination can enter the plasma box during the deposition and the impurities do not contaminate the deposited layers Furthermore, because the surfaces in contact with the plasma are smooth and made exclusively of aluminum, fluorine based plasma can be used to clean the plasma box after any deposition process, and no manual cleaning procedure is required

In our NE D200 Unaxis system, an air-cooled solid-state RF generator is used, which can generate an RF frequency of 13.56 MHz and have the maximum power of 300W The system is optimized to rapidly produce a high vacuum (<3.7×10-3 mTorr at 280oC) and high pumping flow rates at working pressures Also, the gas box is equipped with six gas lines, including SF6, N2, He, NH3, N2O, and SiH4 The SF6 gas is used for the plasma cleaning The He and N2 gases are used for powder removal and purging after the deposition The SiH4 and N2O gases are the reaction sources for SiO2 deposition, while the SiH4 and NH3 gases are for the Si3N4

deposition

For all the PECVD depositions in this work, the chamber temperature was

280oC because this temperature is optimized for this system to rapidly produce a high vacuum as mentioned above For the Si3N4 deposition, the flow rates for SiH4 and

NH3 were 28 sccm and 54 sccm, respectively The chamber pressure was 500 mTorr during the deposition and the RF power was 200 W The reaction can be expressed as: 3SiH4 + 4NH3 → Si3N4 + 12H2 (2.2)

For the SiO2 deposition, the flow rates of SiH4 and O2 were 30 sccm and 400 sccm, respectively The chamber pressure was 730 mTorr during the depositon and the RF power was 100 W The reaction can be expressed as:

SiH4 + O2 → SiO2 + 2H2 (2.3)

To obtain the growth rates and the refractive indices of the SiO2 and Si3N4

films grown by this system, a calibration was conducted SiO2 and Si3N4 films with different thicknesses were grown The thicknesses of these films were measured by the step-profiler, and the growth rates were therefore obtained With the thicknesses known, the refractive indices of the two materials grown by PECVD can then be measured using an ellipsometer The calibrated refractive index for the Si3N4 grown in the above conditions is about 1.97 at the blue wavelength region with a growth rate of

~ 1 nm/s The calibrated refractive index for the SiO2 grown in the above conditions is about 1.45 at the blue wavelength region with a growth rate of ~ 3.3 nm/s If compared to the theoretical refractive index values (~ 2.00 for Si3N4 and ~ 1.46 for SiO2 in the blue wavelength region), the refractive indices of the Si3N4 and SiO2

materials grown by our PECVD system are slightly smaller This is mainly due to the deviation of the actual stoicheometry The silicon nitride and silicon oxide films are

usually denser and with higher refractive indices in the silicon-rich case [Martin1991; Modreanu1998] In our PECVD deposition, the smaller refractive indices, and therefore the lower density of the silicon oxide and silicon nitride films, could be attributed to the deposited nitrogen-rich silicon nitride and oxygen-rich silicon oxide Nevertheless, we will still refer to the silicon nitride and silicon oxide materials deposited by our PECVD system as Si3N4 and SiO2 in this thesis

In this work, the AR and DBR coatings in the GaN-based SESAM structure were fabricated with SiO2 and/or Si3N4 films grown by this PECVD system, which will be discussed in detail in Section 5.1 In fact, the exact values of refractive index (n) and physical thickness (d) of these films are not important; but their product (n×d), which is called optical thickness of a film, directly determines the center wavelength

of a certain film Hence, in the calibrations for the AR and DBR coatings, firstly, according to the above calibrated growth rates and refractive index values, single layer of SiO2 or Si3N4 with a nominal quarter-wavelength optical thickness was deposited on a random flat substrate (Here the quarter-wavelength means a quarter of the required center wavelength.) For example, for a center wavelength of 425 nm, the nominal thickness for a quarter-wavelength SiO2 layer is 73 nm With the calibrated growth rate of 3.3 nm/s for SiO2, the deposition time is set at 22 s Assuming that the substrate is sapphire, the reflectance spectrum of this single-layer SiO2 film on the sapphire substrate is then measured by a spectrophotometer, which operation principle will be described in Section 2.3.2 Because sapphire (n=1.67) has a higher refractive index than SiO2 (n=1.45), the obtained spectrum would show a minimum reflectance

near 425 nm, whereas if the substrate is made of material with lower refractive index than the single-layer film material, a maximum reflectance would be observed near

425 nm The actual wavelength at the minimum reflectance for the single-layer SiO2

film is therefore obtained, which could be deviated from 425 nm Subsequently, the PECVD deposition time was adjusted according to the amount of wavelength deviation from 425 nm, and the above steps are repeated until the wavelength at the minimum reflectance arrives at the 425 nm Same procedures can also be conducted for calibrating a single quarter-layer Si3N4 centered at 425 nm With the calibrated growth rate of 1 nm/s for Si3N4, the initial deposition time can be set at 54 nm for a nominal quarter-wavelength thickness of 54 nm Finally, with the calibrated deposition times for quarter-wavelength Si3N4 and SiO2 layers centered at 425 nm, alternative layers of SiO2 and Si3N4 can be deposited to fabricate a DBR coating centered at 425 nm Single quarter-layer AR coatings, or two-layer quarter/quarter AR coatings can also be fabricated with SiO2 or/and Si3N4 In addition, DBR and AR coatings centered at other wavelengths can also be calibrated and fabricated using the similar procedures

2.3 Characterization techniques

2.3.1 Photoluminescence (PL)

PL spectroscopy is an important non-destructive technique to evaluate the optical properties of semiconductor materials, especially those direct-bandgap materials such as III-nitrides [Perkowitz1993]

Photoluminescence is a photon emission process resulting from optical excitation When a semiconductor is optically excited by photons with energy higher than the bandgap energy ( , the electrons can be excited from the valence band of the semiconductor into its conduction band, and the non-equilibrium electrons and holes are generated Then the non-equilibrium electrons and holes, and other quasi-particles will ultimately recombine through various transitions, which can be radiative (emission of photons) or non-radiative (emission of phonons) The PL spectroscope techniques measure and analyze the energy distribution of the emitted photons from a semiconductor after optical excitation

)

g E v

h >

There are several radiative recombination mechanisms, including (i) band-to-band transition, (ii) free electron to acceptor transition, (iii) free hole to donor transition, and (iv) donor-acceptor pair transition, as indicated in Fig 2.4

Figure 2.4 Different radiative recombination mechanisms, (i) band-to-band transition, (ii) free electron to acceptor transition, (iii) free hole to donor transition, and (iv) donor-acceptor pair transition ED is the donor binding energy and EA is the acceptor binding energy

The most common radiative transition in semiconductors is the band-to-band transition, which occurs between the states in the conduction and valence bands, with the energy difference known as the bandgap When working with a new compound semiconductor, bandgap determination by analyzing the PL spectrum can be particularly useful If the shape of the band-to-band PL spectrum from a direct-bandgap semiconductor is calculated, it is found that the PL spectral shape is given by

1/ 2 for

I (hω ∝ ω −) (h E ) exp[− ωh / k T] hω >Eg, (2.4) where is the emitted photon energy, Eg is the bandgap and kB is the Boltzmann constant According to the calculation from Eqn 2.4, the typical shape of a band-to-band PL spectrum is shown by the dotted curve in Fig 2.5, while the solid curve in Fig 2.5 shows the band-to-band PL spectrum measured from an actual sample As can be observed, the experimental curve shows deviation from the calculated curve especially on the lower photon energy side of the spectrum This is because an actual sample does not have perfect crystal quality The existence of a few defects will result in the emissions at lower photon energies than the bandgap energy, and therefore broaden the PL spectrum on the lower photon energy side

ω

h

Figure 2.5 Calculated (dotted curve) and experimental (solid curve) band-to-band

PL spectra for an n-type InSb sample [Teo2004]

Radiative transitions in semiconductors also involve the free-to-bound transitions, including the free electron to acceptor transition and the free hole to donor transition, which are related with the defect levels The photoluminescence energy associated with these defect levels can be used to identify specific defects and their binding energy, and the amount of photoluminescence can be used to determine the concentration of certain defects For some semiconductors, they may even contain both donors and acceptors Under the equilibrium, some of the electrons from the donors will be captured by the acceptors As a result, the sample would contain both ionized donors and acceptors The energy of the photon emitted from the donor-acceptor pair transition will be related with the bandgap of the semiconductor, the respective binding energies of donor and acceptor levels, as well as the coulomb interaction energy determined by the distance between the ionized donor and acceptor

Besides the radiative transitions mentioned above, there are also radiative transitions related with excitons, including free exciton recombination and bound exciton recombination For the free exciton recombination, its binding energy can be obtained according to the hydrogen-atom model, while for the exciton bound to a certain defect level, the binding energy is determined by the bandgap, free exciton binding energy, and the bound exciton binding energy, as illustrated in Fig 2.6

Figure 2.6 Illustration for free excition transition and bound exciton transition Eb is the free exciton binding energy and Ebx is the bound exciton binding energy

In addition to the radiative transitions, there are also non-radiative transitions associated with deep defect levels, which presence is normally detrimental to the material quality and the device performance Thus, material quality can be measured

by analyzing the amount of radiative recombination In addition, because the existence of shallow defects can cause the broadening of the band-to-band PL peak or even result in the multiple peaks, the crystal quality of the certain material can then be

estimated by the FWHM of the PL spectrum The origin of the additional PL peaks can also be analyzed according to their emission energies and FWHM values These techniques were used in this work to study the morphology related optical properties

of InGaN/GaN quantum wells

The PL measurements in this work were performed using three different setups For the room-temperature PL measurements, a Renishaw 2000 micro-PL setup

or an Accent Rapid Photoluminescence Mapping System (RPM 2045) was used, with the 325-nm He-Cd laser for excitation The analysis can be done at various excitation power densities for both systems, and the spatial resolution of the micro-PL setup is 4

μm In material study, because some defects and structural irregularities manifest themselves only at low temperature, PL spectra obtained at low temperature are also important The above micro-PL system can perform the low-temperature PL measurements at temperatures down to 77 K, using the liquid nitrogen for cooling To achieve lower temperatures, another low-temperature PL system was used The schematic setup of the low-temperature PL system is shown in Fig 2.7 The sample is located in a cryostat connected to a compressor, and the measuring temperature can be adjusted from about 4 K to the room temperature A 325-nm He-Cd laser is used as the exciting source The beam is modulated by a chopper at a low frequency (280 Hz), and is focused to a small spot (0.001 cm2) on the sample surface This set-up is in a back-scattering geometry, i.e., the resulting PL is collected from the same surface as the surface excited by the incident laser This arrangement allows for separate focusing of the laser beam and the imaging of PL The resulting luminescence is then