Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

GaAs₁₋xBix light emitting diodes : a new long wavelength semiconductor light source Lewis, Ryan B. 2008-03-10

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata


24-ubc_2009_spring_lewis_ryan_b.pdf [ 1.01MB ]
JSON: 24-1.0067053.json
JSON-LD: 24-1.0067053-ld.json
RDF/XML (Pretty): 24-1.0067053-rdf.xml
RDF/JSON: 24-1.0067053-rdf.json
Turtle: 24-1.0067053-turtle.txt
N-Triples: 24-1.0067053-rdf-ntriples.txt
Original Record: 24-1.0067053-source.json
Full Text

Full Text

GaAsi_BixLight EmittingDiodesA New Long Wavelength Semiconductor LightSourcebyRyan B. LewisB .Sc., Daihousie University, 2006A THESIS SUBMITTED IN PARTIAL FULFILMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMaster of Applied ScienceinThe Faculty of Graduate Studies(Engineering Physics)The University Of British Columbia(Vancouver)October, 2008©Ryan B. Lewis 2008AbstractGaAs1_Bi is an exciting new semiconductor material, which has beenproposed as a new material for infrared light emitting devices. Recent advancements in the growth ofGaAsi_Bifilms have made it possible toproduce GaAsi_Bi light emitting diodes for the first time. Throughoutthis research we have grown, fabricated and characterized GaAsi_Bi lightemitting diodes. Similarly structured InGai_As light emitting diodes werealso produced and characterized for comparison to the GaAsi_Bi devices.Strong electroluminescence was obtained from GaAs1_Bi devices, showing two emission peaks, one corresponding to the GaAs1_Bi layer and theother to the GaAs cladding. Emission from InGai_As devices was about100 times brighter than from GaAsi_Bi devices.Temperature dependent electroluminescence and photoluminescence measurements of a GaAs1_Bi light emitting diode were made and showed someunusual results. The wavelength of the peak in the electroluminescence fromthe GaAs1_Bi was independent of temperature in the range 100 K to 300 Kwhile the GaAs peak shifted with temperature as expected. Photoluminescence measurements on the same structure show temperature dependence ofthe peak wavelength similar to the temperature dependence of GaAs.IITable of ContentsAbstract iiTable of Contents iiiList of Tables vList of Figures viAcknowledgements ix1 Introduction 12 Growth and Properties of the GaAs1_Bi Alloy 82.1 Molecular Beam Epitaxy Growth of GaAs Alloys 82.2 Growth of the GaAsi_Bi Alloy 92.3 Doping and the p-n Junction Diode 122.4 Heterostructure Design 152.5 Light Emitting Diode Growth and Characterization 163 Post-Growth Fabrication of Light Emitting Diodes 253.1 Ohmic Contacts 253.2 Mesa Etching 30111Table of Contents4 Electrical Characterization 324.1 Current-Voltage Measurements of GaAs1_Bi Diodes . . . . 324.2 Current-Voltage Measurements of InGa1_As Diodes . . . . 365 Optical Characterization 385.1 Electroluminescence (EL) and Photoluminescence (PL) . . 385.2 Photoluminescence Measurements 415.3 Electroluminescence Measurements 435.4 Temperature Dependent Electroluminescence 496 Conclusion 526.1 Future Work 53Bibliography 54ivList of Tables2.1 Light Emitting Diode Growths232.2 Selected Light Emitting Diode growth conditions24VList of Figures1.1 Energy bandgap vs. lattice constant for several semiconductoralloys[1] 31.2 Energy bandgap vs. lattice constant for several semiconductoralloys including the GaAs-GaN and GaAs-GaBi systems. . . S1.3 The section of the periodic table containing elements used incommon semiconductors 62.1 [004] X-ray rocking curves for threeGaAsi_Bi epilayers. Afit to each curve is shown with dotted lines[2] 112.2 Schematic diagram p-n junction at equilibrium (no applied bias). 142.3 Schematic diagram of a forward biased p-n junction 152.4 [004] rocking curve of aGaAsi_Bi LED sample (r1965) containing 1.89’o bismuth. A fit to the data is shown in red . . . . 192.5 [004] rocking curves for twoGaAsi_BiLED samples (r1895and r1917) containing 0.9% and 5.5% bismuth shown in redand blue 202.6 [004] rocking curve of aInGa1_As LED sample (r1929) containing 18% indium. A fit to the data is shown in red 21viList of Figures3.1 Schematic diagram of a metal to n-type semiconductor interface in the absence of surface states273.2 Schematic diagram of a metal to n-type semiconductor interface with surface states283.3 Schematic of an LED chip after metalization and etching.Typical thickness for i-GaAs layers was 25 nm, typicalGaAsBiQW thicknesses were 30 nm to 50 nm and typical p-GaAsthickness was 1 /mum. The top 300 nm of the p-GaAs layerwas removed by etching314.1 Current-Voltage curves for fourGaAsi_Bi light emitting diodesat 300 K. The red line indicates a fit to the data withn = 2.26ideality factor334.2 Current-Voltage curves forGaAsi_Bi LED r1895. Hollowdata points (low leakage values on reverse bias) correspondtodots from an unetched device fabricatedsoon after growth.Solid data points (high leakage) correspondto measurementsmade on the same sample 2 months later, with andwithoutdipping the sample in HC1 to remove possible oxideand alsofor a new device fabricated from the wafer with etchedmesas(solid pink triangles)354.3 Current-Voltage curves forInGai_As LEDs from sample r1929 375.1 Possible non-radiative transitions in semiconductors41viiList of Figures5.2 Photoluminescence spectra forGaAsi_Bi LED structure r1965over a temperature range of 8 K to 300 K. The inset showsthepeak emission energies as a function of temperaturefor boththe GaAs andGaAsi_Bi peaks. A fit to the data using theVarshni equation is also shown (dashed line)425.3 Electroluminescence spectra for aGaAsi_Bi light emittingdiode from sample r1965 for various injectioncurrent densitiesat 300 K. Room temperature photoluminescence isshown forcomparison445.4 Room temperature electroluminescence spectra from threeGaAsi_BiLEDs at 100 A/cm2injection current. The black and redspectra were fabricated from growth r1895 and contain1% bismuth. The blue spectra (from r1917) contained5.59’o bismuthin theGaAsi_Bi layer 475.5 Room temperature electroluminescence spectrafor an InGai_Aslight emitting diode for various injection current densities.485.6 Electroluminescence spectra from GaAsBi LEDr1965 at 50 A/cm2injection current density for temperaturesranging from 100 Kto 300 K50viiiAcknowledgementsI would like to thanks my supervisor Tom Tiedje for his inspiration, guidanceand for sharing so much of his knowledge and experience with me. Thanksto the rest of the MBE group of being so willing to help me out and for beingand so entertaining. Special thanks to Dan Beaton for growing my samplesand for helping me with latex, without you this would have been a shortthesis.I’d also like to thank my parents for keeping in such close contact andfor providing so much support and encouragement. Thanks to Jeff Dahn atDalhousie for encouraging me to continue my academic careere in condensedmatter physics. And of course thanks to Dayna for putting up with me andcheering me up while I was writing this thesis. Your support has meant somuch.ixChapter 1IntroductionIt is hard to imagine a world without semiconductors. These special materials, group IV alloys, 111-V alloys, Il-VI alloys and recently I-Ill-VT2alloysare an integral part of the technology-driven world that we live in. The Ill-Vclass of semiconductors dominate applications for the microwave frequencyintegrated circuits used in cell phones, light emitting diodes (LEDs), diodelasers for data transmission, reading and recording for DVD and CD playersand high efficiency solar cells. ITT-V’s are ideally suited for optical applications because their direct bandgap allows for very efficient generation oflight.GaAs and InP are arguably the most important Ill-V semiconductors.InP-based lasers are extensively used in fiber optic data transmission. GaAshas a high electron mobility, good for making high frequency transistors, italso has a higher breakdown voltage than silicon, allowing for higher powerdevices to be made. Despite these advantages, silicon is much more usedfor integrated circuit manufacturing. Reasons for this are that: silicon ismore abundant and easier to process; silicon dioxide, which is one of thebest known insulators is easily incorporated into silicon circuits; the higherhole mobility of silicon allows for fabrication of faster p-channel field effecttransistors, required for complementary metal-oxide-semiconductor (CMOS)1Chapter 1. Introductionlogic, which makes GaAs-based logic circuits have much higher power consumption. The biggest advantage of GaAs is that it can be easily alloyed withother ITT-V systems to achieve a wide range of bandgap energies, which allowsfor easy fabrication of double heterostructures (DH’s). DHs are formed bysandwiching a low bandgap semiconductor material between a semiconductorwith a higher bandgap. This structure is extremely usefull for optoelectronicdevices.The ability to grow high quality films with energy bandgaps lower thanthe 1.42 eV GaAs bandgap on GaAs substrates is of much interest for deviceapplications. The silica fibers used for optical data transmission have nodispersion at 1.3 tm and are most transparent at 1.55 jim, which correspondto bandgaps of 0.83 eV and 0.80 eV. The solar cell industry would also like togrow a 1.0 eV material on GaAs for use in high-efficiency multi-junction solarcells, thus much effort has been made in developing such materials. Commonbandgap lowering materials used to alloy with GaAs are InAs and GaSb,however because indium is larger than gallium and antimony is larger thanarsenic TnGai_As andGaAsi_Sb alloys have a larger lattice constantsthan pure GaAs. This limits the thickness and compositions that can begrown pseudomorphically on GaAs before the strain causes dislocations toform. Fig. 1.1 shows the energy bandgap as a function of lattice constantfor several semiconductor alloys, including the GaAs-InAs and GaAs-GaSballoys.The heaviest group III and V elements, thorium and bismuth have beenlargely neglected as candidates for alloying with GaAs. These elements aredifficult to incorporate into the lattice due to their large size and tendency2Chapter 1. IntroductionThis figure has been removed due to copyright restrictions. The figureshowed a plot of the bandgap of several semiconductor alloys as a functionof lattice constant and was obtained from [1].Figure 1.1: Energy bandgap vs. lattice constant for several semiconductoralloys[1].to surface segregate[3j [4]. First reports of bismuth incorporation into GaAscame from metal-organic vapor phase epitaxy (MOVPE)in 1998 from Oeand Okamoto, who were able to growGaAsi_Bi with concentrations upto x = O.02[5]. Molecular beam epitaxy (MBE) growth was first presented5 years later[4][6]. Incorporation of small amounts of bismuth produces alarge reduction in the bandgap of GaAs[7]; 88 meV per percent bismuth,which is seven and four times greater than what is achievable with indiumor antimony respectively[8] for equal incorporated amounts.The portion of the periodic table, corresponding to elements found insemiconductors is shown in Fig. 1.3. The red and blue elements canbe described as alloying elements with GaAs, while the green elements: nitrogenand bismuth seem to behave more like impurities when incorporated, rather3Chapter 1. Introductionthan forming an alloy. The two ternary alloys GaNAsi_ and GaAsj_Biare complimentary in that GaNAs1_reduces the bandgap by lowering theconduction band minimum (CBM), while GaAsi_Bi lowers the bandgap byraising the valence band maximum (VBM) [8]. In theGaAsi_Bi alloy, theBi6p level is resonant with the VBM[9], whereas in the case of the GaNAsi_alloy, the N2p levels are deep in the valence band and the unoccupied N2santibonding orbital is resonant with the CBM[10j. Since nitrogen incorporation decreases the GaAs lattice size and bismuth increases it, the quaternaryGaNAsi_Bi alloy has been proposed as strain compensating alloy, allowinglow bandgap GaNAsi_Bi to be lattice matched to GaAs[11]. Fig. 1.2 showsthe energy bandgap vs. lattice constant for selected 111-V semiconductor alloys, including the GaAs-GaN system and the GaAs-GaBi system. The blueline corresponds to the GaAs-GaBi alloy.Most properties, including bandgap to a first approximation follow Vegard’s law[12] when alloying ITT-V semiconductors, however this is not true forthe incorporation of nitrogen. For example, the bandgap of GaN is higherthan bandgap of GaAs, however incorporation of small amounts of nitrogeninto the GaAs lattice results in a reduction of the GaAs bandgap by thehuge amount of 200 meV per percent nitrogen[13][14]. GaBi has not beensynthesized, so it’s lattice constant and bandgap have only been determinedtheoretically by density functional theory.It has been proposed that GaAsi_Bi alloys have an anomalously temperature insensitive bandgap[4]. The temperature dependence of theGaAsi_Bibandgap has recently been measured, however quite different values wereobtained[5] [6] [15]. Photoluminescence and photoreflectance measurements4Chapter 1. IntroductionFigure 1.2: Energy bandgap vs. lattice constant for several semiconductoralloys including the GaAs-GaN and GaAs-GaBi systems.on a sample containing 2.6% bismuth by Yoshida, et al. found the temperature dependence of the GaAs1_Bi bandgap to be 1/3 that of GaAs[5],however other measurements found a similar temperature dependence toGaAs [6]As the heaviest non-radioactive element, bismuth alloying creates an unusually large spin orbit splitting. Incorporation of bismuth into Ill/V alloyscould provide the large spin-orbit coupling needed for spintronic devices.TheGaAsi_Bialloy has recently been shown to have very wide photo-luminescence (Pb) spectra, suggesting the grown films may contain bismuth4304.0 5.0 5.5 6.0Lattice Constant (A)5Chapter 1. IntroductionFigure 1.3: The section of the periodic table containing elements used incommon semiconductors.clusters or compositional variations throughout the film. A broad spectrumGaAsi_rBi,, light emitter could provide the desired light source for opticalcoherence tomography (OCT) [16] [17]. OCT is a medical imaging techniquethat uses interferometry of low coherence light to image tissue cross sectionsat depths of a few millimeters, with micron resolution. The resolution of anOCT system is determined by the coherence legnth of the light source used.For a gaussian beam the coherence legnth is given in Eq. 1.1, where)*-,is thecentral wavelength of the gaussian spectrum, and LS\ is the full width at halfmaximum of the spectrum.lilA IVA VA VIA5 67 8B CN 08oron Caton gen. Ozygen13 14 15Al SIP118 AIMiimXn n PtIOSphOU16S30 31 32 33 34Zn:Ge As SeZinc Gai G.mainr ksnc Seeium48 50 5152Cd In Sn Sb TeCad Th Anniony T.aunum80 81 82 83 84Hg TI Pb Bi Po& fl*bsn L. BnIUth Poonk,m6Chapter 1. Introduction(1.1)For maximum imaging depth the light source must operate in the nearinfrared range(850 — 1300 nm) where tissue is most transparent. Thesedemands perfectly overlap with the characteristics of the GaAsi_Bi alloy,thus OCT is a very promising application of aGaAsi_Bilight emitters.Throughout this research we have fabricated and characterizedGaAsi_Bilight emitting diodes. To our knowledge, these results represent the firstGaAsi_BiLEDs ever made. We discuss in detail the growth and fabrication processes that we have developed. Current-Voltage, temperaturedependent photoluminescence and electroluminescence results are also presented and discussed for this exciting new semiconductor.7Chapter 2Growth and Properties of theGaAsi_BiAlloy2.1 Molecular Beam Epitaxy Growth ofGaAs AlloysGallium Arsenide is probably the most important material for solid statelight emitting devices. GaAs is advantageous because of it’s high electronmobility, ease of manufacturing, availability of 150 mm wafers and abilityto be alloyed with other elements to modify the GaAs bandgap. GaAs, asmany 111-V semiconductors has a direct bandgap, allowing for much moreefficient optical devices to be made, compared to elemental semiconductorslike silicon and germanium. The direct bandgap means transitions betweenthe conduction band minimum (CBM) and valance band maximum (VBM)can occur without either the absorption or emission of phonons, to conservemomentum.Molecular beam epitaxy (MBE) is a common method for the preparationof GaAs based semiconductor materials in thin film form. MBE is a process,where beams of atoms or molecules (usually from thermal evaporation) are8Chapter 2. Growth and Properties of the GaAsi_Bi Alloysimultaneously incident on a heated substrate in ultra-high vacuum (UHV).Individual beams can be turned off or on in a fraction of a second by the useof shutters. This process allows for precise control of compositions and layerthicknesses on the sub-monolayer scale and the ability to abruptly change thecomposition of the layer being grown. Typically, GaAs is grown at rates ofabout 1 pm per hour at temperatures of about half the melting point of GaAsor 500°C to 650°C. The fact that growth takes place so much colder thanthe melting point allows for metastable compounds, not found in nature tobe created. Low temperature also minimizes the number of thermodynamicdefects in the grown material.2.2 Growth of theGaAsi_BiAlloyUnder normal GaAs growth conditions, bismuth tends to surface segregateand does not incorporate into the GaAs lattice. At these conditions bismuthbehaves as an ideal surfactant, as it does not incorporate. Bismuth has beenused as a surfactant in the growth of several GaAs-based material systems,such as GaAs and InGai_As where it has been shown to improve surfacesmoothness and photoluminescence [18] [19] [7].In the case of GaNAs1_andInGaNAsi_ it hasbeen shown to result in smoother surfaces, enhancednitrogen incorporation, and increases photoluminescence by reducing defects[3][19j.Incorporation of bismuth into the GaAs lattice requires atypical growthcondidions, to reduce the tendency of bismuth to surface segregate and reduce the competition for group V sites. To achieve this, low growth temper9Chapter 2. Growth and Properties of the GaAs1_Bi Alloyatures and low V:III ratios (usually 1:1 to 4:1 compared to typical ratios ofabout 10:1) are required. High bismuth incorporation of up to 10% has beenachieved in the temperature range of 270°C-320°C[1]. Low growth rates ofabout (1 nm/mm) are used in order to have more control over the excessbismuth on the surface and minimize the likelihood of bismuth accumulation in the form of droplets[1]. Droplets are very undersirable as they leadto local variations in the amount of bismuth coverage and hence the localcomposition. Droplets also increase the surface roughness. In the case oflow growth rates, the rate of bismuth incorporation is less, so lower bismuthfluxes can be used. In this case most of the incident bismuth flux is evaporated, rather than incorporated, thus allowing for more control over theamount of surface bismuth present. Low temperature growth also leads toa larger critical thickness before dislocations form[1], which allows for highstrain, small bandgap, epi-layers to be grown on GaAs without relaxation.Fig. 2.1 shows [004] X-ray rocking curves for three GaAsi_Bi epilayers for x values of 1.4%, 5% and 10%[1]. Fits to each data set are shownas dotted lines. The sharp peak in each spectra corresponds to the [004]GaAs substrate peak and the smaller satelite peak on the left correspondsto the GaAsi_Bi epilayer. Dampled pendellosung fringes are visible inall the curves, indicating high structural quality. These interference fringesalso allow for determination of layer thicknesses. Composition is determinedfrom the separation in epilayer and substrate peaks using the known shiftof 300 arcsec per percent bismuth incorporation in the [004] X-ray peak forincorporation up to a few percent. The shift in lattice constant with bismuthhas been determined by comparing Rutherford backscattering (RBS) compo10Chapter 2. Growth and Properties of the GaAs1_BiAlloysitional data to X-ray diffraction data on samples with up to 3.1% bismuth byTixier et al. [3]. Since the lattice constant of zincblend GaBi is unknown, onecould not use Vegard’s law to determine composition from X-ray data alone,however the work by Tixier et al. has provided a theoretical prediction ofthe GaBi lattice constant of 0.6336 nm based on an extrapolation of concentrations up to 3.1%[3]. Thus bismuth compositions can now be determinedby X-ray diffraction, in comparison to this work.D(‘3Cl)ci)C0(‘3U-4000 -2000 0 2000 4000Theta (arcsec)Figure 2.1: [004] X-ray rocking curves for three GaAsi_Bi epilayers. A fitto each curve is shown with dotted lines[1].11Chapter 2. Growth and Properties of theGaAsi_Bi Alloy2.3 Doping and the p-n Junction DiodeIn the case of a semiconductor, the covalent bonding results in heavy overlapof the atomic wavefunctions between neighboring atoms. When two atomsare covalently bonded together their atomic energy levels split to producetwo energy states, the lower energy state is called the bonding state andthe upper the antibonding state. For n atoms covalently bonded this splitting produces n energy levels, which, for large crystals form energy bands[20]. The electronic energy level distribution can be found by solving theSchroedinger equation for a periodic potential, corresponding to the appropriate atomic potentials. At zero temperature, the electrons of the crystalpopulate the energy states by filling up levels from lowest to highest. Thehighest filled energy band is called the valence band and the lowest unoccupied band is called the conduction band. Direct bandgap semiconductorsare semiconductors where the valence band maximum (VBM) and conduction band minimum (CBM) occur at the same value of electron momentum. Compound semiconductors, such as GaAs, InSb and GaN are directbandgap semiconductors. Silicon, germanium and AlAs are examples of indirect bandgap semiconductors, which have an offset in momentum betweenthe CBM and VBM. A direct bandgap is required for efficient light generation as photons are created by downward transitions from the conductionband to the valence. Transitions in indirect bandgap semiconductors usuallyinvolves the creation or destruction of phonons in addition to photons.A pure semiconductor at zero temperature has all energy states in thevalence band occupied and all conduction band states vacant. The energyat which there is a 50 percent probability that the state is occupied is called12Chapter 2. Growth and Properties of the GaAs1_Bi Alloythe Fermi energy. The occupation of energy states as a function of energy isgiven by the Fermi function, given in Eq. 2.1, where Ef is the Fermi energy.=+ 1(2.1)The Fermi energy for an intrinsic semiconductor is close to mid-gap. Theaddition of small amounts of impurity, as low as 1 ppb can greatly shift theFermi energy, because there are no allowed energy levels between the bands.Donor dopants are impurity atoms that have occupied energy levels close tothe conduction band minimum, which shift the Fermi level towards the conduction band, resulting in an n-doped material. At non-zero temperatures,electrons can be excited from the donor atom into the conduction band,populating the conduction band. Similarly, acceptor impurities are atomsthat have unoccupied levels near the valence band maximum, which shift theFermi level down, towards the valence band, resulting in a p-doped material.These levels can accept electrons from the valance band.In the case of a p-n junction, p and n-doped materials are put in intimatecontact with each other. At equilibrium (under no electrical bias) the Fermienergy must be equal throughout both materials. This causes band bendingnear the junction, a result of the redistribution of charge in the region closeto the junction, known as the depletion region. This redistribution of chargecauses a built in voltage to be produced, which equals the initial differencein Fermi energies of the two materials. Fig. 2.2 shows a schematic drawing ofthe valance and conduction band energies of a p-n junction at equilibrium.Applying a forward bias causes a flattening of the bands and a shrinkingof the width of the depletion region. Biasing the diode results in a non13Chapter 2. Growth and Properties of the GaAsi_Bi AlloyFigure 2.2: Schematic diagram p-n junction at equilibrium (no applied bias).equilibrium condition and produces two separate quasi-Fermi levels, one oneach side of the junction [21]. Fig. 2.3 shows a hypothetical p-n junctionunder forward bias. The figure shows an area near the junction, where conduction electrons and holes exist together in the same space, allowing recombination to take place and light to be created. Under high enough forwardbias voltages it’s possible to create a population inversion at the junctionmaking it possible for lasing to take place.The diode current, I of an ideal diode with an applied voltageVdis givenby the diode equation shown in Eq. 2.2 where I is the saturation current,n is the ideality factor of the diode andVT is the thermal voltage. Thethermal voltage is given in Eq. 2.3, where k is the Boltzmann constant,T is the temperature in Kelvin and e is the electron charge. The idealityfactor of diodes usually ranges between 1 and 2, depending on the type ofdiode. For devices where recombination in the depletion region is negligible,i.e electrons and holes can be assumed to slip through the depletion regionwithout recombining, the theoretical ideality factor is 1. In this case electrons14Chapter 2. Growth and Properties of theGaAs1_Bi Alloycleotrons—_-------UiC0 -UiE-3holesFigure 2.3: Schematic diagram of a forward biased p-n junction.are injected directly into the p-region and holes into the n-region. In the caseof an LED double heterostructure (DH), ideally all the injected carriers willrecombine in the depletion region radiatively [20] and thus the ideality factorwould be 2. This is the case because carriers only have to travel half wayacross the built in potential before recombining. In many practical cases therecombination in the depletion region can be dominated by non-radiativerecombination.I=I8(e’-” — 1) (2.2)Vj-= (2.3)2.4 Heterostructure DesignModern LEDs and laser diodes typically can have an array of complicatedstructures, to improve confinement and overlap of the carriers and photons15Chapter 2. Growth and Properties of theGaAsi_BiAlloyand to increase light output. Low index (high bandgap) AlGai_As claddinglayers are used to increase carrier confinement. The active region of theseefficient light emitting devices typically consists of multiple low bandgapquantum wells (QWs) spaced with a GaAs layers and this region is typicallysandwiched between A1Gai_As cladding layers. Typical thicknesses of theundoped active region containing the QWs is 100 nm to 200 nm[20]. Thesimplest DH structure that one can imagine in one where there is simplyone low bandgap QW layer inside the undoped active region and there is noA1Gai_As cladding. Such a device is relatively simple to make and doesnot require as much optimization as a more complicated structure. Such astructure was used in making the GaAsi_Bi LEDs in our lab because ofthe relative difficulty in growing GaAsi_Bi layers. In growing InGai_AsLEDs a similar structure was adopted, but instead of one QW, the activeregion contained three InGai_As QWs.2.5 Light Emitting Diode Growth andCharacterizationLED samples were grown in a VG-V8OH molecular beam epitaxy (MBE)system on [100] GaAs substrates using effusion cells for gallium, bismuthand indium, a dual stage cracker for As2,along with a gas injection systemfor CBr4 for p-type doping. n-type doping was achieved using a Si effusioncell for samples r1962 and later, and a SiBr4 gas injection system for earliersamples . The substrate temperature was monitored throughout the growthprocess by optical bandgap thermometry[22] with an accuracy of 5°C. Beam16Chapter 2. Growth and Properties of theGaAs1_Bi Alloyequivalent pressures were measured using a retractable ion gauge. Growthconditions for GaAs were: V:III ratios of 8:1; growth rate approximately1iim/hr;arsenic cell temperature 400°C; gallium cell temperature 950°C;and substrate temperatures between 550°C and 580°C in most cases.The active region of the bismuth LED n-i-p structure (r1965) consistedof a 30 nm GaAsi_Bi layer sandwiched between two 25 nm undoped GaAsspacer layers. Thicknesses may have varied for different samples. The diodeswere grown on n-doped (2 x1018cm3) [100] GaAs substrates (with exception of sample r1810 grown on p-type substrate). A 1000 nm n-dopedbuffer was grown first at standard GaAs growth conditions with a dopingconcentration of 5 x1017cm3.The growth was often interrupted betweenthe n-doped layer and the undoped layer to adjust growth conditions as follows: new substrate temperature of 300°C, arsenic cell temperature of 350°Cand gallium cell temperature of 850°C for r1965. The temperature of theGa cell was lowered to achieve a growth rate of 0.1 nm/hr and the arseniccell temperature was lowered to allow for better control over the As2 fluxat low group V over pressure. Growth interruptions lasted for 10 minutesand were necessary for growths using the SiBr4 since using SiBr4 requiredadditional time in the i-region to switch to the CBr4. This delay was forthe gas lines to be pumped free of the SiBr4 before they could be filled withCBr4 for p-doping could start. The addition of the elemental silicon effusioncell allowed future growths to avoid this growth interrupt. Samples r1970and later did not have growth interrupts. Some early LED growth attemptshad an additional growth interrupt at the i to p interface as well (r1917 andr1930). The undoped region was comprised of 30 nm of GaAsiBi with17Chapter 2. Growth and Properties of the GaAsi_Bi Alloy25 nm of GaAs on either side. This entire region was grown at low growthrate (0.1 um/h). The two surrounding GaAs regions were grown at standardAs2 overpressure, while the GaAsi_Bi layer was grown with the As2 over-pressure lowered to nearly stoichiometric levels (2.5:1 for r1965) to enhancebismuth incorporation. Bismuth flux was present at the substrate for theentirety of the intrinsic region, as bismuth will not incorporate until the As2flux is lowered, even at low substrate temperatures. No growth interruptionwas used for the transition from intrinsic to p-doped layers in most cases.The growth rate and substrate temperature were ramped back to standardconditions while still growing, causing a small region (25 nm) where the pdoping was non-uniform in the case of no second growth interrupt. 1000 nmof p-doped (5 x io’ cm3)GaAs was grown followed by 100 nm of highly pdoped (approximately 5 x1018cm3)GaAs. The highly doped capping layerwas used so that ohmic contacts could be more easily achieved, althoughhigher doping in the 1019/cm3range would have been preferred.Layer thicknesses and compositions were measured by high resolution Xray diffraction (XRD) using a Philips Xpert diffractometer. Rocking curveswere measured for the [004] GaAs diffraction peak over ranges of 2° and 4°.Fig. 2.4 shows a [004] rocking curve for the most luminescent GaAsi_BiLED (r1965). The red curve in the figure corresponds to a fit to the data usingBede RADS Mercury, which models diffraction patterns using the dynamicaltheory of diffraction. The fit shown corresponds to a 3Onm GaAsi_Bi layerwith x = 0.018±0.004. The split off peak from the GaAsi_Bi layer had lowintensity (compared to similar GaAsi_Bi epilayers) due to it being buriedunder about 1im of GaAs. This effect combined with the peak not being fully18Chapter 2. Growth and Properties of theGaAsi_BiAlloyseparated from the GaAs substrate peak resulted in the large uncertainty inthe GaAsi_Bi composition. Small pendellosung fringes can be seen in thedata, which correspond to reflections from the top GaAs-layer-GaAs1_Bi-layer interface. Fitting these fringes we find that the thickness of the toplayer is 890 nm. Two other GaAsi_Bi n-i-p structures, one containingabout 0.9% Bi and the other 5.5% Bi are shown in Fig. 2.5. Table 2.1summarizes all the LED structures that were grown.Cl)CU)CFigure 2.4: [004] rocking curve of a GaAsi_Bi LED sample (r1965) containing 1.8% bismuth. A fit to the data is shown in redInGai_As LEDs were grown for comparison with the GaAsi_Bi LEDs.The n-type and p-type layers of these devices were similar to the structure discussed above, however the intrinsic region contained either a single-1000 0 10009 (seconds)19Chapter 2. Growth and Properties of the GaAs1_Bi Alloy>Cl)CCFigure 2.5: [004] rocking curves for two GaAsi_Bi LED samples (r1895and r1917) containing 0.9% and 5.5% bismuth shown in red and blue.InGai_As QWin r1810 or 3InGai_As QWsin r1929. r1810 had nogrowth interrupts while r1929 had one interrupt at the n-i interface. Forr1929 x = 0.18, each QW was about 5 nm thick, spaced by about 19 nm ofGaAs, resulting in an intrinsic region thickness of about 90 urn. Thicknessesare approximate because a good fit to the data could not be obtained. TheXRD pattern is shown in Fig. 2.6. The figure contains a fit to the data, whichdoes line up with all the peaks in the curve, but does not correctly model therelative heights. The discrepancies between the XRD data likely comes fromsmall differences in the thicknesses and compositions of the layers, whichwere not account for because a supercell model was used. The composition-1 -0.5 0 0.59 (degrees)20Chapter 2. Growth and Properties of theGaAsi_BiAlloyof the QWs was obtained by electroluminescence (EL), discussed in chapter4. The intrinsic region of r1929 was grown with Ga and In cell temperaturesof 950°C and 800°C, respectively at a growth rate of about 1 pm/hour. Thesubstrate temperature was 580°C and there was a growth interrupt at thep-type to intrinsic interface. At the time of this growth n-doping was donewith gas source SiBr4,so the growth interrupt was necessary to pump outthe gas lines (as discussed above). The growth interrupt lasted 8.5 mm.Cl)Cci)CFigure 2.6: [004] rocking curve of a InGai_As LEDtaming 18% indium. A fit to the data is shown in redsample (r1929) con-Table 2.1 summarizes LED growths and characteristics of all grown samples. Many samples contain little or no bismuth and were not fabricatedinto LED’s. Getting bismuth to incorporate proved to be very difficult. The-3000 -1000 1000 30000 (seconds)21Chapter 2. Growth and Properties of the GaAs1_Bi Alloyconditions for bismuth incorporation are very precise and perhaps if detailedcalibrations of fluxes were done before each growth then more reproduciblecompositions of GaAsi_Bi layers would have been obtained. Table 2.2gives more detailed growth information from the most significant samples.Beam equivalent pressures (BEPs) are given as read from ion gauge andare based on flux calibrations done weeks or even months before the actual growths. BEPs have not been corrected for the sensitivities of differentchemical species.22Table2.1:Summaryofalllightemittingdiodesamplesgrown,includingfailedattempts.SampleGrowthBiorInELWavelengthCommentsLog#Interrupts{%](nm)r18100In[3.6]medium900veryresistiver18520In[01NANAdidnotrectifyr18750In[21jNANAveryleakey,overdopedp-layerr18881Bi[0.5]NANAnottested,wantedmoreBir18951Bi[0.9]NoneNAonlydefectemissionr19011Bi[0]NANAnottested,noBir19172Bi[5.5]Veryweak1300mostlydefectemissionandGaAsr19291Inx3[18]Strong1025verystrongEL,goodIVr19302Bi[NA]NANAdroppedinMBE-postgrowthr19621Bi[NA]NANAcoveredinGaduringgrowthr19651Bi[1.8]Strong986seetextr19700Bi[0]NANAnotyettestedr19821GaAsNANAhighTGaAsnotyettestedr19850Bi[0]NANAnotyettested0 0€0 Cb03Chapter 2. Growth and Properties of the GaAsi_Bi AlloyTable 2.2: Summary of growth conditions of significant samples.Sample Substrate P[Ga] P[Asj P[In or Bi]Log#Temp. (°C) (torr) (torr) (torr)r1895 315 3.0 x iO 2.9 x 1O [Bi]3 x 1Or1917 305 1.8 x iO 1.6 x 1O [Bi]6 x iOr1929 580 9.4 x 1O 4.9 x10—6[In]3 x 1Or1965 310 0.8 x iO 0.7 x iO [Bi]7 X 1024Chapter 3Post-Growth Fabrication ofLight Emitting Diodes3.1 Ohmic ContactsOne challenge faced in post-growth fabrication of semiconductor electricaldevices is making reliable ohmic contacts. An ohmic contact is an electricalcontact, which has a linear I-V responce curve, and ideally should have thelowest series resistance possible. Contacts should be easily reproducible andstable over the usable temperature range of the device. This seemingly simpletask has been the focus of an enormous amount of research over the pastseveral decades. For more information and selected review articles on ohmiccontacts Modern GaAs Processing Methods by Williams [23] contains muchuseful information, also see the review article by Shen[24]Normally, when a metal is put in intimate contact with a semiconductor,a depletion region forms in the semiconductor, which bends the bands sothat the Fermi level is equal in the metal and the semiconductor. One wouldexpect that the voltage across the barrier (q5) would simply be the differencein the work function of the metal@m)and the electron affinity(x)for thecase of an n-type semiconductor as shown in Eq. 3.1. In the case of a p-type25Chapter 3. Post-Growth Fabrication of Light Emitting Diodessemiconductor, it would be expected that the barrier voltage would be thedifference between the and the sum ofxand the bandgap energy (Eg),as shown in Eq. 3.2 and illustrated in Fig. 3.1. Fig. 3.1 shows a schematicof band bending for an ideal metal to n-type semiconductor interface asdescribed above.b—n type = — X(3.1)b-ptype=qm—(X+Eg) (3.2)Based on this reasoning, by choosing metals with different work functionsit should be possible to achieve a large range of values for A metal-semiconductor combination where cbb = 0 should result in an ohmic contact,based on this logic. In practice, most metals have the same barrier height ofabout 0.8 V when put in contact with n-doped GaAs, even though the abovediscussion would predict that barrier heights from 0.07 V to 0.57 V shouldbe possible [23]. The reason is that the surface states of the semiconductoractually set the barrier height. Fig. 3.2 shows a schematic of a more realisticmetal to n-type semiconductor interface.Current can flow across the metal-semiconductor interface by either thermionicemission, or quantum mechanical tunneling through the barrier, which canbe enhanced by applying a large electric field (field emission). Thermionicemission is the main mechanism for current flow through a Shottky diode,while current flow through ohmic contacts is usually due to tunneling. Thecurrent density for field emission through a barrier of height q has the formof Eq. 3.3, where e is the magnitude of the electron charge and E00 is given26Chapter 3. Post-Growth Fabrication of Light Emitting DiodesFigure 3.1: Schematic diagram of a metal to n-type semiconductor interfacein the absence of surface Eq. 3.4. Here 7. is Plancks constant, N is the doping concentration, c isthe dielectric constant andm*is the effective mass[25j. It is worth notingthat as the doping is increased the tunneling current increases exponentially.This is attributed to narrowing of the depletion region/barrier.J cexpE00(3.3)whereE00 = (3.4)To make a good ohmic contact it’s necessary for the surface layer tobe highly doped (usually in the 1019cm3 range for n-type and 1020cm3)range for p-type[23]. To achieve this high level of doping the contact ma-Ips27Chapter 3. Post-Growth Fabrication of Light Emitting Diodes1EEESurface StatesFigure 3.2: Schematic diagram of a metal to n-type semiconductor interfacewith surface states.terial usually contains an element, which will diffuse upon annealing intothe semiconductor, producing a highly doped surface layer. In the case ofAuGe n-contacts, which were used for n-type contacts here, this elementis germanium[23]. For contacting n-type GaAs, AuGe-based contacts havebeen the most successful, even though they are somewhat inconsistent andthe resistivity depends strongly on how well optimized the annealing conditions are. Typical contact resistivities of AuGe-based contacts range from0.8 x10—6cm2 to 4 x10—6cm2[23].Another method for improving ohmic contacts is to grade the compositionof the surface layer by alloying, into a low bandgap semiconductor, such asInAs. This method is not widely used for GaAs as many complications, suchas lattice matching arise[23].Many different material combinations are used to make ohmic contacts toGaAs. For contacting the p-doped side of our LED’s Ti/Pt/Au contacts were28Chapter 3. Post-Growth Fabrication of Light Emitting Diodesused [26] [27]. These contacts were selected because of their stability over athe wide annealing range of 420 C to 530 C and extremely low demonstratedcontact resistance of 2.8 x10—8flcm2[26]. Small circular Ti/Pt/Au p-typeohmic contacts with a diameter of 0.32 mm were deposited through a metalshadow mask using e-beam evaporation. Ti/Pt/Au thicknesses of about50 nm/100 nm/200 nm were used. Each circular p-type contact defined asingle device on top of the wafer. n-type contacts were made by evaporationof Ni/AuGe/Au [23] [28] onto the entire back side of the sample, forminga common contact with all the top contacts on the sample. Ni/AuGe/Authicknesses of about 25 nm/100 nm/200 nm were used. Proper annealingtemperatures for the n-type contact on the back were ignored because of thelarge surface of the contact. After deposition, the wafer was annealed at450°C for 20 seconds after deposition to improve the contact conductivity.The resistivity of Ti/Pt/Au contacts was tested by depositing a line ofthe 0.32 mm diameter metal dots onto an p-doped (about 2 x1018cm3)GaAs wafer cleaved to be about 1 mm by 25 mm. Contact resistances ofabout one ohm were found by measuring the resistance between dots asa function of separation distance, corresponding to a specific resistivity ofabout 5 x iO lcm2. It is expected that much lower contact resistivitiescould be obtained if higher doping concentrations were used, but these valueswere deemed adequate for test LEDs. Current voltage curves were confirmedto be linear over the current range of interest and dot-to-dot uniformity wasexcellent.29Chapter 3. Post-Growth Fabrication of Light Emitting Diodes3.2 Mesa EtchingAfter contacts were deposited and annealed, the top 300 nm of highly dopedGaAs around the contacts was etched off using aH2S04:H0wet etchwith volume ratios of 4:1:5 [23], which removed 5tm/min for a room temperature solution. The highly doped top layer was removed to minimizecurrent spreading. This etch was selected because it did not seem to damage the Ti/Pt/Au contacts, while some other etches removed the metal alltogether. The depth of the etch was measured with profilometry. It was discovered that etching through the intrinsic region resulted in high leakage andno light emission, attributed to non-radiative surface recombination. Aftermesas were etched, the sample was cleaved into sizes of about 5 mm by 3 mmand bonded to a small piece of copper using silver epoxy. Contact was madeto the dots by wire bonding the dots to terminals, which were glued to thecopper piece. Fig. 3.3 shows a schematic of the LED chip after metalizationand etching. The dashed lines indicate possible growth interrupts (depending on the sample). Most early samples were not etched, resulting in poorcurrent-voltage characteristics.30Chapter 3. Post-Growth Fabrication of Light Emitting DiodesFigure 3.3: Schematic of an LED chip after metalization and etching. Typicalthickness for i-GaAs layers was 25 nm, typical GaAsBi QW thicknesses were30 nm to 50 urn and typical p-GaAs thickness was 1 /mum. The top 300 nmof the p-GaAs layer was removed by etching.n-contact (Ni/AuGe/Au)31Chapter 4Electrical Characterization4.1 Current-Voltage Measurements ofGaAsi_BiDiodesCurrent-Voltage (I-V) measurements were made with a Keithley 220 currentsource and a Keithley 197A voltmeter, using a 2-probe method. Accordingto Keithley specification sheets the current source had an output resistancegreater than1014Q, absolute accuracy of sourced currents less than 0.1%over the current ranges used and 100 ppm noise in the source. The voltmeterhad a resistance greater than 1 GQ in the 2 V range and a resolution of10 jtV. Both the current source and voltmeter were interfaced to a computerthrough an IEEE 488 interface, and controlled using an in-house-made Lab-View program. I-V measurements were made over the current ranges from10 [LA reverse bias to 10 mA forward bias, usually with 10 data points takenper decade and a minimum step size of 100 nA. Unless otherwise mentioned,I-V measurements were made at room temperature.Fig. 4.1 shows current-voltage measurements of several devices from sample r1965, containing 1.8% bismuth. r1965 was the only GaAsi_Bi LEDto show strong electroluminescence and it thus highlighted here. Dot to dotuniformity was excellent, as the curves have excellent overlap.32Chapter 4. Electrical Characterization1x102.:‘:::‘:::;‘:1x101x10lxi 01lxlfY61x107I • •I_ • • I • • I-1.5 -1 -0.5 0 0.5 1 1.5Potential (V)Figure 4.1: Current-Voltage curves for four GaAs1_Bilight emitting diodesat 300 K. The red line indicates a fit to the data with n = 2.26 ideality factor.33Chapter 4. Electrical CharacterizationThe leakage current was approximately 5 tA at a reverse bias of 1 V.Fitting the forward bias data over the current range i0 A to iO A to thediode equation gives a saturation current (Is) of 7.1 A/cm2with a standarddeviation of 0.4 1A between dots. The fit also gave ideality factor of 2.26.Series resistances of 100 Q with standard deviation of less than 10 weremeasured at high currents. The reason for the high series resistance is unknown. The resistance is believed not to be due to the ohmic contacts, ascontact resistance is expected to be less than 1 Q as inferred from measurements on the p-doped wafer, which had similar doping concentrations.Fig. 4.2 shows current voltage data for another GaAsi_Bi sample (r1895),which was found to emit some light. The hollow data points with low reversebias leakage current are from different diodes for a device prepared soon aftergrowth, where no etching was done to remove the top highly doped GaAslayer. The sample was remeasured after about 2 months (red solid circles)and reverse bias leakage was an order of magnitude greater. The sample wasthen dipped in HC1 to remove any oxide that may have formed and remeasured (black solid diamonds), where it showed the same level of high leakage.A second sample was fabricated from growth r1985, where the top highlydoped GaAs layer was removed, but still showed the high leakage (pink solidtriangles) observed before. This very unusual degradation has no obviousexplanation.34Chapter 4. Electrical Characterization1x102 • • I1x1031x104•......D 1x105\C.)•‘••.Potential (V)Figure 4.2: Current-Voltage curves for GaAsi_Bi LED r1895. Hollow datapoints (low leakage values on reverse bias) correspond to dots from an unetched device fabricated soon after growth. Solid data points (high leakage)correspond to measurements made on the same sample 2 months later, withand without dipping the sample in HC1 to remove possible oxide and alsofor a new device fabricated from the wafer with etched mesas (solid pinktriangles).35Chapter 4. Electrical Characterization4.2 Current-Voltage Measurements ofInGai_AsDiodesFig. 4.3 shows I-V curves for three 1n0,18Ga082As diodes fabricated fromsample r1929 at the same time. The top highly doped layer of GaAs wasremoved by etching. The leakage current at —1 V was observed to be about0.5 ILA (an order of magnitude less than for the GaAsi_Bi sample r1965).Dot to dot uniformity was very good on reverse bias and for forward biasfor voltages less than 0.4 V. At higher voltages the diodes behaved veryinconsistently. Only the red curve in Fig. 4.3 showed the expected “rollingover” on the semilog plot due to the series resistance dominating the I-Vshape, the red and black curves appeared to roll over at 0.5 V, but thencurved back up before rolling over again. One possible explanation for thisis that irreversible changes were taking place in these samples causing thedevices to fail. The diodes were only tested once so this explanation has notbeen confirmed. All the diodes were observed to be strong light emitters.36Chapter 4. Electrical Characterizationlxi02lxiQ..3Z ixiO4C1-5oixi0lxi.6lxi-1.5 -1 -0.5 0 0.5Potential (V)1 1.5Figure 4.3: Current-Voltage curves for InGa1_As LEDs from sample r1929.37I II,. ..+ .+ .• .a+ .‘+ .&+ .+ .+ a+ a+ a• .+ •+ .• .• ‘•+ ‘.+ .+ ‘.•4a...... . . ...1. .. .1. • •.I.. • .1. •. .1. •.a.•• • • • I • ./ .1..1Chapter 5Optical Characterization5.1 Electroluminescence (EL) andPhotoluminescence (PL)Luminescence is the emission of photons of light when an atom, molecule, orcrystal system decays from an excited state to a lower energy state. Typesof luminescence are classified by the method of excitation. Here we discusselectroluminescence (EL), which is the emission of photons, by a materialin responce to an applied voltage. We also discuss Photoluminescence (PL),which is the absorption and then re-emission of photons by a material.Electrons can transition from an occupied initial state to a vacant finalstate, where the occupation probability of an energy state is given by thevalue of the Fermi function. The absorption and stimulated emission ratesR12 and J-?21 and the spontaneous emission rate R8 thus can be written inthe form in Eq.5.1, where R, and R0 are the transition rates of stimulatedand spontaneous processes if all state pairs are available.fandfare theFermi functions for the valence and conduction bands. Using Einsteins coefficients A and B as rate constants for spontaneous and stimulated processesrespectively R0 = A and R,. = BW(zi), where W(v) is the radiation spectraldensity [20].38Chapter 5. Optical CharacterizationR12 = Rrfv(lfc)(5.1)21=Rrfc(lfv)(5.2)R3 = R0f(1— f)(5.3)In the case of PL in semiconductors, light with energy greater than thebandgap is incident on the semiconductor. Electrons in the valence bandabsorb photons and are excited to the conduction band, creating electron holepairs, which can recombine either radiatively, or by a non-radiative process.The main types of non-radiative recombination processes in semiconductorsare defect, surface and auger recombination, these are briefly discussed atthe end of this section. Photoluminescence measurements give informationabout the electronic properties, such as material bandgap. The high energyside of the emission spectum is attributed to thermal excitation of electronsto higher energy levels in the conduction band, and similarly for holes inthe valence band. The shape of this tail can be modeled by a Boltzmanndistribution for relatively high temperatures (T > 100 K), as indicated inEq. 5.4, where Ih(E) is the intensity at photon energy E, k is the Boltzmannconstant and T is the absolute temperature.Ih(E) = Ae(5.4)Emission below the gap is possible due to thermal fluctuations in thelattice, and structural inhomogeneities. This low energy side of the peakcan be modeled using the product of an Urbach edge [29] and a Boltzmanndistribution, as shown in Eq. 5.5, where 1(E) is the intensity at photon39Chapter 5. Optical Characterizationenergy E, c is the 0 K absorption coefficient at the bandgap energy E9 andE0 is the Urbach parameter.E-E9 -f(E) = geEoekT(5.5)Light emission from LEDs is primarily a spontaneous emission process.From Eq. 5.1 we see that the rate of spontaneous emission between twolevels is proportional to the product of the probability that the upper stateis occupied and the probability the lower state is vacant. When a voltageis applied across an LED it causes a separation of quasi-Fermi levels equalto the applied voltage, which causes the occupation of a given level in theconduction band to increase and a decrease of the occupation of energy levelsin the valence band, thus increasing the spontaneous emission rate.Non-radiative recombination can be a significant, or even the dominantform of loss in a device depending on material quality and device design.There are three main types of non-radiative recombination; defect, surfaceand Auger recombination as shown in Fig. 5.1. Defect recombination is dueto impurity atoms or defects in the crystal, which have mid-gap energy levels.In this case the electron can fall from the conduction band to the defect leveland then recombine with a hole, thus depleting the conduction band withoutproducing photons. Surface recombination is similar to defect recombination,except the mid-gap levels are surface states. Auger recombination requirestwo electrons in the conduction band. One electron falls to the valence bandbut the energy is used to push the other electron up to a higher energy level,thus depleting the conduction band without producing light.40Chapter 5. Optical Characterization__zN_• •EvSurface AugerFigure 5.1: Possible non-radiative transitions in semiconductors.5.2 Photoluminescence MeasurementsAs a semiconductor is cooled, it is expected that the bandgap will shiftto higher energy. The shift in bandgap comes from a combination of atemperature-dependant dilation of the lattice and a temperature dependent electron-lattice interaction[30J. The temperature dependence can bedescribed by the Varshni equation shown in 5.6, where a and /3 are constants, Eg is the bandgap energy at absolute temperature T and E0 is thebandgap energy at T = 0 K. For GaAs, a = 8.871x104eV/K, /3 = 572 Kand E0 = 1.5216eV[31].aT2E9(T) = E0-I-/3+T(5.6)Fig. 5.2 shows photoluminescence spectra for a series of temperaturesfrom 8 K to room temperature (300 K) for GaAs1_Bi sample r1965. The532 nm green pump laser and collecting optics were focused on the area be-Defect41Chapter 5. Optical CharacterizationU).1>U)a)1200 1300Figure 5.2: Photoluminescence spectra for GaAs1_Bi LED structure r1965over a temperature range of 8 K to 300 K. The inset shows the peak emissionenergies as a function of temperature for both the GaAs and GaAs1_Bipeaks. A fit to the data using the Varshni equation is also shown (dashedline).800 900 1000 1100Wavelength, nm42Chapter 5. Optical Characterizationtween the metal dot contacts on the top of the sample. A clear peak wasobserved for both GaAs (870 nm) and GaAsi_Bi (975 nm), along witha wide low energy tail which is attributed to shallow defect states in thebandgap. The GaAsi_Bi peak emission wavelength at room temperatureagrees with the expected shift in bandgap energy for 1.8 % bismuth incorporation, 0.16 eV[6]; this estimate for the bismuth content is consistent witha fit to the split off peak seen in the[0041XRD rocking curve which gavex = 1.7 ± 0.2% for GaAsi_Bi. Both the GaAs and GaAsi_Bi peaksblue shifted with decreasing temperature (inset of Fig. 5.2) and followed theVarshni equation[30]:In the fit in the inset of Fig. 5.2, E0 for GaAs is 1.48 eV and E0 = 1.32 eVfor GaAs0982Bi0,018 with = 0.36 meV/K and /3 = 356 K in equation 5.6 forboth fits. The change in E0 corresponds to 1.8% bismuth, based on 88 meVband gap reduction per percent bismuth incorporation[6]. Temperature dependent PL measurements were not made on other samples.5.3 Electroluminescence MeasurementsElectroluminescence (EL) measurements were made at room temperatureunless otherwise stated. The light was collected from the periphery of thetop metal dot, since the metal dot was opaque. EL spectra from GaAsi_Bilight emitting diode r1965 are shown in Fig. 5.3 for forward bias currentdensities of 50 A/cm2,75 A/cm2 and 100 A/cm2.Two clear peaks are seen: GaAs at 870 nm and GaAsi_Bi at 987 nm,along with a low energy tail. With increasing current density a slight blue43Chapter 5. Optical CharacterizationCl)CD>Cl)Cci)-I-.CFigure 5.3: Electroluminescence spectra for a GaAsi_Bi light emittingdiode from sample r1965 for various injection current densities at 300 K.Room temperature photoluminescence is shown for comparison.800 900 1000 1100 1200Wavelength, nm44Chapter 5. Optical Characterizationshift in the peak emission wavelength of the GaAsi_Bi band to band peakwas observed, attributed to a further separation of quasi-Fermi levels athigher pumping currents. The shape of the spectra did not change. TheGaAsi_Bipeak intensity was found to be superlinear with increasing current density, possibly due to defect recombination saturation, resulting in ahigher fraction of additional carriers recombining radiatively.The room temperature PL is also shown in Fig. 5.3 for comparison. PLpeaks are blue shifted relative to the EL peaks and the low energy emissionfrom the defect states is much stronger in the EL spectra. Both of these observations can be explained by the differences in the way carriers are injectedfor PL and EL. In the former case carriers are injected with energies largerthan the bandgap, while in the latter case electrically injected carriers haveenergies close to the band edges and are thus restricted to low energy states.This results in an increasing tendency for recombination from higher energystates to occur in the case of PL.Earlier LED samples performed more poorly than r1965 LEDs. Fig. 5.4shows EL spectra from three GaAsi_Bi LEDs at 100 A/cm2 injection current. The black and red data are from two diodes (fabricated separately)from sample r1895, which contained about 1% bismuth. The blue data isfrom r1917, which contained 5.5% bismuth. EL intensity was very weak,compared to sample r1965 discussed above. All spectra show a peak at870 nm, corresponding to GaAs and broadband emission at wavelengths exceeding 1000 nm, corresponding to emission from defects. Long wavelengthemission intensities in Fig. 5.4, whether from defects or the GaAsi_Bilayer, are all more than an order of magnitude smaller than the emission45Chapter 5. Optical Characterizationfrom the GaAsi_Bi layer from sample r1965 at the same injection current. The long wavelength emission from r1895 could not have come fromthe GaAsi_Bi layer since XRD measurements from Fig. 2.5 showed theGaAsi_Bi layer had x = 0.01. Based on 88 meV reduction of the bandgapper percent bismuth{6], the GaAsi_Bi peak should have been at 930 nm.Sample r1917 was found by XRD measurements to have x = 0.055 in theGaAsi_Bi, which would put the expected GaAsi_Bipeak at 1325 nm.Looking at the EL spectra in Fig. 5.4 it appears that the long wavelengthemission (1000 nm to 1300 nm) from r1917 is composed of two broad peaks,the longer of which is at about 1300 nm, hence it is possible that some of thisemission came from the GaAsi_Bi layer. r1917 also had two growth interrupts, which is expected to have increased the amount of defects resulting infurther loss in the EL from the GaAsi_Bi layer.Fig. 5.5 shows EL from the InGai_As LED (1929) with x=0.18 forcurrent densities of 12 A/cm2,25 A/cm2,37 A/cm2 and 62 A/cm2. Thespectra show a small GaAs peak at 870 nm and a large peak at 1025 nm fromthe InGai_As QWs. As observed with the GaAsi_Bi LED, the peak ofthe InGai_As emission slightly blue shifted with increasing current density,due to further separation of the quasi-Fermi levels. No long wavelength tailfrom defects was observed in this sample. Unlike the GaAsi_Bi LED, whichshowed super-linear light emission, emission form the InGai_As device wassub-linear, as can be seen in Fig. 5.5. This suggests that there was less nonradiative defect recombination to saturate. The integrated room temperatureintensity of the InGa1_As device was about 100 x higher than for the r1965GaAsi_Bi device at the same current.Quantum wells are known to be46Chapter 5. Optical CharacterizationC,)CD-Q(‘3>C’)G)CFigure 5.4: Room temperature electroluminescence spectra from threeGaAsi_Bi LEDs at 100 A/cm2 injection current. The black and red spectra were fabricated from growth r1895 and contain 1% bismuth. The bluespectra (from r1917) contained 5.5% bismuth in the GaAsi_Bi layer.800 1000 1200Wavelength, nm47Chapter 5. Optical CharacterizationDCl)C1100Figure 5.5: Room temperature electroluminescence spectra for anInGa1_As light emitting diode for various injection current densities.900 1000Wavelength, nm48Chapter 5. Optical Characterizationmore efficient than bulk layers and it is expected that had the GaAsi_Bidevice also contained 3 similarly sized quantum wells that it would have beenbrighter than the existing GaAsi_Bi design.5.4 Temperature DependentElectroluminescenceFig. 5.6 shows the temperature dependence of the r1965 GaAs1_Bi LEDEL spectra at 50 A/cm2 from 100 K to 300 K.As the temperature decreased, a clear blue shift in the peak wavelengthof the GaAs in agreement with PL results was observed. Both the emissionfrom the GaAs1_Bi peak and the longer wavelength shallow defect statesincreased with decreasing temperature. However, emission from the defectstates also increased relative to the GaAs1_Bi band to band peak. At 100 Kthe intensity of defect emission surpassed the emission of the GaAsi_Biband to band peak. In contrast to the PL measurements shown in Fig. 5.2,the intensity of the GaAs peak decreased as the device was cooled. This resulted from a greater carrier confinement in the smaller bandgap GaAs1_Bilayer. No shift in peak wavelength of theGaAsi_Biemission was observedover this temperature range. This can be explained by two competing processes, the increase in the bandgap at lower temperature and the increasedtendency for emission to come from lower energy states at lower temperatures. These two processes combined result in the observed temperature independent peak emission wavelength of the GaAsi_Bi band to band electroluminescence. The temperature sensitivity of PL measurements can be49Chapter 5. Optical Characterization(I)CD>U)Ca)CFigure 5.6: Electroluminescence spectra from GaAsBi LED r1965 at50 A/cm2 injection current density for temperatures ranging from 100 Kto 300 K.800 900 1000 1100 1200Wavelength, nm50Chapter 5. Optical Characterizationexplained by the low density of states of the GaAsi_Bi impurity like statesin the bandgap. It is possible that as the temperature is lowered PL emissionof these lower energy states does not increase because the electrons-hole pairsrecombine before they have a chance to find the deeper impurity-like states.51Chapter 6ConclusionGaAsi_Bi is an exciting new semiconductor with promising applicationsas a new infrared light source. The characteristics of theGaAsi_Bi appearto be well matched to the requirements for optical coherence tomography.We have developed a method for the growth and fabrication ofGaAsi_Bilight emitting diodes, based on a simple n-i-p structure containing oneGaAsi_Bilayer in the intrinsic region. n-i-p structures with up to5.5% bismuth in theGaAsi_Bi layer have been realized. Strong light emission was obtainedfrom a sample containing 1.8% bismuth with emission centered at 987 nm.ThisGaAs1_Bi emission peak was found to be independent of temperatureover the temperature range 100 K to 300 K. The long wavelength emissiondid increase, which changed the shape of the spectra.The temperature insensitivity of theGaAsi_Bi electroluminescence canbe explained by two competing processes; the increase in the bandgap withdecreasing temperature and the tendency for emission to come from lowerenergy states as temperature is decreased. Photoluminescence measurementson the same device showed a temperature dependence of theGaAsi_Biband to band transition consistent with measurements by Francoeur [6]. Boththe electroluminescence and photoluminescence spectra have large spectralwidths which is a reflection of the impurity-like nature of the states associated52Chapter 6. Conclusionwith bismuth incorporation. Difficulties incorporating bismuth into the n-i-pstructure are an unresolved issue as well as the poor performance of many ofthe GaAsi_Bi layers.This demonstration of a GaAsi_Bi based LED opens up a new class ofmaterials for long wavelength semiconductor light sources with broad emission spectra.6.1 Future WorkMuch work must still be done to reliably growGaAsi_BiLED structureswith the desired amount of bismuth. The fact that many of the growths didnot show strong luminescence from theGaAs1_Bi layer also requires furtherinvestigation. GaAs n-i-p samples grown at different growth conditions weregrown but have not yet been fabricated and characterized, this will have tobe done to investigate whether the low temperature GaAs or theGaAsi_Bilayer is the source of the impurity emission. NewGaAs1Bidevices shouldalso be grown and fabricated in attempt to obtain strong electroluminescencefrom a GaAsi_Bi light emitting diode with 1.0 eV bandgap. This will showthat GaAsi_Bi devices can exceed the wavelength range ofInGai_Asdevices. Improvements to this rudimentary test device could be made by:optimizing the thicknesses of theGaAsi_Bi and undoped GaAs layers; optimizing contact design; the use of an A1GaAs double heterostructure design;and removing the growth interrupt.53Bibliography[1] H. Foil. Christian-Albrechts-University of Kiel,[2] X. Lu, D. Beaton, T. Tiedje, R. Lewis, M.B. Whitwick. Effect of MBEGrowth Conditions on Bi Content of GaAsi_Bi. Applied Physics Letters, 92, 2008.[3] M.J. Antoneli, C.R. Abernathy, A. Sher, M. Berding, M. Van Schiifgaarde, A. Sanjuro K. Wong. Growth of Ti-Containing ITT-V Materialsby Gas-Source Molecular Beam Epitaxy. Journal of Crystal Growth,188:113—118, 1998.[4] S. Tixier, M. Adamcyk, E.C. Young, J.H. Schmid, T. Tiedje. Surfactantenhanced growth of GaNAs and InGaNAs using bismuth. Journal ofCrystal Growth, 251:449—454, 2003.[5] K. Oe, H. Okamoto. New Semiconductor Alloy GaAsi_Bi Grownby Metal Organic Vapor Phase Epitaxy. Japanese Journal of AppliedPhysics Part 2, 37:L1283—L1285, 1998.[6] J. Yoshida, T. Kita, 0. Wada, K. Oe. Temperature Dependenceof GaAsl-xBix Band Gap Studied by Photoreflectance Spectroscopy.Japanese Journal of Applied Physics Part 1, 42:371, 2003.54Bibliography[7] S. Francoeur, M.J. Seong, A. Mascarenhas, Sebastien Tixier, MartinAdamcyk, Thomas Tiedje. Band Gap of GaAsi_Bi, 0<3.6%. AppliedPhysics Letters, 82:3874—3876, 2003.[8] S. Adachi. Physical Properties of 111-V Semiconductor Compounds. Wiley, 1992.[9] T. Tiedje, E.C. Young, A. Mascarenhas. Growth and Properties of theDilute Bismide Semiconductor GaAsi_Bi a Complementary Alloy tothe Dilute Nitrides. International Journal of Nanotechnology, 5:963—983,2008.[10] l.A. Janotti, S.H. Wei, S.B. Zhang. Theoretical study of the effects ofisovalent coalloying of Bi and N in GaAs. Physical Review B, 65:115203,2002.[11] 1.W. Shan, W. Walukiewicz, J.W. Ager, E.E. Haller, J.F. Geisz, D.J.Friedman, J.M. Olson, S.R. Kurtz. Band Anticrossing in GaInNAs Alloys. Physical Review Letters, 82:1221, 1999.[12] A. Mascarenhas, Y. Zhang, J. Veerley, M.J. Seong. Overcoming Limitations in Semiconductor Alloy Design. Superlattices and Microstructures,29:395, 2001.[13] L. Vegard. Die Konstitution der Mischkristalle und die Raumfullung derAtome. Z. Physik, 5:17, 1921.[14] U. Tisch, E. Finkman, J. Salzman. The anomalous bandgap bowing inGaAsN. Applied Physics Letters, 81:463—465, 2002.55Bibliography[15] E. Nodwell, M. Adamcyk, A. Ballestad, T. Tiedje, S. Tixier, S.E. Webster, E.C. Young, A. Moewes, E.Z. Kurmaev, T. van Buuren. Tight-Binding Model for the X-ray Absorption and Emission Spectra of DiluteGaNxAsl-x at the Nitrogen K edge. Physical Review B, 69:15520, 2004.[16] K. Oe. Characteristics of Semiconductor Siloy GaAsl-xBix. JapaneseJournal of Applied Physics Part 1, 41:2801—2806, 2002.[17] D. Huang, E.A. Swanson, C.P. Lin, J.S Schuman, W.G. Stinson, W.Chang, M.R. Hee, T. Flotte, K. Gregory, C.A. Puliafito, J.G. Fujimoto.Optical Coherence Tomography. Science, 254:1178—1181, 1991.[18] J.M. Schmitt. Optical Coherence Tomography (OCT): A Review. IEEESelected Topics in Quantum Electronics, 5:1205, 1999.[19] M.R. Pillai, S.S Kim, S.T. Ho, S.A. Barnett. Growth ofInGaiAs/GaAsheterostructures using Bi as a surfactant. Journalof Vacuum Science and Technology B, 18:1232—1236, 2000.[20] E.C. Young, S. Tixier, T. Tiedje. Bismuth Surfactant Growth of theDilute Nitride GaNAsi_. Journal of Crystal Growth, 279:316—320,2005.[21] L.A. Coidren, S.W. Corzine. Diode Lasers and Photonic Integrated Circuits. Wiley Series in Microwave and Optical Engineering, 1995.[22] M. Balkanski, R.F. Wallis. Semiconductor Physics and Applications.Oxford, 2000.56Bibliography[23] S.R. Johnson, C. Lavoie, T.Tiedje. Semiconductor Substrate Temperature Measurement by Diffuse Reflectance Spectroscopy in MolecularBeam Epitaxy. Journal of Vacuum Science and Technology B, 11:1007—1010, May 1993.[24] R. Williams. Modern GaAs Processing Methods. Artech House, 1990.[25] T.C. Shen, GB. Gao, H. Morkoc. Recent Developments in Ohmic Contacts for ITT-V Compound Semiconductors. Journal of Vacuum Scienceand Technology B, 10:2113—2132, 1992.[26] S.M. Sze. Physics of Semiconductor Devices. John Wiley and Sons,1981.[27] G. Stareev. Formation of Extremely Low Resistance Ti/Pt/Au OhmicContacts to p-GaAs. Applied Physics Letters, 62:2801—2803, 1993.[28] A. Katz, C.R. Abernathy, S.J. Pearton. Pt/Au Homic Contacts to Ultrahigh Carbon-doped p-GaAs Formed by Rapid Thermal Processing.Applied Physics Letters, 56:1028—1030, 1990.[29] W.D. Edwards, A.B. Torrens, W.A. Hartman. Specific Contact Resistance of Ohmic Contact to Gallium-Arsenide. Solid-State Electronics,15:387, 1972.[30] S.R. Johnson, T. Tiedje. Temperature Dependence of the Urbach Edgein GaAs. Journal of Applied Physics, 78:5609—56 13, 1995.[31] Y.P. Varshni. Temperature Dependence of the Energy Gap in Semiconductors. Physics, 334:149—154, 1967.57Bibliography[32] M.D. Sturge. Optical Absorption of Gallium Arsenide between 0.6 and2.75 eV. Physical Review, 127:768—773, 1962.58


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items