UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Lateral wet oxidation of AIGaAs and its applications in high index contrast distributed feedback ridge… Chen, Peng 1997

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1998-0090.pdf [ 7.31MB ]
Metadata
JSON: 831-1.0088508.json
JSON-LD: 831-1.0088508-ld.json
RDF/XML (Pretty): 831-1.0088508-rdf.xml
RDF/JSON: 831-1.0088508-rdf.json
Turtle: 831-1.0088508-turtle.txt
N-Triples: 831-1.0088508-rdf-ntriples.txt
Original Record: 831-1.0088508-source.json
Full Text
831-1.0088508-fulltext.txt
Citation
831-1.0088508.ris

Full Text

LATERAL WET OXIDATION OF AlGaAs AND ITS APPLICATIONS IN HIGH INDEX CONTRAST DISTRIBUTED FEEDBACK RIDGE WAVEGUIDES by PENG CHEN BSc, Xiamen University, P. R. China, 1994 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENT FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES (Department of Physics and Astronomy) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA « December 1997 © Peng Chen, 1997 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department The University of British Columbia Vancouver, Canada Date DE-6 (2/88) II Abstract A high temperature wet oxidation technique has been developed to laterally oxidize high aluminum content AlxGa t.xAs (x=98%) sandwiched between GaAs and low aluminum content AlxGai_xAs (x=70%). The oxides can be used to produce optical confinement and the current confinement in optoelectronic devices. To characterize the products of this high temperature oxidation, atomic force microscopy has been use to study the induced structure change. No significant vertical expansion has been detected demonstrating that this technique is particularly useful for distributed reflector devices. The composition of the oxidation product was also studied by x-ray photoelectron spectroscopy revealing that the high temperature oxidation produces arsenic poor A10x. We explained that the As is" lost by thermal evaporation of AsH3 during the high temperature oxidation process. By studying the binding energy of Al 2p core levels in the oxide, it has been found that the high aluminum content AlxGai_xAs (x=98%) oxidizes to an A10x while the low aluminum AlxGaj.xAs (x=70%) produces A1203. ' ' , To apply the lateral oxidation to distributed feedback waveguides, we used electron beam lithography and wet etching to fabricate test waveguides. By optimizing the process parameters, 100 nm linewidth gratings with spacing greater than 280nm have been achieved. A multipass writing technique was developed to attenuate the electron beam writing noise and achieve superior uniformity. The surface grating ridge waveguides have been fabricated on the laterally oxidized AlxGai_xAs layer successfully. iii Optical diffraction measurements have been conducted on second-order distributed-feedback waveguides containing a buried oxide with the grating pitch from 300 nm to 500 nm. The dispersion relationships of photon energy versus propagation constant have been determined for 440 nm and 460 nm grating waveguides. The experiments give the effective indices of the waveguides to be 1.7 for the both fundemental TE modes and fundemental TM modes, which show the guided modes were strongly modulated by the buried oxide. I V Table of Contents Abstract ii Table of Contents iv List of Table vi List of Figures vii Acknowledgment x Chapter 1 Introduction 1 Chapter 2 Application of Wet Oxidized AlGaAs 5 2.1 Formation of Partially Oxidized AlGaAs 5 2.2 Current Confinement 8 2.2.1 Selectively Oxidized Vertical-Cavity Lasers 8 2.2.2 GaAs on Insulator Technology 10 2.3 Optical Confinement 12 2.3.1 Distributed Bragg Reflector of High Index Contrast 12 2.3.2 High Index Contrast Distributed Feedback Structure 16 2.3.2.1 Coupling Coefficient 18 2.3.2.2 High Index Contrast Grating by Lateral Oxidation 22 2.4 Summary 26 Chapter 3 Lateral Oxidation of AlGaAs 27 3.1 Lateral Oxidation Process 27 3.2 Calibration of the Oxidation Rate 28 3.3 Structure Change of the Oxide 31 3.4 Vertical Expansion 32 3.5 XPS Composition analysis 34 3.5.1 XPS Measurements 34 3.5.2 Composition Analysis of Oxidized AlGaAs 35 3.5.3 Oxidation States of High Temperature Oxidized AlGaAs 40 Chapter 4 Waveguide Fabrication 44 4.1 Fabrication Procedures of waveguides 44 4.2 Grating Fabrication 45 4.2.1 Sample Preparation 47 4.2.2 Development 49 4.2.3 Etching 50 4.2.4 Electron Beam Writing 52 V 4.2.4.1 Writing System Components 52 4.2.4.2 Pattern Resolution 53 4.2.4.3 Writing Current 53 4.2.4.4 Accelerating Voltage 54 4.2.4.5 Pattern Rotation 55 4.2.4.6 Dose Tests 56 4.2.5 Improvement of Gratings Uniformity 60 4.2.5.1 Noise Sources and Their Characteristics 60 4.2.5.2 Noise Reduction 62 4.3 Structural Study of the Grating Surface by X-ray 65 4.4 Ridge Definition 69 4.4.1 E-beam lithography using PN-114 70 4.4.2 Ridge Alignment 71 4.5 Thinning 74 Chapter 5 Characterization of High Index Contrast DFB Waveguides 76 5.1 Sample and Experimental Setup 76 5.2 Measurement and Analysis 79 Chapter 6 Conclusion 83 References 86 List of Tables Table 3-1. Sputtering time of XPS measurements 36 Table 3-2. Binding energy for aluminum 2p core levels 41 Table 4-1. Recommended sample preparation and development conditions 49 for high uniformity 100 nm linewidth grating Table 4-2. Recommended ranges for the writing parameters to get 100 nm 58 linewidth gratings and less than 100 nm diameter dots Table 4-3. Recommended process for developing pattern using PN114 71 vii List of Figures Figure 2-1. Partially oxidized high aluminum content AlGaAs mesa, (a) sideview 7 of the mesa before oxidation, (b) sideview of the mesa after oxidation, (c) Nomarski microscope topview of partially oxidized mesa. Figure 2-2. Sketch of an oxide-confined VCSEL showing oxidized layers and the 9 resulting current apertures on each side of the optical cavity. Figure 2-3. (a) Sideview of a GaAs transistor on oxidized AlAs. (b) Epitaxial 11 layer structure of the device. Figure 2-4. Schematic drawing of a distributed Bragg reflector, where k i , k R , and 13 kx are wave vectors for incidence, reflection, and transmission light respectively. Figure 2-5. Reflectance spectrum of 4-period Bragg reflector. The layers are a 15 quarter-wave thick at a wavelength of Ao=980 nm. The top curve is for unoxidized AlAs/GaAs. The lower curve is for oxidized AlAs/GaAs. Figure 2-6. Schematic illustration of a distributed-feedback semiconductor laser. 18 Different refractive indice on opposite sides of the grating result in a periodic index perturbation that is responsible for the distributed feedback. Figure 2-7. Schematic illustration of the DFB grating used for lateral oxidization. 22 Figure 2-8. (a) Calculated second order coupling coefficient K for the TE modes 24 as a function of emission wavelength for a square grating shown in figure 2-7 and (b) the ratio of K for an oxidized grating to that of an unoxidized grating. Figure 2-9. Calculated coupling coefficient K as a function of grating depth for 25 oxidized and unoxidized gratings. The order of Bragg diffraction is denoted by m. Here the pitch of the gratings is designed to meet a 980 nm emission wavelength. Figure 3-1. Schematic of the setup for oxidation. 28 Figure 3-2. The top view of the partially laterally oxidized ASU721 in a Nomarski 29 optical microscope. Figure 3-3. The calibration of the oxidation rate for Alo.9gGao.02As. The sample 30 was oxidized at 425°C with water vapor from a water bottle held at 94°C. viii Figure 3-4. SEM micrograph showing a bird's eye view of the flaked off corner 32 of a laterally oxidized sample ASU721. Figure 3-5. AFM micrograph showing a bird's eye view of the partially oxidized 33 mesa on ASU721. The AFM was operated in tapping mode with a 0.5936Hz scanning rate. Figure 3-6. Al and Ga percentage of the oxidized ASU721 versus sputtering time. 36 Figure 3-7. A wide scan XPS spectrum of the oxidized ASU721 sample after 37 spputtering for 90 minutes with Ar +. The spectrum was taken by averaging 10 individual scans. The interval between sampling points is 0.8 eV and the sampling time was 80 ms. The oxygen Auger signal in the graph is indicated by (a). Figure 3-8. A wide scan XPS spectrum of a GaAs substrate after a long time air 39 exposure. The spectrum was taken by averaging 10 individual scans. The interval between two sampling points is 0.8 eV and the sampling time was 80 ms. The Auger signal in the graph is indicated by (a). Figure 3-9. Photoemission spectra showing aluminum 2p core levels for two 42 different sputtering depths C and F. The circles and triangles in the graph are measured points. The solid lines in the graph are mixed Guassian/Lorentzian fits of the data. Figure 4-1. Procedure for grating fabrication. 47 Figure 4-2. Calibration of the etch rate of H ^ O ^ C ^ ^ C - (1:1:72 by volume) 51 on a GaAs substrate. The dots are data measured from experiments. The solid line is a best fit which gives an etch rate of 90.3 nm/min. Figure 4-3. Feature size versus dose for 300nm spacing grating. The zero 57 linewidth for an incident dose less than 2.8 fC/point means that the patterns did not go through. Figure 4-4. (a) SEM micrograph of a ID grating, whose pitch was designed to 59 be 300 nm. (b) SEM micrograph of a 2D square lattice with 290 nm pitch. Figure 4-5. (a) 340 nm pitch grating written in 13 passes. The pattern was written 64 at 1000X magnification, (b) A circular grating was also written by multi-pass writing. Figure 4-6. Atomic force micrograph of a two dimensional surface grating. 66 ix of a surface grating perpendicular to [110] on a (001) substrate,. k{ is the incident beam and ko is the diffracted beam. Figure 4-8. The x-ray rocking curve obtained by (224) grazing output scan. 68 Figure 4-9. (a) The layout of the grating patterns and the alignment marks. 72 (b) Scanning windows and the overlays to match the alignment marks. Figure 4-10! SEM micrograph showinga side view of the DFB ridge waveguide. 74 The pitch of the grating is 300 nm. Figure 5-1. (a) A sketch of the DFB ridge waveguide from the top. (b) SEM 11 micrograph top view of the waveguide. Figure 5-2. Schematic of the diffraction experiment set-up. 78 Figure 5-3. Schematic of the diffraction measurement. 79 Figure 5-4. Dispersion relationship for theTE mode with a 460 mri pitch grating. 81 Data was collected in J. Young's Lab by Paul Paddon. Acknowledgments I would first like to thank my supervisor, Dr. Tom Tiedje, for his generous offer not only the financial support but also great opportunities. His invaluable suggestions and insight should be claimed everywhere through the thesis. I am also indebted to Dr. Jeff Young and his group including Dr. Paul Paddon, Alex Busch, and Vighen Pacradouni. I appreciate Dr. Shane Johnson from Arizona State University for supplying the MBE samples for this work. Dr. Phillips Wong's helps on XPS measurements and results analysis should also be acknowledged. As well, I would like to give a big thanks to Jim Mackenze for his help on equipments maintanance and repairments. Finally, to my parents, for their encouragements and emotion supporting during my studying in Canada. The academic help should be mentioned as well for my father Dr. Yongqi Chen and my sister Dr. Ying Chen's help on proof-reading portions of the thesis. 1 Chapter 1. Introduction Compound semiconductors, led by GaAs, are an important field of microelectronics technology. A tremendous advantage of compound semiconductors is that several pairs of compound semiconductors are miscible so that homogeneous alloys can be formed. These alloys allow device designers and crystal growers to tailor the electronic properties and the chemical composition of the alloys to meet the design targets for particular device applications. Unlike Si, most compound semiconductors are direct bandgap materials, and this makes compound semiconductors suitable for optoelectronic applications. However, unlike Si GaAs has no native oxide which is a major drawback. This limits the types of microelectronic devices which can be fabricated. On the other hand, silicon dioxide (Si02), the native oxide of silicon, can be formed on Si at high temperature in an oxidizing i ambient. The resulting oxide is an extremely versatile and stable material. This unique oxidation technique led to the development of Si-based metal-oxide-semiconductor (MOS) devices, which are the basis of today's integrated circuit (IC) technology. To get around the lack of a native oxide, many studies have been conducted and various solutions to the insulator problem have been proposed. They include chemical vapor deposition (CVD), sputtering of a dielectric such as Si02, silicon nitride (Si3N4), and 2 aluminum oxide (AI.2O3), direct oxidation of GaAs, dry oxidation of AlAs, and plasma oxidation of InGaAs. All of these insulating films have relatively poor insulator-semiconductor interfaces, making them unsuitable for high quality MOS and optoelectronic devices. Another strategy has also been introduced to bypass the problem with the lack of insulator in GaAs. The III-V bandgap engineering is used to create compound semiconductor transistors, for example, the Complementary Heterostructure Field Effect Transistor (TTEMT).cl'21 These kinds of devices require epitaxial growth of wider bandgap semiconductor materials to form heterostructures or quantum well. Aluminum containing materials like AlGaAs, InAlAs, AlGaP are the most commonly chosen materials. However, microelectronics engineers have to face the problem of degradation of devices in this case. For example, the ternary compound AlxGa).xAs decomposes within a few months on exposure to air, and pure binary compound AlAs (x=l) decomposes in a few minutes. The chemical instability of high Al content AlGaAs in air is due to the oxidation of AlGaAs. This oxidation does not give a mechanically stable oxide film, but leads to the corrosion of the layer, and eventually the epi-layer turns into powder. The problem is to find methods to form a high-quality insulating oxide. Recently, it has been discovered that a highly stable oxide can be produced if the oxidation of high aluminum content AlGaAs is carried out in a high temperature environment (~400°C) with a water vapor.'3] With this process, not only can the binary AlAs and the ternary AlGaAs form a high quality oxide, but other aluminum containing compound semiconductors, such as InAlAs, AlGaP and InAlP, can also be oxidized in a similar manner. High quality oxide films have been obtained.[4] 3 These high temperature oxides are insulating and transparent with refractive index n~1.6. Similar to the native oxide of silicon, it should be possible to use the insulating property to confine current. Moreover, since most compound semiconductors are direct bandgap materials and have been widely used in optoelectronic devices, the relatively low refractive index of these high temperature oxides can be used to achieve optical confinement in optoelectronic applications. The main objective of this research is to investigate the wet oxidation technique and apply it to semiconductor waveguides. The study is carried out on the GaAs/AlGaAs system. This thesis consists of 6 chapters. Chapter 2 illustrates how the electrical and optical confinement can be achieved by using this wet oxidation technique. The effective mode theory and coupled wave analysis are applied to the DFB waveguide structures which have a wet oxide grating incorporated in them. A theoretical simulation has shown that with a proper design, wet oxidation can lead to much larger coupling coefficient, which is essential for high performance single frequency lasers. Chapter 3 contains measurement results and their analysis. The process recipe showing the method used to conduct oxidation is described first. Then the oxidation rate calibration conducted at 425°C is given. Atomic force microscopy (AFM) is used to study the surface morphology change due to the oxidation. The composition and structure of wet oxidized AlGaAs have been studied with the x-ray photon electron spectroscopy (XPS). Primarily A10x has been found in wet oxidized AlGaAs. It is compared with the room temperature degraded GaAs which is an As-rich compound. The 4 binding energy of oxidized AlGaAs has shown that wet oxide of high aluminum content AlxGai_xAs (x=98%) is AlO x while the oxide of low aluminum content AlxGai-xAs (x=70%) gives the A1203 Chapter 4 describes the fabrication of DFB waveguides, including the sample preparation, pattern definition by the electron beam lithography and chemical etching, as well as sample thinning and cleaving. The emphasis is placed on the fabrication of ultra-high resolution (-100 nm) and high uniformity DFB gratings. Chapter 5 describes diffraction measurements from guided modes in the DFB ridge waveguide on the oxidized AlGaAs. The dispersion relationships for the fundamental TE and TM modes for waveguides with 440 nm and 460 nm spacing gratings are determined. A low effective index guided mode is found. The thesis concludes in Chapter 6 with a summary of the findings of this study and recommendations for future work. 5 Chapter 2. Application of Wet Oxidized AlGaAs Wet oxidized high-aluminum content AlxGai.xAs (x>90%) has two attractive properties: it is electrically insulating and has a relatively low refractive index. These properties make it useful for electrical (current) and optical confinement. This chapter addresses these two properties and illustrates their applications. First we explain how the AlGaAs is oxidized 2.1 Formation of Partially Oxidized AlGaAs To form wet oxidized AlGaAs an epitaxial layered GaAs/AlGaAs structure is first produced by metalorganic chemical vapor deposition (MOCVD) or molecular-beam epitaxy (MBE). Usually, the AlGaAs layer which is to be oxidized is protected by a GaAs cap layer from air exposure. The fundamental technology for the formation of AlGaAs oxides is the "Frosch steam oxidation process",[5] which has been used to produce Si02 layers on Si wafers. The detailed process is given in Chapter 4. The spectacular feature of this process is that the oxidation of an AlGaAs layer is carried out from the edge of the exposed layer instead of from its surface. Therefore, this oxidation is sometimes referred 6 to as "lateral oxidation". Hereafter the technique will be called either wet oxidation or lateral oxidation. Figure 2-1 demonstrates this lateral oxidation on a mesa structure. In Figure 2-1, a 55 by 55 urn mesa is defined on the epitaxial layer structure ASU721 grown by MBE. In ASU721, 2000 nm Alo98Ga0 0 2 As is first grown on a GaAs buffer, followed by 30 nm of Alo.3Gao.7As and then covered by 40 nm of GaAs. The mesa fabrication begins with e-beam lithography to define the mesa which is then etched by a 1:1:72 mixture of H 2S0 4 : H 2 0 2 : H 2 0 for 55 seconds to expose the high aluminum content AlGaAs layer. The mesa transfer is finished by 1.2% HF etching to expose the edge of the whole layer of 2um of Alo.98Gao.02As. During the etch, the patterned photoresist acts as mask. After the photoresist is removed, the sample is oxidized in a water vapor ambient at 425°C for 20 minutes. Figure 2-l(c) shows the top view of this partially oxidized mesa under a Nomarski optical microscope. The oxidation changes the refractive index as can be seen in Figure 2- 1(c) which clearly shows the difference in color between the unoxidized AlGaAs and the surrounding oxide. It also shows that the lateral oxidation is very uniform. The lateral-oxide defined central square is very similar to the square mesa, indicating that the oxidation rate is the same in the various <110> directions. The oxide layer beneath the GaAs mesa isolates the current injection and limits current flow to the central aperture (current confinement). The dimension of the aperture can be precisely controlled by adjusting the oxidation time. 7 40 nm GaAs 30 nm Alo.3Gao.7As 2 u.m Al 98Ga.o2As GaAs Substrate (a) (c) Figure 2-1. Partially oxidized high aluminum content AlGaAs mesa, (a) sideview of the mesa before oxidation, (b) sideview of the mesa after oxidation, (c) Nomarski microscope topview of partially oxidized mesa. 8 2.2 Current Confinement The first application of wet oxidized AlGaAs was in the production of stripe-geometry edge-emitting lasers by Dallesasse and his co-workers in 1991[6]. In their work, wet oxidized AlGaAs was used to confine the injected current to achieve a lower threshold. Later, several groups took advantage of the insulating property and developed various devices, like vertical-cavity surface emitting laser (VCSEL), [ 7" l l ] index-guided laser array1 J, buned-mesa index-guided edge-emitting laser,1 real-index-confined photonic lattice light-emitting array'13-1 and a new GaAs transistor.'14'151 The most successful work so far is on VCSEL. In this section two basic application areas of current confinement are discussed, i.e., VCSEL and GaAs on insulator (GOI) technology. 2.2.1 Selectively Oxidized Vertical-Cavity Lasers Vertical-cavity surface emitting lasers (VCSELs), in particular the distributed Bragg reflector (DBR) VCSEL has been regarded as an important component of optoelectronic integrated circuits. Although hundreds or even thousands of VCSELs can be integrated on a single chip, a serious problem with heat dissipation still needs to be solved. A VCSEL usually needs up to 10mA threshold current to stimulate laser action, and the power conversion efficiency is below 10%. At present, the standard method to reduce threshold current is ion implantation.[16] In this method, ions are incident on a selective area of sample surface and then penetrate into the sample. The implanted ions create a region of 9 high resistivity that restricts the current flow to an opening in the implanted region. However, the ions will damage the top structure of the VCSEL, which increases optical absorption and causes the threshold current to rise. The ion penetration depth is material and structure dependent. The lateral oxidation technique can overcome the above drawbacks. Preliminary results show that the technique can dramatically improve the performance of DBR-VCSELs'2"1 6 1 Figure 2-2 is a sketch of an oxide-confined VCSEL. Figure 2-2. Sketch of an oxide-confined VCSEL showing oxidized layers and the resulting current apertures on each side of the optical cavity. [From Ref.15] It has several advantages. Firstly, the ion implantation damage mentioned above can be avoided. Secondly, the current apertures immediately surrounding the optical cavity can 10 eliminate sidewall nonradiative recombination present in etched air-post VCSELs, and minimize lateral current spreading outside of the laser cavity. Finally, the smaller refractive index of the oxide layer also induces index-guided optical confinement.'17'18' The oxide-confined VCSEL enhances electrical and optical confinement which should reduce the threshold current/voltage and the concomitant parasitic ohmic heating. An ordinary ion implanted VCSEL usually needs 10 mA or above as a threshold current, while for a lateral oxide-confined laser a threshold current as low as 91pA has been reported'8'. Meanwhile, the highest power conversion efficiency (>50%)'191, lowest threshold current density (90 A/cm2 per quantum well)'101 and the lowest threshold voltage (50mV above photon energy)'20' have been recorded with the oxide-modified VCSEL's. 2.2.2 GaAs on Insulator Technology In figure 2- 1(c), if the oxidation time is prolonged, the central current aperture will become smaller, and eventually disappear when it is long enough. In this situation, the top structure will be electrically insulated from the substrate by the buried oxide. Recently, GaAs transistors have been successfully fabricated on insulators for the first time using this insulation technique1211. Figure 2-3(a) is a sketch of this so-called GaAs on insulator technology. Figure 2-3(b) gives the expitaxial layer profile used to fabricate the device. Compared with the current heterojunction transistors, this kind of transistor is very much like a traditional silicon MOS transistor. The technology bypasses the 11 quantum well growth which in general simplifies the epitaxial growth. On the other hand, there is no AlGaAs layers involving in the carrier movement. This avoids the influence of air corrosion of AlGaAs on the device performance. The last, but the most important advantage is that the oxidized AlAs reduces the current leakage to the substrate. Measurements have shown that current leakage through 50 nm oxide to the substrate is less than 5 fA/u,m2, f2 l ] which implies that the substrate leakage of this MOSFET is negligible. These results indicate that the lateral oxidation is a useful technique for reducing the power dissipation of transistors which can make this technology attractive for low power electronics. Gate n-channel (n-GaAs) 50 nm oxidized AlAs " t ; , , ' ?',J'/'V/''I V'''' GaAs substrate (a) 20 nm InAs/GaAs n+ ohmic cap 3 nm AlojGao.jAs 200 nm n-GaAs 10 nm Alo.3Gao.7As 50 nm AlAs 200 nm GaAs Buffer GaAs substrate (b) Figure 2-3. (a) Sideview of a GaAs transistor on oxidized AlAs. (b) Epitaxial layer structure of the device. [Ref. 21] 12 2.3 Optical Confinement The essence of dielectric optics is to use various materials with different refractive index to modulate an optical field and guide light propagation. The larger the index contrast of the different materials, the stronger modulation can be obtained. In a GaAs/AlGaAs system, the largest index contrast An is about 0.5. However, if AlGaAs is laterally oxidized, the refractive index for the oxidized AlGaAs changes to about 1.5,[22] an index contrast up to 2.0 is achievable. In the following, distributed Bragg reflector and distributed feedback structures are used to demonstrate that the performance of devices can be dramatically improved with such a high index contrast. 2.3.1 Distributed Bragg Reflector of High Index Contrast High-reflectivity distributed Bragg reflectors (DBRs) are used in a wide variety of optoelectronic devices, including VCESELs and resonant cavity devices[23]. Figure 2-4 is a sketch of a DBR. For a semiconductor DBR, because of the small difference in refractive indices (3.5-3.0) of typical epitaxial semiconductors such as GaAs/AlAs, many pairs of the constituent materials must be grown to achieve high reflectivity (e.g. 34 pairs for achieving 99% reflectivity). In addition, the bandwidth for the high reflectivity spectrum is only about 13 90 nm. However, growth of many pairs of the constituent materials needs a long period of time. This often results in thickness imperfections in DBR layers which will reduce the reflectivity and spectral bandwidth. Now, the problem can be overcome by using selectively oxidizing high aluminum content AlGaAs. The oxidized AlAs and GaAs can form larger index contrast DBR (1.5/3.5)1221 comparing with AlAs/GaAs pairs (3.0/3.5) which will greatly reduce growth time and improve performance. Figure 2-4. Schematic drawing of a distributed Bragg reflector, where kR, and kx are wave vectors for incidence, reflection, and transmission light respectively. 14 A distributed Bragg reflector consists of alternating quarter-wave layers of two different materials. The reflection coefficient r of such a Bragg reflector is given by [ 2 4 ] r= Kal + r " (2-1) 1 + ra i r w where is the reflection coefficient of the N pairs of alternating quarter-wave layers. The evaluation of r N is based on the matrix optics methods'241. rai is the Fresnel reflection coefficient between the air and medium 1, expressed as [24] k -k = ( 2 _ 2 ) k +k K<vc + Klx with kax and k/x being the normal components of the wave-vector in the air and medium 1, respectively. Figure 2-5 shows the calculated reflectance of the above Bragg reflector with 4 pairs of quarter-wave layers. The peak reflectance of such a structure can be adjusted to a desired spectral regime by properly tailoring the layer thickness. Let tj, t2 and ni, n2 be the thicknesses and indices of refraction of the layers, respectively. They can be designed to match the desired wavelength Ao as follows: _ 1. (2-3) 15 16 The peak reflectance R can be obtained from 2 R = (2-4) V J where ns represents the index of refraction of the substrate, na is that of the air, and N is the number of layer pairs. From equation (2-4), one can demonstrate how the lateral oxidation enhances the reflectance. For instance, a 4-pair AlAs-oxide/GaAs DBR can achieve a reflectance of 99.8%, while the reflectance is only 68.6% for an unoxidized one. If one needs a reflectance of 99.8%, 34 pairs of alternating layers are required for an unoxidized AlAs/GaAs DBR. Another important parameter is the width of the high reflectance spectrum. The full width at half maximum of the high reflectance spectrum can be approximated as For the ten layers oxidized DBR the spectral width is 1012 nm, while for a similar unoxidized structure the spectral width is 166 nm. Equation (2-5) indicates that the bandwidth is dependent only on the index contrast. 2.3.2 High Index Contrast Distributed Feedback Structure AX = — A sin n \n2 — n, I (2-5) n2 +n. Optical communication systems with high transmission rates require semiconductor lasers that emit light predominantly in a single longitudinal mode and that can be 17 modulated at high speed. For semiconductor lasers of traditional Fabry-Perot type, the feedback is provided by facet reflections whose magnitude remains the same for all longitudinal modes. The discrimination of different longitudinal modes in such a laser can only be made by the gain spectrum. However, the gain spectrum is usually much wider than the longitudinal-mode spacing, resulting in poor mode discrimination. The most successful way to improve the mode selectivity is to make the feedback frequency-dependent so that the cavity loss is different for different longitudinal modes. As the name implies, the feedback necessary for lasing action in a DFB laser is not localized on the cavity facets but distributed throughout the cavity length. This is achieved through the use of a grating which is so etched that the thickness of one layer varies periodically along the cavity length. The resulting periodic change in the refractive index provides feedback by means of the backward Bragg scattering, which couples the forward- and backward-propagating waves. Mode selectivity of the DFB thus results from the Bragg condition. The Bragg condition tells us that coherent coupling between counterpropagating waves occurs only for those whose wavelengths satisfy the following relation'251: mk _ _ A = (2-6) 2neff where A is the grating period, X the wavelength of the emitted light from the laser medium, neff is the effective refractive index of the structure, and m is the order of the Bragg diffraction induced by the grating. By choosing A properly, a device can be made to provide the distributed feedback only at a selected wavelength. Figure 2-6 is a sketch of a DFB laser. 18 Y - Z — A c t i v e Figure 2-6. Schematic illustration of a distributed-feedback semiconductor laser. Different refractive indices on opposite sides of the grating result in a periodic index perturbation that is responsible for the distributed feedback. Today, the DFB laser is an indispensable component in the modern optical fiber communication system. The following analysis shows how the introduction of the lateral oxidation might be used to improve the performance of DFB structure and simplify their fabrication. 2.3.2.1 Coupling Coefficient The DFB's performance has been studied extensively. The details can be found in, for example, Ref. [26] to Ref.[31]. Here only the most important parameter, the coupling coefficient K , in the DFB structure will be discussed We chose the planar waveguide to simplify the analysis. The coupling coefficient represents the strength of the distributed feedback. It determines the longitudinal-mode of a DFB laser. In the transverse electric (TE) mode, the coupling coefficient is given by25 19 2n*ff [j2(y)dy In the above §(y) is the electrical field distribution in the guided mode propagating in the laser, which can be obtained by solving the time-independent wave equation: V2<$> + E(y,z)kfo = 0 (2-8) where k0 - oVc and CO is the mode frequency. Since the DFB structure has a periodic change in refractive index, the dielectric function &(y,z) is also a periodic function. Assuming that the grating is written along the z-axis, then the dielectric function can be written as e(v,z) = e(y) + Ae(y,z) (2-9) where e(y) is the average value of £(y,z) over grating region which is only y dependent in the planar waveguide, and the dielectric perturbation Ae is nonzero only in the grating region. Because Ae is periodic in the z direction with a period A, it can be expanded into a Fourier series as A£(y,z) = X A U v ) e ' T " . (2-10) In the above equation the Fourier coefficient Aemfor the mth-order Bragg diffraction can be calculated from A e ™ (?) = T J " A £ ( y , z ) e x p ( - ^ ^ ) J z (2-11) A J o A and substituted into equation (2-7) to obtain K. The value of K depends on several grating parameters such as shape, depth, duty cycle, and period of the corrugation in.relation to 20 the wavelength. Furthermore, k is also affected by the composition of the cladding layers. The evaluation of K for GaAs lasers has been discussed in many papers.[32 3 6 ] Here, a rectangular-shaped corrugation profile Ae(y,z) is used to obtain an analytical expression for Aem(y). Consider the geometry, shown in figure 2-7 for a typical DFB semiconductor laser. Let the width of the grooves be Aj. Integration of the right hand side of equation (2-11) yields , sin( y A ) Aem(y) = (n\-n\\ ^ - (2-12) mn inside the grating region,, and Aem(y) = 0 for the region without gratings. Substituting equation (2-12) into equation (2-7), one obtains an expression for the coupling coefficient K: n* - n2 s i n ( m 7 C 7 V ) K = * o r g ^ - ^ — ^ (2-13) lneff mil where T g is the confinement coefficient in the grating region, defined as I* . §\y)dy T g = h ^ (2.14) From equation (2-13), several features are noteworthy. First, the coupling is sensitive to the grating aspect ratio. For example, for the first-order grating (m=l), when Aj=A/2, K reaches the maximum for a symmetric first order grating, but K=0 for a symmetric second-order grating (m=2). It can also be seen from equation (2-13) that the first order 21 grating gives the maximum coupling. For the 980 nm system ( GaAs/AlGaAs/InGaAs), the period for the first order grating is about 140 nm (see equation (2-6)). This means that the grating groove width should be 70 nm to achieve the maximum coupling. However, fabrication of such small features is extremely difficult. Instead, the second order grating is selected to make the fabrication easier. However, to make K acceptable, deep gratings are needed to include as much energy as possible in the grating region. All of the above difficulties imply that the fabrication of a high quality DFB laser is not an easy task and make a DFB laser expensive. A possible strategy to improve coupling is to increase the refractive index contrast in the grating region. In equation (2-13), the coupling coefficient K linearly depends on the index contrast An=nrn2. Usually, in the AlGaAs/GaAs/InGaAs system different aluminum composition AlGaAs is used to form the grating, which makes the index contrast fairly small (<0.44). Attempts were also made to improve the contrast by using AlGaAs and air, i.e. by directly writing the grating on the surface without re-growth. But both experiments and theoretical analysis have shown that no improvement could be made. The main reason is that the low refractive index of air (n=l) "pushes" the guiding mode all the way down below the grating region and leads to a small confinement factor Tg in the grating region. However, the introduction of lateral oxidation into the DFB structure might improve the coupling effect and simplify the fabrication. The theoretical analysis of the DFB structure associated the lateral oxidation is shown in the next sectioa 22 2.3.2.2 High Index Contrast Grating by Lateral Oxidation Since the refractive index of wet oxidized AlGaAs is about 1.5,t22i it is possible to fabricate very high index contrast grating. Figure 2-7 is a sketch of this kind of grating. The fabrication begins with writing a grating on the surface of an epitaxial sample, where the grating is written through AlAs layers. Then, a 70% AlGaAs cladding layer is grown on top. A ridge is subsequently defined on the grating region by wet chemical etching to expose the edge of the AlAs grating layer for oxidation. A * o- > GaAs A l A s - >] A | « c : - Alo.7Gao.3As • ^ > r ^ i r ^ r ^ ^ A k ^ ^ o . 3 A s Alo.7Gao.3As G a A s Substrate Figure 2-7. Schematic illustration of the DFB grating used for lateral oxidization. In Figure 2-7, dg is the thickness of AlAs layer. It is noted that before oxidation Ae = n\ - nf=0.9 at wavelength A.=980 nm, and that after oxidation the difference in the dielectric constants increases to 8.44. If the active layer is 100 nm thick, then the distance 23 from the active layer to the grating region is set to 100 nm and dg = 50 nm. The dimensions of this waveguide are chosen so that the DFB waveguide supports only the fundamental transverse mode. To calculate K for different emission wavelengths, one has to solve the wave equation (2-8) and perform an integration of equation (2-14) to obtain the confinement factor in the grating region. For the oxidized grating, the theoretical values of the second order coupling coefficient K 2 for the TE mode is plotted in Figure 2-8(a) as a function of emission wavelength varying from 700 nm to 1.20 um using dg = 50 hm. To get the maximum coupling for the second order grating Aj=3A/4 is used. Figure 2-8(b) is the ratio of the coupling coefficient K of an oxidized waveguide to that of an unoxidized waveguide. 400 e 350 c o o so "a. O U 300 -250 TE mode 0.7 0.8 0.9 1.0 1.1 Wavelength ( pm) 1.2 1.3 (a) 24 TE Figure 2-8. (a) Calculated second order coupling coefficient K for the TE mode as a function of emission wavelength for a square grating shown in figure 2-7 and (b) the ratio of K for an oxidized grating to that of an unoxidized grating. The K for the TE mode with a 50 nm deep oxidized grating is over 250 cm"1 which is about 5 times large than the unoxidized one. In Figure 2-9, the K values versus the thickness of oxidized AlAs for three orders (i.e., m= 1,2,3) are calculated. For the purpose of comparison, the K values for the first three orders of unoxidized grating are also plotted in the figure. Here, Ai=A/2 is used as the duty factor for m=l and m=3, and Ai=3A/4 for m=2 to get maximum coupling. What is 25 impressive in Figure 2-9 is that the third order Bragg grating using the lateral oxidation can achieve up to about 100 cm"1 coupling coefficient which is larger than the first order coupling for the unoxidized one. This suggests the fabrication of a DFB can be greatly simplified. For example, in the 980 nm AlGaAs/GaAs/InGaAs systems, the pitch for the third order grating is about 420 nm which can be easily realized by the holographic technique and wet etching. Figure 2-9. Calculated coupling coefficient K of TE mode as a function of grating depth for oxidized and unoxidized gratings. The order of Bragg diffraction is denoted by m. Here the pitch of the gratings is designed to meet a 980 nm emission wavelength. 26 2.4 Summary In this chapter, four examples were used to demonstrate how the performance of semiconductor devices can be improved using the lateral oxidation technique and how to realize novel device structures in the AlGaAs/GaAs system. The examples range from active to passive devices and from transistors to lasers. It is clear that the wet oxidization technique has a wide application in GaAs engineering. Moreover, it has been shown that almost all Al-containing III-V's are able to use this technique and might be applied to HI-V device integration.[37] As the author believes, similar to the influence of oxidation technique on the Si technology, the lateral oxidation is poised to make an major impact on n i - V electronics and optoelectronics. However, a good understanding of this oxidation technique and its oxides is important in the development of devices. For instance, appreciation of the difference between the room temperature oxidation and high temperature oxidation helps find and develop the optimal oxidation environment; knowledge of the structure of the buried oxide is critical to understand the change in the optical and electrical properties of devices; the thickness change of oxidized AlGaAs epitaxial layers changes the center wavelength and the bandwidth of the DBR reflectors; information on structure changes due to oxidation is also useful in studying induced strain as well as the interface quality between the oxides and semiconductors. All of these will be addressed in later chapters. 27 Chapter 3 Lateral Oxidation of AlGaAs This chapter describes the procedures used to conduct the lateral oxidation and discusses measurements of the properties of the oxide. In this work an atomic force microscope was used to study the structure changes due to the oxidation process. X-ray photoelectron spectroscopy was used to investigate changes in the composition associated with oxidation and also to illustrate the difference between high temperature oxidation and room temperature oxidation. 3.1 Lateral Oxidation Process In this work the samples were oxidized in a Lindberg tube furnace in a water-vapor ambient. The oxidation temperature was 425°C. The whole setup is shown in Figure 3-1. A quartz tube is placed into the furnace to hold the sample. A N 2 carrier gas was used to transport the water vapour into the quartz tube during the oxidation The flow rate of the gas was 1.8cm3/sl38] as measured by a flowmeter, and manually controlled by a pressure regulator. The nitrogen is bubbled through a water fdled bottle heated to 94°C. The temperature is measured by a thermal couple temperature sensor and controlled manually with a variable voltage transformer. To avoid condensation of the water vapor before it reaches the sample, the part of the quartz tube outside the furnace is heated by heating tape to ensure that the temperature 28 is over 100°C. Since the carrier gas may contain trace amounts of AsH 3 after oxidation, the residual gas is then collected into a gas sink for safety. Heater Figure 3-1. Schematic of the setup for oxidation. i, To make this oxidation recipe reproducible, the following operating procedures were used. First, turn on the furnace and raise temperature to 425°C. Then, seal the sample into the quartz tube and initiate the flow of N 2 . Adjust the power to the heater so that the water temperature rises to 94°C. When the temperature of the furnace and the heater are stabilized at the desired value, place the quartz tube into the furnace and begin to count time. Upon reaching the selected oxidation time take the quartz tube out of the furnace and turn off the water heater and the furnace. Open the N 2 valve to get a high flow rate for quick cooling of the oxidized sample. 3.2 Calibration of the Oxidation Rate An oxidized sample can usually be checked by an optical microscope or electron microscope directly from its top provided the cap layer is not too thick. Since the oxidized AlGaAs has a 29 low refractive index of 1.5,[22J it can be distinguished visually from unoxidized material. Under an optical microscope, oxidized AlGaAs has a different color than unoxidized AlGaAs. Figure 3-2 is the top view of the partially laterally oxidized sample taken with a Nomarski optical microscope. The sample used here is ASU721, whose layer structure is shown in Figure 2-1(a). Cleaved edge Oxidized part Boundary Unoxidized part Figure 3-2. The top view of the partially laterally oxidized ASU721 in a Nomarski optical microscope. In Figure 3-2 the boundary between the oxidized area at the edge and the unoxidized area in the center is clearly shown because of their reflectance difference. The oxidized edge can be seen under an electron microscope. Due to the conductivity contrast between the insulating oxide and semiconducting AlGaAs, the oxidized part appears darker than the unoxidized 30 part. By measuring the width of the oxidized region, one can calibrate the oxidation rate as a function of the oxidation conditions. Figure 3-3 shows the width of the oxidized region as a function of time for oxidation. 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 20 40 60 . 80 100 120 140 160 Oxidation Time (min) Figure 3-3. The calibration of the oxidation rate for Alo.9sGao.02As. The sample was oxidized at 425°C with water vapor from a water bottle held at 94°C. Figure 3-3 shows that the lateral oxidation rate is constant with time. A linear fit gives an oxidation rate of 37.6 |im per hour. However, the rate greatly depends on the aluminum composition. K. D. Choquette[20] and J. -H. Kim. [ 3 9 1 reported that the oxidation rate for Alo.88Gao.12As is over 100 times slower than that for AlAs. The rate is also structure 31 dependent. For example, the lateral oxidation will stop if the thickness of the oxidizing AlGaAs is less than 8 nm. [ 3 9 ] 3.3 Structure Change of the Oxide Oxidation will obviously change the structure and composition of the buried AlGaAs. This needs special consideration in device applications. For example, epitaxial AlGaAs has the same crystal structure as the GaAs substrate and can be cleanly cleaved along the <110> crystalline directions. But, we found that the oxidized AlGaAs cannot be cleanly cleaved using mechanical methods. In optoelectronics applications, clean cleaved edges are used as mirrors in Fabry-Perot optical cavities. Therefore the mechanical method should not be used to cleave the oxidized AlGaAs. Alternatively, one could cleave the sample before the oxidation is conducted. The oxidation-induced structure change also introduces large strain1401 in the epitaxial layers. Before the oxidation, the lattice mismatch between the grown AlGaAs and the GaAs substrate is smaller than 0.13% which assures small strain and strong binding of the epitaxial layer to the substrate. However, the oxidation changes the zinc-blende structure of AlGaAs completely.[40] This change might reduce the binding between the oxide and the structures sandwiching it. If no measures are taken to release the strain during the pattern design, the laterally oxidized AlGaAs will become fragile or even flake off from substrate. To avoid this 32 problem, short oxidation times (<1 hour) are prefered to obtain a good quality non-flaking sample. Figure 3-4, shows a flaked comer of an oxidized piece of ASU721. Figure 3-4. SEM micrograph showing a bird's eye view of the flaked off corner of a laterally oxidized sample ASU721. 3.4 Vertical Expansion Lateral oxidation changes the sandwiched AlGaAs into an oxide. The investigation of the induced thickness change of the oxidized AlGaAs is important for some applications. For example, if the lateral oxidation is applied in distributed Bragg reflector (DBR) structures, changes in the thickness and thickness uniformity will significantly affect the reflectivity and spectral bandwidth. Obviously, the direct measurement of the oxidation induced thickness 33 change is needed to determine the extent of the layer expansion. Due to lack of knowledge of the vertical expansion, people sometimes assume that a significant change in thickness is due to oxidation and obtain a good fit to the measured reflectivity of a laterally oxidized D B R by reducing the thickness by ~10% t 2 2 1. To check this result, atomic force microscopy (AFM) scanning of the surface of a lateral oxidized structure was conducted. Figure 3-5. A F M micrograph showing a bird's eye view of the partially oxidized mesa on ASU721. The A F M was operated in tapping mode with a 0.5936Hz scanning rate. 34 The structure used is a 55 um x 55 um mesa fabricated on ASU721, shown in Figure 2-1. Detailed information on the sample and its structure can be found in Chapter 2. Since the buried Alo.9sGao.02As layer that was 2 um thick was oxidized, the oxidized edge should be lower than the central unoxidized AlGaAs by about 200 nm according to the above fitting assumptions. However, our AFM scanning result, given in Figure 3-5, does not indicate a significant step between the oxidized area and unoxidized area. The maximum step is less than 5 nm which is 0.25% of the original thickness. An understanding of this small amount of vertical expansion is not yet clear, it might relate to the induced composition change and structure change of buried oxide. We therefore conducted xray photoelectron spectroscopy (XPS) to identify the composition of buried oxide. 3.5 X P S Composition Analysis It is mentioned in Chapter 2 that high quality oxide can be produced only when the oxidation of AlGaAs is conducted at high temperature; if it is done at room temperature, the oxide is mechanically unstable and becomes powder. To find possible reasons for this difference, x-ray photoelectron spectroscopy (XPS) analysis was performed on the high temperature oxidized ASU721. From an analysis of the core level binding energy spectrum, the elements in the oxide and their percentages can be identified. Furthermore, from the characteristic chemical shifts of the core levels information about chemical bonding can also be obtained. 35 3.5.1 XPS Measurements All XPS spectra were measured by the MAX200 spectrometer1411 at an operating pressure of about 4.5xl0~9 torr, and the emitted photoelectrons were obtained from a lxl mm2 area on the sample. The polychromatic MgKa excitation source was operated in 15 eV and 20 mA. Wide scan spectra were obtained with the pass energy set at 192 eV. The percentage of the elements was obtained by integrating the corresponding peak areas after background subtraction and sensitivity factor correction. In this work, A12p Ga3d and Ols emission electrons were used to determine the percentage of Al, Ga, O, respectively. The Cls electron at 285 eV was the reference used to calibrate the other core level peaks. The trace amount of carbon comes from contamination during air exposure. A detailed description of the machine and operation procedures is contained in Y-L Leung's MSc thesis.'421 3.5.2 Composition Analysis of Oxidized AlGaAs The sample used in this experiment is oxidized ASU721. The oxidation condition is the same as that described earlier. Specifically, a sample with many defects was selected to accelerate the oxidation process. Oxidation was conducted for a long time (>6 hrs). During the oxidation process, the sample was repeatedly checked with an optical microscope until the whole sample was oxidized. Since the ASU721 sample has three epitaxial layers (see Figure 2-1), in order to study the oxidation of Alo.ggGao 02AS, Ar + was used to sputter off the top layers. The spectrum was recorded at different sputtering depths labeled from A to I, shown in Table 3-1. Ar + Sputtering Time(mins) Chemical Type Code in Text As Received GaAs A 20 GaAs B 40 Al appears, but Ga rich C 60 Al rich D 70 Al rich E 90 Al oxide F 120 Al oxide G Table 3.1. Sputtering time of XPS measurements 80-60-SP 40 CO <u o l -°- 20' 0 -5K— Ga Percentage -H— Al Percentage 0 I — 20 40 60 ~80~ 100 120 Sputtering Time (min) Figure 3-6. Al and Ga percentage of the oxidized ASU721 versus sputtering time. 37 In this sample, the sputtering rates for GaAs and oxidized AlGaAs are unknown. However, the sputtering depths can be estimated by tracing the relative percentage of Ga and Al. The Al and Ga contents are quite different in the three epitaxial layers (see Figure 2-1). Figure 3-6 shows the percentage of Al and Ga at different sputtering times. It is clear from the figure that after 70 minutes of Ar + bombardment, the Al dominates the spectrum. It is therefore believed that spectra F and G are the spectra for the oxidized Alo.9sGao.02As. 2500 r 2000 -1500 G 3 5 1000 500 0 Ols \ A12s A12p Ga3d 02s 1400 1200 1000 800 600 400 200 0 Binding Energy (eV) Figure 3-7. A wide scan XPS spectrum of the oxidized ASU721 sample after spputtering for 90 minutes with Ar +. The spectrum was taken by averaging 10 individual scans. The interval between sampling points is 0.8 eV and the sampling time was 80 ms. The oxygen Auger signal in the graph is indicated by (a). 38 Figure 3-7 presents a wide scan spectrum (E) of oxidized high aluminum content AlGaAs. It provides information on the chemical composition of the high temperature oxide. In figure 3-7, the photoemission peaks of O, Al, and Ga are apparent, but there is no detectable As signal. This indicates that the high temperature oxidation of Al$ 98Gao.o2As has resulted in an Al oxide dominated product. This result is unlikely to be obtained from the room temperature oxidation of AlGaAs or GaAs, which we would expect to produce a AsO x rich compound. . This may explain why the high temperature oxidation produces a stable and dense film, while the room temperature oxidation generates powder. At room temperature, the oxidized products of AlGaAs contain As rich compound oxides (A10X; AsOy...). AsO y is mechanically and thermal not stable, which will result in a powder product. If the oxidation is conducted at high temperature and in water vapor ambient, several products are also obtained. A possible reaction is AlGaAs + H20 -» AlOx + GaOy + AsOz {evaporated) + AsH3 {gas) At the temperature of the high temperature oxidation (425°C), however, the As reaction products will be volatile while the A10x and GaOy will not be. Therefore, for high Al content AlGaAs, the high temperature oxidation gives only Al dominant oxide A10x and produces a mechanically stable and useful film. For the purpose of comparison, Figure 3-8 gives the wide scan spectrum of GaAs after long time air exposure, which generates a very thin oxide film on the GaAs surface. The spectrum clearly shows the existence of As and the photoemission peaks of O, Ga and C. 39 Binding Energy (eV) Figure 3-8. A wide scan XPS spectrum of a GaAs substrate after a long time air exposure. The spectrum was taken by averaging 10 individual scans. The interval between two sampling points is 0.8 eV and the sampling time was 80 ms. The Auger signal in the graph is indicated by (a). 3.5.3 Oxidation States of the High Temperature Oxidized AlGaAs 40 Another interesting point to look into is the binding energy of the aluminum, which can help identify the chemical state. From the binding energy one can tell if the aluminum is oxidized, and moreover, what kinds of oxide it forms. Two kinds of oxidation states have been reported for the high temperature oxidation. One is amphorous AI2O3 and the other is crystalline A1203 [ 4 3 1 There are four major phases of A1203: a- P-, 8- and y- phase.'44' If the 2p characteristic binding energy of aluminum oxide is analyzed in the XPS spectrum, there is about 0.9eV energy difference between amphorous Al 2 0 3 (75.4eV) and crystalline Al203(74.3eV),'44'451 but less than 0.2eV difference among four phases of crystalline A1203. The XPS spectrum can resolve a 0.9eV chemical shift, but not the small shifts (0.2eV) for the different crystalline phases. Thus, XPS can help in answering if the buried oxide of AlGaAs is crystalline or amorphous. In the data the spectra E to G show the aluminum signal, and the percentage of aluminum in spectrum E and F is much less than Ga. It is therefore believed that spectrum E and F come from the 40 nm Alo.3Gao.7As layer, and the rest come from aluminum dominated Alo.98Gao.02As. The binding energy scale of the 1200 eV XPS survey spectrum was then expanded to 20.5 eV centred around the 72.25 eV to resolve Al 2p photoemission peak. The peak position of the spectra C to G is given in Table 3-2. Figure 3-8 plots the Al 2p peak for spectra C and G for purposes of comparison. From Table 3-2, the Al 2p binding energy for 40 nm Alo.3Gao.7As (C, D) is 74.6 eV, which is close to the binding energy of A1203 (74.32 41 eV) rather than the AlGaAs (-73.6 eV). This suggests that the Alo.3Gao.7As is also oxidized. Since the oxidation rate for low aluminum content AlGaAs is extremely slow, it is believed that the oxidation was conducted vertically rather than laterally. When the Al 0 9 8Gao 0 2As layer under the Alo.3Gao.7As is oxidized, the bottom side of 40nm Alo.3Gao.7As will be open to oxidation, although the oxidation rate is very low for Alo.3Gao.7As. Long time oxidation (>4hrs) makes it possible to oxidize this Alo.3Gao.7As layer to a thickness of 40nm. Spectrum C D E F G Al 2p Energy (eV) 74.6 74.6 75.4 75.4 75.4 Table 3-2. Binding energy for aluminum 2p core levels Table 3-2 also shows that the binding energy of the aluminum rich layer and that of the aluminum poor layer are different. The spectra E, F, G give 75.4eV binding energy for the Al 2p peak, which corresponds to the amphorous A1203. That suggests that either crystalline or amphorous A1203 will be the high temperature oxidation products for AlGaAs, depending on the aluminum content and how the sample is processed. Our measurements reveal that low aluminum content AlGaAs prefers the crystalline state. This may be due to the presence of Ga, which might resist the thermal disorder of A1203. It is also possible that the smaller chemical shift for the lower Al content AlGaAs is due to the incomplete oxidation of low aluminum content AlGaAs. In this situation, the presence of the Ga may also inhibit the thermal disorder. 42 o U A SpecuxmCfor Oxidized Al 0 3GaQ ?As • SpectajmFfbr Oxidized Al 0 ggGaQ 02As 1 B _ . 0.000 75 i 80 70 Binding Energy (eV) Figure 3-9. Photoemission spectra showing aluminum 2p core levels for two different sputtering depths C and F. The circles and triangles in the graph are measured points. The solid lines in the graph are mixed Guassian/Lorentzian fits of the data. 43 In the Figure 3-8, the XPS data was fitted by the mixed Gaussian/Lorentzian functions1451 as: /(£) = =—--Exp{(l-M)[ln2-(E-E0)2 /$2]} (3.1) [l + M ( £ - £ 0 ) 2 / p 2 ] Where f(E): is the fitting function E : the corresponding binding energy Eo : the peak centre (3 : a parameter from which the actual FWHM is calculated M : the mixing ratio (1 for a pure Lorentzian peak; 0 for a pure Gaussian peak) H : the peak height 44 Chapter 4. Waveguide Fabrication The major aim of this thesis is to develop a technique for producing high contrast distributed feedback (DFB) ridge waveguides of high quality. This includes grating fabrication, the definition of the ridge, as well as the lateral oxidation. This chapter discusses the general processing techniques except the lateral oxidation, which has been discussed in the previous chapter. In this chapter the emphasis is on the e-beam lithography . 4.1 Fabrication Procedures of waveguides Electron beam (e-beam) lithography and wet chemical etching have been used to fabricate the waveguides. The fabrication procedures are described below. The basic process steps are as follows. . Grating fabrication: sample cleaning; resist spin-on and bake; patterning grating and the alignment mark by e-beam writing; resist development; pattern transfer by wet etching. Ridge definition: sample cleaning; resist spin-on and pre-bake; 45 grating pattern alignment; > ridge defination by e-beam writing; sample post-bake and development; wet etching to form the ridge waveguide. Tfiinning: sample protection; sample thin and cleave. Oxidation: lateral oxidation to form low index contrast layer, m this work, the positive photoresist consisting of 950K molecular weigh polymethyl-methacrylate (PMMA) was used to make ultrafine grating lines (-lOOnm). The negative photoresist PN114 acted as a wet etch mask for defining the ridge. To protect the sample from contamination introduced during the thinning, a PN114 film was also used to protect the patterned surface. 4.2 Grating Fabrication In this work, one dimensional DFB gratings and two dimensional arrays of dots were created by e-beam lithography. A Hitachi S-4100 field emission scanning electron microscope (SEM) driven by the Nanometer Pattern Generation System version 7.5 was used to accomplish this task. A CAD graphic generation software DesignCAD was used to design the lithography patterns. A detailed description of the machine and the software can be found in the corresponding manualsf46Jt8] and Alex Busch's thesis/491 Figure 4-1 shows the fabrication procedures. Developed photoresist Sample Removed etching mask Figure 4-1. Procedure for grating fabrication. 4.2.1 Sample Preparation Cleaning the sample initially and keeping it clean during the process are particularly important for nanofabrication. Because the photoresist thickness is only about 150 nm and a typical feature size in the DFB gratings is around 100 nm, any contamination particles, whose size is usually on the order of micrometers, will damage the grating patterns. All the processes, therefore, are 48 highly recommended to be conducted in a dust-free environment (e.g. laminar flow hood). To get rid of as much contamination as possible, the sample was first immersed in acetone for about 10 minutes, then rinsed by acetone and methanol. The solvents were blown off the sample by clean nitrogen to avoid evaporation residues. After cleaning, a drop of 950K molecular weight PMMA was placed on the sample using a capillary pipette. The sample was then spun at 9000 rpm for one minute. At this speed, the PMMA film on the sample was 150 nm thick.1491 The uniformity of the film can be judged by its color. However, the uniformity is sensitive to the size and shape of the sample. It is very difficult to get an even and uniform film for a small sample. But it has been found that the uniformity of gratings in width is not very sensitive to the thickness uniformity of the photoresist. If the spin-speed is reduced to 7500 rpm, which produces films about 200 nm thick, 100 nm wide grating lines are also obtainable. After being spun, the sample was baked at 170°C for about 2 hours. Then the sample is ready to be patterned in the SEM. In principle, the sample cleaning procedure, the sample shape, the rotation speed, the pre-bake time as well as the pre-bake temperature will all influence the pattern quality. However, it has been found from this research that the pattern quality is quite robust and insensitive to these pre-treatment parameters. Table 2-1 lists the recommended development operation ranges for the above parameters. Basically, high uniformity ultra-fine grating lines (-100 nm) can be guaranteed if the operation conditions are in the recommended range. The most important factor which governs the quality of the patterns is the e-beam writing. 49 One of the major endeavors in this work have been made to optimize this process, which will be discussed in detail in section 4.2.4. Process Parameters Recommendation Condition Sample Size 3mmx5mm ~ quarter of 2 inch wafer Spinning Speed 7500rpm ~ 9500rpm Bake Time SOminutes ~ 3 days Bake Temperature 165°C~ 185°C Developer MIBKTPA 1:3 by volume Developing Temperature -25 °C Developing Time 75 seconds ~ 95 seconds Rinser IPA Rinse Time ~25 seconds Photoresist Stripper Acetone Stripping Time ~ 3 minutes Table 4-1. Recommended sample preparation and development conditions for high uniformity 100 nm linewidth grating 4.2.2 Development After patterning, the sample was developed in a mixture of methyl-isoburyl-ketone (MEBK) and isopropyl alcohol (IPA) 1:3 by volume for about 85 seconds, and then in pure IPA for 20 seconds. The sample was then rinsed in deionised (DI) water. The remaining liquid was blown off by dry nitrogen. The development time is not critical. Usually, the patterned sample was checked by the Nomarski optical microscope with a resolution of about 0.5um. With experience, any feature larger than lum can be easily identified It is also possible to check grating pattern features on a 0.1 urn scale if the grating pattern occupies a 50 reasonably large area (e.g. 10x15 um). In the case that the patterns are not well developed and are invisible, longer development time is needed. In most situations, invisible patterns after development mean something went wrong during the e-beam writing. Sometimes, the SEM is also used to check the developed pattern. It is not difficult to find and focus a pattern in which the e-beam pattern has been developed completely because of the large conductivity contrast between the photoresist and the semiconductor. However, the SEM inspection is not recommended in the real production of devices. Exposing patterns to an electron beam will leave carbon contamination on the sample surface, which will resist the etching. This is one of the reasons why each step needs careful testing to accomplish high-yield fabrication. 4.23 Etching To transfer the grating patterning from the photoresist to the sample, a mixture of dilute H 2S0 4 and H 2 0 2 was chosen because this etchant does not attack the PMMA mask layer. To obtain a smooth etching edge, a very dilute solution, H 2 S0 4 : H 2 0 2 : H 2 0 (1:1:72 by volume) was used. Since the grating features are about 100 nm and the spacing is about 300 nm, a precise calibration is needed to prevent the etch from undercutting and the undercut washing out the whole pattern. Also, the etchant preparation greatly affects the etch rate and morphology of the etched surface due to the heat of disasociation of the peroxide.'501 Accordingly, the etchant is prepared by the following procedure: 1ml 38% hydrogen peroxide is added to 32 ml of DI water; lml pure H 2S0 4 is added to 40 ml of DI water. After the second solution cools down to room temperature, the etchant is made by mixing these two solutions together. To ensure a constant 51 etch rate, the temperature should be maintained at 25°C during etching. The volume of etchant should be large enough for the desired etching time and the sample size. Figure 4-2 is the calibration result of this recipe on the GaAs substrate. From Figure 4-2, one can see the etching rate is almost constant. The best fit gives an etch rate of 90.3 nm/min. One should keep in mind that the etch rate is also affected by the composition of the sample. For example, the etching rate for AlxGai_xAs is different from GaAs151"53], and the difference depends on the aluminum content x. For this reason, the above calibration is useful only to etch GaAs and low aluminum content AlGaAs. After etching for a specific period of time, the sample was rinsed using large volumes of DI water to neutralize the acid and then blown dry using N 2 . Time (second) Figure 4-2. Calibration of the etch rate of ^SO^rfeCfcrfcO (1:1:72 by volume) on a GaAs substrate. The dots are data measured from experiments. The solid line is a best fit which gives an etch rate of 90.3 nm/min. 52 After etching, the sample was immersed in acetone for about 10 minutes at room temperature to dissolve the unexposed PMMA. 4.2.4 Electron Beam Writing 4.2.4.1 Writing System Components Electron beam (or e-beam) lithography systems expose patterns by direct writing on photoresist. Compared with other lithography methods, it has great flexibility in pattern design because a mask is not required. This advantage together with its superior resolution (the diameter of the e-beam can be focused to 1 nm.) make e-beam lithography attractive for nanofabrication. The e-beam lithography system in this lab consists of a Hitachi S4100 field emission scanning electron microscope (SEM), Nanometer Pattern Generation System (NPGS) which drives the electron beam, and a computer-aided-design package DesignCAD for pattern design. The resolution of the e-beam driving system is 16 bits, which allows one to expose at most 65,536x65,536 points at in a given area. The allowable exposed area, called the field-of-view, is determined by the magnification in the SEM. The SEM allows a field-of-view of up to 3 mm by 3 mm. However, a 100 nm linewidth grating can be fabricated successfully only when the field-of-view is smaller than 290 pmx290 urn, corresponding to a magnification of 300X. This area is more than enough to fulfill most of our application requirements. 53 4.2.4.2 Pattern Resolution The experiments conducted by the author have shown that a mininum linewidth of about 50 nm can be obtained for an isolated line or a 70nm linewidth in a grating with a spacing of 300nm. For these small lines, the result is not highly repeatable and there are several problems and restrictions on the process parameters. For example, a magnification of larger than 1000X has to be set, which limits the length of the grating to 90 urn; because the samples are small, the photoresist is not uniform, and the uniformity of the gratings is not very good. However, if we look for gratings with linewidths of 100 nm, most of these problems are not serious. Highly repeatable, uniform, large-scale grating can be achieved. Not only were ID and 2D gratings fabricated using this method, but circular gratings were also successfully made. The grating spacing can vary from 250 nm to isolated lines. High uniformity can be achieved for any kind of pattern as long as the mininum feature size is larger than 100 nm. Patterns with smaller features can also be fabricated using the same methods if thinner photoresist is used. Furthermore the author has achieved a 1:1.5 width-to-height aspect ratio. With 75 nm thick photoresist, one can make patterns with 50 nm features and good repeatabality. This fulfills the requirements for many optoelectronic devices. 4.2.4.3 Writing Current A slow writing speed is the major disadvantage of e-beam lithography. Writing speed basically depends on the beam current as well as the threshold dose of the photoresist. In the 54 SEM system used, the maximum emission current is 20 mA. This is further reduced by two level current apertures, resulting in an actual writing current of no more than 100 pA. For most situations, only the two smallest pinholes in the second aperture were used to avoid beam aberrations, which further reduced the writing current to less than 7 pA. The writing current hitting the sample surface was collected by a Faraday cup linked via a BNC connector to a Keithley Model 617 Electrometer. To achieve a more stable current during the writing, the SEM tip was flashed for 2 hours before conducting lithography. However, the writing current continues to drop during writing, and therefore very long time writing is not recommended. The current was normally measured just before writing each pattern to reduce this unavoidable systemmatic error. Other techniques to improve reproducibility will be described in detail later. 4.2.4.4 Accelerating voltage The electron beam energy is also a major consideration for writing small features. The accelerating voltage of the S-4100 field emission SEM can be adjusted from 0.5 kev to 30 kev. For 150 nm thick 950K molecular weigh PMMA, theoretical analysis and experiments have shown that the threshold electron energy to go through the PMMA is about 3 kev.'54] The author has successfully fabricated gratings with 500 nm spacing using various electron energies from 5 kev to 30 kev. Usually, two strategies are used to accomplish high resolution lithography. One is to use a very high energy electron beam, making the penetration depth as large as possible to avoid the back scattering effects in the photoresist.'55'561 The other is to 55 use very low energy electrons whose penetration depth is only slight larger than the thickness of the photoresist to make back-scattering as small as possible.'57"621 Our study has shown that a suitably low accelerating voltage gave such a small writing current that the writing time became unrealistically long. Thus, the first strategy was adopted in this work. An acceleration voltage of 20 kev was used instead of 30 kev because the 30 kev writing generated a high noise level. The noise might come from the second aperture which has been left unbaked for a long time. The emission current was preset to maximum value of 20 |lA to increase the writng speed. 4.2.4.5 Pattern Rotation During the writing, another system error to be considered is the difference between the writing direction of the electron beam and the motion of the sample accomplished with the stage. This difference comes from the magnetic field in the objective lens and scanning coils that control the scanning beam. In practice, this angle between the motion of the sample and writing direction might lead to the misalignment between patterns and the <110> crystallographic direction, along which the crystal cleaves. There are two ways to avoid this problem. One is accomplished by adjusting the working distance to minimize the angle-difference. This is to adjust the objective lens current so that the image is just at the focus with that specified working distance. The other way is to fix a working distance and rotate the pattern during the design to match the above mentioned two directions under a preset 56 working distance. In the case that the acceleration voltage is 20 keV, the working distance is 17 mm when the angle difference is minimal. 4.2.4.6 Dose Tests Due to the proximity effect caused by electrons scattering into neighbouring pixels, the pattern resolution depends on the specific pattern shape and layout. To determine an allowable dose range for production of -100 nm linewidth lines and -60 nm diameter 2D arrays a series of dose tests under different writing parameters and grating pitches were conducted. Figure 4-3 summarizes the grating linewidth obtained for a number of different incident doses. In this figure, the threshold dose is 3.1 fC/point for a center-to-center distance of 30 nm, which gives the lowest linewidth of about 70 nm. However, such small linewidths are not very reproducible. The guaranteed turn-out linewidth is about 100 nm with an incident dose >3.6fC/point. The pitch for the grating in Figure 4-3 is 300 nm. We expect better resolution for this larger grating due to the proximity effect being less important. The magnification used to write the patterns in Figure 4-3 is 1000X, however, it has been found that as long as the magnification is greater than 300X, the magnification does not have a significant influence on the pattern resolution. The dose test has also been conducted with 2D dot arrays. We found that the dot diameter is quite insensitive to the incident dose compared with the ID grating. For example, a ~60nm diameter dot turns out reliably with an incident dose bewteen 15fC/point and 25fC/point. 57 Table 3-2 gives the recommended ranges of the writing parameters for generating a linewidth oflOOnm. 3001 250-200-•S 150 -a •J 100. 50 H 0 T " 3 4 5 6 Incident Point Dose(fC/point) -i r 7 Figure 4-3. Feature size versus dose for 300nm spacing grating. The zero linewidth for an incident dose'less than 2.8 fC/point means that the patterns did not go through. If one follows the recipe given in Table 4-1 and Table 4-2, good grating patterns should be guaranteed. It should also be mentioned that proper operation of the SEM can not be overlooked. It includes careful adjustment of the focus to get a clear image, and precise alignment of the e-beam through the column to the sample. Usually, a gold standard is used to accomplish this work. The detailed procedures can be found in the SEM manual.'461 It 58 should also be noted that the sample should be placed as horizontal as possible on the holder to avoid any focus changes during writing. SEM Writing Parameters Recommanded Range Acceleration Voltage '20keV Emission Current 20uA Pinhole size in aperture #2 #3, #4 for ID grating #2, #3, #4 for 2D grating Stabilization Period 2hrs ~ 6hrs after tip flashed Center-to-center Distance 300 A Magnification >X300 Scanning Mode #4 Condenser Lens #15 Working Distance 17mm Dose for writing ID gratings 3.6-4.0 fC / point (spacing >300nm) Dose for writing 2D gratings 15-25 fC/point (Spacing >250nm) Table 4-2. Recommended ranges for the writing parameters to get 100 nm linewidth gratings and a less than 100 nm diameter dots. Figure 4-4 shows examples of a ID grating and a 2D grating written using the above recipe. From the figures, one can see that the linewidth and the diameter of the gratings meet our requirements well, but the uniformity of the gratings could be improved. In the following section, we discuss methods used to improve the quality of the fabrication. I i ? W i l l i I 2 1 0 2 5 2 1 2 0 . 0 k V X 1 5 . 0 K 2 . 0 0 v m 59 (a) (b) Figure 4-4. (a) SEM micrograph of a ID grating, whose pitch was designed to be 300 nm. (b) SEM micrograph of a 2D square lattice with 290 nm pitch. 60 4.2.5 Improvement of Grating Uniformity The uniformity of the grating has a significant influence on DFB laser performance. Inhomogeneity of the gratings leads to different couplings in different grating regions. This can reduce the side mode discrimination of lasers and degrade its linewidth. In our e-beam writing system, the linewidth of the grating has been assured as long as the magnification is set > x300. However, from Figure 4-4(a), one can see that the standard deviation of grating lines with respect to their centers (the mean positions) is about 1/5 of its linewidth. This grating was written at 600x magnification without using any noise reduction techniques. Therefore, an effort has been made in this work to improve the uniformity of the gratings. In our e-beam writing system, the variation can be reduced by writing the grating patterns at higher magnification (>1000). This suggests that there is some systematic variation coming from the electron beam driving system. When the magnification is increased, the systematic error is scaled down through the microscope. However, writing gratings at higher magnification only reduces part of the variation, but puts further limitation on the writing area. Therefore, other methods need to be developed to improve uniformity. 4.2.5.1 Noise Sources and Their Characteristics Theoretically, many noise filtering techniques are based on the law of large numbers in probability theory. Let Xj (i=l,2,..., N) be N independent measurements of a random variable X. Then, the mean value of the outcomes is : 61 The law of large numbers states that the probability that deviates from <x> goes to zero as N->°° Thus lim p(\yN-<x>\>£) = 0 (4-2) when N is large enough, yN behaves like a Gaussian random variable with a variance where the ox is the standard deviation of Xj . This suggests that if a repeating procedure can be designed, variations in the pattern due to noise can be averaged out. The more repetitions, the better the uniformity that can be achieved. To develop this repeating procedure, it is important to know the sources and some of the characteristics of the noise which affects the pattern. In our system, we found that the noise comes from the following sources: (1) Mechanical disturbance from the building and equipment due to poor mechanical isolation. This type of disturbance affects the position of the electron beam during writing; (2) Variations in the writing current. (3) 60 Hz AC pickup from power supply. The first two sources usually generate random noises, while the third one is synchronized with the 60 hz power. In Figure 3-4(a), the distortion in the grating shows some periodicity. If one divides the period by the writing speed, one finds that the noise frequency is about 60Hz. 62 Therefore it is believed that the periodic noise comes from the coupling between the beam position control signal and the 60Hz AC power supporting the microscope or from AC power lines in the building. For random noise, the repeating procedure is easy to realize. For example, if the writing current is reduced, the writing speed will be slowed down. To get enough dose, the exposure time for each point should be longer, which will generate a longer noise sequence at each point. Thus, one will get a better chance to average them out. However, time averaging cannot average periodic noise (i.e. pickup). The efficiency of the time averaging method in removing noise depends on the ratio of the intensity of the random noise to the periodical noise. 4.2.5.2 Noise Reduction To reduce the periodic variation, a new method is introduced in this work namely multi-pass writing. The idea is^  instead of writing a line in a single pass, the line is written repeatedly. However, the total dose for each line is maintained the same as in single pass writing. For example, if 4fC/pt is the point dose to be used in writing a single grating line in one pass, for 12 times pass writing, the point dose should be set to 4/12=0.333 fC/pt. Since the writing time only depends on the integrated exposure, multiple pass writing does not increase the writing time. Mathematically, each pass is represented by Yi(.t) = Yi0+Asm(aa + ^ ) + ^ (t) (4-4) r 63 where Yi0 is the desired writing trace for each pass. The second term on the right side is the electric pickup, and £,-(f) is the random noise, which we will assume has. a gaussian distribution. Obviously, if a grating line is written in a single pass, the periodical noise can not be ruled out. The variation depends only on the amplitude A. However, if a grating line is written in the multi-pass fashion, one will have a good chance to obtain an independent initial phase factor (j), for each pass, i.e., (<)),• •<(>.) = 0. Thus, the effect of the pickup introduced in each pass will be independent, and will be reduced by the multiple pass averaging. Of course, the random noise will also be reduced. Theoretically, the more passes, the better the uniformity that will be obtained. In addition, we have found that the multi-pass writing also increases the linewidth. This might come from drift in the controlling computer's digital-to-analog converter (DAC) board. The control voltage for the electron beam position can drift slightly from the intended value for each pass and cause them not overlap with each other. This misalignment between two passes is dependent on the field-of-view. For example, when writing a pattern at 1000X magnification, the field-of-view is 98.63 u\m, so that a 1.5 bit DAC error for our 64K resolution DAC will give 4.5 nm distance between two successive passes.'This misalignment will be worse for lower maginification writing since a 1.5 bit DAC error represents a larger distance at the lower magnification. Therefore the uniformity must be balanced with resolution. In practice, 10 passes is found to give a reasonable compromise between good uniformity and good pattern resolution. With this procedure, above 280 nm pitch gratings can be fabricated with -120 nm linewidth and superior uniformity. (a) Figure 4-5. (a) 340 nm pitch grating written in 13 passes. The pattern was written at 1000X magnification, (b) A circular grating was also written by multi-pass writing. 65 The improvement in uniformity with multi-pass writing is illustrated in Figure 4-5(a). The grating was written in 13 passes. Comparing this with single pass writing shown in Figure 4-4(a), the noise induced variation has been totally eliminated. Actually, the uniformity for any pattern can be improved in the same manner as long as it is made by writing individual lines. In Figure 4-5(b), a superior uniformity circular grating was fabricated by mult-pass writing. 4.3 Structural Study of the Grating Surface by X-ray High resolution x-ray diffraction is a standard method to probe the structure of crystalline materials. For monochromatic x-ray and particular incident directions, the x-ray reflected from successive crystal planes can generate constructive interference, which are called Bragg peaks. Atomic structure information can be obtained by analyzing these Bragg peaks. In this work, we applied x-ray diffraction analysis to structures on the mesoscopic length scale. Like all the traditional optical characterization techniques, x-ray diffraction analysis is non-destructive and can be done routinely. It can be applied to gratings with periodicities from the nanometer range to the micrometer range with a high accuracy. These performance advantages are not shared by cross-sectional scanning electron microscopy which is destructive and limited in resolution, and by transmission electron microscopy which is also destructive and involves complex sample preparation. It is also possible to determine the width of the grooves and to characterise the quality of the sidewalls[65] if a suitable intense x-ray beam is available. 66 In this section, we present the diffraction features of a two dimension (2D) photonic structure formed on a monocrystalline semiconductor with a periodicity - 500 nm. The sample was prepared by e-beam lithography and wet chemical etching. The details of the sample preparation have been described in earlier parts of this chapter . Figure 4-6 is an atomic force micrograph showing a bird's eye view of such a structure. The direction of this square lattice has been aligned to the {110} crystal directions. Figure 4-6. Atomic force micrograph of a two dimensional surface grating. 67 The x-ray diffraction patterns were collected by using a computer controlled high-resolution double crystal x-ray diffractometer. A 5 kW rotating anode with copper target (lcuKai=0.154 nm) is used to generate x-ray. An asymmetrically channel cut (100) Ge crystal was used as a monochromator and a single crystalline Si was used as a collimator. The horizontal divergence of the x-ray beam reflected by the Ge crystal was calculated to be 11 mrad for the (400) CuKal reflection. The diffraction patterns of the surface are measured in the vicinity of the asymmetrical (224) GaAs reflection. In Figure 4-7(a), we show the geometry of x-ray diffraction at a high angle of incidence. The reciprocal structure of such a grating is shown in figure 4-7(b). •41 • Transverse scan qx 1 (a) (b) Figure 4-7. (a) Reciprocal space structure of the grating in the vicinity of the (224) reciprocal lattice point. A transverse scan was conducted, (b) Geometry of the measurement of (224) reflection in real space of a surface grating,which shows a compressed diffracted beam As shown in figure 4-7(b), if we carry out a transverse scan, satellite diffraction rods are expected in the plane of diffraction if this plane is perpendicular to the direction of the 2D grating e.g. (110). Inspection of the geometry of the (224) reflection shown in Figure 4-7(b) shows that this reflection is ideal to improve the resolution of a detector monochromized by a slit. Our incident 68 beam width is about 0.5 mm. Due to the asymmetry, the diffracted beam is compressed 8.35 times. In other words, the theoretical resolution is improved 8.35 times compared with the symmetrical geometry. 600 Relative incident angle (arcseconds) Figure 4-8. The x-ray rocking curve obtained by (224) grazing output scan. Figure 4-8 shows the rocking curve obtained with the transverse scan perpendicular to the [110] direction. From the angular separation between the satellite peaks, the period of the gratings can be calculated using the following relation'661 d = A,cos(8B -\\f)/ Acosin29B (4-5) 69 Here X is the x-ray wavelength, 8B is the Bragg angle of the crystal reflection, \\f is the asymmetry angle (i.e. the angle between the atomic planes used for diffraction and the (001) plane), and Aco is the angular separation between the satellite peaks. Here, the glancing exit angle, 0B-\|/ is 4.75° for the (224) reflection. The average satellite spacing here is found to be 55.6 arcsec. This leads to a value of 566.2 nm for the period. This value is in good agreement with the value of 562 nm ± 2 nm determined using diffraction of 488 nm Ar ion laser light. Compared with optical diffraction, x-ray method can generate multiple orders diffraction, and more structure information for gratings such as the width of the grooves and the quality of the sidewalls can be obtained,[65] which is impossible for light diffraction. Meanwhile, the period determined by optical methods should be compared with the designed value of 500 nm revealing the systematic error of the lithography system. 4.4 Ridge Definition After the gratings were fabricated on the sample, a ridge has to be defined in the grating region to expose the edge of the high aluminum content layer for oxidation. In this work, two methods were used to fabricate ridges. A simple method is used to fabricate test structures for examining the optical properties of a DFB. Long (>500 urn) and wide lines are fabricated on both sides of the grating region. The high aluminum content layer underneath the grating region can therefore be oxidized through these open lines. But in most situations, "real" ridge structures are needed for practical applications, like the production of lasers. In this case a 70 second method is used. It involves several process steps: pattern alignment, e-beam lithography, pattern transfer by etching and liftoff. These will be discussed below. 4.4.1 Electron beam Lithography Using PN-114 To define ridges on a sample, a negative photoresist PN-114 is chosen for the e-beam lithography. For a negative resist, the developer gets rid of the photoresist which is not exposed. So, only ridges themselves need to be to be written and the writing process is very fast. As in the case of the grating fabrication, the sample was first cleaned with acetone and methanol. Then it was baked at 120°C for 20 minutes to remove moisture. After baking, PN-114 was coated on the sample by spinning it at a speed of 8000 rpm. The resultant thickness of the film is about 0.8 um. [ 6 3 ] The sample was again softbaked at 120°C for 2 minutes. After that the sample was ready to be patterned in the electron microscope. Again, the pattern was written by NPGS and Hitachi S-4100 field emission SEM. The PN114 is much more sensitive than the PMMA with a threshold dose of 5 u.C/cm2. Because the photoresist is very thick, it is difficult to get features smaller than lu\m. But this feature size is still good enough to make waveguide ridges since the required width for ridges is about 10p.m. Table 4-3 lists the recommended process recipe for PN114. After the exposure, the sample was moved to an oven for a hardbake. The hardbake was done at 110°C for 5 minutes. It should be pointed out that the hardbake temperature is critical. It was found that the patterns do not turn out if the 71 bake temperature is below 105°C. After the hardbake, the sample was developed in a solution of tetramethylammonium hydroxide (TMAH) for 80-90 seconds, and then rinsed with DI water and blown dry with clean nitrogen. Process Parameters Recommended Conditions Softbake temperature •120°C ' Softbake time 2 minutes Exposure Dose 5-7 uC/cm2 Size for Aperture #l-#4 Accelaration Voltage 20 Kev Hardbake Temperature 110±1°C Hardbake Time 5 minutes Development time 80-90 s Table 4-3. Recommended process for developing patterns using PN114 The next step is to etch the ridges. For etching the high aluminum content AlGaAs, 1.2% HF was chosen. The etch rate of 1.2% HF is very fast, about 12 um/min at room temperature. Since the height of the ridges is on the order of two microns, a precise calibration is not necessary. After the etching, the PN114 was lifted off using microstrip 2001 photoresist stripper to clear the top of the ridge. The sample was immersed in 80°C stripper for 10 minutes to finish liftoff. 4.4.2 Ridge Alignment The ridges must be accurately located in the grating regions. Usually, the area of a grating is less than 30 um X 200 um, therefore precise positioning of the ridge is needed. The Hitachi S4100 has a manual translation stage, which can position a specific area (e.g. grating regions) with an error less than 2 um. However, because the length of the ridges is greater 72 than 500 um, they have to be written at a much lower magnification than that used in writing the gratings. In addition, the center of the field-of-view of the SEM doesn't overlap exactly under different magnifications. This causes a system error which can be as large as tens of microns depending on the difference in the magnification. Therefore, a more precise method is needed to position the ridges. The method used in this work is described next. Alignment Marks (a) Figure 4-9. (a) The layout of the grating patterns and the alignment marks, (b) Scanning windows and the overlays to match the alignment marks. 73 To precisely align ridges with the gratings, alignment marks were made on the sample when fabricating the gratings, such as those shown in Figure 4-9(a). The alignment marks were written on the four sides of the grating patterns. After etching, the alignment marks will permanently reside on the sample. To carry out the alignment procedure, an alignment file is first created using DesignCAD and NPGS. The alignment file is made up of a set of polygons (windows) on selected parts of the screen. Also, each window must also have overlays that are made up of single-pass lines, circles, etc. They correspond to the alignment marks, shown in Figure 4-9(b). When conducting the alignment, the SEM/lithography system is first turned to microscope mode, and the manual stage is used to roughly position the grating region. Then, the system is quickly turned to lithography mode and the alignment file is run. When NPGS runs the alignment file, the beam continuously scans over the polygon windows. The alignment marks are visible if they are in the window regions. The beam is turned off to avoid exposing the area of interest if it is in between the windows. When the alignment marks are positioned in each window, the overlays can be moved to match the visible alignment marks precisely. As the overlays are superimposed on the images of the marks, the NPGS will calculate a general transformation matrix that will correct for x, y errors of the microscope, as well as sample rotation, and offset. The transformation matrix is subsequently used in pattern writing to give accurate positioning of the ridges. As mentioned before, the alignment procedure should be conducted under the same magnification as that used in the e-beam writing to avoid the center mismatch in the field-of-view. The detailed description of the pattern alignment can be 74 be found in the NPGS manual. Figure 4-10 gives a SEM micrograph of the side view of the finished state-of-the-art DFB ridge waveguide. The pitch of the grating is 300nm. Figure 4-10. SEM micrograph showing a side view of the DFB ridge waveguide. The pitch of the grating is 300 nm. 4.5 Thinning After patterning, the only step left to finish fabrication is sample thinning and cleaving. A semiconductor waveguide needs optically smooth facets on both ends to form an optical cavity. This is achieved by mechanically cleaving the semiconductor wafers along {110} crystal planes. A typical length of a laser cavity is several hundred microns, while the typical thickness of GaAs 75 down to about 100 u\m to get reasonable length-thickness aspect ratio. In this work, thinning is done by mechanical grinding using sandpaper. Since the grating is fabricated on the sample surface, extra care should be taken to protect the grating patterns from contamination during thinning. To protect the patterned surface, a 0.8|im thick film of negative photoresist PN114 was spun on the patterned sample at a speed of 8000 rpm. The sample was then moved into a furnace and heated up to 170°C for about 5 minutes. At this temperature, the photoresist will expose and form a hard protecting film. The sample was then mounted on a specially designed chuck using paraffin wax, and manually polished to the required thickness. A detailled description of these processes can be found in Rich Morin's MASc thesis.[64] After thinning, the sample was immersed in Petroleum Ether for about 30 minutes to remove the remaining wax. This step is important because residual wax can make the subsequent photoresist removal procedure very difficult. Acetone and methanol were also used to clean the sample. The exposed PN114 was removed using Microstrip 2001 positive resist stripper, to clear the protection film. The stripper was maintained at 80°C and the sample was immersed for 10 minutes to complete the photoresist removal. A longer stripping time may be needed in some cases to remove the protecting film. 76 Chapter 5. Characterization of High Index Contrast DFB Waveguides Direct measurements of basic DFB wavguide properties is important for the determination of the operation characteristics of devices. To test the effect of buried oxides in DFB structures, measurements of guided modes diffraction from a surface grating DFB waveguide were conducted. These measurements were to determine the dispersion relation1671 for transverse electric (TE) modes and transverse magnetic (TM) modes. It was expected that the measured dispersion relations would reveal the effect of the buried oxide on the guided modes. Since the waveguide has the second order DFB surface texture, the dispersion relation should exhibit grating-induced features, which can be used to estimate the coupling coefficient of these surface textures and will show the effect of a high index contrast grating. 5.1 Sample and Experimental Setup The sample used to fabricate the waveguide is the ASU721 sample, as shown in Figure 2-1(b). 15 pm x 30 urn gratings were fabricated on its surface and the grating lines were aligned parallel to the 110 crystal direction. The pitch of the gratings varies from 300 nm to 500 nm. To open the underneath Alo.98Gao.02As for oxidation, 700 (im long and about 2 u\m wide lines were drawn on both sides of the grating region. Figure 5-1(a) is a sketch of this waveguide, 77 waveguide, and Figure 5- 1(b) shows a S E M micrograph top view, where the pitch of the grating is 460nm. Figure 5-1. (a) A sketch of the DFB ridge waveguide from the top. (b) S E M micrograph top view of the waveguide. 78 The experimental setup is shown in Figure 5-2 and the data was collected by Dr. Paul Paddon. The light source is a Ti-Sapphire tunable laser with an output power of 0.5 Watt and the wavelength tuning range from 750nm to 850nm. The wave plate makes the laser output circularly polarized. The light is then linearly polarized by the polarizer. The orientation of polarizer can be adjusted for either the TE or the TM polarization in the. waveguide. By adjusting the X-Y-Z stage, the light can be focused on the front edge of the waveguide to excite the guiding mode. The CCD camera 1 is used to monitor the guiding mode, while the CCD camera 2 is used to monitor the diffracted light. Ti-Sapphire tunable laser CCD camera 2 ] Wave plate CCD Camera 1 a Polarizer Aperture focus lens X-Y-Z stage Waveguide sample Figure 5-2. Schematic of the diffraction experiment set-up. Once the coupling to the waveguide is optimized, the direction of the diffracted light from the surface grating can be measured by CCD camera 2. By tuning the laser, the diffraction angles 79 ft can be determined for different laser output frequencies. The dispersion relationship of the guiding mode can then be established from these diffraction angles. 5.2 Measurement and Analysis yt GaAs substrate Figure 5-3. Schematic of the diffraction measurement. Figure 5-3 is a drawing of the measurement. Since the frequency of the incident laser co is known, the propagation direction of the incident light and diffraction light from the waveguide can be determined by kdiff = CO / c , where c is the speed of light in air, and co is the operating frequency of the laser. Based on the coupled mode theory, the angle 0 of the diffracted light from the gratings is related to the guided mode propagation constant p as[68] P = neffk = £sin0 + £ c (5-1) Where the neg is the effective index of the waveguide, 9 is the diffraction angle from the normal as shown in Figure 5-3, kc is the grating vector 2n/A, and A is the pitch of the 80 grating. The propagation constant P is equal to the grating vector kG in the case of the normal diffraction (i.e., 9 = 0). Thus, the effective index of the guided mode neff can be computed from n*-~k ( 5 - 2 ) where E represents the photon energy at normal diffraction. Here, wavenumbers (cm-1) are used as the unit for frequency on the y-axis. The effective indices for the TE and T M modes with 440nm pitch and 460nm pitch grating have been determined. However, a more accurate determination of the effective index n,,ff is not expected in this experiment. That is because the surface grating is relatively short (20um), the diffracted beam has relatively broad angular spread which makes a precise measurement of the diffraction angle difficult. For a slab waveguide, the effective index can be approximated as[69' » « = Z r A (5-3) j where Fj and «, are the confinement factor and the refractive index of the jth layer material, respectively. The effective index is 1.7, which is much lower than the effective index of the traditional III-V compound semiconductor waveguide (about 3.2). This indicates that most of the mode energy is confined in the lower diffractive index mediums (i.e., air and A!Ox) due to the introduction of the buried oxide. From diffraction measurements shown in Figure 5-3, the dispersion relationship between the working frequency of the laser co and the propagation 81 constant P of the waveguide can be determined. As a example, Figure 5-4 shows the dispersion relationship for TE mode of the 460 nm pitch DFB waveguide. 138001 13600-13400-13200-^ 13000-i 12800-r l 12600-12400-12200-12000 460nm Gratings TE mode • • -1500 -1000 -1 r •500 0 — i 1 1 1 1 1 1— 500 1000 1500 2000 Ak^ k-kpCcm"1) Figure 5-4. Dispersion relationship for the TE mode with a 460 nm pitch grating. Data was collected in J. Young's Lab by Paul Paddon. 82 In the above figures, the x-axis represents the difference between the wavevector of the incident light and the wavevector (2 71 / A) of the grating. The unit of photon energy in y axis is cm"1 , the same as the unit of wavevector. 83 Chapter 6. Conclusion Since 1990, when the lateral oxidation of AlGaAs was demonstrated for the first time by Holonyak and co-workers at the University of of Illinois, there has been considerable work done in this area. Inspired by a number of successful applications of this technique, some people believe that new opportunities for the III-V compound semiconductor industry are on the horizon. In this thesis, research work on lateral oxidation for the last seven years have been reviewed. The three most successful applications of this technique have been introduced, they are the vertical cavity surface emission laser, GaAs on insulator technique, and distributed Bragg reflector. We also propose a new device, namely, high index contrast distributed feedback lasers, which takes advantage of current confinement and optical confinement produced by the lateral oxidation process. The theoretical simulation results show an over two times higher coupling efficiency can be achieved by this new type of device than that in traditional distributed feedback structures. Together with the current confinement property, the high index contrast distributed feedback laser has promise for high performance and low cost optical communication systems. The major part of this thesis described our efforts in the past year to accomplish this objective. Part of the work 84 has been done on the material research, and the other part of the work is directed toward the fabrication of high index contrast distributed feedback laser. First, we developed a process to realize lateral oxidation and calibrate the oxidation rate for Al 0 98Gao.o2As. A reproducible linear oxidation rate has been obtained showing that the precise control of the lateral oxidation is possible. Atomic force microscopy has been conducted on a partially oxidized mesa to investigate the vertical expansion due to lateral oxidation. Less than 5 nm vertical expansion was detected on an oxidized 2 um thick Al 0 98Gao.o2As layer. The results will be useful for distributed feedback devices which require precise knowledge of the thickness of quarter-wave stacks. X-ray photoelectron spectroscopy (XPS) was used to study the oxidized AlGaAs in different oxidation environments. We have found that the good quality oxide from high temperature oxidized AlGaAs has negligible As content, in contrast to the poor quality AlGaAs oxidation product which is As rich compound. Also by analyzing the core level chemical shifts of the A12p peak, we found that the chemical composition of the high temperature oxidation product of Al0.9gGao.02A.S is consistent with the multi-phase A10x rather than single phase A1203 crystal In the device fabrication, we have successfully made a second order DFB ridge waveguide with a surface grating which is a passive prototype of high index contrast DFB laser. By optimizing the e-beam lithography, we are able to fabricate both one dimensional grating and two dimensional photonic lattice structures with the spacing 85 greater than 280 nm. Moreover, by applying the novel multipass writing technique, superior uniformity gratings can be fabricated reproducibly. Associated with the lateral oxidation, we fabricated a series of test DFB ridge waveguides on the buried oxide with grating spacings from 300 nm to 500 nm. Diffraction measurements were conducted to determine the dispersion relationship and effective index for the guided modes. It has been found that the effective index of the TE and T M modes is 1.7. References 1. T. Mimura, M. Fukuta; IEEE Trans. On Electron Devices 27, 1147 (1980) 2. D. Fulkerson, S. Baier, J. Nohava and R. Hochhalter, Solid State Electronics 39, 461 (1996) 3. J. M. Dallesasse, N. Holonyak, Jr., A. R. Sugg, T. A. Richard, and N. El-Zein, Appl. Phys. Lett. 57, 2844 (1990) 4. H. Takenouchi, T. Kagawa, Y. Ohiso, T. Tadokoro and T. Kurokawa, Electron. Lett. 32, 1671 (1996) 5. See, for example, Adir Bar-Lev, Semiconductors and Electronics Devices 2nd Edition, Prentice Hall International, 1984 6. J. M. Dallesasse and N. Holonyak, Appl. Phys. Lett. 58, 394 (1991) 7. D. L. Huffaker, D. G. Deppe, K. Kumar and T. J. Rogers, Appl. Phys. Lett. 65, 97(1994) 8. D. L. Huffaker, J. Shin and D. G. Deppe, Electron. Lett. 30, 1946 (1994) 9. K. D. Choquette, R. P. Schneider, K. L. Lear and K. M. Geib, Electron. Lett. 30, 2043 (1994) 10. K. D. Choquette, K. L. Lear, R. P. Schneider, Jr., and K. M. Geib, Appl. Phys. Lett. 66,25(1995) 11. K. D. Choquette, K. L. Lear, R. P. Schneider, Jr., K. M. Geib, J. J. Figiel and R. Hull, IEEE Photon. Tech. Lett., 7, 1237 (1995) 12. J. M. Dallesasse, N. Holonyak, Jr., D. C. Hall, N. Elzein, A. R. Sugg, S. C. Smith and R. D. Buraham, Appl. Phys. Lett. 58, 394 (1991) 13. M. J. Ries, E. I. Chen, and N. Holonyak, Jr., Appl. Phys. Lett. 68,2035 (1996) 14. E. I. Chin, N. Holonyak, Jr., and S. A. Maranowski, Appl. Phys. Lett. 66, 2688 (1995) 87 15. P. A. Grudowski, R. V. Chelakara, and R. D. Dupuls, Appl. Phys. Lett. 69, 388(1996) 16. K. Tai, R. J. Fischer, K. W. Wang, A. Y. Cho, Electron. Lett. 25, 1644 (1995) 17. D. L. Huffaker, J. Shin, and D. G. Deppe, Appl. Phys. Lett. 66, 1723 (1995) 18. K. L. Lear, K. D. Choquette, R. P. Schneider, Jr., and S. P. Kilcoyne, Appl. Phys. Lett. 66,2616(1995) 19. K. L. Lear, K. D. Choquette, R. P. Schneider, Jr., S. P. Kilcoyne, and K. M. Geib, Electron. Lett. 31, 208 (1995) 20. K. D. Choquette, r. P. Schneider, Jr., K. L. Lear, and K. M. Geib, Electron. Lett. 30,2043(1994) 21. P. Parikh, P. Chavarkar, U. Mishra, personal communication 22. M. H. MacDougal, H. Zhao, P. D. Dapkus, M. Ziari, and W. H. Steier, Electron. Lett. 30, 1147(1994) 23. A. Dodabalapur, T. Y. Chang, Appl. Phys. Lett., 60,929(1992) 24. Pochi Yeh, Optical Waves in Layered Media, A Wiley-Interscience Publication, 1988 25. Jens Buus, Single Frequency Semiconductor Lasers, SPJE Optical Engineering Press, 1991 26. Yoshiaki Nakano, Kunio Tada, IEEE J. Quantum Electron., 10, 2017(1988) 27. H. Kogelnik and C. V. Shank, J. Appl. Phys., 5,2327 (1972) 28. Govind P. Agrawal, Niloy K. Dutta, IEEE J. Quantum Electron., 6, 534 (1985) 29. William Streifer, Donald R. Scifres, Robert D. Burnham, IEEE J. of Quantum Electron., 4,134(1977) 30. Amnon, Yariv, IEEE J. of Quantum Electron., 9, 919 (1973) 31. Rudolf F. Kazarinov and Charles H. Henry, IEEE J. of Quantum Electron. 2, 144 (1985) 88 32. William Streifer, Donald R. Sciffes, Robert D. Burnham, IEEE J. of Quantum Electron., 11, 867 (1975) 33. Horng J. Luo and Peter S. Zory, IEEE J. of Quantum Electron., 1,10 (1994) 34. Horng J. Luo and Peter S. Zory, IEEE Photon. Technol. Lett., 9,614 (1990) 35. William W. rigrod, D. Marcuse, IEEE J. of Quantum Electron., 11, 673 (1976) 36. T. Makino, D. M. Adams, IEEE Transactions Photon. Technol. Lett., 11, 963 (1991) 37. R. D. Dupuls, Compound semiconductor, 1, 32 (1997) 38. Gas Mixer Calibration done by R. Liang, and R. Krahn, 1991 39. J. -H. Kim, D. H. Lim, K. S. Kim, G. M. Yang, K. Y. Lim, and H. J. Lee, Appl. Phys. Lett., 69, 3357 (1996) 40. S. Guha, F. Agahi, B. Pezeshki, J. A. Kash, D. W. Kisker, and N. A. Bojarczuk, Appl. Phys. Lett., 68, 906 (1996) 41. LEYBOLD MAX200 Spectrometer Manual, LEYBOLD LTD, 1991 42. Ying-Ling Leung M. Sc. Thesis ( University of British Columbia, Vancouver, . 1992) 43. S. Guha, F. Agahi, B. Pezeshki, J. A. Kash, D. W. Kisker, and N. A. Bojarczuk, Appl. Phys. Lett. 68, 906 (1996) 44. D. Briggs, M. P. Seah, Practical Surface Analysis, John Wiley & Sons, 1983 45. C. D. Wagner, H. A. Six, W. T. Jansen, J. A. Taylor, Appl. Surf. Sci. 9, 203 (1981) 46. Hitachi Ltd., Instruction Manual for Model S-4I00 Field Emission Scanning Electron Microscope, 1991 47. Nabity, J.C., Nanometer Pattern Generation System, Version 7.4, J.C. Nabity Lithography Systems, Bozeman, Montana, USA (March 1994) 48. DesignCAD 2-D Reference Manuals, Version 6, American Small Business Computers, Pryor, OK, USA, 1992 49. Alex Busch M. Sc. Thesis ( University of British Columbia, Vancouver, 1995) 89 50. Ralph Williams, Modern GaAs Processing Methods, Artech House, Boston, 1990, Pg. 101-106 51. L. A. Coldren, J. L. Merz, Electron. Lett., 21, 559 (1985) 52. Kelly Kenefick, J. Electrochem. Soc. Solid-State Sci. & Technol., 129,2380 (1982) 53. Gregory C. DeSalvo. Wen F. Tseng, and James Comas, J. Electrochem. Soc, 139, 832 (1992) 54. S. A. Rishton, S. P. Beaumont, and C. D. Wilkinson, Microcircuit Eng. 341 (1982) 55. Dr. S. dimming, S. Thorns, S. P. Beaumont, and J. M. R. Weaver, Appl. Phys. Lett. 68,114(1996) 56. H. G. Craighead, P. E. Howard, L. D. Jackel, and P. M. Mankiewich, Appl. Phys. Lett., 42, 38 (1985) 57. Y. -H. Lee, R. Browning, N. Maluf, G. Owen, and R. F. W. Pease, J. Vac. Sci. Technol. B, 10, 3094 (1992) 58. P. A. Peterson, Z. J. Radzimski, S. A. Schwalm, and P. E. Russell, J. Vac. Sci. Technol. B, 10,3088(1992) 59. M. A. McCord, and T. H. Newman, J. Vac. Sci. Technol. B, 10, 3083 (1992) 60. R. Steffen, F. Faller, and A. Forchel, J. Vac. Sci. Technol. B, 12, 3653 (1994) 61. T. J. Stark, K. M. Edenfeld, D. P. Griffis, Z. J. Radzimski, and P. E. Russell, J. Vac. Sci. Technol. B, 11, 2367 (1993) 62. C. W. Lo, M. J. Rooks, W. K. Lo, M. Isaacson and H. G. Craighead, J. Vac. Sci. Technol. B, 13, 812 (1995) 63. Manoj Kanskar (Private Communication) 64. R. Morin M. A Sc. Thesis ( University of British Columbia, Vancouver, 1995) 65. Qun Shen, C. C. Umbach, B. Weselak, and J. M. Blakely, Phys. Rev. B, 48, 17967 (1993) 65. A. T. Macrander and S. E. G. Slusky, Appl. Phys. Lett., 56,443 (1990) 90 66. For a general reference see Photonic Crystals, J. D. Joannopoulos, R. D. Meade, and J. N. Winn, Princeton University Press, Princeton, NJ, 1995 67. Paul Paddon (Private Communication) 68. G. P. Agrawal and N. K. Dutta, Long-Wavelength Semiconductor Lasers, Van NostrandReinhold, New York, 1989 69. M. Kanskar, P. Paddon, V. Pacradouni, R. Morin, A. Busch, Jeff F. Young, S. R. Johnson, Jim Mackenzie, and T. Tiedje, Appl. Phys. Lett. 70, 1438 (1997) 70. P. M. Bell, J. B. Pendry, L. Martin Moreno, and A. J. Ward, Computer Physics Communications, 85, 306 (1995) 71. Sajeev John, Physics Today, 32, May 1991 72. C. FL Chen, L. H. Chen and Q. M. Wang, Electron. Lett., 32, 697, (1996) = 73. H. L. Cao, Y. Luo, Y. Nakano, K. Tada, M. Dobashi, and H. Hosomatsu, IEEE Photonics Technology Letters, 4, 1099 (1992) , ^ 74. R. D. Martin, S. Forouhar, S, Keo, R. J. Lang, R. G. Hunsperger, R. C. Tiberio, and P. F. Chapman, IEEE Photonics Technology Letters, 7, 244 (1995) 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

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"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0088508/manifest

Comment

Related Items