UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Reactions of the {100} face of gallium arsenide with molecular and atomic bromine Salusbury, Ian McKenzie 1990

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

Item Metadata


831-UBC_1990_A6_7 S34.pdf [ 3.45MB ]
JSON: 831-1.0059736.json
JSON-LD: 831-1.0059736-ld.json
RDF/XML (Pretty): 831-1.0059736-rdf.xml
RDF/JSON: 831-1.0059736-rdf.json
Turtle: 831-1.0059736-turtle.txt
N-Triples: 831-1.0059736-rdf-ntriples.txt
Original Record: 831-1.0059736-source.json
Full Text

Full Text

R E A C T I O N S O F T H E {100} F A C E O F G A L L I U M A R S E N I D E WITH M O L E C U L A R A N D ATOMIC BROMINE By IAN McKENZIE SALUSBURY B . S c . (Hons.), University of Bath,  1988  A T H E S I S S U B M I T T E D IN P A R T I A L F U L F I L L M E N T O F T H E REQUIREMENTS FOR T H E D E G R E E O F M A S T E R O F SCIENCE in T H E F A C U L T Y O F G R A D U A T E STUDIES Department  We  of  Chemistry  accept this thesis as conforming to  the  required  standard  T H E UNIVERSITY O F BRITISH C O L U M B I A July ©  Ian  1990  M c K e n z i e Salusbury,  1990  In  presenting this  degree at the  thesis  in  University of  partial  fulfilment  of  of  department  this thesis for or  by  his  or  requirements  British Columbia, I agree that the  freely available for reference and study. I further copying  the  representatives.  an advanced  Library shall make it  agree that permission for extensive  scholarly purposes may be her  for  It  is  granted  by the  understood  that  head of copying  my or  publication of this thesis for financial gain shall not be allowed without my written permission.  Department  of  CHgfAf ST^V  The University of British Columbia Vancouver, Canada  D a t  e  DE-6 (2/88)  -3vU_»f  ^  Abstract  T h e reaction of gallium arsenide {100} atomic  bromine was studied at temperatures  and at pressures of bromine between 0.1  with molecular and between  100  and 40 Torr.  and  225°C  Samples of  G a A s were placed on a silicon platform within a Pyrex reactor flow system and the etch rate was determined change of the sample.  by profilometry  or weight  A t o m i c bromine was produced by a 2450 M H z  microwave discharge and the samples were dtched downstream. atomic  concentration was  measured by an isothermal  The  calorimetric  detector. Pressure dependence studies for molecular Br2 etching that below limiting  1-2  showed  Torr of bromine, a first order reaction was rate-  whereas above this pressure a half order reaction was rate-  limiting.  Temperature  dependence studies for the low  high pressure regimes  gave  activation  energies  and  pressure and  pre-exponential  values for the two respective rate controlling reactions.  The first  order reaction was found to have an activation energy of 29.2 kJ m o l cm  - 2  s  - 1  _ 1  and a pre-exponential value of (3.4 Ton" . 1  ± 4.4) x 1 0  2 1  ±4.0  molecule  The activation energy for the half order reaction was  found to be 8.4 ± 0.7 kJ m o l " l O * * molecule c m ~ s " 1  2  1  1  with a pre-exponential of (6.4  ± 1.3) x  Torr / . 1  2  T h e activation energy for atomic etching was calculated to be 12.9 ± 0.9 kJ m o l ' cm  - 2  s  _ 1  1  and the pre-exponential, (7.1  Torr . 1  ii  + 2.0) x 1 0  2 0  atom  Table  of  Contents  Abstract  ii  List of tables  v  List of figures  vi  Acknowledgments  viii  1 INTRODUCnON 1.1 H I S T O R I C A L P E R S P E C T I V E  1  1.2 C R Y S T A L S T R U C T U R E O F G A L L I U M A R S E N I D E  3  1.3 G A L L I U M A R S E N I D E IN E L E C T R O N I C D E V I C E S  8  1.4 P R O C E S S I N G T E C H N O L O G Y  11  1.5 E T C H I N G P R O C E S S E S  1 3  1.6 G A S - S U R F A C E I N T E R A C T I O N S  17  1.6.1  Adsorption 1.6.2a  17  Langmuir-Hinshelwood  mechanism  1.6.2b Eley-Rideal mechanism  19 21  1.7 P R E V I O U S S T U D I E S O N G A L L I U M A R S E N I D E E T C H I N G  21  1.8 P U R P O S E O F S T U D Y  26  2 EXPERIMENTAL 2.1  MATERIALS  27  2.2 M O L E C U L A R B R O M I N E E T C H I N G 2.2.1  Apparatus for molecular  28 etching  28  2.2.2 Procedure for molecular etching  30  2.3 A T O M I C E T C H I N G 2.3.1  35  Apparatus for atomic  etching  2.3.2 Procedure for atomic etching  iii  35 40  3 RESULTS  4  3.1 E T C H R A T E D A T A  42  3.2 E R R O R A N A L Y S I S  4 6  3.3 K I N E T I C A N A L Y S I S  4 8  3.3.1  L o w pressure region  5 2  3.3.2  H i g h pressure region  64  3.4 A T O M I C E T C H I N G  64  3.5 A N A L Y T I C A L R E S U L T S  64  DISCUSSION 4.1 P R O P O S E D M E C H A N I S M  6 9  4.2 K I N E T I C A N A L Y S I S O F M E C H A N I S M  7 1  4.3 C O N C L U S I O N S  7 7  References  7 8  iv  List Table 1.1  of Tables  Physical properties of silicon and gallium arsenide  9  Table 3.1  List of etch rates of GaAs (100) with B r at 125°C  43  Table 3.2  List of etch rates of GaAs (100) with B r at 145°C  44  Table 3.3  List of etch rates of GaAs (100) with B r at 183°C  44  2  2  2  Table 3.4 Gradients of In etch rate versus In pressure of B r for the low pressure region  55  Table 3.5 Gradients of In etch rate versus In pressure of B r for the high pressure region  55  Table 3.6 Gradients of etch rate plotted against pressure of B r for the low pressure region  59  Table 3.7  59  2  2  2  Arrhenius data for low pressure region  Table 3.8 List of etch rates for a B r pressure of 0.83 Torr, at several temperatures  61  Table 3.9 List of etch rates for a B r pressure of 1.0 Torr, at several temperatures  61  Table 3.10 Gradients and intercepts of the graphs of etch rate versus (pressure of bromine) / , for the high pressure region.  65  Table 3.11 List of etch rates for GaAs (100) with Br, at a B r pressure of 0.83 Torr  65  2  2  1  2  v  2  List  of  Figures  Figure 1.1 T h e zincblende lattice observed perpendicular to a [111] axis and along a [110] axis Figure 1.2  Schematic representation  on the {111}, {100} Figure 1.3  and {110}  4  of the surface bonding  planes of G a A s  Crystallographic etching of {100}  6 GaAs  7  Figure 1.4  Schematic diagram of microelectronic processing  12  Figure 1.5  Profiles of (a) anisotropic and (b) isotropic etches  14  Figure 1.6  Reactors for plasma etching  16  Figure 1.7  Potential energy profiles for adsorption with  (a) zero E c and (b) large E c  2 0  Figure 2.1 apparatus  Schematic diagram of molecular bromine  Figure 2.2  Sample platform  3 1  Figure 2.3  Geometry of V-shaped groove formation  3 3  Figure 2.4  Schematic diagram of atomic etching apparatus  3 6  Figure 2.5  Isothermal calorimetric  3 8  Figure 2.6 calorimetric  Wheatstone detector  Figure 3.1  2 9  bridge  detector  arrangement  for  isothermal 3 9  Graph of etch rate against pressure of B r ^ at 1 2 5 ° C  Figure 3.2 T y p i c a l etch rate profiles obtained profilometry Figure 3.3  etching  4 5  from 4 7  Graph of In etch rate against In pressure of Br2  at 1 2 5 ° C  4 9 vi  Figure 3.4  Graph of In etch rate against In pressure of B r  2  at 1 4 5 ° C Figure 3.5  5 0 Graph of In etch rate against In pressure of B r  2  at 1 8 3 ° C Figure 3.6  5 1 Graph of In etch rate against In pressure of B r  2  for the low pressure region at 1 2 5 ° C Figure 3.7  5 3  Graph of In etch rate against In pressure of B r  2  for the high pressure region at 1 2 5 ° C Figure 3.8  5 4  Graph of etch rate against square root of B r  2  pressure for the high pressure region at 1 2 5 ° C  5 6  Figure 3.9 Graph of etch rate against pressure of B r 2 for the low pressure region at 1 2 5 ° C  5 8  Figure 3.10 Arrhenius plot for the low pressure rate constants  6 0  Figure 3.11 0.83  Arrhenius plot for etch rates at a B r  2  pressure of  Torr  Figure 3.12  6 2 Arrhenius plot for etch rates at a B r  2  pressure of  1.0 Torr  6 3  Figure 3.13  Arrhenius plot for the high pressure  rate constants Figure 3.14 Figure 4.1  6 6 Arrhenius plot for atomic etching rate constants  Potential energy profile for the  and G a A s (100)  reaction  6 7  bromine 7 5  vii  Acknowledgments  I would like to thank Professor E . A . Ogryzlo for his guidance and encouragement throughout the course of this work. I am grateful to Professor D . Chong for making his linear least squares computer program available for our group's use. M y thanks also go to the Electrical and Mechanical shops for their assistance and to the glass blowers, Steve and Sean, for their help and patience. T h e kinetic analysis given in the discussion section was first developed by Z . H . Walker and my thanks go to him for its use. The advice and encouragement of Mark, Paul and Zane in the lab is gratefully acknowledged. They helped to make my time at U . B . C . as enjoyable as it has been. Finally, I would like to express my gratitude to my family and friends for their support and interest over the last two years.  viii  1  INTRODUCTION  1.1 H I S T O R I C A L P E R S P E C T I V E The invention of the bipolar transistor by J . Bardeen, W . H .  Brattain  and W . Shockley of the B e l l Telephone Laboratories, in 1948, was the beginning of a second industrial revolution.  1  Microelectronic technology  has provided the driving force behind the rapid post-war science, engineering  and the  information  advances in  industry.  The properties of semiconductors such as silver sulphide, have been investigated since the last century and in the early days of radio the detector in all receivers was the crystal rectifier, crystal.  usually a galena  This situation changed with the introduction of cheap  2  thermionic  (or  vacuum)  tubes in the early  twentieth century  all electronic equipment became based on vacuum tubes.  and almost  They were  used as switches or amplifiers to magnify electrical signals but their size, lack of reliability  and high power dissipation restricted their usefulness.  3  The first transistor was made from germanium because of the ease of growing good single crystals.  However, germanium was affected by  operating  due to its relatively  temperature  limitations  and so alternative materials were sought.  small bandgap  The larger band gap and  superior oxide of silicon prompted research into its use and by 1952, preparation  the  of single crystals of silicon had been reported , with the first 4  silicon transistors becoming commercially available The transistor provided low power, switches and amplifiers  two  years later.  small and robust electronic  and quickly replaced the vacuum tube. The 1  demand from the computer industry was a major transistor production.  driving force for  The next stage in microelectronic  was the production of the integrated circuit (IC), became commercially available in  1960.  A  development  the first of which  monolithic integrated  circuit  or chip is defined as the "microstructure of many circuit elements inseparably associated on or within a continuous substrate" .  Instead of  5  the assembly of components one by one and the consequent complex wiring, only a series of photographic processes determines of several million components on a wafer.  the locations  The advantages of  integration  are reflected by the change in the maximum number of components per chip since 1960.  A t that time one or two devices were put on each chip  and that has now risen to 16 million devices per chip, with over 30,000 components  mm .  This level of integration is known as Very Large Scale  Integration ( V L S I ) .  The levels of integration are loosely defined by these  figures  - 2  : 100  components for standard integrated  components for standard L S I VLSI.  circuits,  10,000  and 1 million components for  standard  The next stage in development has been named ultra large scale  integration  (ULSI).  6  The predominance of silicon in microelectronic device fabrication is due to a number of factors.  Si has a suitably high band-gap ( l . l e V ) so  that it can operate throughout a reasonable temperature can be used at temperatures of 7 5 ° C for germanium.  7  range.  Silicon  up to 2 0 0 ° C , which compares to a maximum  A l s o Si is the second most abundant element in  the Earth's crust and so is inexpensive and in an inexhaustible supply. The material properties of Si enable large single crystals, essentially dislocation free, to be grown and then annealed, sawn and polished with well established techniques.  Finally, S i forms a stable oxide when  exposed to oxidising agents at high temperatures, which can be utilised as 2  a dielectric layer or a protective mask.  Desired electronic properties can  be produced by inserting impurities (by diffusion or ion implantation) into selected areas of the wafer surface, from which oxide has been stripped. Despite silicon's long established technology and many capabilities, the variety of tasks for which semiconductors are required demands a variety of semiconductors to perform them. still used in high voltage transistors.  For example, germanium is  The potential for using III-V  compounds (that is compounds formed between elements of Groups IIIA and V A of the periodic table) was realised in the early stages of microelectronics research.  During the 1950's there was a considerable  research effort in this field , however because of the physical properties 8  of the materials and processing problems it is only within the last decade that these materials have found commercial use.  They enjoy processing  speeds unequalled by silicon and have many optoelectronic The  applications.  most important examples are gallium arsenide (GaAs) and indium  phosphide  (InP).  1.2 C R Y S T A L STRUCTURE OF G A L L I U M ARSENIDE Gallium arsenide crystallises in the zincblende (or sphalerite) structure.  This structure consists of two interpenetrating face centred  cubic lattices with one of the sublattices displaced one quarter of the way along the main diagonal of the other.  Figure 1.1 shows this configuration  : one sublattice consists of gallium atoms and the other arsenic atoms . 9  The  bonding is polar covalent (32% ionic character)  concentrated on the As atoms.  10  with the charge  Hereafter, { }, ( ) and [ ] will refer to  families of planes, specific planes and directions of planes, respectively. The  most significant feature of the GaAs structure is that the structure is 3  Gallium atoms Arsenic atoms  Figure 1.1 The zincblende lattice observed perpendicular to a [111] axis and along a [110] axis.  4  asymmetric ; the [111] axis is polar.  The (ill) plane in the figure is  called the Ga ( 1 1 1 ) face or A face and the ( 1 1 1 ) plane is the As ( 1 1 1 ) face or B face.  If we consider the Ga ( 1 1 1 ) face we can see that two possible  surface states exist.  An As atom could be connected to one Ga atom  below, leaving the As atom with three 'dangling' bonds, that is three unpaired electrons.  However, this state appears too energetically  unfavourable to be found as a surface state in a cleaved crystal. Alternatively, there can be a Ga atom attached by three bonds to arsenic atoms in the layer below, so that the Ga has one bond coming out of the plane.  Therefore, if the Ga atoms in this face are neutral, they have no  free electrons as their three valence electrons are used for bonding. If we now consider the As ( 1 1 1 ) face, we can see that the favoured state involves As atoms connected to three Ga atoms, leaving one non-bonding pair of electrons if the atom is neutral.  So each As atom has two free  electrons since only three of their five valence electrons are used in bonding (figs 1.2a and 1.2b).  Thus the As {111} faces are more readily  etched and at temperatures below rapidly from their surface.  770°C,  evaporation takes place more  Also oxidation of As {111} planes occurs more  quickly. The {100} planes consist of either all Ga or As atoms (fig. 1.2c). The [100] direction is not polar and so all {100} planes behave identically. {100} planes are found to etch more slowly than As {111} planes and faster than Ga {111} planes.  11  This disparity in etch rates causes the  phenomenon of crystallographic etching.  For example, when the {100}  plane is etched with an appropriately aligned striped mask in place, Vshaped (fig. 1.3a) or reverse mesa shaped grooves form (fig. 1.3b) . This 12  feature has been utilised in the fabrication of VMOS (vertical metaloxide-semiconductor) devices.  5  {110}  d)  Figure 1.2 Schematic representation of the surface bonding on the {111}, {100} and {110} planes of GaAs 6  [100]  [111]  (TiQ)l  (a) O  Gallium atom  #  Arsenic atom  [111]  (b)  Figure 1.3 Crystallographic etching of {100} GaAs, which results i (a) V-shaped groove or (b) reverse mesa formation  7  The  {110}  cleavage planes intersect the {100}  in the {110}  planes at right angles and  face an equal number of Ga and As atoms occur (fig. 1.2d).  Each Ga atom is bound to two As atoms in the plane and a third bond is directed into the interior of the crystal.  1.3 G A L L I U M ARSENIDE IN ELECTRONIC DEVICES The development of GaAs integrated circuits was dependent on the production of high quality substrates and this proved problematic.  Some  methods used in producing and then treating silicon crystals are sometimes not viable for compounds.  For example, in the Czochralski  technique for crystal growth a seed crystal is slowly withdrawn from a 'melt'.  It has proven difficult to prevent decomposition of a GaAs melt  since the arsenic component is more volatile than the gallium.  Similarly,  annealing processes to remove defects cause selective vaporisation of arsenic.  New techniques such as liquid encapsulated Czochralski (LEC)  have been developed to prevent the vaporisation of arsenic.  13  GaAs is also more brittle than Si and so is more difficult to process. These combined problems mean that it is still not possible to produce GaAs wafers that are as large as Si wafers. In contrast to Si, the native oxides of Ga and As are not suitable as dielectrics or passivating layers.  Therefore, other means must be devised  to form insulating structures. GaAs has many useful properties, which make these problems worth overcoming (Table 1.1).  14  GaAs has a higher electron mobility and  drift velocity than Si and so the switching speed is faster. switching speed reported for any transistor is 250 x 10  9  The fastest switches per  second, which was achieved by a transistor made of layers of aluminium indium arsenide and gallium indium arsenide on an indium phosphide 8  Table  1.1  Physical properties of silicon and gallium  Silicon Crystal Lattice  structure constant (A)  Atomic/molecular Density  ( g cm-3)  Band gap (eV) T y p e of band gap Drift mobility (cm V s- ) Electrons Holes 2  1  wt  Gallium  arsenide  Arsenide  Diamond  Zincblende  5.43  5.64  28.09  144.63  2.33  5.32  1.12  1.42  Indirect  Direct  1500 450  8500 400  1  9  base.  In contrast, the fastest reported switching speed for a silicon  transistor is 75 x 1 0 already  9  switches per s e c o n d .  been exploited in some mainframe  This speed advantage has  15  and supercomputers.  The  Cray 3 supercomputer, the first of its kind to be based on G a A s , will have a clock speed as high as 300  megaHertz.  16  H i g h electron mobility is also important noise operation. be particularly signals. dishes.  for high frequency, low  The lower noise operation of G a A s circuits turns out to valuable for the detection of television and  microwave  Therefore, G a A s detectors are widely used in satellite receiving Stronger fields than 10,000 V  cm  - 1  elicit  additional physical  interactions and G a A s loses its advantage over silicon. Band-gap size determines semiconductor can operate.  the range of temperatures  Therefore, G a A s integrated  function at higher temperatures  than S i .  at which a  circuits can  A l s o the band-gap in G a A s can  be engineered by alloying with another element.  F o r example,  aluminium arsenide has a much wider band gap than G a A s . substitution to form A l G a i _ x  x  So partial  A s , produces band gaps directly  proportional to the mole fraction x of aluminium. Another major  advantage of G a A s and all other direct band-gap  semiconductors is the optoelectronic capabilities, which the direct bandgap offers.  In  direct band-gap semiconductors, the conduction band  minimum is located at the same momentum value as the valence band maximum.  Thus an electron can more readily make a transition from the  conduction band to the valence band by radiating a p h o t o n .  17  This  process can also be reversed so that a photon is absorbed and converted into electrical energy. B y converting light energy into electrical energy, G a A s can act as a photodetector and can be integrated into high speed G a A s electronic  circuitry. 10  Such transitions cannot occur efficiently  in indirect  semiconductors without an accompanying nonradiative collision, to conserve m o m e n t u m .  band-gap  reaction, such as a  Since G a A s can efficiently  18  convert  electrical energy into light it is used to make light emitting diodes ( L E D ) and diode lasers. size.  The wavelength  of the light depends on the band-gap  Diode lasers emit radiation at near infrared wavelengths  and the  output can be shifted by alloying with aluminium, indium or phosphorus, by the  so-called band-gap engineering mentioned previously.  are widely  Such lasers  used in optical fibre based communications.  1.4 P R O C E S S I N G T E C H N O L O G Y Microelectronics processing involves several stages (fig. crystal is grown and then annealed by heating, for semiconductors, such as silicon.  1.4).  Firstly, a  elemental  This minimises defects in the crystal.  Then the crystal is sawn, polished and cleaned to produce wafers (up to 20 c m in diameter). by doping with  Next the desired electrical properties are introduced  impurities.  Pattern generation on the wafer is accomplished by laying down a photoresist. light.  This is patterned by applying a mask and then exposing it to  Photoresists can create negative or positive images.  In the former  case, bonds crosslink on exposure to light whereas in the latter, bonds are broken.  The developed photoresist must then be removed by hot  oxidising agents (H2SO4-H2O2), solvents or an oxygen plasma. Once the desired structures have been etched into the wafer resist can be removed.  T o produce metal oxide semiconductor integrated  circuits, various metal and oxide layers must be layed down. requires the repetition  the  of the pattern  1 1  generation  This usually  and etching stages.  X GiETID  Melt  Saw, polish and clean  Crystal growth  Masking and etching stages  Oxidation, diffusion, ion implantation and photoresist laydown  Ultraviolet radiation  '•Negative photoresist •(bonds crosslink) Si02 " Silicon  V///////////A  Expose photoresist  IZZ3  1 I  Y////////////A Etching  Wafer probe Dice Package Test  1  V////////////A Remove resist with oxygen plasma or selective solution. Figure 1.4  Schematic diagram of microelectronic processing 12  The process under consideration in this thesis is etching. many  applications  including patterning  insulating materials,  of  Etching has  semiconductor, conductor  and  stripping the ingot, cleaning and polishing the  wafers.  1.5 E T C H I N G P R O C E S S E S Etching can be defined as the removal of solid material by chemical reactions or physical processes.  Chemical etching processes fall into  two  categories : wet etching and dry etching.  Wet etching is performed by  simple immersion in the etchant solution.  A s for any chemical reaction,  the rate depends on the  temperature  control is critical for attaining dominant  of the reactants  optimum rates.  and so  temperature  Wet etching has been the  etching technique in silicon technology but undercutting  of  the  mask can occur.  Ideally, an etch should be anisotropic (fig.  1.5a), which  produces vertical  walled  which  etch figures, rather  restricts the resolution of features (fig.  1.5b).  than isotropic, A  1% solution of bromine in  methanol is a common etchant for G a A s and because of the  selective  etching of certain crystal faces some degree of anisotropy can be Dry  attained.  etching encompasses plasma etching, ion etching, laser assisted  etching and molecular gas phase etching.  In ion etching processes, such  as sputtering and reactive ion etching (RIE), ion bombardment major role in the etching.  plays a  Laser assisted etching can be performed in the  liquid or gas phase and involves photochemical etching.  The  irradiation  of a semiconductor with photons that can excite electrons above the band gap causes the production of electron-hole pairs and this may open an alternative photochemical reaction reaction. 1 9  2 0  13  pathway  to that of the  thermal  (b)  Figure 1.5  Profiles of (a) anisotropic and (b) isotropic etches 14  Plasma etching uses molecular gases, usually containing one or more halogen atoms.  The molecules are dissociated in a radio frequency  discharge at pressures between 0.01  and 10 Torr, to produce a partially  ionised gas composed of ions, electrons, molecules and free radicals.  21  The temperature in plasmas, as measured by the kinetic energy of the gas molecules and ions is only 5 0 - 1 0 0 ° C and so a plasma can be considered as an "ensemble of highly reactive particles in a relatively cool  medium".  22  After the collision of the etchant with the semiconductor  surface, these species react and the products desorb and subsequently diffuse in the etch chamber. planar reactors.  Plasma etching is carried out in barrel or  Barrel type plasma etching is already used in LSI  processing (fig. 1.6a).  Wafers are loaded into a barrel shaped quartz  chamber and rf power is applied to electrodes attached outside the chamber.  In a planar reactor, the wafers are placed at one of the  electrodes and ion bombardment results in the etch rate being greater on the axis between the electrodes (fig. 1.6b).  So the anisotropy achieved in  planar reactors is higher than in barrel reactors but there is also a lack of homogeneity in the etching and poorer reproducibility. offers a number of advantages over wet etching.  Plasma etching  Plasma processes can  be highly automated as reactant or effluent can be monitored by spectrometric methods to determine the degree of etching.  The highly  anisotropic etches that can be achieved by the ion bombardment processes in planar reactors, facilitates the formation of sub-micron sized features. In plasma and ion etching processes, the active species have usually been assumed to be the atomic and ionic particles.  However, previous  studies in this lab have shown that molecular etching often makes a significant contribution to the overall process. 15  23  Gas phase chemical  Vacuum chamber  Gas flow controller  Electrodes  Rf power supply] '//////////////<  Wafer  Exhaust system (b)  Figure 1.6 Plasma etching reactors (a) barrel 1 6  (b) planar  etching is a relatively simple process involving the flow of a gas (often a halogen) at various temperatures and pressures, over a wafer. reactive species are the gaseous molecules.  The only  In contrast, plasma and ion  etching often involve several reactive species, making the study of individual reactions difficult.  Therefore, little is known about the  mechanisms involved in etching in such systems.  1.6 GAS-SURFACE INTERACTIONS Gas surface reactions can be classified into four types : i) Heterogeneous  catalysis  ii) Oxidation iii) Chemical vapour deposition (CVD) iv) Etching Chemical etching shares many of the fundamental characteristics of heterogeneous catalysis, apart from the actual reaction step.  In chemical  etching the substrate reacts with the species from the gaseous phase rather than catalysing some reaction between gas phase molecules, which is the situation in catalysis.  Therefore, most of the theories and models  developed for heterogeneous catalysis can be applied to chemical etching with the minimum of adaptation.  The other classes of gas-surface  reactions listed differ from catalysis and etching because deposition on the surface is involved rather than the formation of gaseous products. A l l gas-surface reactions involve some form of adsorption : 1.6.1  Adsorption  When a gaseous phase atom or molecule becomes bound to the surface of a solid it is said to be adsorbed.  Adsorption may be physical or  chemical adsorption, which are referred to by the shortened terms physisorption and chemisorption, respectively. 17  Physisorption is the  weaker  of the two interactions  between  and involves van der Waals forces  the surface and the adsorbed molecule.  The energy released is  of the order of the enthalpy of condensation, usually 30 kJ m o l " less.  A physisorbed molecule has a residence half-life  2 4  seconds at room temperature.  or  1  of about  The time of a collision between  10"  8  a molecule  and a surface can be considered to be of the order of a vibrational frequency  (10-  entrapment  of the molecule in a shallow potential  relatively  1 2  seconds) . 25  brief period.  Therefore, physical adsorption involves  the  energy well for a  Chemisorption involves chemical bonds  between  adsorbed molecules and surface atoms that are comparable in strength covalent bonds and so the heat of adsorption is usually between 500 kJ m o l  - 1  .  2 6  However chemisorption can be much weaker  and the distinction between made.  physi- and chemisorption is often  200  to  and  than this not easily  Chemisorption only occurs in the first monolayer of adsorbate and  once the surface is saturated, physical adsorption takes place on top of the  chemisorbed When  monolayer.  gaseous molecules are adsorbed they  non-dissociative adsorption is often  thought  may  dissociate  although  to precede the dissociation.  Dissociative adsorption is seen in many catalytic processes.  For example,  hydrogen is adsorbed in the form of atoms on many metal surfaces and methane fragments  adsorbs dissociatively on some metals to produce CH3 and H  atoms . 27  If  adsorption is dissociative, an  or C H  2  activation  energy for the process can be observed, which is attributable to the bond length of the molecule increasing in order for chemisorption to occur. the bonds form between  the surface atoms and the adsorbed species, the  heat of adsorption is evolved and the potential minimum  As  when these bonds are at full strength.  adsorption an energy barrier  energy curve reaches a For non-dissociative  may still have to be overcome even 1 8  though  the intramolecular  bonds are not broken.  This barrier  can arise from  the  reconstruction of the surface to accommodate the incoming molecules. T y p i c a l energy profiles for adsorption are shown in fig.  1.7.  28  These are schematic diagrams and the actual potential energy curve w i l l be a composite of the two curves shown in each diagram. If 1.7a)  the activation energy for chemisorption ( E )  is very slight (fig.  c  the potential energy of the system may not rise above that of the  free molecule and this results in rapid chemisorption. may be large (fig.  1.7b).  Alternatively,  Ec  Therefore, to achieve chemisorption the  potential energy of the system must surpass that of the free molecule and so the process will be much slower. Generally, adsorption is found to be an exothermic process, as illustrated in fig. 1.7. entropy  However, endothermic adsorption is possible if  of the system increases during the adsorption process.  feasible if a molecule A2  motions'.  motions' and the two  movement,  Two  models have been most widely : the Langmuir-Hinshelwood  Rideal mechanism (1.6.2b).  mechanism (1.6.2a) and the  Eley-  These will be discussed only as they apply to  Langmuir-Hinshelwood  mechanism  This mechanism comprises four fundamental  (2)a)  'translational  used to explain gas-surface  processes.  1.6.2a  (1)  atoms each  2 9  Models for etching reactions  etching  adsorbed A  which is a total of four  1.6.2  reactions  This is  is adsorbed as two atoms since the free molecule  has only three 'translational have two-dimensional  the  steps :  Transport of reactant molecules to surface.  Non-dissociative adsorption of gas phase  19  reactants.  D = Dissociation energy of diatomic molecule He = Heat of chemisorption Hp = Heat of physisorption Ec = Activation energy for chemisorption  Figure 1.7  Potential energy profiles for adsorption with (a) zero Ec and (b) large Ec  20  and/or b) Dissociative adsorption of this adsorbed gas. (3) adsorbed (4)  Reaction between  adsorbed atoms and solid surface to  form  products. Desorption of the product m o l e c u l e s .  30  Diffusion can be a limiting factor for porous catalysts such as zeolites or in higher pressure systems than the ones used in classical investigations.  Langmuir  on tungsten filaments activation energy.  used low  pressure systems to study  and the process was found to have a measurable  A s diffusion in the gas phase has no activation  it was concluded that diffusion was not the rate limiting pressure  systems.  this model not all the reactant  adsorbed molecules. two  molecules are strongly adsorbed to occur between  molecules : a strongly adsorbed molecule,  usually dissociatively adsorbed, and a molecule either  degree  adsorbed layer.  of mobility  strongly  Therefore reaction is proposed to take place  types of reactant  or from a weakly  step in low  mechanism  since reaction was considered unlikely  between  energy,  3 1  1.6.2b E l e y - R i d e a l In  adsorption  of the reactants  from the gas phase  This model incorporates a higher than the  Langmuir-Hinshelwood  mechanism and has been used to account for the diffusion of hydrogen in various  metals.  32  1.7 P R E V I O U S S T U D I E S O N G A L L I U M A R S E N I D E E T C H I N G In  III-V semiconductor etching, the volatility  be rate l i m i t i n g .  33  Group III  of the products can  fluorides are involatile  at temperatures  than 3 0 0 ° C and so fluorine or fluorine generating plasmas problems. this  3 4  However,  temperature  create  corresponding chlorides are more volatile  and there are  numerous 21  studies of  less  below  chlorine-containing  plasma etching for these semiconductors.  Group III and V bromides are  also volatile but few studies have reported using bromine or brominegenerating plasmas.  Iodine compounds are involatile.  Apart from the halogens, hydrogen and hydrogen halides are the most commonly used gases in plasma etching studies.  35  Also a variety of  other etching techniques, such as reactive ion etching (RIE), laser assisted etching and molecular beam assisted etching have been applied to GaAs. Only those studies using pure halogen gases, in either atomic or molecular form, will be discussed in the following review of the literature. Ibbotson et al  36  reported crystallographic etching of GaAs (100)  with bromine and chlorine plasmas.  The apparatus used was a parallel  plate, radial flow reactor and concentrations of Br and Br2 were measured by absorption spectroscopy.  Reactions were carried out between 0.1 and  0.3 Torr, at a power of 0.5 W cm" and at a temperature of 100°C. Etch 2  rates were measured from the etch depth or by monitoring the emission from elemental gallium and by etching through samples completely.  It  was found that the dominant reactants were Br atoms and ion bombardment did not significantly influence the etching rate.  This  conclusion was based on the observation that the reaction probability of bromine atoms was independent of discharge frequency.  If ion  bombardment enhanced the gasification reaction, this probability would increase as ion energy increased.  Etching was found to be  crystallographic, which is unusual in low temperature, low pressure plasmas.  The relative etch rates for the planes were found to follow this  order: {111}B > {100} > {110} >{111}A. Etch rates were directly proportional to Br atom concentration and ranged from 20 to 70 u,m min* . 1  With pure molecular bromine and no plasma, a much lower etch  rate of 2 u.m min- (at 0.3 Torr, 100°C) was measured although the etched 1  surface was noted to be very smooth.  In a related study by the same  workers the apparent activation energy for GaAs etching in a chlorine plasma was found to be 43.9 kJ m o l . -1  37  They proposed a general  reaction scheme for III-V semiconductor etching in a chlorine plasma : CI + GaAs(surface) xCl-GaAs(surf)  7  —>  x  GaCl (ads) y  (1)  > AsCl -GaAs(surf)  (2)  > GaCl -GaAs(surf)  (3)  x  yCl-GaAs(surf) AsCl (ads)  Cl-GaAs(surf)  y  *  AsCl (g)  (4)  x  > GaCl (g)  (5)  y  So at steady state, chlorine atoms adsorb on the substrate surface and react with the substrate, generating gallium and arsenic chlorides, of unknown formula.  The observed etch rate for GaAs in a chlorine plasma  was many orders of magnitude less than the calculated evaporation rate of gallium trichloride (20,000 n.m min  -1  at 70°C).  This volatilisation was  estimated from the escape of Ga2Cl6 molecules from a complex dimeric liquid.  Assuming that the vaporisation of this dimer from a GaCl3(l) layer  on the GaAs surface occurs at a similar rate they excluded vaporisation as a possible rate-limiting step.  If GaCl3 is strongly adsorbed the formation  of this dimer could be rate-limiting : 2GaCl (ads)  > Ga Cl  Ga Cl (ads)  > Ga Cl (g)  3  2  6  2  2  6  (ads)  6  The desorption of the dimer was considered to be fast.  They also  proposed that the reaction of the etchant species with the GaAs surface (reactions (2) and (3) in the scheme above) may be relatively slow and so rate determining. In the case of indium phosphide, they proposed that step (5) above was rate limiting.  The activation energy for etching was very similar to  the heat of sublimation of InCl3 and the absolute etch rate was consistent  with the thermodynamically calculated InCl3(s) evaporation rate (45 u,m min-1 at 250°C). Ha  3 8  '  3 9  studied both molecular chlorine etching and downstream  atomic chlorine of GaAs {100}.  In the downstream system, the gas flows  through a microwave discharge and the resulting mixture of atoms and undissociated molecules then flow over the sample.  This is different than  the conventional plasma etching system wherein the sample is located on or between the electrodes.  Downstream etching eliminates etching due to  bombardment by accelerated  ions.  The etching by molecular chlorine was found to be first order at low pressures of chlorine.  A n activation energy of 100.3 kJ mol" was 1  measured by determining the etch rate over the temperature range from 25-160°C.  At pressures above 15 Torr the reaction appeared to reach a  limiting rate with an activation energy of 57.3 kJ mol" .  Atomic chlorine  1  etching showed a first order dependence on the partial pressure of chlorine atoms and an activation energy of 37.6 kJ mol" . 1  molecules displayed crystallographic etching. and  Both atoms and  Reaction products, GaCl3  ASCI3 were detected mass spectrometrically and provided one  method of measuring the etch rate; profilometry and laser interferometry being the other methods used.  A mechanism was proposed to account for  the observed pressure dependencies.  The first step involved rapid  chlorination of the surface Ga and As atoms to form surface bound monoor dihalides : Cl (g) + Ga-As(s) —> 2  The  GaCl (surf) + AsCl (surf) x  x  (1)  second step, rate controlling at high pressure, produced readily  vaporised ASCI3 and surface bound GaCl3 : GaCl (surf) + AsCl (surf) —> x  x  24  AsCl (g) + GaCl (ads) 3  3  (2)  The rate limitation  encountered at high pressures was shown not to  be caused by the build-up of a GaCl3 layer.  Since higher etch rates could  be attained when atomic CI was the etchant, they considered that the reaction was not limited by any process occurring after the products were formed.  The low pressure rate-limiting  reaction was thought to be  the formation at reaction sites of products, such as G a C l g .  The etch rate  2  was believed to reach the maximum  once these sites were saturated, that  is at high pressures. A  study by Furuhata et a l  4 0  considered C l  G a A s under high vacuum conditions ( C l x IO  -3  Torr.)  plane.  2  2  chemical dry etching of  pressure between 8 x I O  They found three distinct reaction regimes for the  Below 1 5 0 ° C , etch rates showed a 0.6  and 1  -5  {100}  order dependency on the  pressure of chlorine and the activation energy for the process was calculated to be 42 kJ m o H . be little temperature 67 kJ m o l "  1  Between 150 and 4 5 0 ° C , there appeared to  dependence.  was measured.  A b o v e 4 5 0 ° C , an activation energy  They attributed  these observations to the  rate being limited by the desorption of various G a C l products are known to be more volatile).  of  x  products (as A s C l  x  Below 1 5 0 ° C they suggest the  etch rate is limited by GaCl3 desorption and that above 4 5 0 ° C the rate is limited by G a C l desorption.  T h e relative etch rates of G a A s planes was  found to differ  slightly from that proposed by Ibbotson et al in that the  rates for  and {110}  {100}  Balooch and O l a n d e r molecular  were found to be equivalent. 41  studied the reaction of {100}  chlorine, with and without  energetic argon ions.  G a A s with  simultaneous bombardment  They proposed that at 2 7 ° C ,  ion  bombardment  accelerated the etch rate by a factor of six and attributed the rapid removal of a scale of relatively non volatile G a C l reacting  surface.  Without  ion bombardment, 25  the low  by  this mainly to 3  from  thermal  the  desorption  rate constant of GaCl3 molecules in the surface scale was thought to cause accumulation on the GaAs substrate and the limiting of the reaction rate. The dimerisation step proposed by Ibbotson and Ha was not mentioned. Ibbotson et al observed that at a temperature of 2 5 ° C ,  samples  being etched in a bromine plasma were totally coated by a liquid layer, believed to be a gallium bromide reaction product and etching was negligible.  Ha reported the formation of a liquid layer at temperatures  between 25 and 7 0 ° C and high pressures of chlorine (over 15 Torr), which resulted in an unusual etch profile : etch depth was greatest at the sides of the profile. Morphology of the etch was found to depend on the temperature: Ibbotson et al reported a rough to smooth transition at 125°C in a chlorine plasma whereas Ha found that shiny surfaces were produced at 160°C and above for molecular chlorine etching.  1.8 PURPOSE OF STUDY The objective of the study described in this thesis was a determination of the pressure and temperature dependencies etching of GaAs {100}  of the  by molecular and atomic bromine.  Since molecular etching effects are often neglected in ion and plasma etching studies, a secondary objective was to determine whether there is a significant contribution by molecular bromine to such etching processes.  26  2 EXPERIMENTAL  2.1 M A T E R I A L S The bromine used in these studies was A . C . S . reagent (Aldrich) with a quoted purity of 99.5%.  grade  T o fill a reservoir, it was first  evacuated, the inlet was immersed in bromine and the tap was opened. In order to prevent contamination by water, calcium oxide was used as an in situ drying agent.  The bromine was then frozen using an acetone  and solid carbon dioxide bath and the air was pumped off. thawing twice  with a heating  gun, this freeze-pump-thaw  After  cycle was  repeated  more. The gallium arsenide samples were cut from 5 c m diameter  wafers.  G a A s {100}  was the subject of the study since this plane is the  most widely used in industry.  Both undoped and silicon doped wafers  were used and were supplied by B e l l Northern Research. samples had a doping concentration of 2 x 1 0  1 6  T h e Si doped  atom c m - . Sample chips 3  were generally rectangular and had a surface area of 0.1 Some G a A s samples were masked with silicon nitride  to 0.2  (S13N4)  cm . 2  stripes, by  the U . B . C . Centre for Advanced Technology in Microelectronics. stripes were 20  {100}  to 50  micrometres wide,  and aligned with the [TlO] considered as the (100)  or [IlO]  plane.  27  micrometres  apart  directions, if the surface is  Wafer  mm.  spaced 100  The  thicknesses were 0.25  and 0.5  2.2 M O L E C U L A R B R O M I N E E T C H I N G 2.2.1  Apparatus for  molecular  etching  The reaction system used (shown in schematic form in fig. was constructed mainly of Pyrex glass. inner diameter,  1.8  c m outer diameter  connecting tubing was 1.0 c m I.D.  2.1)  The reactor was made of 1.5  cm  tubing and was 18 cm long.  and 1.3 c m O . D . .  Most  Connections between  sections of glassware were mostly made using A c e glass O-ring joints 1.5 cm I.D..  Cajon Ultratorr  O-ring joints were used for some glass-glass  connections and all metal-glass connections.  The bromine reservoir  connected to the system via a Matheson 602, with glass float. the meter  The flowmeter  volume (V)  15 c m tube flowmeter,  in standard cubic centimetres  per  The calibration was carried out by first calculating the  of the system from the dimensions of the components.  was found to equal 150 c m . 3  flow  was  was calibrated so that each reading on  corresponded to a flow  minute (seem).  of  This  Next, bromine was admitted at a certain  reading and the pressure increase in the system was measured  connecting the pressure gauge (Edwards  by  H i g h V a c u u m , 0-1000 Ton-  range) to a chart recorder (Hewlett Packard 6 8 0 M strip chart The time for the pressure increase (dP/dt) was then related  recorder).  to the  flow  of gas (dn/dt), using the ideal gas law:  dn dt  =  dP  V  dt  RT  Flow rates were in the range of 10 to 70 seem.  Flow was adjusted  so that the supply of gas exceeded that consumed in the etching reaction, by a factor of thirty or more. A 1400)  Sargent-Welch  two-stage  Duo-seal rotary  with a pumping speed of 25 litres m i n  - 1  vacuum pump  was used in  (model  conjunction  with a liquid nitrogen cold trap, which functioned as a cryogenic pump. 28  Sample platform Resistive heater  L  R  e  a  c  t  o  r  Temperature  Flowmeter To rotary pump  t  Pressure gauge  Bromine reservoir  Liquid nitrogen cold trap  Figure 2.1  Schematic diagram of molecular bromine etching apparatus  29  The vacuum pump was connected to the glassware with 3cm O . D . heavy wall rubber  tubing.  The samples were held by a glass spring onto a piece of oxidised silicon, which acted as a heat sink (fig. 2.2).  thermally  The silicon was  mounted on a flattened portion of a thin walled glass tube, inside which was placed a chromel/alumel thermocouple. holder design proved unsuccessful.  A n alternative  This involved the  sample  thermocouple  being attached directly to the silicon platform with Torr Seal adhesive, which is a low vapour pressure resin suitable for vacuum systems. arrangement  would initially  temperature bromine. platform  reading.  appear to give a more  However,  Furthermore,  accurate  the resin was quickly corroded by  temperature  readings from  the  agreed closely with those measured in the  2.2.2  This  new  sample  alternative  set-up.  Procedure for molecular etching  GaAs  samples were immersed in concentrated hydrochloric acid  for thirty seconds to remove any thick oxide layer and then washed with deionised water.  Samples were then mounted on the  platform,  under a stream of nitrogen, and then the sample holder was placed in the reactor.  The system was evacuated to approximately  10  milliTorr  and a resistive heater was used to heat the reactor to the required temperature.  Experiments  25 and 2 2 5 ° C . steady flow  out  at  temperatures  between  Next, bromine was flowed through the system at a  and pressure and the time of exposure was recorded.  was usually 1 to 3 minutes. adjusted with either cryogenic  were carried  This  T h e flow and pressure of bromine could be  the flowmeter  valve or a valve before  the  pump.  The etch rate was measured in one of two ways.  For low etch  rates a masked sample was used and the etch depth was measured by 30  Gallium arsenide sample  Figure 2.2  Sample platform  3 1  profilometry.  The profilometer (Tencor alpha-step 200) scans along the  surface with a very fine needle tip.  The silicon nitride mask was  aligned with the [110] or [llO] direction, with respect to the (100) plane.  Due to the different reactivity of the various planes of the GaAs  structure, the mask can produce two etch profiles.  As stated in section  1.2, when the mask is aligned with the [110] direction, the faster etching As (111) planes are exposed on the etching of the (100) plane and a reverse mesa shape is produced: the profile has outward sloping walls.  Whereas when the mask is aligned with the [TlO] direction, the  slower etching Ga (111) planes become exposed and cause the formation of V-shaped grooves.  This crystallographic etching causes problems  when trying to measure the etch rate of the (100) face.  If reverse mesa  formation occurs, the undercutting removes the mask once a limit of (ytan6)/2 has been passed (fig. 2.3), where y = distance between stripes and 0 = angle between (100) and (111) planes, 5 4 . 7 ° .  Similarly, when  V-shaped grooves form, the flat bottom of the profile (corresponding to the (100) plane) disappears when this limit is reached.  Therefore a new  method of etch rate determination was required to enable  measurement  of larger etch depths. As etch rates were so high, relating the weight loss of the samples to the etch rate of the {100}  plane was found to be an accurate method.  The contribution to weight loss from the edges of the wafer was experimentally determined by sandwiching two wafers etching them at previously used conditions.  together and  The weight loss of the  bottom wafer was due to etching from the edges only.  For a sample of  weight 0.0281 g, the weight loss from the edges was 0.0023 g. expected weight loss from the {100}  The  surface was 0.0047 g (calculated  from the etch rate at these conditions determined by profilometry). 32  Mask  x = (y tan 0) / 2 x = 0.71 y  Figure 2.3  Geometry of V-shaped groove formation  33  Therefore, the proportion of the apparent etch rate of the {100}  plane  due to the loss from the edges :  0 0023 % o f total w e i g h t loss, f r o m edges = 6  u  ,  u  u  J  x 100%  0.0070 = 33%  This is within experimental error of the 40% expected from geometric calculations.  These results showed that the weight loss from  the edge planes could be compensated for by geometric calculations, as the etch rate of the edge plane was comparable to the etch rate of the top surface.  The weight loss due to etching of the edges is  proportionately  less for larger samples and so correction factors  calculated for a range of sample sizes.  were  These varied between 0.55  0.7 for the 0.5 mm thick wafer and was about 0.8 for the 0.25  and  m m thick  samples. Another method of etch rate determination that can be used for etch depths less than that specified above, is laser interferometry. involves the detection of a laser beam reflected from a partially surface.  This masked  A s the wafer is etched the light reflected from the etched  surface interferes  with that reflected from the masked area.  When  the  path difference is a multiple of half the wavelength of the laser light, the  light  waves interfere  is at a maximum.  constructively and the photodetector response  However, in our studies it was found that the  intensity of the reflected beam fell sharply as soon as the wafer  was  exposed to bromine and so the etch rate could not be monitored by this technique.  This may have been due to (a) the shape of the etch  trenches (b) the mask being too widely spaced or (c) the rapid formation of a poorly reflecting surface.  34  2.3 A T O M I C E T C H I N G 2.3.1  Apparatus for atomic  etching  The reactor used in these experiments  had a discharge region  constructed of 1/2 inch O . D . quartz tubing, to withstand the high temperature  of the discharge (fig. 2.4).  discharge area to the Pyrex glass tubing.  Graded seals connected the The distance from the  discharge region to the sample platform was 13 c m . be short enough to prevent  This was found to  significant atom recombination but at  the  same time not cause overheating of the sample by heat generated  by  the discharge.  A light trap was also incorporated between the sample  platform and discharge so that light from the discharge was not detected in any attempt to monitor the glow due to bromine combination. between A  A nitrogen gas inlet (Linde nitrogen)  the flow  meter  atom  was positioned  and the discharge region.  Microtron 200 Watt, 2450 M H z  microwave generator was used  and the output was directed into the discharge region by an 1/4  wave coaxial cavity.  to cool the discharge area. handheld Tesla c o i l .  adjustable  Compressed air was blown through the cavity The discharge was triggered using a  The main advantages of a microwave discharge  over a radio frequency discharge are that it is a very intense localised discharge with no tendency to spread and can be used at pressures of up to several T o r r .  4 2  The reactor was treated with phosphoric acid to 'poison' the and so reduce the degree of recombination of bromine a t o m s . the reactor was washed with a 5% solution of H3PO4 and then and left to stand for one hour.  43  walls  First drained  This produces a mildly acidic surface.  35  Joint enabling positioning of ied coil Isothermal calorimetric detector  Compressed air  Wheatstone bridge  Temperature  Quartz  Discharge cavity  To liquid nitrogen trap and rotary pump.  Microwave generator  Figure 2.4  Schematic diagram of atomic etching apparatus  36  H3PO4 and  Then the discharge area was coated with concentrated  heated  in a flame to produce a white coating of phosphorus pentoxide. The absolute concentration of bromine atoms was measured with an isothermal calorimetric p r o b e .  4 4  The probe usually consists of a  platinum coil which is maintained at a given temperature current through it.  by passing a  The probe is placed in the gas flow as near as  possible to the sample platform.  When the discharge is in operation  atoms recombine on the coil and so heat it.  The electrical power  supplied must be reduced to maintain it at the original A platinum coil (99.99% purity, 0.25  temperature.  m m diameter and 15 c m  long) was spotwelded to two stainless steel rods (fig. 2.5).  The coil was  nickel-plated electrolytically as this has been found to produce the most efficient  surface for bromine atom recombination.  The rods were  wrapped in teflon tape to prevent reaction of bromine with the metal and then set in a Pyrex tube, using Torr Seal for the vacuum seal.  To  minimise contamination by the resin it was only used at the end of the tube, 10 c m away from the coil.  The probe was held in place by an O -  ring arrangement and its position in the reactor was  adjustable.  The resistance of the coil and the current passing through it were measured using a Wheatstone bridge set-up (fig. 2.6). was balanced with bromine flowing over the coil.  The current  maintained the coil at a temperature of about 1 3 5 ° C . the  standard resistor was measured to determine  through the coil (ii)  First the bridge  The voltage across  the  and the resistance of the probe (R)  measuring the resistance of the variable resistor.  current passing was found by  Then the discharge  was started and the bridge was balanced again, by reducing the current. This reduced current value (i ) 2  is then used to calculate the atom flow :  37  Figure 2.5  Isothermal calorimetric detector  38  DETECTOR  Figure 2.6  Wheatstone bridge arrangement for isothermal calorimetric detector  A t o m flow =  (li.lolR  %  - i  — g atom s  where D = dissociation energy of bromine = 193.9 kJ mol"  1  4  5  The complete removal of atoms by the detector cannot be guaranteed unless two detectors are used in series. was  operated  at the  optimum temperature  However, the coil  (determined  by running  the  coil over a range of temperatures) and was coiled so as to occupy as much of the reactor's cross-section as possible. concentration which we  Furthermore, the  obtained compared well with concentrations  measured by chemical titration of bromine atoms with N O C 1 in similar systems.  4 6  2.3.2  Procedure for atomic etching  The measurement of atom flow could not be carried out during an etching experiment minutes.  as balancing the Wheatstone  Therefore, the atom flow  experiments  bridge took  several  was measured before and after  and an average value was  taken.  The glow emitted by the recombination of bromine atoms was considered as a way of monitoring the atom flow during each experiment.  The response of a photodetector to the glow could be  calibrated using the isothermal calorimetric probe.  However,  the  cadmium sulphide detector proved too insensitive to measure the It  was important  to avoid contaminating the system with water  since the phosphoric acid and phosphorus pentoxide would absorb water.  glow.  readily  This would be released again when the reactor is heated  and the G a A s surface would be oxidised.  This phenomenon was  observed with the aid of a microscope set-up over the reactor. Interference fringes were seen as an oxide layer formed when  40  the  discharge was started.  The temperature  was raised and when  it  reached a certain point ( > 1 0 0 ° C ) the bromine atoms penetrated parts of the oxide layer and etching of the wafer could clearly be seen. Therefore, the reactor was only opened to the atmosphere after a positive pressure of nitrogen had been attained.  The sample could then  be loaded (after immersion in concentrated hydrochloric acid and rinsing with deionised water) as nitrogen flowed through the reactor. When  the  system had reached the required temperature,  bromine  was  admitted at maximum flow, in order to maximise the number of atoms reaching the sample.  The sample was exposed to molecules for up to  half a minute before the discharge was started.  41  3  RESULTS  3.1 M O L E C U L A R B R O M I N E E T C H R A T E D A T A E t c h rates were measured at three temperatures and 1 8 3 ° C (tables 3.1, 3.2, 3.3). in fig. 3.1.  : 125°C,  145°C  The data for the 125 ° C series is plotted  A t temperatures below 1 0 0 ° C , etching was very slow and  above 2 0 0 ° C etch rates were too high and the possibility of insufficient reactant supply was encountered.  In  order to avoid the rate being  limited by such factors, the ratio of bromine supplied to that required for the observed etch rate was kept as high as possible.  This was done  by maintaining maximum flow rates to keep this ratio thirty to one or greater. Therefore, because of the above limitations temperature  range with this particular  the usable  apparatus was normally  125°C  to 1 8 3 ° C , although temperatures as high as 2 2 5 ° C could be used if the pressure was kept very low at the same time.  Low  temperature  experiments were run as long as possible to give a reasonable etch depth. The pressure of bromine in the system could be varied from Torr to 20 Torr.  The lower end of the range was determined by  instrumentation limits : 0.1  Torr was the lowest pressure for which a  consistent measurement could be maintained. of insufficient bromine flow recur. temperature  A b o v e 20 Torr, problems  H i g h pressure combined with high  conditions resulted in the complete etching away  thinnest G a A s  0.1  of  samples, when measurable etch times were used.  the  Table 3.1  List of etch rates of GaAs (100) with Br at 125°C 2  Etch rates from profilometry Pressure (Torr) 0.1 0.2 0.3 0.4 0.4 0.5 0.5 0.6 0.7 0.8 0.8 0.9 1.0 1.1 1.2 1.4 2.0 2.6 2.8 3.3 Etch rates  Etch rate (\im min" ) 0.8 1.7 1.9 2.8 2.9 4.2 3.4 5.6 5.4 7.1 7.1 5.1 8.8 9.1 11.1 13.0 17.9 20.5 17.1 21.4 1  m weighing method 1.0 4.4 5.12 7.0 9.6 11.0 17.3 26.0  10.8 29.3 30.6 40.0 40.9 36.1 57.1 60.4  43  Table 3.2  List of etch rates of GaAs (100) with B r at 145°C 2  Pressure (Torr) 0.2 0.5 0.8 I. 0 2.3 2.3 2.3 5.3 7.0 II. 0 15.0  Table 3.3  Etch rate (u.m min- ) 2.6 7.2 13.8 15.1 16.5 21.6 20.0 36.7 41.0 51.0 51.5 1  List of etch rates of GaAs (100) with B r at 183°C 2  Pressure  (Torr)  Etch rate (jim min" ) 25.8 37.8 34.1 58.9 55.7 43.9 49.8 53.5 62.3 71.5 76.0 72.2 62.4 89.9 81.4 77.4 89.0 1  0.83 1.0 1.0 2.5 2.9 3.3 5.7 7.0 10.0 10.4 12.2 14.4 16.4 18.0 18.2 19.0 19.6  44  10  20  Pressure of bromine (Torr)  Figure 3.1  Graph of etch rate against pressure of Br at 125°C 2  Profilometry  was used for etch depth determination  at pressures below  4 Torr, for the 1 2 5 ° C series and a typical etch profile is shown in fig. 3.2a.  F o r higher etch rates, that is those at elevated temperatures  and  pressures, the use of masked samples was no longer feasible because of the limitation imposed by the crystallographic nature of the etch. creates a V-shaped profile (fig. 3.2b)  This  or a reverse mesa, both of which  cause complete removal of the mask once a certain etch depth has been passed. A l l other etch rates were measured by weight loss.  Compensation  for weight loss from the edges of samples was made by calculating surface areas and then applying a correction factor to the apparent etch rate.  The etch rates determined by this technique compared well with  those measured with a  profilometer.  T h e morphology of the etch was not observed to change from rough to smooth at any specific temperature. generally  gave  However, high etch rates  smoother etches.  3.2 E R R O R A N A L Y S I S The etch rates determined were weighted reciprocal of ten percent of the etch rate.  when plotted, with  the  In this way the high etch rate  data, which shows greater scatter, takes on a lower weighting.  The  figure of ten percent was arrived at by repeating several etching experiments under the same conditions. difference in the measured etch rate.  This is due to errors of about ± 2 %  in the pressure measurements, ± 2 % in ±5%  These showed 5 to 10 %  temperature  measurement  and  in the weighing technique. A  linear least squares analysis was performed on all etch rate-  pressure data with errors accounted for in both the x and y v a l u e s .  47  Figure 3.2  T y p i c a l etch profiles obtained from  47 _  profilometry  The errors in the gradients obtained in these graphs were then used as error bars in the subsequent Arrhenius plots.  T h i s provides an  estimation of the errors (with 95%  confidence) in the activation energies  and pre-exponentials  the  In  derived  from  graphs.  order to account for the possibility of irreproducibility  caused by defects in the wafer simultaneously.  samples, two  samples were etched  The difference in measured etch rates was 8%,  an overall error of ± 4 % , which is within the experimental the weight loss technique.  being  giving  uncertainty  A l s o the etch rates of the silicon doped G a A s  samples, which have a dislocation density an order of magnitude than undoped sample, were no different rates.  of  than the undoped G a A s  higher etch  This seems to confirm that surface defects do not play a  significant role in determining  bulk etch rates.  3.3 K I N E T I C A N A L Y S I S The order of a reaction with respect to a particular gas phase reactant is the power to which the partial pressure of that reactant is raised in the rate law.  Order can be determined by plotting  natural  logarithm of the rate against natural log of the pressure of that reactant. For example, if the rate law is given by :  then, In (rate) = In k +n In P  B r 2  Therefore, the gradient of the line obtained by plotting In (etch rate) against In (pressure of B r ) is equal to the order of the reaction 2  with respect to B r . 2  Such plots are shown in figs 3.3, 3.4 and 3.5 for  1 2 5 ° C , 1 4 5 ° C and 183 ° C , respectively.  48  1 0 1 In pressure (Torr) Gradients of lines are 1 and 0.5, approximately  Figure 3.3  Graph of In etch rate against In pressure of Br at 125°C 2  c S £ o  O  -2  -1  0 1 In pressure (Torr)  Gradients of lines are 1 and 0.5, approximately  Figure 3.4  Graph of In etch rate against In pressure of Br at 145°C 2  50  5  1  El  4 -  H  B B  _  B B  %  •  1  1  1  a  Q  H  BP %  B  H  •  3  2  -  •  1  1.0  Figure 3.5  0.0  i  2.0 1.0 In pressure (Torr)  <  3.0  Graph of In etch rate against In pressure of Br at 183°C 2  These graphs show two distinct regions. there  appears to be a transition  A t a pressure of 1 to 2 Ton-  between  two pressure dependencies.  these two regions are plotted separatedly (figs 3.6, 3.7),  If  it can be seen  that the low pressure region has an order of 1 and the high pressure region has an order of 0.5.  Gradients for the three temperature  are presented in tables 3.4 and 3.5.  series  The anomalously high value for the  low presssure, 1 8 3 ° C readings can be attributed  to the gradient  being  based on only three points. A  graph of etch rate against the square root of pressure should  give a straight line for a half order reaction.  Figure 3.8 shows such a  graph for the high pressure region of the 1 2 5 ° C data.  The data for the  1 2 5 ° C and 1 4 5 ° C series give small negative intercepts but that for the 1 8 3 ° C shows a small positive  intercept.  Therefore we can represent the data by two empirical equations : A t pressures below 1 to 2 Torr of bromine, where a is a first order rate constant :  Etch rate = a ( P  B r 2  )  (3.1)  A t pressures in excess of 1-2 Torr, where b is a half order rate constant and c is the negative intercept of the line drawn through the high pressure points on the half order graph :  1/  Etch rate = b ( P  3.3.1  B r 2  )  / 2  - c  (3.2)  L o w pressure region  B y plotting etch rate against pressure of bromine we can determine  the rate constant in u.m m i n 52  - 1  Torr , - 1  at each  temperature  -3  -2  -1 In pressure  Gradient =  Figure 3.6  (Torr)  1.04  Graph of In etch rate against In pressure of B r for the low pressure region at 1 2 5 ° C  53  1  0  2  0  1  2 In pressure (Torr)  Gradient = 0.46  Figure 3.7  Graph of In etch rate against In pressure of Br for the high pressure region at 125°C  54  2  Table 3.4  Gradients of In etch rate versus In pressure of Br2 for the low pressure region  Temperature ( ° C )  Gradient  125  1.04  145  1.12  183  1.49  Table 3.5  Gradients of In etch rate versus In pressure of B r for the high pressure region  Temperature ( ° C )  Gradient  125  0.46  145  0.58  183  0.47  2  1  Figure 3.8  2 Square root of pressure (Torr)  Gradient =  13.53  Intercept =  -2.3  Graph of etch rate against square root of pressure of B r for the high pressure region at 125°C  56  2  (Fig. 3.9, Table 3.6).  If a single reaction step is rate limiting, the etch  rate will exhibit an Arrhenius temperature dependency corresponding to this slow step.  From the Arrhenius equation :  k = Ae -E/RT In k = In A - E / R T where A = pre-exponential factor, E = activation energy and T = absolute temperature.  The rate constants for etching experiments are  most conveniently expressed in terms of the units used for measuring etch depth and pressure. against l / T .  A and E can be determined by plotting In k  Figure 3.10 shows the resultant graph and the values  obtained from this are 33.8 * 1.0 kJ m o l (1.03 ± 0.3) x 1 0  22  molecules c n r s 2  _1  -1  for the activation energy and  Torr-  1  for the pre-exponential  factor. Similar graphs can be constructed using (etch rate/pressure of bromine), which is equivalent to rate constant a in equation 3.1.  Two  such graphs were plotted for data at 0.83 Torr and 1.0 Torr (Tables 3.8, 3.9 ; figs 3.11, 3.12).  These graphs show close agreement and give an  average activation energy of 27.0 kJ m o l .  Data obtained from all these  -1  Arrhenius plots is seen in Table 3.7.  The average activation energy  calculated from these three sets of data is 29.2 ± 4.0 kJ mol-1 and the average pre-exponential is (3.4 ± 4.4) x 1 0  21  molecule s-  1  cm  - 2  Torr . -1  As these experiments were carried out at constant pressure, we actually derive an activation enthalpy from our data.  However, the  difference is only R T for a first order reaction and 1/2 R T for a half order reaction and this is within the experimental error.  57  0.2  0.4  0.6 0.8 Pressure (Torr)  1.0  Gradient = 9.26 ± l . i ^ m min-1 Torr-1  3.9  Graph of etch rate against pressure of B r for the low pressure region at 125°C 58  2  Table 3.6  Gradients of etch rate plotted against pressure of B r  2  for the low pressure region  Temperature  (°C)  Gradient (}im  - 1  125  9.26  145  16.36  183  34.56  Table 3.7  Pressure  min  Torr ) 1  Arrhenius data for the low pressure region  (Torr)  Activation  energy  (kJ m o l )  Pre-exponential (molecules  - 1  s  _ 1  factor cnr  2  0.83  25.4 * 2.6  (9.31  ± 9.7) x 10 <>  1.0  28.6 ± 4.0  (1.97  ± 4.4) x 10  2  1  0 to 1 Torr => k values f r o m gradients  33.8 ± 1.0  (0.94  ± 0.3) x 10  2  2  59  2  Torr* ) 1  0.U021  0.0022  0.0023  0.0024  1/Temperature  0.0025  0.0026  (K" ) 1  Gradient = -4060 ± 115 K Intercept = 12.4 + 0.3  Figure 3.10  Arrhenius plot for the low pressure rate constants  60  Table 3.8  List of etch rates for a B r  pressure of 0.83  2  Torr, at several  temperatures  Temperature  Table 3.9  Etch rate (nm  (°C)  117  9.4  129  10.0  151  12.3  171  23.6  202  37.3  231  47.1  List of etch rates for a B r  2  mirr ) 1  pressure of 1.0 Torr, at several  temperatures  Temperature  Etch rate (u.m min" ) 1  (°C)  112  5.45  122  11.0  125  8.8  127  10.8  142  15.1  166  19.9  193  30.0  61  c  JS  g  2  1  '  1  —-«  0.U019 0.0020  w  1  1  0.0021  «  -•  "  •  .  1  0.0022 0.0023 0.0024 1/Temperature (K" )  I 0.0025  1  ci  Gradient = -3052 ± 308 K Intercept = 9.94 ± 0.7  Figure 3.11  Arrhenius plot for etch rates at a B r of 0.83 Torr 62  2  pressure  %  0.0021  0.0022  0.0023 0.0024 0.0025 1/Temperature (K" )  0.0026  1  Gradient = -3443 ± 478 K Intercept = 10.9 ± 1.2  Figure 3.12  !  Arrhenius plot for etch rates at a B r of 1.0 Torr  2  pressure  3.3.2 High pressure region From the empirical equation 3.2, the constants b and c can be obtained by plotting etch rate against the square root of pressure (Table 3.10) .  The gradient b is a half order rate constant.  Assuming an  Arrhenius equation can be used to describe the temperature dependence then a plot of In b against 1/T (fig. 3.13) gives an activation energy (8.4 10  1 8  ±  0.65 kj mol* ) and pre-exponential factor ((6.4.± 1.3) x 1  molecules  cnr  2  s  _1  T o r r * / ) for the half order rate constant. 1  2  3.4 ATOMIC ETCHING Atomic etching was carried out at 0.83 Torr of bromine (Table 3.11) , a pressure at which a series of molecular etches had been carried out.  Therefore compensation for the etching due to bromine molecules  could readily be carried out.  The concentration of atoms obtained  varied between 12 and 15%.  Assuming that the atomic etching reaction  is first order, the rate constant can be calculated by first correcting for the etching by molecules and then dividing the atomic etch rate by the partial pressure of bromine atoms.  Figure 3.14 shows an Arrhenius plot  in which the In of the rate constant is plotted against the reciprocal of temperature.  This gives an activation energy of 12.9  a pre-exponential factor of (7.1 ± 2.0) x 1 0  20  atom s  ±  _1  0.9 kJ mol" and 1  cm  - 2  Torr . - 1  3.5 A N A L Y T I C A L RESULTS After several etching experiments,  a brown deposit appeared  downstream of the sample platform, on the periphery of the heated region of the reactor.  This was dissolved in acetone and analysed by  electron ionisation mass spectrometry.  64  The most prominent peaks were  Table 3.10 Gradients and intercepts of the graphs of etch rate versus (pressure of bromine) / , for the high pressure region 1  Temperature ( ° C )  2  Gradient  Intercept  b  c  125  13.53  -2.3  145  15.03  -5.7  183  18.99  4.00  Table 3.11  List of etch rates for GaAs (100) with Br, at a Bt2 pressure of 0.83 Torr  Partial pressure of  Temperature ( ° C )  T». * /T- \ Br atoms (Torr)  Etch rate (corrected) (um min ) ' -1  r  0.12  89  30.2  0.15  102  43.0  0.10  178  75.7  0.16  204  95.8  0.12  206  98.9  0.12  229  104.2  65  U.U021  0.0022  0.0023  0.0024  1/Temperature  0.0025  0.0026  (K" ) 1  Gradient = -1014 ± 78 K Intercept = 5.2 ± 0.2  Figure 3.13  Arrhenius plot for the high pressure rate constants  66  2 "g5 0 L J5 5U.U018  •  1  •  0.0020  1  •  '  •  0.0022 0.0024 1/Temperature ( K )  1  0.0026  _1  Gradient = -1548 ± 104 K; Intercept = 9.9 ± 0.2 ;  Figure 3.14  Arrhenius plot for atomic etching rate constants  •  0.0028  those due to gallium tribromide found for  the equivalent  and its fragments.  N o peaks were  arsenic compounds.  A n etched wafer was analysed by X-ray analysis ( E D X ) using a scanning electron microscope (Hitachi S-579). gallium was found to be 52.3% corresponds to 50.6 wt%  and that for arsenic 47.7%.  gallium and 49.6 wt%  of the analysis is given as within 0.5 wt%  arsenic.  This  A s the accuracy  N o traces of  This analysis probes to a depth of about 1  68  of  it can reasonably be concluded  that the sample surface remained equimolar. were detected.  T h e atomic percentage  bromine \im.  4  DISCUSSION  4.1 PROPOSED MECHANISM The reaction of GaAs {100} pressure dependencies.  with bromine displays two distinct  At bromine pressures below  1-2 Torr, the  reaction is first order with respect to the concentration of bromine. When the pressure exceeds 1-2 Torr, the reaction becomes half order. A mechanism involving several steps is needed to account for this behaviour. The adsorption of molecular chlorine on the {110}  surface of GaAs  has been studied by a number of spectroscopic techniques.  48  The  surface was found to be rapidly saturated with chlorine, which is initially non-dissociatively  adsorbed.  The chlorine subsequently  dissociates and is chemisorbed on the GaAs surface. occurs for bromine on {100}  Assuming this  GaAs also, then a brominated surface is  produced, in a rapid Langmuir type process.  A physically adsorbed  layer of bromine molecules may form initially but this should be a rapid process with no significant activation energy and hence not rate controlling. is 1 or 2. {100}  The surface will consist of GaBr  x  and A s B r  x  species, where x  From steric considerations, the surface Ga and As atoms on a  surface are most likely to be monosubstituted. In order to explain the rate laws observed in the present study  we propose the following rate controlling steps :  Br (g) 2  k  l  ^2Br(ads)  69  (4.1)  Br(ads)  The  »- P r o d u c t s  (4.2)  first step is the reversible dissociative adsorption of bromine  molecules on the already halogenated gallium arsenide surface. second step (reaction 4.2)  The  represents the reaction of one of these  adsorbed atoms with the halogenated surface to produce gaseous product or at least some intermediate a non rate-determining  step.  which yields gaseous products in  T h e mechanism can be considered a  variation of the E l e y - R i d e a l mechanism since it postulates that a mobile, weakly  adsorbed layer of bromine atoms is involved in the product  formation. The products are likely to be ASBT3 and G a 2 B r g . stable  arsenic b r o m i d e .  49  AsBr3 is the only  AsBr3 is volatile and it should desorb rapidly.  Therefore the desorption of the arsenic product should not limit the rate of reaction.  Ga2Brg is the most stable gallium bromide near room  temperature but it is less volatile than A s B r 3 . the  gallium  bromide ( 2 7 9 ° C ) it has been shown that 70% of the bromide  exists as the d i m e r .  50  The vapour pressure of GaBr3 is given b y  log P * = (-14,300/4.57T) A t the melting point of 5.4 Torr.  A t the boiling point of  5 1  :  + 8.554  GaBr3 ( 1 2 2 - 1 2 4 ° C ) , the vapour pressure is  F r o m the kinetic theory of gases :  Re = a ( M / 2 T I R T ) 1 / 2 P * where  1.0.  5 2  a  = coefficient of condensation  M  = molecular weight  and  R , T have their usual meanings.  The  coefficient of condensation typically falls between 0.1  Taking a  as 0.1  and  gives the minimum evaporation rate, which is 70  calculated as 0.028 g c m  s .  - 2  This is equivalent to an etch rate of 3100  _1  Um m i n , which is two orders of magnitude higher than the etch rates -1  observed at this temperature.  So the desorption of G a B r g 2  molecules  from a GaBr3 liquid layer can be dismissed as a possible rate determining step, in the same way as Ibbotson et al discounted G a C l 6 2  vaporisation as rate limiting, in their studies using a chlorine plasma. The atomic etching experiments gave faster rates than molecular etching (though this could only be explored for the low pressure region) and so, assuming the products of the reaction are the same, this supports the argument that product desorption is not rate limiting Diffusion limitations can arise at the pressures used in this study. However, it was found firstly that no levelling off of etch rate was seen at pressures up to 40 Torr at 125°C. found to influence the etch rate.  Secondly, the flow rate was not  In a diffusion limited reaction the rate  is determined by how fast the reactant is delivered to the surface and this can be dependent on flow.  4.2 KINETIC ANALYSIS OF MECHANISM. If equations 4.1 and 4.2 represent the rate controlling steps :  Br (g) ^ = ^ k = - 2 B r ( a d s ) -i  (4.1)  Br(ads)  (4.2)  2  k  ± Products  Then the etch rate is given by the rate of formation of products :  Etch rate =  d  (  P  r  o  d  u  c  dt 71  t  )  = k [Br(ads)] 2  (4.3)  where  [Br(ads)] = concentration of adsorbed bromine  atoms.  During the steady-state etching reaction :  2 k P - 2 k _ [ B r ( a d s ) ] - k [Br(ads)] = 0  (4.4)  2  x  1  2  where P = pressure of bromine.  Solving this quadratic equation for [Br(ads)]  1/2 [Br(ads)]=  -k  2 2  4 k.  - +  Substituting from (4.5)  Etch rate  =—— 4 k.  1 +  x  k 2  2  +16k.!kiP 1  (4.5)  1  4 k.  into (4.3)  ^k.^jP  •2  gives  ,1/2 (4.6)  4k.  /  In the low pressure region : A s (1 + x ) / 1  2  -  (1 + x/2) when x is small, then at low pressures :  ^ k . ^ P  1 +  •2  xl/2  \  +  8k  l i  k P  1  J  (4.7)  /  This approximation arises from the back reaction in the equilibrium  between  B r atoms (4.1)  the  bromine  molecules and the  weakly  adsorbed  being negligible relative to the rate of reaction 4.2 for the  low pressure region.  B y substituting into 4.6 from 4.7, we find :  Etch rate = 2 k ! P  (4.8)  We can compare this etch rate equation with the empirical equation 3.1 : Etch rate = a ( P  Br2  )  (3.1)  Then it can be seen that a = 2kj.  Therefore from the values  obtained for this constant a (section 3.3.1) the rate controlling reaction step can be presented in the form of an Arrhenius equation, for the pressure range 0-2 Torr and temperatures from 100 to 2 0 0 ° C : kj = 4.1 x 105 ^im min-1 Torr-1 e(-29.2 kJ/mol)/RT  At high pressures: 1  6  >  k  k  2  k  - l » !  (4.9)  - - ^ - i  (4.10)  2  Substitution from 4.9 into 4.6 yields :  Etchrate = k  2  —  p  1  /  2  l -i/ k  4  k  The mechanism predicts a low pressure first order region and a high pressure half order region, which is observed experimentally. will be recalled that at high pressures the data was found to fit the equation empirical equation 3.2 :  Etch rate = b ( P  B f 2  )  1 / 2  -c  (3.2)  Comparing this with equation 4.9 :  /1 \ 1/2  b=k and  73  ' 2  k  l  (4.11)  It  The negative intercept c, is characteristic of half order reactions that result from a dissociative  surface adsorption.  53  It follows from the values obtained for b (section 3.3.2) that : / k i ^ -1 - 1 / (-8.4kJ/mol)/RT —^k =1.73ummin Torr / 2 e 2  (4.13)  Ik.  where the composite rate constant is the half order rate constant that applies at pressures over approximately 2 Torr. If the mechanism described by equations 4.1 and 4.2 is correct for the etching of GaAs by molecular bromine, then the reaction occurs by the dissociative adsorption of B r on the GaAs surface. 2  The potential  energy barrier for that process is given by the activation energy of k j , which we find to be 29.2 kj mol" . 1  At high pressures, where reaction (4.1) is reversible the activation energy for the composite half order rate constant gives the height of the potential energy barrier (relative to Br (g)) for the second step in the 2  reaction.  This value is 8.4 kJ mol" . 1  This information is summarised in  the potential energy diagram shown in figure 4.1.  The position of the  intermediate Br(ads) in the second step and 2Br(ads) in the first step remains undetermined by this data.  However, if Br(ads) can be formed  directly from bromine atoms in the gas phase, that is : Br(g)  2  > Br(ads)  (4.14)  then the activation energy for that process fixes the position of Br(ads) on the potential energy diagram.  Our experiments yield an  Potential Energy ^kJ/mol 194  2Br(g)-  Br(g)  97  29.2  J—  Products Figure 4.1 Potential energy profile for the bromine and GaAs (100) reaction  75  activation energy of 12.7 kj m o l  -1  for that atom reaction, placing Br(ads)  at -4.3 kJ and 2Br(ads) at -8.6 kJ. It is worth noting that because the atomic etching experiments were performed in the low pressure region, where the etch rate is first order with respect to the bromine pressure, the back reaction is slow compared to reaction 4.2.  Hence the etch rate is simply given by the  sum of the individual etch rates for molecular and atomic etching : Etch rate = 2ki[Br ] + k [Br] 2  (4.15)  3  From the data presented in section 3.4, the Arrhenius plot of the atomic etching experiments yields :  k  3  = 1.9 x 105 nm min-l Torr-1 (-12.9 kJ/mol)/RT e  The bond energy of bromine is 194 kJ m o l . -1  (4.16)  The activation  energy for the dissociative adsorption of bromine on gallium arsenide which was obtained is only 29.2 kJ mol" . 1  Much of the energy needed  to dissociate the molecule must therefore be provided by the formation of Br(ads)-Ga and Br(ads)-As bonds.  From the position of Br(ads)  relative to Br(g) in fig. 4.1, these bonds are about 100 kJ mol" . 1  The pre-exponential values  measured also provide information  about the reaction mechanism when compared to the collision frequency.  In the low pressure region, the pre-exponential factor A  which we have determined for reaction 4.1, is a factor of 10 larger than the collision frequency with the surface. reaction 4.1  This could be understood if the  is preceded by pre-adsorption equilibrium described by the  equation : Br (g) 2  <=t  Br (ads) 2  (4.17)  Therefore, k^ is a composite rate constant given by k{ K .  4.3 CONCLUSIONS The reaction of bromine with GaAs {100}  was found to be first  order at the lowest pressures used (below 1-2 Torr) and half order at higher pressures.  This behaviour is consistent with a mechanism which  involves the reversible dissociative halogenated GaAs surface.  adsorption of bromine on a  This is followed by reaction with the  halogenated surface to give the product ASBT3, which desorbs rapidly, and  GaBr3 or its dimer Ga2Br6, which then probably desorbs more  slowly but would not be rate determining under our conditions.. Our results are consistent with the occurrence of at least the following two rate controlling steps :  Br (g) 2  .  k  Br(ads)  l  ^2Br(ads)  (4.1)  > Products  We find that the rate constant  (4.2)  for the first  reaction in the  forward direction can be represented by the following Arrhenius equation : k i = (2.5 ± 3.2) x 105 nm min-1 T o r r l e("29.2 ± 4.0 kJ/mol)/RT The composite rate constant and activation energy for the half order reaction are given by ki\ —M k  /  2  k  2  = 1.73 + 0.35 nm min  -1  - / (-8.4±0.7kJ/mol)/RT Torr / e 1  2  -ii  The rate constant for the reaction with atomic bromine is given by the following Arrhenius equation : k  3  = (1.9 ± 0.5) x 105 n m min-1 T o r r l e(-l2.9 ± 0.9 kJ/mol)/RT  77  References  1 A . D. Milne, M O S Devices : Design and Manufacture, John Wiley and Sons, New York, p l (1983) 2 P. F . Kane and G . B. Larrabee, Characterisation of Semiconductor Materials, McGraw-Hill Book Company, New York, p2 (1970) 3 J. Millman and A . Grabel, Microelectronics, 2nd Ed., McGraw-Hill Book Company, New York, p8 (1987) 4  G . K . Teal and E . Buehler, Phys Rev., 8J7_, 190 (1952)  5 Y.Tarui, VLSI Technology : Fundamentals and Applications, SpringerVerlag, Berlin, p3 (1986) 6 D . K . Ferry, R. O. Grondin and L . A . Akers, "Two Dimensional Automata in VLSI" in Submicron Integrated Circuits, Editor R. K . Watts, John Wiley and Sons, New York (1989) 7 J. Millman and A . Grabel, Microelectronics, 2nd E d , McGraw-Hill Book Company, p9 (1987) 8 H . Welker and H . Weiss in Solid State Physics, vol. 3, Editors F. Seitz and D . Turnbull, Academic Press, New York, p l 1956 9 S. K . Ghandhi, VLSI Fabrication Principles : Silicon and Gallium Arsenide, John Wiley and Sons, New York, p l 2 (1983) 10 J. C . Phillips, Phys. Rev. Lett., 22, 705 (1969) 11 H . C. Gatos and M . C. Lavine, J. Electrochem. Soc, 101, 427 (1960) 12 Y . Tarui, Y . Komiya, Y . Harada, J. Electrochem. S o c , 118, 118 (1971) 13 M . Yokogawa, "LEC GaAs" in Bulk Crystal Growth Technology, editor T. Ikoma, Gordon and Breach Science Publishers, New York (1989) 14 J. W . Mayer and S. S. Lau, Electronics Materials Science : For Integrated Circuits in Si and GaAs, Appendix, Macmillan Publishing Company, New York (1990) 15  Anon, New Scientist, p36, 7th April, 1990  78  16 M . H . Brodsky, "Progress in Gallium Arsenide Semiconductors", Scientific American, p68, February 1990 17 D. A . Ross, Optoelectronic Devices and Optical Imaging Techniques, Macmillan Press, L o n d o n , p6 (1979) 18 R. J , Chaffin, Microwave Semiconductor Devices : Fundamentals and Radiation Effects, John Wiley and Sons, New Y o r k , p l l - 1 4 (1973) 19  F. A . Houle, Basic Mechanisms in Laser Etching and Deposition, A p p l .  Phys. 20  A 4 1 , 315-330 (1986)  C . I. H . Ashby, A p p l . Phys. Lett., 4 5 , 892 (1984)  21 D . J . Elliot, Integrated Circuit Fabrication Technology, p 359-363, M c G r a w - H i l l Publishing Company (1989) 22 S. K . Ghandhi, V L S I Fabrication Principles : Silicon and Gallium Arsenide, John W i l e y and Sons, New Y o r k , p499 (1983) 23  J . H . H a , P h . D. thesis, Department of Chemistry, U . B . C . (1989)  24  H . F. Winters, J . A p p l . Phys., 4, 10 (1978)  25 p. w. Atkins, Physical Chemistry, 3rd E d . , Oxford University Press, O x f o r d , p772 (1986) 26  Ibid  27 R. P. H . Gasser, A n Introduction to Chemisorption and Catalysis by Metals, Oxford University Press, Oxford, p l 7 7 (1985) 28  J . E . Lennard-Jones, Trans. Farad. S o c , 28, 333 (1932)  29 J . M . Thomas and W . J . Thomas, Introduction to the Principles of Heterogeneous Catalysis, Academic Press, L o n d o n (1967) 30  D. L . Flamm, V . M . Donnelly and D. E . Ibbotson, J . V a c . Sci. Technol.  B l ( l ) , 23 (1983) 31  I. Langmuir, Trans. Farad. S o c , 17, 607 (1922)  32  J . F. L y n c h and T . B . Flanagan, J . Phys. Chem., 77., 2628 (1973)  79  33 R. H . Burton, R. A . Gottscho and G . Smolinsky, "Dry Etching of Group III-V Compound Semiconductors" in Dry Etching for Microelectronics, Editor R. A . Powell, Elsevier Science Publishers (1984) 34 V . M . Donnelly, D. L . Flamm and D. E . Ibbotson, J. Vac. Sci. Technol. A 1, 626 (1983) 35 G . Smolinsky, R. P. Chang and T. M . Mayer, J. Vac. Sci. Technol., 18, 12 (1981) 36 D . E . Ibbotson, D . L . Flamm and V . M . Donnelly, J. Appl. Phys., 54, 5974 (1983) 37 V . M . Donnelly, D. L . Flamm, C . W . Tu and D. E . Ibbotson, J. Electrochem. S o c , 129, 5974 (1983) 38 J. H . Ha, Ph. D . Thesis, Department of Chemistry, U . B . C . (1989) 39 J. H . Ha, E . A . Ogryzlo and S. Polyhronopoulos, J. Chem. Phys., 89, (5), 2844 (1988) 40 N . Furuhata, H . Miyamoto, A . Okamoto and K . Ohata, J. Elect. Mat., 19, 201 (1990) 41 M . Balooch and D. R. Olander, J. Vac. Sci. Technol. B 4, 794 (1986) 42  K . R. Jennings, Quart. Rev., 11, 237 (1961)  43 E . A . Ogryzlo, Can. J. Chem., 39, 2556 (1961) 44 L . Elias, E . A . Ogryzlo and H . I. Schiff, Can. J. Chem., 21, 1680 (1959) 45 R. C . Weast, C R C Handbook of Chemistry and Physics, 61st E d . , F-222 (1980-1981) 46  Z . H . Walker, unpublished results.  47  D . C . Chong, unpublished results  48 R. H . Williams and I. T. McGovern, "Adsorption on Semiconductors" in The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis, Editors D . A . King and D . P. Woodruff, vol. 3, p298, Elsevier Science Publishers, Amsterdam (1984)  80  49 Bromine and its Compounds, E d . Z . E . Jolles, Academic Press, New Y o r k , p222 (1966) 50  Ibid, p l 3 1  51  K . Wade and A . J . Bannister, "Aluminium, gallium, indium and  thallium"  in Comprehensive Inorganic Chemistry, v o l . 1, Executive  editor  A . F . Trotman-Dickenson, Pergamon Press, L o n d o n , p993 (1973) 52 V . M . Donnelly, D. L . Flamm, C . W . T u and D. E . Ibbotson, J . V a c . S c i . Technol. B 4, 794 (1986) 53  C . J . Smithells, C . E . Ransley, Roy. Soc. Proc. A , 1 5 0 , 172 (1935)  81  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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


Related Items