Search

Our objective is to propose a neural cell segmentation algorithm for fluorescent mi- croscopic images of mice brains. We received a dataset from the Laboratory of Neu- rochemistry, Institute of Physiology of the Czech Academy of Sciences in Prague. We conducted various image segmentation experiments to identify a method that can most accurately segment neural cells. Our thesis will present the challenges linked with the segmentation of biological images. Then, we will describe the details of our experiments and the evaluation metrics used for measuring our methods' accuracy. 1

The aim of the thesis is to implement a procedural method which transfers a natural image into a pencil drawing-like style. Our project is written in C++. It uses libraries like OpenCV for image processing and Eigen for linear algebra computations. Since neural networks are frequently questioned, as to whether or not they are better than procedural methods for artistic style reproduction, this work presents also a detailed comparison of both of these approaches. We have re-implemented a selected method for procedural generation of pencil drawing style, bringing several modifications. We compare results of the method with a recently released code for neural network-based drawing generation. The result of this subjective comparison indicates that neural networks maybe be better suited for the generation of pencil-like hatching texture to reproduce shading. On the other hand, the procedurally generated outlined produced by the implemented approach provide more natural renderings.

Reflectance difference spectroscopy (RDS) is a surface analysis technique that was invented in 1985 by Aspnes. Here, we give a summary of its application to crystal growth techniques which gave new and valuable real time information about the growth process. Also, this information was obtained in-situ, in molecular beam epitaxy (MBE) and organometallic chemical vapor deposition (OMCVD) crystal growing setups which are routinely used to produce high quality device structures. The application of RDS to OMCVD allowed us to develop a “textbook” like model of growth kinetics, which includes two independent microscopic mechanisms, i.e. adsorption (at -26 kcal/mol) of the reacting molecule (trimethylgallium (TMG) in the case of GaAs), followed by its decomposition (at 39 kcal/mol) on the growing GaAs surface. Our model includes an effect called steric hindrance, associated with the large size of the TMG molecule. This study represents the first direct quantitative evaluation of the catalytic effect of the GaAs surface for the decomposition of TMG. We discuss implications of the model both for growth in the ALE mode as well as for conventional OMCVD growth and comment on the relative importance of surface and gas phase reactions. The application of RDS to MBE revealed remarkable details about the complex intermediate steps that surfaces undergo during growth and enabled to extract directly surface dielectric functions. Finally, applications of the technique as well as results obtained in a number of laboratories where RDS is currently being developed are discussed.

This paper summarizes results of our investigations of growth on (001) and (110) GaAs by atmospheric pressure organometallic chemical vapor deposition (OMCVD). We follow evolutions of surface species to a sensitivity of 0.01 monolayer (ML) on a time scale of 0.1 s under alternating flows of trimethylgallium (TMG) and arsine (AsH3) as functions of partial pressure, sample temperature, and surface orientation. The reaction of TMG with an AsH3-saturated (001) surface is rate-limited by competition between desorption and decomposition of TMG molecules chemisorbed to surface lattice sites via an exclude-volume mechanism, while the reaction of AsH3 with the TMG-saturated (001) surface is essentially instantaneous. In contrast, TMG reacts essentially instantaneously with the AsH3-saturated (110) surface while the AsH3 reaction with the TMG-saturated (110) surface is the rate-limiting step. However, the latter rate is not intrinsic to the AsH3-surface reaction but appears to be determined by desorption of adsorbed species that block active sites.

This paper summarizes results of our investigations of growth on (001) and (110) GaAs by atmospheric-pressure organometallic chemical vapor deposition (OMCVD). We follow evolutions of surface species to a sensitivity of 0.01 monolayer (ML) on a time scale of 0.1 s under alternating flows of trimethylgallium (TMG) and arsine (AsH3) as functions of partial pressure, sample temperature, and surface orienta-tion. The reaction of TMG with an AsH3-saturated (001) surface is rate-limited by com-petition between desorption and decomposition of TMG molecules chemisorbed to surface lattice sites via an excluded-volume mechanism, while the reaction of AsH3 with the TMG-saturated (001) surface is essentially instantaneous. In contrast, TMG reacts essentially instantaneously with the AsH3 -saturated (110) surface while the AsH3 reaction with the TMG-saturated (110) surface is the rate-limiting step. However, the latter rate is not intrinsic to the AsH3-surface reaction but appears to be determined by desorption of adsorbed species that block active sites.

Using the double‐modulation extension of reflectance‐difference spectroscopy, we obtain surface dielectric anisotropy spectra under steady‐state conditions for several reconstructions encountered in molecular beam epitaxy on (001) GaAs and follow their evolution with temperature. The changes observed with temperature are gradual, generally not exhibiting distinct boundaries. Marked changes in anisotropy are observed during thermal desorption of the native oxide, indicating that the process is not simple but involves surface roughening.

We show that the principles that govern the preparation and properties of “transparent” platinum films apply also to films of palladium, rhodium and rhenium. Films of these metals, of 42–60 nm thickness, having a metal volume fraction between 0.3 and 0.5, transmit more light than equivalent, dense metal films of identical metal loadings per unit area. The films are prepared by photoelectrodeposition onto p-InP (100) photocathodes, from ∼5×10 −5 M solutions of the metal ions in 1 M HClO 4 , under mass-transport limited conditions, at deposition rates of ∼ 2 nm/h. The “transparent” rhodium, rhenium and palladium films exhibit their normal catalytic behavior and have normal crystal structures. While the “transparent” rhodium and palladium films, like the bulk metals, are stable in air, the rhenium films oxidize over a period of days to form films that are either truly amorphous or consist of crystallites of less than 1 nm diameter. The spectroellipsometrically measured dielectric functions of these films in air are analyzed, in the framework of the Bruggeman effective medium approximation, to yield film thicknesses, metal volume fractions, and mean depolarization factors. The respective depolarization factors of Rh, Pd, and Re indicate dendritic, particulate, and platelet microstructures, consistent with the structures observed by transmission electron-microscopy. With the spectroellipsometrically derived, substantially divergent, depolarization factors for the different microstructures, one obtains consistent relationship among film thicknesses, metal volume fractions, and the optical properties. Absorption in tenuous films, of constant metal loading but different microstructures, tends to different finite values strongly dependent on microstructure, in the limit of infinite metal dilution.

A method is described for fabricating and evaluating metal–semiconductor (Schottky diode) samples for electroreflectance measurements over a wide spectral range. This technique has been used successfully on a variety of semiconductors. The completed samples have been cycled from 300 to 6 K maintaining very stable characteristics and allowing excellent reproducibility in electroreflectance data.

We report visible‐near UV reflectance–difference (RD) spectra for (110) GaAs and InP. In contrast to those for (110) Si and Ge, chemisorption, physisorption, and surface many‐body screening contributions are much weaker, or even absent. Aqueous solutions simply roughen the (110) GaAs surface on a few‐A scale, generating microstructural features whose net lengthwise orientation along 〈001〉 is consistent with the formation of pits with (111)A stop‐etch‐plane boundaries. The dominant RD contribution for both (110) GaAs and InP has the line shape of bulk spatial dispersion. A calculation of this effect for the E1 spectral regions in the standard two‐dimensional M0 critical‐point model (bands parallel along 〈111〉) yields results that have the wrong sign and are an order of magnitude too small. Both discrepancies can be eliminated if the full three‐dimensional parabolic nature of these E1 critical points is included, provided, as expected, that they are of symmetry type M1. These results show that optical aniso...

Complex refractive‐index and dielectric function data are given for anodically grown native oxides on GaP, GaAs, and GaSb from 1.5 to 6.0 eV. The absorption thresholds for the oxides on GaAs and GaSb occur at 5.0 and 4.2 eV, respectively, with weak tails extending 0.5 eV to lower energy. The absorption threshold in GaP is not seen but is estimated to be near 7 eV. A comparison of these data with each other and with those for SiO2 and GeO2 suggests a more complex bonding behavior than a simple amorphous phase.

Using reflectance–difference spectroscopy to follow surface coverage on a submonolayer scale, alternating trimethylgallium (TMG) and arsine flows to separate their surface‐chemical effects and to eliminate gas‐phase interactions, and low (330–430 °C) temperatures to eliminate gas‐phase pyrolysis, we isolate and identify the different factors that kinetically limit organometallic chemical vapor deposition (OMCVD) growth on (001) GaAs by their time, pressure, and temperature dependences. When an AsH3‐stabilized surface is exposed to TMG, the relative coverage by Ga‐containing reacted species increases linearly in time to 80% coverage and exponentially thereafter, and exhibits a pressure dependence of classic Langmuir adsorption‐isotherm form. We derive a model that quantitatively describes these features in terms of reversible excluded‐volume chemisorption and subsequent irreversible decomposition of TMG at (001) surface lattice sites. Analysis of the data yields −26 and 39 kcal/mol for the chemisorption en...

We report values of pseudodielectric functions $〈\ensuremath{\epsilon}〉=〈{\ensuremath{\epsilon}}_{1}〉+i〈{\ensuremath{\epsilon}}_{2}〉$ measured by spectroscopic ellipsometry and refractive indices $\stackrel{\ifmmode \tilde{}\else \~{}\fi{}}{n}=n+ik$, reflectivities $R$, and absorption coefficients $\ensuremath{\alpha}$ calculated from these data. Rather than correct ellipsometric results for the presence of overlayers, we have removed these layers as far as possible using the real-time capability of the spectroscopic ellipsometer to assess surface quality during cleaning. Our results are compared with previous data. In general, there is good agreement among optical parameters measured on smooth, clean, and undamaged samples maintained in an inert atmosphere regardless of the technique used to obtain the data. Differences among our data and previous results can generally be understood in terms of inadequate sample preparation, although results obtained by Kramers-Kronig analysis of reflectance measurements often show effects due to improper extrapolations. The present results illustrate the importance of proper sample preparation and of the capability of separately determining both ${\ensuremath{\epsilon}}_{1}$ and ${\ensuremath{\epsilon}}_{2}$ in optical measurements.

We report the first real-time, in situ spectroscopic measurement of a semiconductor surface [(001) GaAs] during crystal growth by organometallic chemical vapor deposition (OMCVD). Reflectance-difference spectroscopy (RDS) is used to enhance the ordinarily low sensitivity of optical measurements to surface effects and thereby to obtain information about the state of the surface for various sample temperatures and ambient conditions during OMCVD growth. The maximum reflectance-difference (RD) signals that we observe fall in the same 2–3 eV energy range as those seen in molecular beam epitaxy (MBE). This suggests that RD measurements can be used to follow relative surface coverages of Ga and As in OMCVD as well as in MBE. Time- and temperature-dependent studies show that this maximum signal closely follows the kinetics of the last rate-limiting step in the decomposition of trimethylgallium (TMG), thereby confirming its sensitivity to surface chemistry (relative coverage of As and Ga.) Using this signal, we follow changes of the OMCVD (001) GaAs growth surface with ambient composition and growth history as well as substrate temperature. Our results indicate that RDS will be very useful in providing insights into, and real-time measurement and control of, crystal growth by OMCVD.

Extremely uniform, microscopically smooth, large-grained polycrystalline Si films saturation doped with phosphorous to a carrier concentration $n=3.3\ifmmode\times\else\texttimes\fi{}{10}^{20}$ ${\mathrm{cm}}^{\ensuremath{-}3}$ were prepared by a combination of low-pressure chemical-vapor-deposition and heat-treating processes to obtain samples suitable for measuring the small effects of large impurity concentrations on the above-band-gap dielectric properties of Si. From 1.5 to 2.5 eV, the effects of heavy doping are described by the Drude free-carrier model with lifetimes determined by scattering from lattice vibrations. Above 3 eV, the dielectric function spectrum of heavily doped Si is remarkably similar to that of undoped material except that the ${E}_{1}$ and ${E}_{2}$ structures are broadened and shifted to lower energies. The impurity-induced broadening is much larger for ${E}_{1}$ than for ${E}_{2}$, indicating that a metastable excitonic resonance is well developed for the former but not for the latter. Present and previous data of the authors, together with previous data of Vi\~na and Cardona, are consistent with an approximately linear shift of the threshold energy with impurity concentration. This indicates that the dominant mechanism responsible for the shifts is the effect of the random impurities on the crystal potential. While the possibility of electron-electron and electron-ion interactions on the electron and hole self-energies cannot be ruled out, comparison with recent calculations of Berggren and Sernelius show that these effects must be much smaller at higher interband transitions than at the fundamental indirect band gap. Using our data as a reference, we show that the anomalous absorption observed by Jellison and co-workers for As-implanted and laser-annealed $c\ensuremath{-}\mathrm{S}\mathrm{i}$ is due to microscopically rough surface layers on their samples. Consequently, the differences that they observe between P and As dopants must involve differences in recrystallization and regrowth. Dielectric-function spectra for amorphous Si films doped in situ with phosphorus are essentially identical within microstructural effects to those of undoped material, indicating no significant influence from heavy doping. At energies below 3.2 eV, the data expressed as ${(\ensuremath{\alpha}nE)}^{\frac{1}{3}}$ or as ${({\ensuremath{\epsilon}}_{2})}^{\frac{1}{2}}$ both vary nearly linearly with energy, suggesting nondirect transitions between simple parabolic bands with momentum matrix elements linearly proportional to $E$ or with dipole elements independent of $E$.