A Highly Synthesizable TDC- and DCO-less Fractional-N PLL and An On-chip Learning Neuromorphic Autoencoder with Current-Mode Memory Read

조화숙 POSTECH 2018 국내박사

This thesis presents a highly synthesizable TDC- and DCO-Less fractional-N PLL and an on-chip learning neuromorphic autoencoder with current-mode memory read are proposed. Firstly, a highly synthesizable TDC- and DCO-less fractional-N PLL is proposed with the phase-locked direct digital synthesizer (PLDDS) driven by a free-running oscillator. With the PLDDS designed only by register-transfer level descriptions of hardware description language, the proposed architecture achieves low in-band phase noise by eliminating nonlinear blocks in the loop. In addition, a selfreferenced recursive phase interpolator is also proposed to suppress quantization noise dominating out-band phase noise. The implemented PLL in 28 nm CMOS process offers favorable area of 0.0047 mm2 and generates seamless fractional frequencies down to fREF. Secondly, an IC implementation of on-chip learning neuromorphic autoencoder unit is proposed in a form of rate-based spiking neural network. With a currentmode signaling scheme embedded in a 500x500 6b SRAM-based memory, the proposed architecture achieves simultaneous processing of multiplications and accumulations. In addition, a transposable memory read for both forwardand backward propagations and a virtual lookup table are also proposed to perform an unsupervised learning of Restricted Boltzmann Machine. The IC is fabricated using 28 nm CMOS process and is verified in a three-layer network of encoder-decoder pair for training and recovery of images with 2D 16x16 pixels. With a dataset of 50 digits, the IC shows a normalized root mean square error of 0.078. Measured energy efficiencies are 4.46 pJ per synaptic operation for inference and 19.26 pJ per synaptic weight update for learning, respectively.

Energetic Particle driven Alfven eigenmodes in KSTAR plasmas : KSTAR 플라즈마의 에너지 입자 구동 Alfven 고유 모드

Syed Haider Raza, Rizvi POSTECH, Department of Physics 2017 국내박사

In tokamak plasma, energetic particles (EP) generated by the external heating systems or the fusion reactions can lead to the excitation of global instabilities. Alfven eigenmodes (AEs) are such modes that destabilized via inverse Landau damping, when their EP drive exceeds the total damping, and can degrade the fast ion confinement and, thus, the self-heating. Furthermore, these fast particles can cause the severe damage to the reactor wall. So understanding the stability properties and their induced fast ions losses of such AEs are essential for any future fusion devices. We have presented the detailed study of Alfven eigenmode in tokamak plasmas, with the help of simulations and experimental results for the KSTAR tokamak. This dissertation is divided into two parts, in first part we have presented our study on the AEs based on theory and numerical simulations. For the theory we have focused on the physics of Alfven Eigenmodes and their excitation based on three type of approaches including the MHD, the reduced kinetic-MHD model and the gyrokinetic model. Where for the numerical study based on MHD and reduced kinetic-MHD model we have developed the eigenvalue code, Kinetic Alfven Eigenmodes Solver (KAES), to study the characteristics of AEs (more specifically toroidicity induced Alfven eigenmodes (TAEs) and beta-induced Alfven eigenmodes (BAEs)) and their damping and energetic particle drive characteristics. In this part we have also presented our results for the radially trapped AEs which are formed due to the coupling of AEs and the kinetic Alfven waves (KAWs). In the second part of the thesis we have focused on the more realistic tokamak plasmas (with general geometry and realistic plasma equilibrium), like KSTAR equilibrium. We presented the experimental observation and particle-in-cell (PIC) simulations of the TAEs destabilized by the NBI's driven fast particles and their characteristics found in KSTAR tokamak. We have found the simultaneous excitation of multiple discrete TAEs for a single toroidal mode number observed in KSTAR plasmas. The excitation and characteristics of these modes are studied by using a global gyrokinetic particle-in-cell simulation code which agrees well with the experimental observation. It is shown that compared to a single core localized mode, the excitation of multiple modes is difficult. Moreover we have also presented the effect of AEs on the phase space dynamics and particle transport in tokamak. This thesis provides the detailed physics understanding of the Alfvenic instabilities, from the theoretical aspects to the experimental observations, and from the linear characteristics to the phase space dynamics of the Alfven eigenmodes.

Prediction of typhoon tracks using a generative adversarial network with observational and meteorological data

Ruettgers, Mario POSTECH 2019 국내석사

본 연구는 Generative Adversarial Network (GAN)이 위성영상과 재해석 데이터 이미지를 입으로 써서 태풍의 경로를 예측한다. 이미지들은 1993년부터 2017년까지 한국에 영향을 끼친 태풍들이 포함되고 학습 데이터와 테스트 데이터로 나뉘어 있다. GAN은 학습 데이터를 통해서 6시간 간에 해당하는 태풍의 발달을 학습하며 테스트 데이터를 통해서 학습되지 않은 태풍의 중심과 구름의 이동을 예측한다. 예측된 태풍의 중심에 대한 에러를 정량적으로 측정하였다. 전체 태풍 예측 결과에서 65.5 %의 태풍은 80 km 이내의 오차를, 31.5 %의 태풍은 80 - 120 km의 오차를 보였다. 단 3.0 %의 태풍만 120 km 보다 큰 오차를 보였다. 평균 오차는 67.2 km으로 나타났다. 예측된 구름의 이동 정성적으로 평가되었다. GAN은 구름 구조발달에 대한 경향을 예측할 수 있다. 추후에는 한국에 영향을 끼친 태풍들 뿐만 아니라 아시아에서 발생한 모든 태풍의 데이터를 고려하여 학습하고자 한다. Tracks of typhoons are predicted using a generative adversarial network (GAN) with observational data in form of satellite images and meteorological data from a reanalysis database. Time series of images of typhoons which occurred in the Korean Peninsula in the past are used to train the neural network. The trained GAN is employed to produce a 6-hour-advance track of a typhoon for which the GAN was not trained. The predicted image favorably identifies the future location of the typhoon center as well as the deformed cloud structures. The errors between predicted and real typhoon centers are measured quantitatively in kilometers. 65.5 % of all typhoon center predictions have an error of less than 80 km, 31.5 % lie within a range of 80 - 120 km and the remaining 3.0 % are above 120 km. The overall error is 67.2 km, compared to 95.6 km when only observational data are used as input. The cloud structure prediction is evaluated qualitatively. It is shown that the GAN is able to predict trends in cloud motion. It is found that adding physically meaningful meteorological data to satellite images improves the sharpness of predicted images.

Application of divergence-free symmetric positive semi-definite matrices to gas dynamics

하헌진 POSTECH 2019 국내석사

이 논문에서 자유발산 대칭 준 양부호행렬 함수의 여러가지 수학적 성질을 중명하고, 이를 이용하여 자유발산 시스템인 압축성 오일러 시스템의 약해의 에너지타입 추정을 연구한다. 이 논문은 Denis Serre의 논문을 읽고, 유체역학을 연구하는 새로운 수학적 접근법을 탐색하는 것에 초점을 두었다. We investigate several mathematical properties of matrix-valued functions which are symmetric, semi-positive definite, and divergence-free for each row(or equivalently, column). Moreover, we study several energy-type estimates of weak solutions to the Euler system of compressible flow. This is a summary of [5] for the purpose of survey to find a new approach to study gas dynamics.

Reliable and Efficient Multi-rate Aware Routing Protocols for IEEE 802.11 Based MANETs

Khalid, Ahmed POSTECH Graduate School 2015 국내석사

Most existing mobile ad hoc network (MANET) routing protocols use the minimum number of hops as a routing metric. These paths tend to contain long range links that may lower both throughput and reliability. Higher throughput can be achieved by using short but effective links. A few protocols have been proposed to address this issue. The main problem with these protocols is that they use receiver signal strength indicator (RSSI) value to assess the link states. In realistic networks, RSSI value is not always accurate and hence using it can result in inaccurate decisions which will degrade the reliability of paths. Some other unaddressed issues in these protocols include: overhead of periodic hello messages, overhead of relaying data, neglecting paths with same throughput but lesser hops and neglecting changes in link states other than link breakage. We propose two reliable and efficient multi-rate aware routing protocols for MANETs: reliable and efficient multi rate aware proactive (REMAP) routing protocol and reliable and efficient multi rate aware reactive (REMAR) routing protocol. REMAP is a proactive protocol which provides a solution to the other proactive protocols. REMAR is a reactive routing protocol and it solves the problems in other traditional and multi rate aware reactive routing protocols. Both the proposed protocols do not make routing decisions based on RSSI value and hence provide more reliable paths. They also decrease the overhead caused by periodic hello messages in discovering link states of the neighbors and the overhead caused by relaying data. REMAP and REMAR ensure that the path chosen by a node is the fastest path and they update the path whenever a change in link state results in another path being faster than the current one. Simulation results show that our proposed protocols can significantly increase the path reliability and throughput and decrease the network delay and overhead.

Thermoelectric and Mechanical Characterization of Synthesized One-Dimensional Nanostructures

Davami, Keivan POSTECH 2013 국내박사

The relatively high cost and limited supply of fossil fuels has caused increased attention to cleaner and cheaper alternative energy sources such as solar, wind, and, as discussed here, thermoelectric energy. A new way of energy production via the conversion of thermal energy to electricity is now achievable through the application of thermoelectric effects. One-dimensional nanostructured materials have outstanding potential to enhance existing conversion devices. As tellurides of Zn, Cd and Pb have become popular for thermoelectric applications, a thorough characterization of their thermoelectric and mechanical properties is needed. In this research, the thermoelectric and mechanical properties of ZnTe nanowires (NWs) are investigated. We produced ZnTe NWs with diameters less than 30 nm via the bottom-up vapor-liquid-solid method. Bandgap engineering of single-crystalline alloy CdxZn1-xTe (0 ≤ x ≤ 1) NWs was achieved successfully through control of growth temperature and a two zone source system in a vapor-liquid-solid process. The effective process parameters on the morphologies of NWs were investigated, and the nanostructures were characterized by SEM and TEM. For the mechanical characterization, an experimental and a computational approach (molecular dynamic simulation) were used to investigate the size effects on Young’s modulus of ZnTe NWs. The mechanical properties of individual ZnTe NWs in a wide diameter range (50-230 nm) were experimentally measured inside a high resolution transmission electron microscope using an atomic force microscope probe with the ability to record in situ continuous force-displacement curves. For the smaller NWs, the computational approach was used. Mechanical characterization showed that neither molecular dynamic simulation nor experimental measurement illustrated evidence of strong size dependency of Young’s modulus of ZnTe NWs. For thermoelectric characterization, a MEMS device with two suspended thermometers was made to measure the thermal conductivity, Seebeck coefficient and electrical conductivity of the individual NWs. After positioning the NWs on the device, the aforementioned parameters were measured for NWs with different diameters. Thermal measurements concluded that the thermal conductivity of ZnTe NWs was significantly lower than that of bulk ZnTe. However, the electrical properties need to be modified in order to obtain a figure a merit comparable with or higher than bulk. The limiting factor for the use of ZnTe NWs in a wider range of applications is their high resistance. This can be removed by optimal doping or any other methods for the modification of their electrical conductivity. Also, more studies need to be done in order to achieve Ohmic electrical contacts with ZnTe NWs for a better investigation of their thermoelectric potential. Overall, it was concluded that ZnTe nanowires have the potential to be used in thermoelectric applications because of their relatively high strength and low and size dependent thermal conductance.

복소 불변량에 관하여: 자기동형사상군, 불변거리 그리고 쌍단면곡률 : On the holomorphic invariants: Automorphism groups, Invariant metrics and Bisectinal curvatures

서애령 POSTECH 2012 국내박사

In this dissertation, we present several new domains that do not admit any complete Kahler metric with negatively pinched bisectional curvature. Moreover, we show the uniform extension of the automorphisms up to the boundary for the strongly pseudoconvex domains with Ck,a boundaries when k>1, an integer and 0<a<1. This allows us to prove that the semicontinuity theorem for holomorphic automorphism groups holds for the class of bounded strongly pseudoconves domains with Ck+2,a boundary.

Application of Photoactive Diruthenium Catalyst for New Organic Reactions: Syntheses of Imines, Isoquinolines, and Vinylstannanes

Gupta, Sreya POSTECH 2014 국내박사

Light had the potential to serve as an inexpensive, abundant, renewable, and nonpolluting reagent for chemical transformations. Thus developing new methodologies by employing metal-catalyst activated by visible light is the pivotal interest in the field of organic synthesis. In this thesis, we investigated important chemical transformation catalyzed by carbonyl bridged diruthenium complexes which is activated by visible light source. In chapter 1, Ru-catalyzed synthesis of N-unsubstituted imines from easily available alkyle azides under the photolytic condition has described. N-unsubstituted imines were generated by removal of N2 gas from azide precursors and subsequent migration of hydrogen atom under the house hold fluorescent light. This novel protocol has shown broad synthetic scope to construct various N-unprotected imines including unprecedented enolizable substrates. Mild reaction conditions used in this protocol even allowed obtaining fundamental physical parameters of N-unprotected imines, such as, existence of dynamic geometric isomers in the solution phase and dissociation energy of imino N-H bond. This novel synthetic route was successfully applied in one-pot imine generation/allylation sequence by using allylboron reagents; directly N-unprotected allylamines were formed without any undesirable side reactions. The above one-pot sequence has shown greater efficiency over prevailing methodologies in terms of enantio- and diastereoselectivity. Furthermore, this protocol also effective to synthesize N-unsubstituted imidates from various azidoethers. This is the first report of imidate synthesis in neutral reaction conditions which allowed constructing imidates having acid-labile functionality. Chapter 2 describes the Ru/Rh-tandem catalysis to synthesize isoquinoline from arylazides and internal alkynes by direct C-H bond activation. N-unsubstituted arylimines generated from the corresponding azides by the ruthenium catalysis were used as a directing group for rhodium-catalyst to activate the C-H bond at ortho-position of aromatic ring. Subsequent addition of various internal alkynes produces N-annulated isoquinolines. The essential merits of this cascade protocol are broad scope of applicable benzylic azides, including those having carbonyl and chiral functionalities. Additionally, this protocol is completely atom-economical as N2-gas is the only the byproduct. Furthermore, N-unsubstituted imidate synthesized by Ru-catalyst II from the corresponding azidoethers, also turned out to be the successful entity for direct synthesis of 1-alkoxyisoquinolines. The mechanism of the above transformation was established through the isolation and characterization (X-ray crystallography structure) of the important unknown intermediates. Previously we reported the hydrosilylation of aldehydes1 by using ruthenium catalyst under fluorescent light condition. We expected this catalyst can activate other metal-hydride species too. In the chapter 3, ruthenium-catalyzed addition of tin hydride to various internal alkynes (hydrostannation) under fluorescent light was described. The reaction was initiated by the formation of ruthenium hydride and tin radical species under the illumination of light. In the termination step, tin-radical and the hydrogen from ruthenium hydride species were added to the alkynes in anti-stereoselective way. In general, control the regio- and setreoselectivity of hydrostannation reaction is the formidable challenge. However, in above ruthenium photocatalyzed system, (Z)-β-selective vinylstannanes were obtained selectively for various internal alkynes and shown better activity over well-known radical initiators such as AIBN (azobisisobutyronitril) or BEt3 (triethylborane). To obtain insight into the hydrostannation mechanism, we conducted various experiments. The outcome of various control experiments strongly suggested the insolvent of radical mechanism and the formation of ruthenium-hydride (VI) as a key intermediate species under the fluorescent light. Additionally, the recyclability of the ruthenium catalyst added extra advantage of this novel methodology. 1 Do, Y.; Han, J.; Rhee, Y. H.; Park. J. Adv. Synth. Catal. 2011, 353, 3363.

Tungsten carbide 합성과 DMFC 응용을 위한 메탄올 산화 촉매로서의 특성 평가

한승현 POSTECH 2008 국내석사

Experimental study on condensation heat transfer of high pressure flowing steam on a bundle of coiled tubes

차명기 POSTECH 2021 국내석사

The coiled tube is used as a heat exchanger in various industries such as refrigeration and air conditioning, chemical plants, and small reactor systems, and is composed of a spiral wound heat exchanger (SWHX). It is used as an evaporator in the field of refrigeration and air conditioning, a reactor in a chemical plant, and a steam generator in a SMR. In addition, in the event of an accident with a SMR, it may be used as a condenser as a part of the residual heat removal system (RHRS) that removes superheated steam generated from reactor core under accident condition. When used as a condenser, condensation occurs on the shell side where the flow of condensate water and steam is free. In case of an accident, film condensation occurs on the tube because it is a high heat flux condition. Several researchers have studied the external condensation of the coiled tube under static steam conditions. Depending on the ratio of the curvature of the coiled tube and the diameter of the tube, acceleration forces due to gravity and centrifugal forces that are not found in the horizontal tube are generated, and the shape of the film thickness changes due to the forces generated. It is reported that the heat transfer coefficient is higher than that of the horizontal tube depending on the degree of change. However, this study did not consider the steam flow required for industrial applications. In addition, since the coiled tube is close to the form of a bundle tube in which several tubes are arranged, the bundle effect should be considered. Therefore, in this study, experiments were conducted under the conditions of two vapor mass flux (7.2kg/m^2 s,12.0kg/m^2 s) and two pressures (1.0MPa, 2.0 MPa) to confirm how the mass flux of steam is reflected in the coiled tubes, and the heat transfer coefficient for each row. By measuring, it was verified how the tube bundle shape affects the heat transfer coefficient in the coiled tube.