Patent analysis for technology forecasting : an aspect of audio guidance devices for blind and visually impaired people

황주미 Greduate School, Korea University 2021 국내석사

(A) study on brillouin optical correlation domain analysis adopting time-domain data processing

류국빈 Greduate School, Korea University 2021 국내박사

최근, 광섬유 기반 광학 시스템을 센서로 활용하기 위한 많은 연구가 진행되고 있다. 광섬유는 전자기파에 간섭하지 않으며, 고온에서 잘 견딜 수 있기 때문에 극한 환경에서도 센서로 동작할 수 있다. 또한, 유연하고 직경이 가늘어 설치가 용이하며, 장기간 유지 관리가 편리하면서도 작은 전송 손실을 이용하여 장거리 측정이 가능하다는 장점도 있다. 특히, 빛이 광섬유를 진행할 때 발생하는 산란을 이용하여 광섬유 자체를 센서로 활용하는 분포형 광섬유 센서에 대한 연구가 활발히 진행되고 있다. 기존 여러 지점을 측정하기 위해서 여러 센서와 여러 장치가 필요한 것과는 달리, 분포형 광섬유 센서는 하나의 시스템으로 여러 지점을 측정할 수 있다는 특징이 있다. 이용하는 산란 및 분석 방법에 따라 여러 형태의 분포형 광섬유 센서 시스템이 존재한다. 빛이 광섬유를 진행할 때 발생하는 레일리, 브릴루앙, 라만 산란을 이용할 수 있고, 시간, 주파수, 상관 영역에서 분석할 수 있으며, 이들의 조합으로 센서의 형태가 구분된다. 세 가지의 산란 중 비탄성 산란인 브릴루앙 산란은 광섬유 고유의 브릴루앙 주파수가 광섬유 주변 온도나 인장 변화에 따라 선형으로 변하는 특성이 있다. 이를 이용하여 구현한 센서는 구조물 안전 진단, 변형 측정, 온도 모니터링 등 다양한 분야에 활용할 수 있다. 특히, 상관 영역에서 분석하는 방법은 높은 공간 분해능으로 공간 선택적 측정할 수 있다는 장점이 있으며, 브릴루앙 광 상관영역 분석 방법(BOCDA, Brillouin optical correlation domain analysis)은 브릴루앙 산란을 이용한 공간 선택적 센싱 방법 중의 하나이다. BOCDA 시스템에서는 특정 지점에서만 유도 브릴루앙 산란이 발생하도록 제어하여 높은 공간 분해능을 구현한다. 이를 위해 광 주파수가 변조된 연속 펌프, 프로브광이 측정 광섬유 내에 서로 반대 방향으로 진행하며 중첩하도록 한다. 주파수 변조된 두 광의 중첩으로 주파수 차이가 일정한 상관점이 주기적으로 발생하고, 두 광의 주파수 차이가 브릴루앙 주파수와 같을 경우 상관점에서만 유도 브릴루앙 산란이 강하게 발생한다. 유도 브릴루앙 산란에 의해 프로브 광이 증폭되고, 증폭된 프로브 광을 분석함으로써 브릴루앙 산란 정보를 얻을 수 있다. 광 변조 주파수를 변화시키며 상관점 위치를 이동할 수 있고, 측정 광섬유 전 범위의 브릴루앙 산란 정보를 측정할 수 있다. 하지만, 측정 광섬유 내 여러 상관점이 존재할 경우, 프로브 광이 어느 상관점에서 증폭됐는지 구분할 수 없기 때문에, 오직 하나의 상관점만 측정 광섬유 내에 존재해야 한다. 즉, 측정 범위가 상관점 간격으로 제한된다. 또한, 공간 분해능과 측정 범위가 trade-off 관계에 놓여있어, 높은 분해능을 유지하면서 측정 범위를 늘리는데 한계가 있다. 본 학위논문에서는 브릴루앙 광 상관 영역 분석 방법에서 시간 영역 분석 방법을 도입한 연구 결과를 제시하고, 시스템의 성능을 향상시키기 위한 연구 결과를 제시한다. 시간 영역 분석 방법은 BOCDA 시스템에서 펌프 광을 기존 연속 광이 아닌 펄스 형태로 이용하고, 펌프 펄스 광과 중첩으로 증폭된 프로브 광을 시간 영역에서 측정하는 방법이다. 시간 영역에서 분석하기 때문에 측정 광섬유 내 여러 상관점이 존재하더라도 어느 상관점에서 브릴루앙 산란이 발생하여 프로브 광에 증폭했는지 분석할 수 있다. 특히, 펌프 펄스 폭을 적절히 조절하면 인접 상관점에서 발생한 브릴루앙 산란 정보가 겹치지 않도록 제어할 수 있다. 광원 주파수 변조의 한 주기가 하나의 상관점의 브릴루앙 산란 정보를 담고 있으므로, 증폭된 프로브 광을 시간영역에서 각 주기마다 분할 분석하면, 여러 상관점에서 발생한 브릴루앙 산란 정보를 각각 얻을 수 있다. 실험을 통해 약 150개의 상관점을 동시에 센싱점으로 이용하여 공간 분해능을 2.5 cm로 유지하면서도 측정 범위는 1.5 km로 늘린 결과를 얻을 수 있었다. 시스템을 최적화하여 10 km까지 측정 범위를 늘릴 수 있었으며, 이에 대한 결과를 제시하였다. BOCDA 방식에서 시간 영역 정보 처리 방식을 도입하여 측정 범위를 향상할 수 있었지만, 여전히 측정 범위가 제한되었다. 광섬유를 진행할 때 발생하는 광섬유 손실과 제한되는 입사 광 세기 때문이다. 브릴루앙 정보를 담고 있는 브릴루앙 이득은 펌프 광의 세기에 비례하는데, 펌프 펄스 광이 광섬유를 진행하며 광섬유 손실에 의해 점차 줄어든다. 따라서, 수십 km의 장거리로 진행하면 브릴루앙 이득 역시 분석할 수 없을 정도로 작아진다. 한편, 브릴루앙 펌프 광의 세기가 너무 크면, 변조 불안정 (modulation instability) 현상에 의해 측정 오류를 나타낸다. 이를 극복하기 위해 1차 라만 증폭 방법을 이용하였다. 브릴루앙 펌프 광과 반대 방향으로 라만 펌프 광을 입사하면, 장거리에서 세기가 작아진 브릴루앙 펌프광은 강하게 증폭하면서도, 브릴루앙 펌프 광 입사단에서는 약하게 증폭할 수 있다. 라만 증폭에 의한 브릴루앙 펌프 광 세기에 대해 이론적으로 계산하였으며, 그 결과를 실제 측정한 결과와 비교하였다. 측정 범위를 50 km로 늘릴 수 있었으며, 7 cm의 공간 분해능으로 광섬유에 가해진 스트레인을 측정한 결과를 보인다. 앞서 제시한 시간 영역에서 정보 처리 방법을 도입한 BOCDA 시스템은 측정 광섬유 양 끝이 모두 시스템에 연결된 형태이다. 실제 파이프, 터널, 철도 등의 환경에 설치할 경우 실제 활용 범위는 절반으로 줄어든다. 이를 해결하기 위해 측정 광섬유 한 끝만 측정 시스템에 연결되어도 측정 가능한 시스템을 제안한다. 기존 양 방향에서 각각 입사하던 펌프, 프로브광을 측정 광섬유 한 끝에서 모두 입사하며, 다른 한 끝에서 반사되어 되돌아오는 프로브광과 펌프 펄스광을 중첩하는 방법이다. 광섬유 끝 단의 반사율, 두 광의 입사 타이밍 및 펌프 펄스 반복률 등에 의해 측정 형태가 달라지며, 이에 대해 정리하였다. 실험을 통해 10 km 측정 광섬유에서 구현한 결과를 보인다. 본 논문에서 제시한 결과는 브릴루앙 산란 기반 광섬유 센서의 성능 향상을 위해 적용할 수 있을 것이며, 장거리 측정이 필요한 거대 구조물 감지 및 진단에 활용할 수 있을 것이다. 또한, 인공지능 및 빅데이터와 결합하여 인공 신경망 구축에도 응용 가능할 것으로 기대한다. Recently, many studies have been conducted on the use of optical-fiber-based optical systems as sensors. Optical fiber can operate as a sensor, even in an extreme environment, because the optical fiber sensors are immune to electromagnetic interference and can withstand high temperatures. In addition, optical fiber is flexible and has a small diameter so it is easy to install. Long-term maintenance of the sensor is convenient, and long distance measurement is possible with a small transmission loss. In particular, various studies have been conducted on distributed optical fiber sensors that utilize the optical fiber itself as a sensor by using the scattering generated when the light travels through the optical fiber. Unlike other sensors that require multiple devices to measure multiple points, a distributed optical fiber sensor can measure multiple points with a single system. There are various types of distributed optical fiber sensors about the scattering and analysis methods used. Rayleigh, Brillouin, and Raman scatterings, generated when light travels through the optical fiber, can be used, and the back-scattered lightwaves can be analyzed in the time, frequency, or correlation domain. Among the three scatterings, Brillouin scattering, which is an inelastic scattering, has a characteristic that the inherent Brillouin frequency of the optical fiber changes linearly according to the temperature variation or strain applied to the optical fiber. Sensors based on the Brillouin scattering can be used in various fields such as safety diagnosis, deformation measurement, and temperature monitoring of structures. In particular, the correlation domain analysis has the advantage of making position-selective measurement with high spatial resolution possible, and Brillouin optical correlation domain analysis (BOCDA) is one of the position-selective sensing methods using Brillouin scattering. In a BOCDA system, the stimulated Brillouin scattering (SBS) occurs only at specific points. Continuous pump and probe waves, whose frequencies are sinusoidally modulated, are injected in opposite directions and interact with each other. The specific points, where the frequency difference between the two waves is constant, occur periodically owing to the interaction of two frequency-modulated waves, and the points are called correlation peaks (CPs). When the frequency difference between the two waves is the same as the local Brillouin frequency of the optical fiber, SBS occurs strongly only at the CPs. The probe wave is amplified by the SBS at CPs, and the Brillouin scattering information can be obtained by analyzing the amplified probe light. Changing the optical modulation frequency makes it possible to move the position of the CPs, and the Brillouin scattering information of the entire range of the sensing fiber can be measured with a high spatial resolution. However, if multiple CP’s exist in the sensing optical fiber, it is impossible to distinguish at which correlation point the probe light is amplified so the measurement range is limited to the interval between CP’s. In addition, because the spatial resolution and the measurement range are in a trade-off relation, there is a limit in increasing the measurement range while maintaining high resolution. In this dissertation, the application of the time-domain data processing to a BOCDA system for enhancement of the limited measurement range was investigated to improve the performance of the system further. In time domain data processing, a pump wave is used in the form of a pulse instead of the continuous wave used in the conventional BOCDA system, and the probe wave amplified by SBS interaction with the pump pulse is measured in the time domain. Even if there are multiple CP’s within the sensing fiber, it is possible to analyze at which correlation point Brillouin scattering occurred. In particular, by adjusting the pump pulse width properly, it is possible that Brillouin scattering information generated at adjacent CPs does not overlap. Because one period of the frequency modulation contains Brillouin scattering information of one CP, by dividing and analyzing the amplified probe light into each period in the time domain, it is possible to obtain Brillouin scattering information generated at multiple correlation points, separately. In the experiment, approximately 150 CPs were simultaneously used as sensing points, and the 2.5 cm-spatial resolution was maintained while the measurement range was increased to 1.5 km. By optimizing the system, it was possible to extend the measuring range to 10 km. In the BOCDA method, the measurement range could be improved by introducing the time domain information processing method, but it was still limited. This is due to the propagation loss in the optical fiber and the limited incident intensity of the Brillouin pump. The Brillouin gain, which contains Brillouin information, is proportional to the pump intensity, but the pump pulse gradually decreases owing to the fiber loss as it passes through the fiber. If the pump pulse travels over a long distance of several tens of kilometers, the Brillouin gain becomes too small to be analyzed. However, if the intensity of the Brillouin pump light is too large, incorrect measurement results are obtained because of the modulation instability. To overcome this, a first-order Raman amplification method is used. When the Raman pump light is incident in a direction opposite to the Brillouin pump, the Brillouin pump can be strongly amplified at the far end of the sensing fiber, but weakly amplified at the near end. These phenomena were calculated theoretically, and compared with the experimental results. In the experiments, the measurement range could be increased to 50 km, and the applied strain to the optical fiber was detected with a spatial resolution of 7 cm. Meanwhile, in the BOCDA system, both ends of the sensing fiber should be connected to the sensor system, so the actual range is reduced by half when installed in such an environment as pipeline or tunnel. To solve this problem, a linearly configured BOCDA system adopting time-domain data processing is proposed. The system can operate even if only one end of the measurement fiber is connected to the measurement system. The pump and probe light are incident from one direction at one end of the sensing fiber, and the incident timing of the pump pulse is adjusted to interact with the probe light reflected from the far end. Measurement of the reflected probe wave that interacts with the counter-propagating pump wave in the time domain enables Brillouin scattering information induced at multiple CPs to be analyzed. The data processing of the system varies depending on the reflectance of the end of the optical fiber, the incident timing of the two lights, and the pump pulse repetition rate. The experimental results implemented in a 10 km sensing fiber are presented. The results in this paper can be used to improve the performance of Brillouin scattering-based optical fiber sensors, and can be used for the detection and diagnosis of large structures requiring long-distance measurements. In addition, they are expected to be applicable to artificial neural networks by combining artificial intelligence and big data technology.

(A) study of package technologies for improving the performance and reliability of III-V compound semiconductor-based light-emitting diodes

김호영 Greduate School, Korea University 2021 국내박사

The demand for energy saving is increasing, and due to the rapid development of LED (Light Emitting Diode) technology, a type of photonic device, next-generation lighting using LED is being applied by replacing existing incandescent bulbs and halogens. III-V compound semiconductor-based light-emitting diodes (LEDs) are becoming more and more widespread in curing and sterilization applications as well as lighting applications such as solid-state lighting, automotive headlights, and micro LED displays. With the development of LED technology and the expansion of the application range, it is necessary to continuously improve performance and cost competitiveness. As semiconductor devices that can provide high output of LEDs are requested, a III-V compound that can increase output by applying a high-power device source research on semiconductor-based optical devices is ongoing. There are several original core technologies in the technology of a III-V compound semiconductor-based optical device, and one of the important ones is packaging material and design technology. Package material is also a particularly important factor because the package material cost accounts for more than 50% of the material cost of the optical device. For this reason, research is being actively conducted on ways to improve the light extraction efficiency of semiconductor devices and improve the optical performance in terms of the package, in LED packages. In addition, research is being conducted on ways to reduce manufacturing cost and improve manufacturing yield by material optimizing and structural changes in LED packages. In addition, there has been a continuing need to improve the durability and reliability of LED packages. Therefore, design technology for the material, structure, and process of LED packages has become a particularly important factor. In this regards, this dissertation concentrates on the verification and design of materials and structures, which are the design factors of LED packages based on group III-V compound semiconductors, which are the most widely used among photonic devices, and focuses on future development directions. The first part deals with the contents of improving optical and electrical characteristics and thermal characteristics and securing reliability and stability through the development of the LED package structure. By applying the via hole type structure to the LED package for middle power, the heat dissipation characteristics are improved when the FC (Flip Chip) LED is applied to allow the high power to be applied, and minimized crack defect on the bonding surface(=die bond line) caused by CTE (Coefficient of Thermal Expansion) mismatch, and solder remelting defects that may occur during the solder reflow process are effectively prevented. In addition, we identified the penetration path of moisture inside the package that occurs during the humidity test, and based on this, we implemented the optimization of the internal structure of the package. The second part deals with the contents of improving the color quality and optical performance and securing reliability and stability by improving the characteristics of the white LED package used as a general lighting and BLU (Back Light Unit) light source. The CCT uniformity was improved through effective sedimentation of the phosphor, the color conversion material of the white LED package, and the reduction of moisture penetration inside the package improved the lifetime of the phosphor deterioration. The third part deals with improving the characteristics of the LED package used as an external light source in the automotive lighting system. By implementing red color using InGaN based chip and phosphor instead of using AlGaInP based chip, we effectively suppressed the thermal drop characteristic due to increase in ambient temperature. In addition, by developing a package structure using a vertical chip with a surface emitting structure instead of a package using a volume emitting structure flip chip, the contrast ratio, which is important for the ADB (Adaptive Driving Beam) light source, is improved, and reliability stability is secured by effectively suppressing silicon cracks caused by side light emission.

Structural and barrier properties of beeswax pickering emulsion film stabilized via polyvinyl alcohol

김민하 Greduate School, Korea University 2021 국내석사

The aim of this study was to investigate the effect of beeswax pickering emulsion on physical, structural and barrier properties by adding it to polyvinyl alcohol (PVA) composite film. Polyacrylic acid (PAA) were used as crosslinking agent to enhance the properties of PVA film. Different content of beeswax (0, 1, 5, 10, 15, 20% w/w) added in PVA/PAA solution to prepare film. The structural properties of the film through Scanning electron microscopy (SEM) and confocal laser scanning microscope (CLSM) showed irregular structure when the beeswax content increased. The addition of beeswax improved the water and oil barrier properties of the film. Oxygen permeability increased gradually from 5% beeswax. In mechanical properties, the tensile strength value showed a small decrease and the elongation increased up to 10% beeswax compared to the control, but both values showed a significant decrease from 15% beeswax. As a result, the PVA composite film with 10% beeswax added has improved moisture barrier properties while maintaining the physical properties or structural properties of the film, which is suitable for packaging.

Structural analysis of 3-Hydroxybutyryl-CoA dehydrogenase from Faecalibacterium prausnitzii A2-165

박정아 Greduate School, Korea University 2021 국내석사

Atopic dermatitis (AD) is caused by T-helper cell type 2 inflammation, the main symptom of the AD is itchy, damaged skin, red and swollen. Recent research suggests that intestinal microbes are closely linked to autoimmune disorders such as AD. The living culture such as living environment and westernized diet change the composition of the intestinal microbiome. As a result, it leads to an imbalance of intestinal microorganisms, and decreases the production of materials that play an essentials role in keeping barriers strong, such as butyric acid and propionic acid. When the fecal samples of AD and non-AD are compared, AD is strongly related to Faecalibacterium prausnitzii strain A2-165 than other bacteria. Also they are connected with metabolites, specially butyrate production. The outcome shows that I determined the crystal structures of apo and complexed with NAD⁺ cofactor of 3-Hydroxybutyryl-CoA dehydrogenase from Faecalibacterium prausnitzii strain A2-165 at 2.75 Å and 2.55 Å, respectively. An Apo crystal belongs to monoclinic space group, P2₁ with unit cell parameter, a= 87Å, b= 126Å, c= 108Å and α= 90Å, β= 110Å, γ= 90Å. Also Rwork and Rfree values are 21.4% and 25.6%. The NAD⁺ complex crystals belong to P2₁ space group and the unit cell parameters are a= 90Å, b= 80Å, c= 128Å and α= 90Å, β= 102Å, γ= 185Å. The Final Rwork and Rfree values are 20.5% and 26.6%. It shows that the target protein exists in six molecules in a dimer state in the solution using multi-angle light scattering and Gel-filtration chromatography. In addition, NAD⁺ binding site are stabilized by hydrogen bonds such as Glu90, Lys95, Asn115, Ser117 residues. More detailed crystallographic analysis is ongoing now.

Strategy for robust and efficient electrocatalyst structure based on template mediated synthesis for acidic water splitting

주진환 Greduate School, Korea University 2021 국내박사

The robust and efficient electrocatalysts for acidic water splitting are crucial for the large-scale deployment of proton exchange membrane water electrolyzer (PEMWE). Oxygen evolution reaction (OER) demands the development of highly robust and efficient catalysts more imperatively due to its corrosive condition and sluggish reaction kinetic. To develop electrocatalysts that fulfill all the criteria for OER, compositional, morphological, and electronic properties of catalysts must be finely controlled at the nanoscale. The nanotemplate-mediated synthetic method using metastable nanotemplates can accomplish the facile synthesis of electrocatalysts with the aforementioned properties for highly robust and efficient electrocatalyst. This thesis presents a unique nanotemplate-mediated synthetic method to obtain hollow dendritic nanostructures appropriate for water electrolysis. In Chapter 1, the comprehensive introduction of water electrolysis, nanotemplate-mediated synthetic method, and their applications in electrocatalysis are introduced. Chapter 2 describes the synthesis of compositionally tunable IrxRu1-x hollow dendritic nanoparticles via additional growth reaction on the Cu2-xS@IrSy core-shell nanoparticles. This work exhibits the metastable nature of Cu2-xS@IrSy nanotemplates in that the in situ-removal of Cu2-xS core during additional growth reaction of iridium and ruthenium at high temperature and a large portion of copper and sulfur alloying into the iridium and ruthenium. The resulting IrxRu1-x hollow dendritic nanoparticles exhibited an outstanding water electrolysis performance by taking over the beneficial doping effect of reactive Cu2-xS@IrSy nanotemplates. In Chapter 3, an additional manganese dopant effect on Ru2Ir hollow dendritic nanoparticles was synthesized via the Cu2-xS template-mediated method, and its OER catalytic properties are reported. The manganese dopant allows the implementation of IrO2/Ru heterostructure by controlling the reduction rate of iridium precursor. These fine-controlled IrO2/Ru heterostructures showed a dramatically enhanced activity and stability in OER compared to simple core-shell and other previously reported Ir- and Ru-based electrocatalysts. Furthermore, the nanotemplate-mediated synthetic route offers the most versatile method appropriate to the scaling their solvothermal synthesis.

Stability boosting charge transfer from heterointerfaced Pt to catalytic RuO2 toward durable oxygen evolution reaction under acidic media

양희수 Greduate School, Korea University 2021 국내석사

With the increasing demand for renewable energy sources, electrolytic water splitting has become a promising technology to overcome environmental issues. In particular, proton-exchange membrane water electrolysis (PEMWE) has been widely developed because it has high proton conductivity and can be operated in low temperature. On the other hand, oxygen evolution reaction (OER) is half reaction of PEMWE and requires a large overpotential. Therefore, OER is one of the factors that lowers the low catalytic stability. Due to the stability issues, RuOx phase has not been considered as a candidate for commercially viable OER catalyst, although RuOx shows a great innate activity toward acidic OER. Therefore, improving stability of RuOx catalysts and the success of PEM technology are closely related. Studies have shown that doping Pt on RuO2 shell solve the stability issues. However, few studies have focused on how Pt and RuO2 interact and affect stability during the reaction. Therefore, we deposited three different types of Ru shells on the Pt-based nanocables: hetero-interface type, core-shell type, and merely doped type. Among them, the hetero-interface type reached a current density of 10 mA cm-2 at an overpotential of 215 mV and maintained their catalytic ability even after 40 h of stability test, surpassing previous work activity.

SSTI mitigation strategy based on delay reconstruction approach for nonlinear time series data

노혁진 Greduate School, Korea University 2021 국내박사

This paper proposes a mitigation strategy using the delay reconstruction on damping controller of flexible AC transmission system when sub-synchronous torsional interaction is detected during the operation. As fast-response power electronics such as renewable energy has increased in power systems, the range to be observed from the stability point of view has been expanded to electromagnetic transients. Previously, severe conditions can be avoided by analyzing at the planning stage to suppress the vulnerability of sub-synchronous resonance. However, as the complexity increases due to the penetration of renewable energy and topology changes in power systems, modelling of planning and analysis for all scenarios has become difficult. Therefore, recently, interest in real-time oscillation analysis based on measurement data has increased. The controller proposed in this paper reconstructs nonlinear time series data to detect sub-synchronous torsional interaction in real-time and generates an alert signal. The protection function that disconnects the affected generated resource can burden the system operation. Therefore, the controller mitigates the oscillation by controlling the current regulator gain and firing angle using the geometric length of the trajectory away from the equilibrium point of the reconstructed space.

Spiking neuron using a bipolar Electrochemical Metallization (ECM) switch and its enhanced spiking properties through filament confinement

김태현 Greduate School, Korea University 2021 국내석사

Neural networks composed of artificial neurons and synapses mimicking the human nervous system have received much attention because of their promising potential in future computing systems. Especially, spiking neural networks (SNNs), which are faster and more energy-efficient than conventional artificial neural networks, have recently been focus of attention. However, because typical neural devices for SNNs are based on complementary metal-oxide-semiconductors that exhibit high consumption of power and require a large area, it is difficult to use them to implement a large-scale network. Thus, a new structure should be developed to overcome the typical problems that have been encountered and to emulate bio-realistic functions. This study proposes a versatile artificial neuron based on the bipolar electrochemical metallization threshold switch, which exhibits four requisite characteristics for a spiking neuron: all-or-nothing spiking, threshold-driven spiking, refractory period, and strength-modulated frequency. Furthermore, unique features such as an inhibitory postsynaptic potential and and the bipolar switching characteristic for changing synaptic weight are realized. Additionally, by using a filament confinement technique, a high on/off ratio (~ 6 × 10^7), a low threshold voltage (0.19 V), low variability (0.014), and endurance over 10^6 cycles are achieved. This research will serve as a stepping-stone for advanced large-scale neuromorphic systems. 인간 신경계를 모방한 인공 뉴런과 시냅스로 이루어진 신경망은 차세대 컴퓨팅 시스템에서의 잠재력을 인정받아 많은 관심을 받고 있다. 특히, 스파이킹 신경망 (SNNs)의 경우 기존의 인공 신경망보다 더욱 빠르고 전력 효율이 높아 더욱 관심이 집중되고 있다. 하지만, 기존의 스파이킹 신경망을 위한 뉴런 소자의 경우 넓은 면적과 높은 전력을 소모하는 CMOS를 기반으로 하고 있기 때문에 이를 사용하여 큰 규모의 망을 형성하는 것이 어렵다. 따라서, 이러한 문제들을 해결하면서도 생물학적 기능을 모방하기 위해 새로운 구조가 개발되어야 한다. 본 논문에서는 양방향 전기화학적 금속화 스위치를 기반으로 한 다목적 인공 뉴런을 소개하고자 한다. 이 뉴런은 스파이킹 뉴런의 필수적인 4가지 특성(1. 양자택일적 스파이킹, 2. 임계값 기반 스파이킹 3. 불응기, 4. 강도 변조 주파수)을 보였다. 또한, 억제성 시냅스 후 전위와 시냅스 가중치 변경을 위한 양방을 스위칭 특성과 같은 특성들을 구현하였다. 추가적으로, 필라멘트 제한 기술을 사용하여 높은 on/off ( ~6x10^7) 비율, 낮은 임계전압 (0.19V), 낮은 변산도 (0.014), 그리고 10^6 회 이상의 내구성을 달성하였다. 이러한 연구는 발전된 대규모 신경망 구현을 위한 디딤돌이 될 것이다.

Solubility of carboxylic acid in various solvents

안은경 Greduate School, Korea University 2021 국내석사

Carboxylic acid is a useful material widely used in field of chemical engineering and produced from oil or biomass resources. It is produced with by-products. Therefore, the separation is an important to obtain carboxylic acid with high purity. Crystallization is one of the promising ways to separate pure carboxylic acid efficiently. Solubility data are essential for operation and design in the crystallization process. In the study, best solvents are selected for the design and operation of crystallization process. The solubility and uncertainty of carboxylic acid and various solvents were measured. We selected reliable data and optimized parameter of the thermodynamic model such as modified Apelblat equation and the NRTL. Binary solubility was correlated by the parameters of the modified Apelblat equation. In cooling crystallization, ethanol and DMF were best solvent for cooling crystallization of adipic acid, succinic acid acid, respectively. Ternary solid-liquid equilibrium for drowning-out crystallization was predicted using NRTL. As a result of the prediction, it was predicted and verified that water could be used for drowning-out crystallization in adipic acid and water mixture.