Earticle

To prevent low yields in the semiconductor industry is crucial to the success of that industry. However, to prevent low yields is difficult because of too many factors to affect yield variation and their complex relation in the semiconductor manufacturing process. This study presents a new efficient detection methodology for detecting abnormal yields including high and low yields, which can forecast the yield level of a production unit (namely a lot) based on yield-related feature variables" behaviors. In the methodology, we use C5.0 to identify the yield-related feature variables that are the combination of correlated process variables associated with yield, use SOM (Self-Organizing Map) neural networks to extract and classify significant patterns of past abnormal yield lots and finally use C5.0 to generate classification rules for detecting abnormal yield lot. We illustrate the effectiveness of our methodology using a semiconductor manufacturing company’s field data.

The power of semiconductor, Korea is continuously constructing semiconductor production line for keeping a front-runner status. however, studies and data about potential risks in semiconductor factory are still short. If fire does not initially suppressed, the fire causes a great damage. To decrease fire risk factors, in addition to fire fighting safety equipment, more important thing is how to design and construct fire protection system. The current fire protection codes about semiconductor factory come under functional law, and this law is short of consideration about particularity of factory. The existing prescriptive fire codes depending on experience compose without evident engineering verifications, thus equipments which is created by the current prescriptive fire code may bring about a variety of problems. For example, the design under the current regulation can not cope with the excessive investments, low efficiencies, and the diversifying construction designs and be applied to the quick changes of new technologies. Ergo, an optimal design for fire protection is to equip fire protection arrangements with condition and environment of production field. Manufacturing factory of semiconductors is a windowless airtight space. And for cleanliness, there exists strong flow of cooperation. Therefore, there is a need for fire safety design that meets the characteristic of a clean room. Accordingly, we are to derive smoke flow according to cooperation process within a clean room and construction plan of an optimal sensor system. In this study, in order to confirm the performance of proposed smoke-exhaust equipment and suggest efficient smoke exhaust device when there is a fire of 1MW of methane in the clean room of company H, we have implemented fire simulation using fluid dynamics computation.

LCD와 반도체 제조공정에서 발생하는 인산, 질산, 초산, Al, Mo 등이 혼재하고 있는 인산계 혼산폐액을 액정제조공정에서 사용할 수 있는 고순도 에칭액으로 재활용하기 위해서 용매추출법, 진공증발법, 확산투석법 및 이온교환법의 각각의 기술적 특성을 살린 혼합 처리공정을 이용하여 고순도 인산회수 기술을 확립하고 상용화 시스템을 개발하고자 하였다. 시험 결과 진공증발에 의해 질산과 초산을 100% 제거할 수 있었고, TOP를 이용한 용매추출에서도 추출 4단, 탈거 6단으로 완벽하게 제거할 수 있었다. 이온교환의 전단계로 적용한 확산투석에서 Al 97.5%, Mo 36.7% 제거할 수 있었고 이온교환공정에서 Al 및 Mo를 각각 1ppm 이하로 정제할 수 있었다.

The waste solution discharged from the LCD manufacturing process contains acids like nitric, acetic and phosphoric acid and metal ions such as Al, Mo and other impurities. It is important to remove impurities less than 1 ppm in phosphoric acid to reuse as an etchant because the residual impurities even in sub-ppm concentration in semiconductor materials play a major role on the electronic properties. In this study, a mixed system of solvent extraction, diffusion dialysis and ion-exchange was developed to commercialize in an efficient system for recovering the high-purity phosphoric acid. By vacuum evaporation, almost 99% of nitric and acetic acid was removed. And by solvent extraction method with tri-octyl phosphate (TOP) as an extractant, the removal of acetic and nitric acid from the acid mixture was achieved effectively at the ratio A/O=1/3 with 4th stage of extraction stage. About 97.5% of Al and 36.7% of Mo were removed by diffusion dialysis. Essentially almost complete removal of metal ions and purification of high-purity phosphoric acid could be obtained by using ion exchange.

현재 구축되어 있는 반도체 공정에서의 상태감시 시스템은 센서 데이터를 수동으로 수집하는 방식으로써 복합 장애 검출이나 실시간 감시에서 한계가 존재한다. 본 논문에서는 영역 온톨로지를 구성하여 시간에 따른 관계망을 형성하는 상황인지 알고리즘을 설계하고 이를 통해 반도체 공정에서 위험요소가 발견되는 부분에 대해서 이벤트를 생성하여 사용자에게 서비스하는 시스템을 제안하며, 이를 구현하기 위해 상황 추론을 위한 다중센서 노드를 설계하고 이를 실험하였다. 실험 결과, 다수의 수집된 데이터에서 시간에 대한 관계가 형성된 내용에 대해서는 시간적 규칙추론이 적용된 이벤트가 발생하였으며 오작동 및 외부의 시간적 요인에서 발생되는 이벤트는 Log로만 데이터를 제공하는 것을 확인할 수 있었다.

The current state monitoring system at a semiconductor manufacturing process is based on the manually collected sensor data, which involves limitations when it comes to complex malfunction detection and real time monitoring. This study aims to design a situation recognition algorithm to form a network over time by creating a domain ontology and to suggest a system to provide users with services by generating events upon finding risk factors in the semiconductor process. To this end, a multiple sensor node for situational inference was designed and tested. As a result of the experiment, events to which the rule of time inference was applied occurred for the contents formed over time with regard to a quantity of collected data while the events that occurred with regard to malfunction and external time factors provided log data only.

The evolution of semiconductor manufacturing technology has accelerated the reduction of device dimensions and the increase of integrated circuit density. In order to improve yield within a short turn around time and maintain it at high level, a system that can rapidly determine problematic processing steps is needed. The statistical process control detects abnormal process variation of key parameters. Expert systems in SPC can serve as a valuable tool to automate the analysis and interpretation of control charts. A set of IF-THEN rules was used to formalize knowledge base of special causes. This research proposes a strategy to apply expert system to SPC in semiconductor manufacturing. In analysis, the expert system accomplishes the instability detection of process parameter, In diagnosis, an engineer is supported by process analyzer program. An example has been used to demonstrate the expert system and the process analyzer.

본 연구는 반도체 제조사업장 직원 중에서 이직을 결정한 후 상사 또는 인사 관리자와 이직면담을 하면서 이직결정을 철회한 직원을 대상으로 이들의 이직결정 철회에 기여한 이직면담과정을 심층적으로 이해하고자 실시되었다. 이를 위해 연구목적에 적합한 연구대상을 섭외하였는데 반도체 제조사업장에 근무하는 직원 중에서 이직면담을 통해 이직을 철회한 경험이 있는 12명의 직원을 연구대상으로 반구조화된 면접질문지를 제작하여 심층 면접을 통해 자료를 수집하였으며 합의적 질적연구방법(CQR)으로 자료를 분석하였다. 연구문제를 중심으로 분석하여 9개 영역과 26개의 중심개념이 도출되었으며 이직을 결심한 이유는 ‘개인적인 사유’, ‘상사 동료 부하직원과의 갈등’ 및 ‘통제가 안 되는 업무 스트레스’라는 3개 영역이 도출되었고, 이직면담과정을 통해 이직결정을 철회한 요인으로는 ‘면담자 자질’, ‘면담자의 태도 및 기법’, ‘정서적, 인지적 변화’, ‘문제의 해결’이라는 4개 영역이, 이직결정 철회 후 변화된 것으로 ‘조직에 대한 시각 변화’과 ‘약해진 자아개념’이라는 2개 영역이 도출되었다. 연구결과를 바탕으로 조직의 인사관리 측면에서 시사점을 논의하였다.

This study aims to explore the reasons why employees of semiconductor manufacturing plants make a turnover decision, and their experiences on the factors that influence the withdrawal of a turnover decision during an exit interview with one’s superior and a human resources department staff in charge. The total number of participants was twelve, and the data was collected by conducting an in-depth interview and analyzed by Consensual Qualitative Research (CQR) method. The study results showed that there were three determinants of making a turnover decision: personal reasons irrelevant to an organization, conflicts with one’s superior, colleagues and subordinates, and a stress due to uncontrollable tasks. Four major factors that influence one’s withdrawal of a turnover decision during an exit interview with one’s superior and a human resources staff in charge were found: the interviewer him or herself, interviewer’s interview skills and attitude, participants’s cognitive and emotional changes, and the problem solving. Following the making of a turnover decision and a withdrawal, the participants experienced the weakening of a self-concept and the changes to their perspectives on an organization. This study is meaningful in the following respects. The research aimed at the members of an organization who actually had a turnover decision, and examined the factors that influenced the withdrawal of a turnover decision through an exit interview process for the first time. This study could serve as the basis when introducing a workplace counseling system or a training program for counseling techniques within an organization or a company, whilst also contributing significantly to the development of interview, leadership and communication training programs for the managers carrying out employee interviews within an organization.

우리는 반도체 제조 공정에서 고속 회전하는 웨이퍼의 비정상 상태를 효과적으로 탐지하기 위한 이미지 세그멘테이션 기반 접근을 제안한다. 기존 CNN 기반 방법은 웨이퍼 전체를 통합적으로 인지하기 어렵고, 반사·난반사 조건에서 패치 영역 구분에 제한이 있었다. 우리는 HSV 색공간 변환 및 ViT 기반 패치 세그멘테이션 모델을 활용하여 불균일 조명 및 왜곡 조건에서 이상 패턴을 검출하였다. 연구에서는 목업(Mock-up) 기반 실험 데이터를 구축하여 웨이퍼 wobbling 조건에 따른 세그멘테이션 성능을 평가하였으며, hard-case 상황에서도 ViT 기반 접근이 가능함을 실헙적으로 보였다. 또한 YOLOv11-seg와의 비교를 통해 CNN 기반 bounding-box 세그멘테이션의 구조적 한계를 확인하였다. 결과적으로 본 연구는 실환경에서 발생 가능한 다양한 광학 조건 및 노이즈 상황에 대응 가능한 세그멘테이션 프레임워크를 제시하며, 향후 반도체 공정 중 웨이퍼 클리닝 실공정 데이터 활용을 통한 확장 실험의 방향성을 제시하였다.

We propose an image-segmentation-based approach for effectively detecting abnormal states in high-speed rotating wafers within semiconductor manufacturing processes. Conventional CNN-based methods struggle to capture the global structural patterns of the entire wafer and face limitations in distinguishing patch regions under reflective and refractive optical distortions. We employ HSV color-space transformation and a ViT-based patch segmentation model to detect abnormal patterns even under uneven illumination and geometric distortions. In this study, we constructed a mock-up experimental dataset and evaluated segmentation performance under different wafer wobbling conditions, demonstrating that the ViT-based method remains effective even in hard-case scenarios. Moreover, through comparison with YOLOv11-seg, we confirmed the structural limitations of CNN-based bounding-box segmentation approaches. Consequently, this study presents a segmentation framework capable of addressing various optical variations and noise conditions that may occur in real environments, and suggests a research direction involving expanded testing with real wafer-cleaning process data.

Bacteria were isolated and identified from an advanced high-purity water system that supplies ultra-high purity water (UHPW) for 16-megabyte DRAM semiconductor manufacturing. Scanning electron microscopic and microbiological observations revealed that the primary source of the bacteria isolated from the UHPW was detached cells from biofilms developed on the pipe wall through which the UHPW, a man-made and extremely nutrient poor environment, was passing. About 63-65% of the bacteria isolated from the UHPW and the pipe wall were Gram-positive, whereas only 10% of the bacteria isolated from the feed water were Gram-positive. The of Gram-positive bacteria and seven genera of Gram-negative bacteria. Strains of the UHPW bacteria effectively adhered to and formed a biofilm on the surface of polyvinyl chloride (PVC) pipe.

Chemical mechanical planarization is one of the core processes in fabrication of semiconductors, which are increasingly used for information storage devices like solid state drives. For higher data capacity in storage devices, CMP process is required to show ultimate precision and accuracy. In this work, 2-dimensional finite element models were developed to investigate the effects of the slurry particle impact on microscratch generation and the phenomena generated at pad-particle-wafer contact interface. The results revealed that no plastic deformation and corresponding material removal could be generated by simple impact of slurry particles under real CMP conditions. From the results of finite element simulations, it could be concluded that the pad-particle mixture formed in CMP process would be one of major factors leading to microscratch generation.

A life-cycle assessment (LCA) is based on the attention given to the environmental protection and concerning the possible impact while producing, making, and consuming products. It includes all environmental concerns and the potential impact of a product's life cycle from raw material procurement, manufacturing, usage, and disposal (that is, from cradle to grave). This study assesses the environmental impact of the ultra pure water process of semiconductor manufacturing by a life-cycle assessment in order to point out the heavy environmental impact process for industry when attempting a balanced point between production and environmental protection. The main purpose of this research is studying the development and application of this technology by setting the ultra pure water of semiconductor manufacturing as a target. We evaluate the environmental impact of the Precoat filter process and the Cation/Anion (C/A) filter process of an ultra pure water manufacturing process. The difference is filter material used produces different water quality and waste material, and has a significant, different environmental influence. Finally, we calculate the cost by engineering economics so as to analyze deeply the minimized environmental impact and suitable process that can be accepted by industry. The structure of this study is mainly combined with a life-cycle assessment by implementing analysis software, using SimaPro as a tool. We clearly understand the environmental impact of ultra pure water of semiconductor used and provide a promotion alternative to the heavy environmental impact items by calculating the environmental impact during a life cycle. At the same time, we specify the cost of reducing the environmental impact by a life-cycle cost analysis.

최근 반도체 공정은 대기업을 중심으로 4차산업에 따른 수요 증가로 메모리 반도체에서 파운드리로의 공정변화를 모색하고 있으며 10나노(nm) 공정에서 3 나노(nm)이하 공정으로의 초미세화 공정개발과 같이 산업이 확대되고 있다. 이러한 반도체 칩을 생산하기 위해 사용되는 원료인 특수가스 및 Precursor(전구체) 등의 특성은 유독성, 자연발화성, 인화성, 부식성 물질이다. 이러한 반도체 원료들은 폐쇄시스템으로 운영이 되어 정상 중에는 누출이 되지 않으나 누출 될 경우 가스박스 내부로 확산되고 가스박스 내부에서 적절한 환기가 되지 않는 경우 가스박스 외부로 확산되어 화재, 폭발을 일으키거나 독성 물질의 누출 등 큰 사고로 이어질 수 있다. 최근 반도체 산업에서 원료가 폐쇄시스템으로부터 누출되어 가스박스 내부 및 외부로 확산되는 사고 사례가 발생하고 있으나 가스박스 내부의 적정환기 시스템의 적용 이외에 적절한 예방대책을 제시한 연구 사례를 찾아 볼 수 없었다. 본 연구에서는 반도체 원료 이송배관의 연결부위인 VCR 피팅에서 원료가 누출되어 가스박스 외부로 확산된 사고사례를 바탕으로 이에 대한 예방대책을 제시하고자 한다.

Recently, in the semiconductor process, large companies are seeking process changes from memory semiconductors to the foundry due to the increase in demand due to the 4th industry. industry is expanding. The characteristics of special gases and precursors, which are raw materials used to produce these semiconductor chips, are toxic, pyrophoric, inflammable, and corrosive. These semiconductor raw materials are operated in a closed system and do not leak to the outside during normal times, but when leaked, they spread to the inside of the gas box, and when proper ventilation is not provided inside the gas box, they spread to the outside, causing fires, explosions, or toxic substances. It can lead to major accidents such as leakage. Recently, there have been cases of accidents in which hazardous materials leaked from the closed system of the semi conductor process and spread to the inside and outside of the gas box. . In this study, we propose preventive measures based on the case of an accident in which raw material leaked from the VCR fitting, which is the connection part of the semiconductor raw material transfer pipe, and spread to the outside of the gas box.

반도체 산업은 집적화와 제품의 원가절감 및 생산수율 향상에 상반되는 초청정 제어 기술의 한 계에 도전하고 있으며, 앞으로 이에 대한 기술축적이 없으면 반도체 산업의 발전은 이루어질수가 없을 만큼중요한 것이라고 사료된다. 초고집적회로의 본격적인 양산시대를 맞이하여, 오염제어 기술에서도 종래의 입자오염이외에, 유기물, 금속 및 가스상 오염물에 대한 복합적인 새로운 계측, 제어 기술이 개발되어져야 할 것이다. 그리고 공기청정화 환경기술뿐만 아니라, 온 . 습도, 기류, 정전기 및 진동제어 관한 극한기술도 예외가 될 수 없으며 큰 비중을 차지하게 될 것이다. 공조환경 제어기술에서 에너지 절감 및 운전비 절감에로의 연구와 함께 온 . 습도 제어 정도에 서도 보다 응답성 빠르고, 정밀도 높은 시스템 개발에 대한 연구가 병행되어야 할 것이다.

목적 본 연구는 반도체 제조공정 전문인력 양성을 위하여 반도체 관련학과 재학생을 대상으로 적용할 수 있는 NCS(National Competency Standards) 기반 교육 프로그램을 개발하고자 수행되었다. 방법 이를 위하여 반도체 분야 내용 전문가와 교육학 분야 교수설계 전문가를 포함하여 개발팀을 구성하고 선행연구 분석을 통해차세대 반도체 제조공정 분야 NCS를 선별한 후, 대안적 교수설계 모형인 RPISD(Rapid Prototyping Instructional System Design) 모형의 절차와 방법에 따라 교육 프로그램을 개발하였다. 결과 본 연구에서 개발한 교육 프로그램의 전체 과정은 크게 이론교육을 위한 집체교육과 실습교육을 위한 소집단교육으로 구성되었다. 먼저 이론교육의 경우 학습자의 선수지식 차이를 보완하기 위해 온⋅오프라인 블렌디드로 설계하였다. 기초 내용은 온라인수업으로 운영하여 학생 개인의 수준에 따라 동영상 강의를 반복 수강할 수 있도록 하고, 심화 내용은 오프라인 수업으로 운영하여실습교육과 연계되도록 하였다. 그리고 실습교육의 경우 프로그램 전반에 걸쳐 팀 기반 프로젝트학습방법을 적용하여 스스로 계획하고 해결안을 구상하는 실천적 활동을 통해 학습하도록 하였다. 또한 직무적합성 검사와 반도체 기업 면접 멘토링 등 취업지원 활동을 프로그램에 포함하였다. 결론 본 연구는 반도체 관련학과 재학생들이 반도체 제조공정 교육과 실무를 경험할 기회가 매우 적은 실정에서 해당 직무역량을개발할 수 있는 프로그램을 개발하였다는 데 의의가 있다. 특히 RPISD 모형을 적용하여 프로그램을 개발함으로써 교육수요자의의견을 반영하였다는 점과, 온⋅오프라인 블렌디드, 팀 기반 프로젝트학습을 활용하여 구성주의적 교수-학습방법으로 수업을 구성하였다는 점에서 더욱 의미가 있다. 이러한 다각도의 교육 프로그램 개발 노력이 반도체 산업에 더욱 다양한 전공의 취업준비생들과일반 이직준비자들이 유입될 수 있는 발판으로 작용할 수 있을 것이라 기대한다.

Objectives This study was conducted to develop an education program based on the NCS(National Competency Standards) that can be applied to students in semiconductor-related departments to train semiconductor manufacturing process experts. Methods To this end, a development team was formed including content experts in the semiconductor field and instructional design experts in the pedagogy field. In addition, NCS in the next-generation semiconductor manufacturing process field was selected through analysis of previous research on job competencies in the semiconductor field. Afterwards, an educational program was developed according to the procedures and methods of the RPISD(Rapid Prototype to Instructional System Design) model, an alternative instructional design model. Results The training program developed in this study consisted of two main components: group training for theoretical training and subgroup training for practical training. Theoretical education was designed as a mixture of online and offline methods to accommodate differences in learners' prerequisite knowledge, allowing students to repeat the course according to their individual level. In addition, content related to equipment operation and advanced learning was conducted offline to increase convergence with practical training. In addition, the entire program was operated as a team-based project learning to promote constructivist learning activities. In addition, programs such as career suitability assessment, counseling, and semiconductor company interview mentoring were also conducted to support employment activities. Conclusions This study is significant in that it developed an educational program that can develop job competencies for job seekers in semiconductor-related departments in a situation where there are very few opportunities to experience the semiconductor manufacturing process. In particular, it is more meaningful in that it was developed systematically by actively reflecting the opinions of education consumers, and that classes were structured using constructivist teaching and learning methods using online/offline blending and team-based project learning.

본 연구에서는 반도체 제조공정에서 발생하는 실리콘 슬러지에 금속산화물인 이산화티타늄을 코팅하였으며, 이를 안료 및 도료의 소재로 응용하였다. 상세히는, 불순물이 제거된 판상형의 실리콘 슬러지 분말에 졸-겔법을 통해 이산화티타늄을 코팅하였다. 안료 및 도료의 소재로서 응용 가능성을 확인하기 위해서 친수성 투명 코팅제에 분산하여 페인트를 제조한 후 유리 기판에 분사하였다. 그 결과, 이산화티타늄의 도입으로 인한 발림성 향상과 하얀색의 발현을 확인할 수 있었다. 본 연구를 통해서 실리콘 슬러지에 이산화티타늄을 코팅함으로써 발림성 및 친수성이 향상됨을 확인하였으며, 이를 통해서 반도체 공정에서 발생하는 실리콘 슬러지가 안료 및 도료의 소재로서 응용 가능성을 확인하였다.

In this study, silicon sludge generated in semiconductor manufacturing process is recycled and applied as materials for pigments and paints. In detail, metallic impurities are removed from silicon sludge to obtain plate-like silicon sludge powder (SW-sludge), which is then coated with titanium dioxide via sol-gel method (TCS-sludge). SW-sludge and TCS-sludge are dispersed in hydrophilic transparent varnish and sprayed onto glass substrates to observe the possibility for the application as materials for pigments and paints. Notably, the applicability of TCS-sludge-based paint is improved compared to SW-sludge-based paint after the titanium dioxide coating. Moreover, the color of TCS-sludge-based paint turns into white. Accordingly, it is confirmed that the applicability and hydrophilicity are improved by the presence of outer titanium dioxide layer. In this regard, it is expected that the recycled TCS-sludge may be a future material for the application as pigments and paints.

A new scheduling algorithm for large scale semiconductor processes is addressed. The difficulties of scheduling for semiconductor fabrication processes are mainly due from repeating production of wafers that experience reentrant flows. Sequence branch algorithm (SBA) is proposed for large real scheduling problems when all processing times are deterministic. The SBA is based on the reachability graph of Petri net of which the several defects such as memory consumption and system deadlock are complemented. Though the SBA shows the solution deviating a little from the optimal solution of mixed integer programming, it is adjustable for large size scheduling problems. Especially, it shows a potential that is capable of handling commercial size problems that are intractable with mathematical programming.

반도체 성능 향상으로 신호를 전달하는 회로의 단위가 마이크로 미터에서 나노미터로 미세화되어 선폭(linewidth)이 점점 좁아지고 있다. 이러한 변화는 검출해야 할 불량의 크기가 작아지고, 정상 공정상태와 비정상 공정상태의 차이도 상대적으로 감소되어, 공정오차 및 공정조건의 허용범위가 축소되었음을 의미한다. 따라서 검출해야 할 이상징후 탐지가 더욱 어렵게 되어, 높은 정밀도와 해상도를 갖는 검사공정이 요구되고 있다. 이러한 이유로, 미세 공정변화를 파악할 수 있는 신규 검사 및 계측 공정이 추가되어 TAT(Turn-around Time)가 증가하게 되었고, 웨이퍼가 가공되어 완제품까지 도달하는데 필요한 공정시간이 증가하여 제조원가 상승의 원인으로 작용한다. 본 논문에서는 웨이퍼의 검계측 데이터가 아닌, 제조공정 과정에서 발생하는 다양한 센서 및 장비 데이터를 기반으로 웨이퍼 제조 결과가 양품인지 그렇지 않으면 불량인지 구별할 수 있는 가상계측 모델을 제안한다. 기계학습의 여러 알고리즘 중에서 다양한 장점을 갖는 XGBoost 알고리즘을 이용하여 예측모델을 구축하였고, 데이터 전처리(data-preprocessing), 주요변수 추출(feature selection), 모델 구축(model design), 모델 평가(model evaluation)의 순서로 연구를 수행하였다. 결과적으로 약 94% 이상의 정확성을 갖는 모형을 구축하는데 성공하였으나 더욱 높은 정확성을 확보하기 위해서는 반도체 공정과 관련된 Domain Knowledge 를 반영한 모델구축과 같은 추가적인 연구가 필요하다.