Search NASA⌕ Search

SEARCH · Search NASA

Results for “semiconductor manufacturing”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 19 records

Applications of Federated Learning in Semiconductor Manufacturing [Poster]

As semiconductor manufacturers explore advanced data analytics and modeling techniques and data hungry machine learning models increase in popularity due to their accuracy in solving generalized problems and ability to learn complex relationships, federated learning emerges as a privacy preserving machine learning technique for preserving data privacy and ensuring intellectual property protection. Federated Learning is a machine learning technique focused on training models using distributed data that never needs to be centrally stored, allowing the use of advanced machine learning techniques without compromising data privacy, and in the semiconductor manufacturing industry advanced machine learning techniques can reduce cost and time, but maintaining data privacy is essential to maintaining a competitive advantage. This paper systematically reviews existing literature on applications of federated learning in the semiconductor manufacturing industry with a focus on identifying common themes, algorithms, and gaps within the literature to drive future research directions. The findings reveal five key themes, including improvements in quality assurance, virtual models, privacy preservation, reliable data practices, and emerging trends and developments. By identifying key themes in literature on federated learning and semiconductor manufacturing and analyzing gaps and discussed methodologies, this study highlights several potential future research directions to expand the application of federated learning techniques in the semiconductor manufacturing domain.

42 ENGINEERING↗

Algae-based sorbents for removal of gallium from semiconductor manufacturing wastewater

Gallium is increasingly used in industry, and particularly as a component of new materials that have entered semiconductor manufacturing. The release of gallium in factory’s wastewater is potentially an environmental issue and a human health concern. Utilization of algae-based sorbents in the treatment of wastewater polluted by heavy metals has been regarded as a potentially efficient and environmentally-friendly technology. Removal of gallium from simulated wastewater using dead freshwater green algae, ( Chlorella sorokiniana ), as a sorbent was studied. Both batch and continuous flow operations were investigated. Mass balance-based process models were developed and validated for both batch and flow systems. These models were found to be valuable for both understanding the process steps as well as for analyzing the experimental data and determining the fundamental kinetics and thermodynamic parameters. Adsorption capacity was found to be 14.1 mg/g at pH of 2.3, increasing to 38.5 mg/g when the pH is raised to 2.8. The effect of pH is attributed to the surface charge interactions between biomass and gallium and indicates that gallium can be acid-washed and desorbed from the biomass. This property allows for capture followed by recovery of gallium from wastewater using algae based biosorbents.

09 BIOMASS FUELS↗

Simulating Energy and Security Interactions in Semiconductor Manufacturing: Insights from the Intel Minifab Model

Semiconductor manufacturing is a highly complex. Fabrication plants must deal with re-entrant flows to support multiple types of wafers being produced simultaneously, each with their own deadlines and specifications. The manufacturing process itself depends upon the ability to control and programmatically adjust a variety of environmental conditions including temperature, humidity, and air quality. In addition, wafer fabrication consumes large amounts of electricity. Emerging technologies may help reduce the energy footprint of such facilities but can introduce cybersecurity risks. Therefore, this paper presents a modeling and simulation framework to quantify tradeoffs between operational measures of performance, energy consumption, and cybersecurity controls. We augment the Intel Minifab model, with the Purdue Enterprise Reference Architecture (PERA) for cybersecurity as well as tool-level energy consumption data from a real-world semiconductor manufacturing testbed. In this manner, we intend to provide stakeholders with a systematic, data-driven approach to evaluate emerging risks within the manufacturing process.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Cleaner Chips: Decarbonization in Semiconductor Manufacturing

The growth of the information and communication technology sector has vastly accelerated in recent decades because of advancements in digitalization and Artificial Intelligence (AI). Scope 1, 2, and 3 greenhouse gas emissions data of the top six semiconductor manufacturing companies (Samsung Electronics, Taiwan Semiconductor Manufacturing Corporation, Micron, SK Hynix, Kioxia, and Intel) were gathered from the publicly accessible Carbon Disclosure Project’s (CDP) website for 2020. Scope 3 emissions had the largest share in total annual emissions with an average share of 52%, followed by Scope 2 (32%) and Scope 1 (16%). Because of the absence of a standardized methodology for Scope 3 emissions estimation, each company used different methodologies that resulted in differences in emissions values. An analysis of the CDP reporting data did not reveal information on strategies implemented by companies to reduce Scope 3 emissions. The use of renewable energy certificates had the largest effect on decarbonization centered on reducing Scope 2 emissions, followed by the deployment of perfluorocarbon reduction technologies to help reduce Scope 1 fugitive emissions. Technology-specific marginal abatement costs of CO2 were also estimated and varied between −416 and 12,215 USD/t CO2 eq., which primarily varied depending on the technology deployed.

42 ENGINEERING↗

An Assessment of Critical Dimension Small Angle X-ray Scattering Metrology for Advanced Semiconductor Manufacturing

Simultaneous migration of planar transistors to FinFET architectures, the introduction of a plurality of materials to ensure suitable electrical characteristics, and the establishment of reliable multiple patterning lithography schemes to pattern sub-10 nm feature sizes imposes formidable challenges to current in-line dimensional metrologies. Because the shape of a FinFET channel cross-section immediately influences the electrical characteristics, the evaluation of 3D device structures requires measurement of parameters beyond traditional critical dimension (CD), including their sidewall angles, top corner rounding and footing, roughness, recesses and undercuts at single nanometer dimensions; thus, metrologies require sub-nm and approaching atomic level measurement uncertainty. Synchrotron critical dimension small angle X-ray scattering (CD-SAXS) has unique capabilities to non-destructively monitor the cross-section shape of surface structures with single nanometer uncertainty and can perform overlay metrology to sub-nm uncertainty. In this dissertation, we perform a systematic experimental investigation using CD-SAXS metrology on a hierarchy of semiconductor 3D device architectures including, high-aspect-ratio contact holes, H2 annealed Si fins, and a series of grating type samples at multiple points along a FinFET fabrication process increasing in structural intricacy and ending with fully fabricated FinFET. Comparative studies between CD-SAXS metrology and other relevant semiconductor dimensional metrologies, particularly CDSEM, CD-AFM and TEM are used to determine physical limits of CD-SAXS approach for advanced semiconductor samples. CD-SAXS experimental tradeoffs, advice for model-dependent analysis and thoughts on the compatibility with a semiconductor manufacturing environment are discussed.

77 NANOSCIENCE AND NANOTECHNOLOGY↗

Intelligent monitoring and control of semiconductor manufacturing equipment

The use of AI methods to monitor and control semiconductor fabrication in a state-of-the-art manufacturing environment called the Rapid Thermal Multiprocessor is described. Semiconductor fabrication involves many complex processing steps with limited opportunities to measure process and product properties. By applying additional process and product knowledge to that limited data, AI methods augment classical control methods by detecting abnormalities and trends, predicting failures, diagnosing, planning corrective action sequences, explaining diagnoses or predictions, and reacting to anomalous conditions that classical control systems typically would not correct. Research methodology and issues are discussed, and two diagnosis scenarios are examined.

Murdock, Janet L.↗

Planning for the semiconductor manufacturer of the future

Texas Instruments (TI) is currently contracted by the Air Force Wright Laboratory and the Defense Advanced Research Projects Agency (DARPA) to develop the next generation flexible semiconductor wafer fabrication system called Microelectronics Manufacturing Science & Technology (MMST). Several revolutionary concepts are being pioneered on MMST, including the following: new single-wafer rapid thermal processes, in-situ sensors, cluster equipment, and advanced Computer Integrated Manufacturing (CIM) software. The objective of the project is to develop a manufacturing system capable of achieving an order of magnitude improvement in almost all aspects of wafer fabrication. TI was awarded the contract in Oct., 1988, and will complete development with a fabrication facility demonstration in April, 1993. An important part of MMST is development of the CIM environment responsible for coordinating all parts of the system. The CIM architecture being developed is based on a distributed object oriented framework made of several cooperating subsystems. The software subsystems include the following: process control for dynamic control of factory processes; modular processing system for controlling the processing equipment; generic equipment model which provides an interface between processing equipment and the rest of the factory; specification system which maintains factory documents and product specifications; simulator for modelling the factory for analysis purposes; scheduler for scheduling work on the factory floor; and the planner for planning and monitoring of orders within the factory. This paper first outlines the division of responsibility between the planner, scheduler, and simulator subsystems. It then describes the approach to incremental planning and the way in which uncertainty is modelled within the plan representation. Finally, current status and initial results are described.

Fargher, Hugh E.↗

Origins of wear-induced tungsten corrosion defects in semiconductor manufacturing during tungsten chemical mechanical polishing

During chemical mechanical polishing (CMP) process optimization, a unique corrosion defect was identified in the tungsten (W) contact plugs, which were investigated to develop practical solutions to improve CMP mass productivity. The corrosion defect occurred in the absence of external incident energy and was observed to occur only at the W plugs connected to the P + /n-well close to the wafer edge. To examine the root cause of the corrosion, a series of experiments was carried out at each W CMP step, using in-line defect inspection. Additional experiments were conducted using on-line inductively coupled plasma mass spectrometry (ICP-MS) to understand pH- and potential-dependent electrochemical dissolution behaviors of W. Experimentally, W corrosion was confirmed to be exacerbated if the surface oxide of the W plugs is removed by the deionized water (DIW) rinsing step under wafer pressurization, then exposed to DIW and alkaline cleaning solution. Although the DIW rinsing step is commonly performed to clean polishing pads right after the main polishing, a small amount of applied pressure (0.5 psi) without slurry supply can lead to corrosion defects. Finally, this wear-induced corrosion of the W plugs can be prevented by removing the applied pressure during the post DIW rinsing step of W CMP.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Development of Numerical Extended Hydrodynamics for Transition-Regime Non-Equilibrium Flows Encountered in Semiconductor Manufacturing Processes

Six months of funding was received for the proposed three year research program (funding for the period from March 1, 1997 to August 31, 1997). Although the official starting date for the project was March 1, 1997, no funding for the project was received until July 1997. In the funded research period, considerable progress was made on Phase I of the proposed research program. The initial research efforts concentrated on applying the 10-, 20-, and 35-moment Gaussian-based closures to a series of standard two-dimensional non-reacting single species test flow problems, such as the flat plate, couette, channel, and rearward facing step flows, and to some other two-dimensional flows having geometries similar to those encountered in chemical-vapor deposition (CVD) reactors. Eigensystem analyses for these systems for the case of two spatial dimensions was carried out and efficient formulations of approximate Riemann solvers have been formulated using these eigenstructures. Formulations to include rotational non-equilibrium effects into the moment closure models for the treatment of polyatomic gases were explored, as the original formulations of the closure models were developed strictly for gases composed of monatomic molecules. The development of a software library and computer code for solving relaxing hyperbolic systems in two spatial dimensions of the type arising from the closure models was also initiated. The software makes use of high-resolution upwind finite-volumes schemes, multi-stage point implicit time stepping, and automatic adaptive mesh refinement (AMR) to solve the governing conservation equations for the moment closures. The initial phase of the code development was completed and a numerical investigation of the solutions of the 10-moment closure model for the simple two-dimensional test cases mentioned above was initiated. Predictions of the 10-moment model were compared to available theoretical solutions and the results of direct-simulation Monte Carlo (DSMC) calculations. The first results of this study were presented at a meeting last year.

Groth, Clinton P. T.↗

Data Mining in Multi-Dimensional Functional Data for Manufacturing Fault Diagnosis

Multi-dimensional functional data, such as time series data and images from manufacturing processes, have been used for fault detection and quality improvement in many engineering applications such as automobile manufacturing, semiconductor manufacturing, and nano-machining systems. Extracting interesting and useful features from multi-dimensional functional data for manufacturing fault diagnosis is more difficult than extracting the corresponding patterns from traditional numeric and categorical data due to the complexity of functional data types, high correlation, and nonstationary nature of the data. This chapter discusses accomplishments and research issues of multi-dimensional functional data mining in the following areas: dimensionality reduction for functional data, multi-scale fault diagnosis, misalignment prediction of rotating machinery, and agricultural product inspection based on hyperspectral image analysis.

42 ENGINEERING↗

A Data-Agnostic, Continuous Machine Learning Framework for Application in High Energy Physics and Beyond: Phase 1 Final Scientific/Technical Report

This Phase 1 effort has focused on the development of continual learning frameworks for use in machine learning, specifically in the applied context of High Energy Physics (HEP). Machine learning (ML) is a transformative technology by which computers, typically through the use of neural networks, are able to perform tasks with proficiency that rivals or surpasses that of human users. Model Degradation & Catastrophic Forgetting are two undesired phenomena which can occur in ML where the performance of a model degrades when either deployed on novel data streams, or trained on novel data which are sufficiently different than the data the models were initially trained on. A natural example where these sorts of effects can be observed is in the performance of detectors in harsh environments, where the detector signature may change over the lifetime of the detector as it ages and deteriorates — precisely what occurs in the experiments conducted in HEP. Real world HEP data is therefore an excellent test-ground and use-case for Continual Learning paradigms, which are techniques used in ML to counteract these problems. Ensemble learning is one such technique, where multiple smaller models are trained on subsets of the overall data and are ensembled together during inference. The intuition behind this technique is that, although there are shifts in the distributions which govern the incoming data streams, these shifts are not expected to be homogeneous or global. If a sufficient diversity in solutions within the various sub-models has been achieved, then at least one sub-model is expected to retain its performance within the overall ensemble. One further strength of this approach is that the architectures of the various models do not need to be identical, and in fact even different modalities of data can naturally be combined in this way. This work focused on applying ensemble learning techniques to derive results using two main datasets, anomaly detection in HEP data & time-series forecasting in semiconductor manufacturing data. Semiconductor manufacturing involves data with surprising similarity to that of HEP (e.g. wafer maps look very similar to digi-occupancy maps) and Cerium Lab’s prominence within the semiconductor industry makes semiconductor manufacturing a natural opportunity for commercialization of this work. Our efforts have led to two strong results. The first is that we evaluated the proposed ensembling techniques using previously proposed machine learning architectures for use in anomaly detection, namely AutoEncoder based models and their derivatives. We also developed new architectures which have not been evaluated in this context before. In fact, this work marks the first use of Vision Transformers for anomaly detection in HEP. Second, we demonstrated that ensemble learning significantly improves model performance in scenarios prone to degradation, validating its effectiveness across both HEP and semiconductor datasets. These results further support ensemble learning as a powerful strategy for mitigating catastrophic forgetting and maintaining robust performance in evolving data environments.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Method for anisotropic etching in the manufacture of semiconductor devices

Hydrocarbon polymer coatings used in microelectronic manufacturing processes are anisotropically etched by hyperthermal atomic oxygen beams (translational energies of 0.2 to 20 eV, preferably 1 to 10 eV). Etching with hyperthermal oxygen atom species obtains highly anisotropic etching with sharp boundaries between etched and mask protected areas.

Koontz, Steven L.↗