Jakey Blue

Integration of scheduling and advanced process control in semiconductor manufacturing: review and outlook

Scheduling in semiconductor manufacturing is of vital importance due to the impact on production performance indicators such as equipment utilization, cycle time, and delivery times. With the increasing complexity of semiconductor manufacturing, ever-new products and demanding customers, scheduling plans for efficient production control become crucial. Scheduling and control are mutually dependent as control requires information from scheduling, for example, where jobs are processed, and scheduling requires control information, for example, on which equipment operations can be processed. Based on a survey of the literature, this article proposes a review and an outlook for the potential improvements by binding scheduling decisions and information coming from advanced process control systems in semiconductor manufacturing.

Translation-Invariant Multiscale Energy-Based PCA for Monitoring Batch Processes in Semiconductor Manufacturing

The overwhelming majority of processes taking place in semiconductor manufacturing operate in a batch mode by imposing time-varying conditions to the products in a cyclic and repetitive fashion. These conditions make process monitoring a very challenging task, especially in massive production plants. Among the state-of-the-art approaches proposed to deal with this problem, the so-called multiway methods incorporate the batch dynamic features in a normal operation model at the expense of estimating a large number of parameters. This makes these approaches prone to overfitting and instability. Moreover, batch trajectories are required to be well aligned in order to provide the expected performance. To overcome these issues and other limitations of the conventional methodologies for process monitoring in semiconductor manufacturing, we propose an approach, translation-invariant multiscale energy-based principal component analysis, that requires a much lower number of estimated parameters. It is free of process trajectory alignment requirements and thus easier to implement and maintain, while still rendering useful information for fault detection and root cause analysis. The proposed approach is based on implementing a translation-invariant wavelet decomposition along the time series profile of each variable in one batch. The normal operational signatures in the time-frequency domain are extracted, modeled, and then used for process monitoring, allowing prompt detection of process abnormalities. The proposed procedure was tested with real industrial data and it proved to effectively detect the existing faults as well as to provide reliable indications of their underlying root causes.

Efficient FDC based on hierarchical tool condition monitoring scheme

Tool condition evaluation and prognosis has been an arduous challenge in modern semiconductor manufacturing environment, especially for the foundry and analog companies with high product-mix and complicated technology nodes. More and more embedded and external sensors are installed to capture the genuine tool status for tool fault identification and, thus, tool condition analysis based on real-time equipment data becomes promising but also much more complex with the rapidly-increased number of sensors. In this paper, the feasibility of Generalized Moving Variance (GMV) technique is validated to consolidate the pure variations within tool Fault Detection and Classification (FDC) data into one indicator. Based on GMV, a hierarchical tool condition monitor scheme is developed by analyzing the GMV within functional clusters of sensors. With the introduction of this hierarchy, abnormal tool condition can be diagnosed and drilled down into sensor level for an efficient root cause analysis.

Equipment Condition Diagnosis and Fault Fingerprint Extraction in Semiconductor Manufacturing

As the high-tech production system gets more complex, Equipment Condition Diagnosis (ECD) in semiconductor manufacturing for Fault Detection and Classification (FDC) is becoming more and more challenging than ever. This paper uses well-known machine learning techniques such as Support Vector Machine (SVM), K-Means clustering and Self-Organizing Map (SOM) to develop an efficient ECD model. The process normality is checked by SVM following by decomposing the process dynamics via K-Means. The abnormal observations are then projected into normal models built by Principal Component Analysis (PCA). Finally, by calculating the contribution values of out-of-control observations, different fault fingerprints with corresponding fault root are extracted again by K-Means. The impact of clustering techniques is investigated by comparing K-Means, SOM, and hierarchical clustering. An empirical study was conducted in collaboration with the leading semiconductor company in France to validate the methodology. The result shows that the proposed approach can effectively detect abnormal observations as well as automatically classify the fault fingerprints to give evident guidelines in explaining the detected faults.

FDC R2R variation monitoring for sensor level diagnosis in tool condition hierarchy

Tool behavior modeling and diagnosis is a big challenge in modern semiconductor fabrication, in particular for the foundry and analog companies with high product-mix and complicated technology nodes. Tool condition monitoring has been practiced by implementing the FDC (Fault Detection and Classification) system and analyzing large amount of real-time equipment data. This paper continues the work of tool condition hierarchy, where the excursions can be detected in one single overall tool indicator and are intuitively drilled down to the level of sensor groups. A R2R (Run-to-Run) variation monitoring technique is developed in order to correlate the tool faults with single sensor and thus completes the diagnostic gap of the hierarchy. The tool condition monitoring then becomes efficient and tool fault diagnosis can be systematically top-down.

The light behavior from Shallow Trench Isolation profiles at Chemical Mechanical Planarization step and correlation with optical endpoint system by interferometry

In the semiconductor manufacturing, the control of Chemical-Mechanical Planarization (CMP) process time for Shallow Trench Isolation (STI) is important. A wafer under- or over-polishing causes leakage and short-circuits making the chips defective. The CMP process control by interferometry is one of the most used systems to monitor the polishing time. In some cases, the interferometry process control is not possible because the wafer patterns cause some unwanted effects such as scattering, diffraction, and absorption. Consequently the signal is affected. In this paper, we apply a theoretical and experimental approach on the light reflected from different STI stacks in order to interpret the observed optical phenomenon. The experimental study is done to get close to the light measurement conditions within the manufacturing environment. With this experiment, we evidence that the trench pattern inside memory zones is responsible for the diffraction effect on the signal. In a production environment, this pattern results in a lower measured intensity when the size of memory area increases. Besides, numerical calculations are performed based on differential method in order to predict the diffracted intensity, which depends on the chip design parameters and the incident wavelengths tuning. By using STI models, this method helps to determine the wavelengths with the highest reflected intensity.

Impact of integrating equipment health in production scheduling for semiconductor fabrication

Abstract Monitoring the Equipment Health Indicator (EHI) of critical machines helps effectively to maintain process quality and reduce wafer scrap, rework, and machine breakdowns. To model and illustrate the integration of EHI in scheduling decisions to balance between productivity and quality risk, this paper presents two mixed integer linear programs to schedule jobs on heterogeneous parallel batching machines. The capability of a machine to process a job is categorized as preferred, acceptable, and unfavorable based on the job requirements. The quality risk of processing a job by a machine is a function of its EHI and the capability level of the machine for the job, which is modeled as a penalty in the objective function of trading-off between productivity and quality risk. The first model is static and assumes constant EHI of machines on the scheduling horizon, whereas the second model considers the EHI dynamics, i.e., the machine condition deteriorates over time based on the scheduled jobs. Numerical experiments indicate the potential applications of using EHI-integrated scheduling approaches to analyze and optimize the trade-off between productivity and quality risk.

Embedded spectroscopic reflectometry metrology on FEOL silicon dioxide trench polishing equipment

The aim of the research is to develop a material thickness measurement method to monitor oxide polishing by Chemical-Mechanical Planarization (CMP) during the realization of the Shallow Trench Isolation (STI). The underlying goal is to build a statistical regulation model of the polishing time on a single platen (the two others platens are monitored by an endpoint signal). In addition to the process parameters (head sweep, platen, and head rotation velocity), input and output polished material thicknesses data are essential to build a run-to-run model for CMP. Therefore, stack layer thickness, before and after polishing, needs to be measured fast enough to maintain the acceptable throughput and to accurately control the polishing time wafer by wafer. In this paper, we describe how spectroscopic reflectometry embedded in the polishing equipment, can meet rapidity and capability requirements in setting up a run-to-run control algorithm to maintain the target thickness for STI CMP.

Equipment health modeling for deterioration prognosis and fault signatures diagnosis

Equipment condition monitoring for deterioration prognosis has drawn lots of attention in semiconductor manufacturing. Deterioration model represents healthy state of equipment and enables manufacturers to avoid equipment breakdown and non-essential maintenances. This research details an automated and intelligent approach to build a healthy state model for equipment deterioration monitoring. Wavelet packet decomposition (WPD) method is utilized first to extract energy-based features. A windowing Gaussian Mixture Model (GMM) then is used to track probability distribution function (PDF) of process to model the deterioration. The proposed model is validated through a real industrial case from a local semiconductor manufacturer.

Run-to-run sensor variation monitoring for process fault diagnosis in semiconductor manufacturing

Tool behavior modeling and diagnosis is a big challenge in modern semiconductor fabrication, in particular, with high product-mix and complicated technology nodes. Tool condition monitoring has been long conducted by implementing the Fault Detection and Classification (FDC) system and analyzing the large amount of real-time sensor data collected during the process. The tool condition hierarchy developed in the previous work proposed that the excursions can be firstly detected by an overall condition indicator and then intuitively traced down to the level of sensor groups. In this paper, a Run-to-Run (R2R) variation monitoring technique is developed in order to correlate the tool excursions with individual sensors, instead of sensor groups, and thus to close the diagnostic gap in the hierarchy. Therefore, the tool condition can be efficiently monitored by one overall indicator and the detected tool faults can be systematically diagnosed at the sensor level.