Metrology Algorithms Research Articles

Data-driven plasma science based plasma etching process design in OLED mass production referring to PI-VM

The production efficiencies of organic light emitting diode (OLED) displays and semiconductor manufacturing have been dramatically improving with the help of plasma physics and engineering technology by utilizing a process monitoring methodology based on physical domain knowledge. This domain knowledge consists of plasma-heating and sheath physics, plasma chemistry, and plasma-material surface reaction kinetics. They were applied to the plasma information based virtual metrology (PI-VM) algorithm with the plasma diagnostics and noticeably enhanced process prediction performance by parameterizing plasma information (PI) in various processes of OLED display and semiconductor manufacturing fabs. PI-VM has shown superior process prediction accuracy, which can trace the states of processing plasmas as an application of data-driven plasma science compared to the classical statistics and machine learning-based virtual metrologies; thus, various plasma processes have been managed and controlled with the help of the PI-VM models. More than this, we have adopted the PI-VM model to optimize the patterning architecture and plasma processes simultaneously. The best combination of the etching pattern structure and plasma condition was adjustable based on the detailed understanding of the angular distribution of sputtered atoms from the etching target surface and their interaction with the plasma sheath based on the PI-VM modeling for etching profile failure prediction.

Surface Quality Optical Measurement of Part Inclined Planes Manufactured by FDM Technology

Fused Filament Deposition (FDM) additive manufacturing technology allows the generation of three-dimensional parts by overlapping layers of an extruded polymer. The staircase effect caused by overlapping is well known and makes the manufactured parts have poor surface quality. This problem has been focused on by numerous works trying to optimize the surface quality by studying the process technological parameters or part geometry. Also, post-processing techniques have been developed that improve this roughness, but which involve an increase in manufacturing time and cost. In this work, a methodology to measure the roughness based on laser confocal microscopy is proposed. To evaluate the methodology, an experimental study is carried out that relates the surface roughness of parts manufactured with FDM with the building orientation and the layer height. The main objective of this study is to compare two roughness measurement methodologies: a mechanical measurement with a conventional contact roughness meter, and an optical measurement with a confocal microscope. The contact roughness meter provides a direct value of the roughness of the part wall profile, while the confocal microscope provides an image of the three-dimensional surface of the part wall, which must be processed. The data from the confocal microscope are evaluated with the Internet-based software Surface Metrology Algorithm Testing System (SMATS) developed by the National Institute of Standards and Technology (NIST) of the USA. The SMATS will provide an average value of the roughness of the analyzed surface. The test results with both methodologies are very similar, with an average difference of 5%. These results show the influence of the printed plane inclination angle on the roughness, which is higher for low values of the angle. It can also be seen that this influence decreases for low-layer heights.

Parallel unambiguous generalized phase-shifting and T-spline fitting algorithms for optical micro-structured surface 3D topography metrology.

3D topography metrology of optical micro-structured surfaces is critical for controlled manufacturing and evaluation of optical properties. Coherence scanning interferometry technology has significant advantages for measuring optical micro-structured surfaces. However, the current research faces difficulties of designing high accuracy and efficient phase shifting, and characterization algorithms for optical micro-structured surface 3D topography metrology. In this paper, parallel unambiguous generalized phase-shifting and T-spline fitting algorithms are proposed. To avoid phase ambiguity and improve the accuracy of the phase-shifting algorithm, the zero-order fringe is determined by the iterative envelope fitting with Newton's method, and the accurate zero optical path difference is determined by a generalized phase-shifting algorithm. In particular, the calculation procedures of the multithreading iterative envelope fitting with Newton's method and generalized phase shifting are optimized with the graphics processing unit-Compute Unified Device Architecture kernel function. Additionally, to fit the base form of optical micro-structured surfaces and characterize the surface texture and roughness, an effective T-spline fitting algorithm is proposed by optimizing the preimage of the T-mesh with image quadtree decomposition. Experimental results show that the surface reconstruction of optical micro-structured surfaces using the proposed algorithm is more accurate, and the efficiency is 10 times higher than that of current algorithms; the time of the surface reconstruction is less than 1s. Compared with the current B-spline method, the accuracy of roughness characterization using the proposed T-spline algorithm is improved by more than 10%.

TEMAS: A Flexible Non‐AI Algorithm for Metrology of Single‐Core and Core‐Shell Nanoparticles from TEM Images

AbstractAn essential application of electron microscopy is to provide feedback to tune the fabrication of nanoparticles (NPs). Real samples tend to follow a size distribution commonly linked to the synthesis process used and in turn to their functional properties. This study presents an algorithm for measuring particle size distributions in electron microscopy images. State‐of‐the‐art methods based on Artificial Intelligence (e.g., Deep Learning) require extensive datasets of labeled images similar to those expected to be analyzed, and extensive supervised re‐training is often required for cross‐domain application. In contrast, the non‐AI algorithm described in this study is accurate and can be quickly set up for measuring new experimental images in different domains. The accuracy of the method is validated quantitatively and comparing graphical and descriptive statistics. Different size distributions are measured on images of platinum and gold nanocatalysts supported on carbon black, amorphous carbon, and titanium dioxide crystals. Also, images of platinum‐iron core‐shell NPs supported on thin amorphous carbon film are successfully analyzed. The limitation of evaluating different algorithms for NPs metrology is the lack of standards that different researchers can use as ground truth. In order to overcome this limitation, the images and the ground truth measurements presented here are shared as an open dataset.

Freeform optics characterization with surface registration and fitting algorithms for optical point-based spatial path 3D topography metrology.

Accurate characterization of the form error for freeform optics is critical for controlled manufacturing and evaluation of optical properties. To solve the difficulty of current surface registration and fitting algorithms, and improve characterization accuracy of the form error of freeform optics for optical point-based spatial path 3D topography metrology, in this paper, improved surface registration and fitting algorithms are proposed, including a B-spline surface description of freeform optics, point orthogonal projection, registration parameter optimization, and B-spline fitting. Limited by the angular characteristics of an optical point-based sensor, the slope and reference frame of freeform optics must be flexibly adjusted, and polynomial surfaces are described and fitted as B-spline surfaces based on B-spline geometric invariance. To efficiently determine corresponding points on B-spline surfaces, the point orthogonal projection algorithm without Bezier subdivision is proposed using the first-order Newton method. Then, the iteration procedure of coordinate adjustment in sphere space using Lie algebra registration parameters is proposed to solve the difficulty of the current registration parameter optimization procedure. To fit a spatial path form error, the least squares B-spline fitting method is proposed to improve the Zernike method. Through simulation experiments, the proposed registration algorithm with good convergence can improve accuracy by one to two orders of magnitude compared with current registration algorithms. Through repeated experiments of a freeform prism, the proposed method can significantly improve the peak-to-valley and RMS accuracies compared with the stylus method, and characterize the mid-high frequency form error (about 200nm) of a freeform prism.

Traceable machine learning real-time quality control based on patient data.

Patient-based real-time quality control (PBRTQC) has gained attention as an alternative/integrative tool for internal quality control (iQC). However, it is still doubted for its performance and its application in real clinical settings. We aim to generate a newly and easy-to-access patient-based real-time QC by machine learning (ML) traceable to standard reference data with assigned values by National Institute of Metrology of China (NIM), and to compare it with PBRTQC for clinical validity evaluation. For five representative biochemistry analytes, 1,195 000 patient testing results each were collected. After data processing, independent training and test sets weredivided. Machine learning internal quality control (MLiQC) was set up by Random Forest in ML and was validated by way of both metrology algorithm traceability and 4 PBRTQC methods recommended by IFCC analytical working group. MLiQC were established. As an example of albumin (ALB) at the critical bias, the uncertainty of MLiQC was 0.14%, which was evaluated by standard reference data produced by NIM. Compared with four optimal PBRTQC methods at critical bias, the average of the number of patient samples from a bias introduced until detected (ANPed) of MLiQC averagely decreased from 600 to 20. The median and 95 quantiles of NPeds (MNPed and 95NPed) of MLiQC were superior to all optimal PBRTQCs above 90% for all test items. MLiQC is highly superior to PBRTQC and well-suited in real settings. The validation of the model from two aspects of algorithm traceability and clinical effectiveness confirms its satisfactory performance.

Deep learning speckle de-noising algorithms for coherent metrology: a review and a phase-shifted iterative scheme [Invited

We present a review of deep learning algorithms dedicated to the processing of speckle noise in coherent imaging. We focus on methods that specifically process de-noising of input images. Four main classes of applications are described in this review: optical coherence tomography, synthetic aperture radar imaging, digital holography amplitude imaging, and fringe pattern analysis. We then present deep learning approaches recently developed in our group that rely on the retraining of residual convolutional neural network structures to process decorrelation phase noise. The paper ends with the presentation of a new approach that uses an iterative scheme controlled by an input SNR estimator associated with a phase-shifting procedure.

A Novel Light Sectioning Vision System for A Three Dimensional Surface Roughness Assessment (Dept.M)

Standard roughness measurement procedures depend heavily on stylus instruments which have only limited flexibility in handling different parts. On the other hand, optical non-contact techniques are very interesting for 3D characterization of sensitive and complex engineering surfaces. In this study, a new ap proach is introduced to measure surface roughness in three dimensions by combining a light sectioning microscope and a computer vision system. This approach has the advantages of being non-contact, fast and cheap. A prototype version of a user interface program, currently named SR3D Vision, has been developed to manage three dimensional surface roughness measurements. A light sectioning microscope is used to view roughness profiles of the specimens to be Measured and the vision system is used to capture images for successive profiles. This program has been totally developed in-house using MatlabTM software to analyze the captured images through four main modules: (Measurement controller, Profile or surface extraction, 2D roughness parameters calculation and 3D roughness parameters calculation). The system has been calibrated for metric units and verified using standard specimens. In addition, the system was used to measure various samples machined by different operations and the results were compared with a commercial software and a web-based surface metrology algorithm testing system. The accuracy of the system was verified and proved to be within ±4.8% compared with these systems.

PROBA-3 mission and the Shadow Position Sensors: Metrology measurement concept and budget

PROBA-3 is a space mission of the European Space Agency that will test, and validate metrology and control systems for autonomous formation flying of two independent satellites. PROBA-3 will operate in a High Elliptic Orbit and when approaching the apogee at 6·104 Km, the two spacecraft will align to realize a giant externally occulted coronagraph named ASPIICS, with the telescope on one satellite and the external occulter on the other one, at inter-satellite distance of 144.3 m. The formation will be maintained over 6 hrs across the apogee transit and during this time different validation operations will be performed to confirm the effectiveness of the formation flying metrology concept, the metrology control systems and algorithms, and the spacecraft manoeuvring. The observation of the Sun’s Corona in the field of view [1.08;3.0]RSun will represent the scientific tool to confirm the formation flying alignment. In this paper, we review the mission concept and we describe the Shadow Position Sensors (SPS), one of the metrological systems designed to provide high accuracy (sub-millimetre level) absolute and relative alignment measurement of the formation flying. The metrology algorithm developed to convert the SPS measurements in lateral and longitudinal movement estimation is also described and the measurement budget summarized.

Cause analysis of the faults in HARC etching processes by using the PI‐VM model for OLED display manufacturing

AbstractHigh‐aspect ratio contact (HARC) etching is a bottleneck step of the high‐definition organic light emitting diode (OLED) display manufacturing processes. HARC process is frequently failed during the mass production, because this requires the high‐energy ion flux and the sidewall passivation, simultaneously. To analyze the cause of HARC process failures, plasma information (PI)‐based virtual metrology (VM) algorithm was developed by using the equipment engineering system and the optical emission spectroscopy data recorded from the fab. Developed PI‐VM could predict the process faults with >90% of the accuracy, and the cause analysis function was also validated. We could suggest a right solution to the failure, and more efficient management of the OLED display manufacturing was possible.

Monte Carlo based analysis of confocal peak extraction uncertainty

Localisation of axial peaks is essential for height determination in confocal microscopy. Several algorithms have been proposed for reliable height extraction in surface topography measurements. However, most of these algorithms use nonlinear processing, which precludes estimating the peak height uncertainty. A Monte Carlo based standard uncertainty analysis model is developed here to evaluate the precision of height extraction algorithms. The key parameters of this model are the vertical sampling deviation and the size of the scanning pitch. Height extraction uncertainty of the centroid algorithm and nonlinear fitting algorithms were calculated using simulations. Our results offer a reference for selecting algorithms for confocal metrology.

Confocal signal evaluation algorithms for surface metrology: uncertainty and numerical efficiency.

Confocal microscopy is one of the dominating measurement techniques in surface metrology, with an enhanced lateral resolution compared to alternative optical methods. However, the axial resolution in confocal microscopy is strongly dependent on the accuracy of signal evaluation algorithms, which are limited by random noise. Here, we discuss the influence of various noise sources on confocal intensity signal evaluating algorithms, including center-of-mass, parabolic least-square fit, and cross-correlation-based methods. We derive results in closed form for the uncertainty in height evaluation on surface microstructures, also accounting for the number of axially measured intensity values and a threshold that is commonly applied before signal evaluation. The validity of our results is verified by numerical Monte Carlo simulations. In addition, we implemented all three algorithms and analyzed their numerical efficiency. Our results can serve as guidance for a suitable choice of measurement parameters in confocal surface topography measurement, and thus lead to a shorter measurement time in practical applications.

소실점 추정 정확도 개선을 통한 SVM 성능 향상

In this paper, we propose an improved single view metrology (SVM) algorithm to accurately measure the height of objects. In order to accurately measure the size of objects, vanishing points have to be correctly estimated. There are two methods to estimate vanishing points. First, the user has to choose some horizontal and vertical lines in real world. Then, the user finds the cross points of the lines. Second, the user can obtain the vanishing points by using software algorithm such as [6-9]. In the former method, the user has to choose the lines manually to obtain accurate vanishing points. On the other hand, the latter method uses software algorithm to automatically obtain vanishing points. In this paper, we apply image resizing and edge sharpening as a pre-processing to the algorithm in order to improve performance. The estimated vanishing points algorithm create four vanishing point candidates: two points are horizontal candidates and the other two points are vertical candidates. However, a common image has two horizontal vanishing points and one vertical vanishing point. Thus, we eliminate a vertical vanishing point candidate by analyzing the histogram of angle distribution of vanishing point candidates. Experimental results show that the proposed algorithm outperforms conventional methods, [6] and [7]. In addition, the algorithm obtains similar performance with manual method with less than 5% of the measurement error.

Image Processing Assisted Tools for Pre- and Post-processing Operations in Additive Manufacturing

Additive Manufacturing (AM) or 3D printing (3DP) process is a layer by layer manufacturing approach for manufacturing parts with complex geometries. The STL (Stereolithography) file format is the de-facto industry standard for manufacturing parts by AM/3DP. The STL file is a triangular mesh representation of CAD geometry in non-uniform rational B-spline (NURBS) surface format and thus is an approximation of actual part geometry. During the conversion of a complex CAD geometry to an STL file, geometric errors are introduced in the model. These drawbacks associated with the STL file may translate into a faulty or inaccurate final manufactured part. This paper presents a novel Image Processing (IP) based Direct CAD Slicer, IPSlicer, which can be used to manufacture components directly from CAD geometry (without converting to STL file). Using sectional image snapshots of a part, captured normal to the build direction and sectional 2D bounding box data, contour points for each section are identified by performing boundary tracing operation followed by application of Contour Mapping Algorithm (CMA). The method slices the actual NURBS geometry and thus parts manufactured by this method have reduced GD&T errors such as flatness, cylindricity, and profile error. In addition, a support removability analysis tool is developed by performing color-based segmentation on sectional image snapshots in conjunction with a pixel traversal approach. Using the segmented images and sectional bounding box, a sintering area and time calculation tool for each layer is also developed. The effectiveness of IPSlicer is verified by virtually manufacturing test components and calculating GD&T errors by the application of computational metrology algorithms on virtually manufactured data. The support removability analysis tool as well as sintering time and area calculator are validated using several test parts.

Capsule Deformation in Pharmaceutical Industry by a Non-Contact Metrology Algorithm

Capsule Deformation in Pharmaceutical Industry by a Non-Contact Metrology Algorithm

Efficient Proximity Probing Algorithms for Metrology

Metrology, the theoretical and practical study of measurement, has applications in automated manufacturing, inspection, robotics, surveying, and healthcare. The geometric probing problem considers how to optimally use a probe to measure geometric properties. In this paper, we consider a proximity probe which, given a point, returns the distance to the boundary of the nearest object. When there is an unknown convex polygon P in the plane, the goal is to minimize the number of probe measurement needed to exactly determine the shape and location of P. We present an algorithm with upper bound of 3.5n + k + 2 probes, where n is the number of vertices and k ≤ 3 is the number of acute angles of P. The algorithm requires constant time per probe, and hence, O(n) time to determine P. We also address the related problem where the unknown polygon is a member of a known finite set Γ and the goal is to efficiently determine which polygon is present. When m is the size of Γ and n' is the maximum number of vertices of any member of Γ, we present an algorithm with an upper bound of 2n + 2 probes with O(1) computations per probe and a O(n'm) preprocessing phase (depending only on Γ).

Application of Algorithms for High Precision Metrology

This paper evaluates the performance of algorithms suitable for processing the measurements from two laser beam metrology systems, with particular reference to the Gaia Basic Angle Monitoring device. The system and signal characteristics are reviewed in order to define the key operating features. The low-level algorithms are defined according to different approaches, starting with a simple, model free method, and progressing to a strategy based on the signal template and variance. The signal model is derived from measured data sets. The performance at microarcsec level is verified by simulation in conditions ranging from noiseless to large perturbations.

Generation of 3D random topography datasets with periodic boundaries for surface metrology algorithms and measurement standards

A procedure is presented to generate three-dimensional (3D) random topography datasets with periodic boundaries for the evaluation of surface metrology algorithms and original data for measurement standards. A non-causal two-dimensional (2D) autoregressive (AR) model, which expresses the surface as a linear weighted summation of AR parameters and topography data in addition to a random noise component, is applied to computationally generate 3D random topography data. By the use of an extension that assumes periodic boundaries, the edges of the generated data become continuous across the boundaries. It has been verified that the spectral properties are not affected by this extension. This technique offers advantages for the evaluation of computational techniques for surface metrology, such as filtrations and spectral analysis since the edge effect can be avoided by assuming periodic boundaries, and inherent effects of the techniques can be evaluated. In addition, for use as a random measurement standard for instrument calibration, it is possible to simply arrange the generated data repeatedly in the measuring window similarly to floor tiles without introducing discontinuous edges at the boundaries of the data.

NN-Based Key-Variable Selection Method for Enhancing Virtual Metrology Accuracy

This paper proposes an advanced key-variable selection method, the neural-network-based stepwise selection (NN-based SS) method, which can enhance the conjecture accuracy of the NN-based virtual metrology (VM) algorithms. Multi-regression-based (MR-based) SS method is widely applied in dealing with key-variable selection problems despite the fact that it may not guarantee finding the best model based on its selected variables. However, the variables selected by MR-based SS may be adopted as the initial set of variables for the proposed NN-based SS to reduce the SS process time. The backward elimination and forward selection procedures of the proposed NN-based SS are both performed by the designated NN algorithm used for VM conjecturing. Therefore, the key variables selected by NN-based SS will be more suitable for the said NN-based VM algorithm as far as conjecture accuracy is concerned. The etching process of semiconductor manufacturing is used as the illustrative example to test and verify the merits of the NN-based SS method. One-hidden-layered back-propagation neural networks (BPNN-I) are adopted for establishing the NN models used in the NN-based SS method and the VM conjecture models. Test results show that the NN model created by the selected variables of NN-based SS can achieve better conjecture accuracy than that of MR-based SS. Simple recurrent neural networks and generalized regression neural network are also tested and proved to be able to achieve similar results as those of BPNN-I.

Accuracy and Real-Time Considerations for Implementing Various Virtual Metrology Algorithms

In the semiconductor industry, run-to-run (R2R) control is an important technique to improve process capability and further enhance the production yield. As the dimension of electronic devices shrink increasingly, wafer-to-wafer (W2W) advanced process control (APC) becomes essential for the critical stages of production processes. W2W APC requires the metrology values of each wafer; however, it will be highly time and cost consuming to obtain actual metrology values from each wafer by physical measurement. Recently, an efficient and cost-effective approach denoted virtual metrology (VM) was proposed to substitute the actual metrology. To implement VM in W2W APC, both conjecture-accuracy and real-time requirements need to be considered. In this paper, various VM algorithms, including back-propagation neural networks (BPNN), simple recurrent neural networks (SRNN), and multiple regression (MR), are evaluated to see whether they can meet the accuracy and real-time requirements of W2W APC or not. The fifth generation TFT-LCD chemical-vapor deposition process is used to test and verify the requirements. Test results show that both one-hidden-layered BPNN and SRNN VM algorithms achieve acceptable conjecture accuracy and meet the real-time requirements of semiconductor and TFT-LCD W2W APC applications.