Data storage capacity of hard drive disks (HDD) has been increased over time. Partly this has been achieved by shrinking the dimensions of magnetic writing device also called the “main Pole”, a key magnetic flux emanating component in the slider that hovers over the magnetic media disk writing bits in nanoscale magnetic domains along the circular tracks. Even though the writer pole device is a single isolated device, its 3D shape and material composition are quite complex and critical dimensions are sub-50 nm. The reliability and accuracy of writing data in bits of magnetic nano-domains depends on the precise control of the pole shape and dimensions. Its fabrication requires over thousand process steps with many of them being various types of metrology steps. Final shape or back-end shape of the pole is influenced by various process steps and related information is captured by the subsequent metrology. Overlay, optical (thin film thickness using ellipsometry, topography using white light interferometry), CD-SEM, CD-AFM and inline FIB cross sections based metrology is commonly used. All this metrology data during fabrication process is called wafer metrology data. The data about the shape of the pole after device fabrication (lapping) is called back-end data and is obtained by cross section using FIB and SEM. Few device chips are sent ahead to determine the lapping parameters which takes time. This paper is about predicting the back-end shape parameters namely, pole height (PH), pole width (PW), and pole angle (PA), based on wafer data so that the backend metrology process requiring send-ahead can be optimized or even eliminated. Predictive models using machine learning and analytics techniques (neural networks, multivariate regression, and principal components analysis) have been studied and results will be presented and discussed in this paper. In this metrology centric data science study various steps have been pursued from the beginning to retrieve, integrate, transform, model, visualize data and understand the outcomes. Wafer data corresponding to more than seventy different parameters was used in this study. It is not possible to have all metrology data for each device as some techniques are destructive, imputation technique based on nearest neighbour data points is used. Model was trained and validated on a set of data and tested on an independent new data. Based on testing on independent data, study concludes that it is possible to train the model on data from few wafers for technology in production and predict the back-end pole shape parameters with acceptable accuracy for upcoming wafers. This could be very useful in reducing cycle time by minimizing the need for send ahead wafer components and optimizing the tuning of lapping process. Correlation of predicted (Y-axis) and measured (X-axis) values of PW parameters for four wafers is plotted in figure 1 at device level, flash field or exposure field level, and wafer level. The correlation between predicted and measured PW is reasonable for making use of the predictive model at the flash field level and wafer level to optimize the need for send ahead and determine the lapping parameters at back-end process.
A key sensor element in a Hard Disk Drive (HDD) is the read-write head device. The device is complex 3D shape and its fabrication requires over thousand process steps with many of them being various types of image inspection and critical dimension (CD) metrology steps. In order to have high yield of devices across a wafer, very tight inspection and metrology specifications are implemented. Many images are collected on a wafer and inspected for various types of defects and in CD metrology the quality of image impacts the CD measurements. Metrology noise need to be minimized in CD metrology to get better estimate of the process related variations for implementing robust process controls.
Though there are specialized tools available for defect inspection and review allowing classification and statistics. However, due to unavailability of such advanced tools or other reasons, many times images need to be manually inspected. SEM Image inspection and CD-SEM metrology tools are different tools differing in software as well. SEM Image inspection and CD-SEM metrology tools are separate tools differing in software and purpose. There have been cases where a significant numbers of CD-SEM images are blurred or have some artefact and there is a need for image inspection along with the CD measurement. Tool may not report a practical metric highlighting the quality of image. Not filtering CD from these blurred images will add metrology noise to the CD measurement. An image classifier can be helpful here for filtering such data. This paper presents the use of artificial intelligence in classifying the SEM images. Deep machine learning is used to train a neural network which is then used to classify the new images as blurred and not blurred. Figure 1 shows the image blur artefact and contingency table of classification results from the trained deep neural network. Prediction accuracy of 94.9 % was achieved in the first model. Paper covers other such applications of the deep neural network in image classification for inspection, review and metrology.
In contrast to semiconductor manufacturing where features are mostly lines or contact holes, the disk drive reader has a complex, nonlinear 3D geometry. Metrology of such geometries is challenging; especially with regard to repeatability of measurements. New methods were needed to keep up with production requirements for metrology regarding uncertainty of critical dimensions (CD). We report a new method developed for CD metrology of the disk drive writer pole. The method demonstrated improved uncertainties compared to the regularly used CD-SEM algorithms and also has capability for side wall angle (SWA) metrology for process control.
The method utilizes multiple steps: a) extract contours from SEM images, b) identify exact locations on a curvilinear feature where CD should be measured, and c) provide CD measurements at these locations. SEM images from a variety of production wafers were used for evaluation of the developed method. Multiple series of SEM images were processed using a software utilizing advanced algorithms without using regular brightness threshold CD-SEM methodologies. It was found that CD repeatability was improved by a factor of three compared to the results of the regular threshold based CD-SEM method.
TEM imaging is used to measure side wall angle; it takes a lot of efforts for sample preparation and the feedback is slow. If extraction of SWA is possible from top down SEM images it would provide instant feedback to manufacturing and reduce cost. Monte Carlo simulations were used to understand the sensitivity of SWA to the trench CD and depth. Height of features was measured using AFM. A method to extract SWA from top down images was developed. Numerous SEM images were processed; results were compared to experiment and analyzed. It was shown that the 3-sigma repeatability of SWA measurement was 0.15 degree. It was also found that the left and the right SWA were different on multiple wafers, the results were very consistent from one image series to another one, at the same time the SWA difference between the left and right walls was considerably larger than the uncertainty of SWA measurement.