Abstract: A computer system providing a function for: extracting, from image data of a semiconductor pattern, coordinate information of a base point for measuring a dimension of a desired location of the semiconductor pattern; and measuring the dimension using the coordinate information of the base point. The computer system includes a training device in which a pose estimation model for outputting coordinate information of at least two base points as a training result is implemented. The training device is trained in advance using teacher data having the image data of the semiconductor pattern as an input and the coordinate information of at least two base points as an output, and extracts, for new image data input into the training device, the coordinate information of at least two base points and a dimension.

Abstract: A computer system for extracting, from image data, coordinate information on base points for measuring a dimension of a desired portion of a pattern, and measuring the dimension by using the coordinate information, the computer system including a preprocessing unit configured to allow training by matching all of the samples by setting a base point insufficient in annotation data as an insufficient measurement portion and shielding the insufficient measurement portion on the image data for the sample in which coordinate values of only a part of the base points are described, the preprocessing unit is also configured to allow learning and is trained in advance by using training data in which the image data is set as an input and the coordinate information is set as an output, and the preprocessing unit extracts the coordinate information and the dimension for new image data input for learning.

Abstract: A dimension measurement apparatus that automatically corrects a deviation of a contour without operator determination. The dimension measurement apparatus includes: a model learning unit that obtains a learning cross-sectional image and learning labels attached to different regions of the learning cross-sectional image and generates a deep learning model for image region division using the learning cross-sectional image and the learning labels; a model estimation unit that applies the model to a newly input target image and labels each independent region; a contour correction unit that detects a contour of each region using the target image and the labels, sets a representative point on the contour of the region, moves each representative point according to a movement rule, and repeats movement of the representative point until contour correction is complete; and a dimension measurement unit that measures a dimension of a device cross-sectional structure using the corrected contour.

Abstract: An imaging device is provided with a sample stage configured to be capable of transferring a sample by at least two drive shafts, and shifting an imaging field of view corresponding to positional information of the sample, the positional information being obtained by a computer system. The computer system includes a classifier which outputs, in response to input of image data of a tilt image where the sample is imaged in an inclined state with respect to a charged particle beam, positional information of one or more feature parts existing on the tilt image. The classifier is trained in advance by using training data in which input is image data of the tilt image, and output is positional information of the feature part. The computer system executes, with respect to new tilt image data inputted into the classifier, a processing to output positional information of the feature part.

Abstract: In order to enable a user to utilize optimum reference processing data for searching for a target processing condition from a large number of stored reference processing data without requiring special knowledge of machine learning, a search apparatus that searches for a manufacturing condition corresponding to a desired processing result of a semiconductor manufacturing apparatus by predicting the manufacturing condition corresponding to the desired processing result using a learning model is configured to generate a learning model by transfer learning using first data and second data, and regenerate, when the generated learning model does not satisfy a predetermined determination criterion, a learning model by transfer learning using the first data and the added second data.

Abstract: To facilitate evaluation of a predicted process shape in process recipe development using machine learning, a process recipe search apparatus that searches for an etching recipe that is a parameter of a plasma processing apparatus set so as to etch a sample into a desired shape displays, on a display device, the predicted process shape of the sample by a candidate etching recipe predicted by using a machine leaning model, by highlighting a difference between the predicted process shape and a target shape.

Abstract: A computer system is configured to: receive input of a plurality of pieces of learning data, a piece of learning data includes a combination of a value of at least one input parameter which is the apparatus condition and a value of at least one output parameter which is a result of processing; analyze a structure of manufacturing steps of the processing in a piece of standard learning data and a structure of manufacturing steps of the processing in a piece of other learning data; execute, based on a result of the analysis, classification processing of identifying a piece of learning data which is aggregable with a same dataset as the piece of standard learning data; generate a dataset; execute processing of training a model; execute search processing of searching for the value of the at least one input parameter; and present a result of the search processing.

Abstract: High-accuracy dimension extraction for a complicated shape possibly appearing through semiconductor processing. A starting point and an endpoint are placed on the periphery of a graphic shape including a combination of multiple ellipses, and a curve unicursally drawn on the periphery between the two points is used as a shape mode, thereby describing a contour of an intended structure.

Abstract: There are provided a charged particle beam device capable of appropriately setting a tilt angle of a sample stage even if an observation surface is larger than the field of view, and a sample observation method using the same.

Abstract: A computer system providing a function for: extracting, from image data of a semiconductor pattern, coordinate information of a base point for measuring a dimension of a desired location of the semiconductor pattern; and measuring the dimension using the coordinate information of the base point. The computer system includes a training device in which a pose estimation model for outputting coordinate information of at least two base points as a training result is implemented. The training device is trained in advance using teacher data having the image data of the semiconductor pattern as an input and the coordinate information of at least two base points as an output, and extracts, for new image data input into the training device, the coordinate information of at least two base points and a dimension.

Abstract: High-accuracy fitting for complex shapes that can occur in semiconductor processing is performed. A shape model is a curved line that is drawn with one stroke from its start point to its end point along a periphery of a figure that is a combination of one or more ellipses and one or more line segments in an xy-plane defined by an x axis and a y axis which are perpendicular to each other, and constraint conditions of a shape model are derived on the basis of a processing dimension function that represents processing dimensions specified on the basis of singular points in the shape model and that has shape model parameters as its variables, and definition ranges of the processing dimensions.

Abstract: A model learning unit learns a prediction model on the basis of learning data, a target setting unit sets a target output parameter value by interpolating between a goal output parameter value and an output parameter value which is the closest to the goal output parameter value in output parameter values in the learning data, a processing condition search unit estimates input parameter values which corresponds to the goal output parameter value and the target output parameter value, a model learning unit updates the prediction model by using a set of the estimated input parameter value and an output parameter value which is a result of processing that a processing device performs as additional learning data.

Abstract: For a machine learning model that receives control parameters of a semiconductor processing device and outputs shape parameters that express a processed shape of a semiconductor sample processed by the semiconductor processing device, an experiment point obtaining learning data is recommended. A contribution of each control parameter to the prediction of the machine learning model is evaluated from feature quantity data that is a value of a control parameter of the learning data used for learning of the machine learning model, and the experiment point is recommended based on a stability evaluation and an uncertainty evaluation of the prediction by the machine learning model in a space defined by the control parameters selected based on the contribution as axes.

Abstract: High-accuracy dimension extraction for a complicated shape possibly appearing through semiconductor processing. A starting point and an endpoint are placed on the periphery of a graphic shape including a combination of multiple ellipses, and a curve unicursally drawn on the periphery between the two points is used as a shape mode, thereby describing a contour of an intended structure.

Abstract: A parameter compression unit compresses first input parameter values so that a parameter restoration unit can restore the first input parameter values, and generates first compressed input parameter values in which the number of control parameters is reduced, a model learning unit learns a prediction model from learning data that is a set of the first compressed input parameter values and first output parameter values that are processing results obtained by giving the first input parameter values, as a plurality of control parameters, to a processing device, and a processing condition search unit estimates a second compressed input parameter values corresponding to target output parameter values by using the prediction model.

Abstract: A model learning unit learns a prediction model on the basis of learning data, a target setting unit sets a target output parameter value by interpolating between a goal output parameter value and an output parameter value which is the closest to the goal output parameter value in output parameter values in the learning data, a processing condition search unit estimates input parameter values which corresponds to the goal output parameter value and the target output parameter value, a model learning unit updates the prediction model by using a set of the estimated input parameter value and an output parameter value which is a result of processing that a processing device performs as additional learning data.

Abstract: Provided is a system that constructs a learning database for a learning model in a short period of time. The system generates an image in which a structure image is similar to an actual image by image processing of the structure image. One or more processors acquire a first structure image and a second structure image different from the first structure image. The one or more processors create a plurality of intermediate structure images indicating an intermediate structure between the first structure image and the second structure image. The one or more processors generate an image by making each of the plurality of intermediate structure images to be similar to an actual image by image processing of each of the plurality of intermediate structure images.

Abstract: A model learning unit learns a prediction model on the basis of learning data, a target setting unit sets a target output parameter value by interpolating between a goal output parameter value and an output parameter value which is the closest to the goal output parameter value in output parameter values in the learning data, a processing condition search unit estimates input parameter values which corresponds to the goal output parameter value and the target output parameter value, a model learning unit updates the prediction model by using a set of the estimated input parameter value and an output parameter value which is a result of processing that a processing device performs as additional learning data.

Abstract: A dimension measurement apparatus that automatically corrects a deviation of a contour without operator determination. The dimension measurement apparatus includes: a model learning unit that obtains a learning cross-sectional image and learning labels attached to different regions of the learning cross-sectional image and generates a deep learning model for image region division using the learning cross-sectional image and the learning labels; a model estimation unit that applies the model to a newly input target image and labels each independent region; a contour correction unit that detects a contour of each region using the target image and the labels, sets a representative point on the contour of the region, moves each representative point according to a movement rule, and repeats movement of the representative point until contour correction is that measures a dimension of a device complete; and a dimension measurement unit cross-sectional structure using the corrected contour.

Abstract: To efficiently search a processing condition of giving a desired target processing result, there is provided a processing condition determination system including a processing apparatus that processes a sample, a processing monitor system that monitors the state of the processing in the processing apparatus, and an analysis system that sets the processing condition of the processing apparatus of giving a target processing result, wherein the system includes a processing condition and result database that stores a set of an explanatory variable that is a processing condition under which the processing apparatus processes a sample and an objective variable that is the processing result obtained by the processing apparatus' processing the sample, and when the processing apparatus processes the sample under the processing condition set using the correlation model derived from the database and it is determined that a probability of failure occurrence becomes high, based on the monitor data of the processing monito