The present disclosure describes a method for controlling radiation conditions and an example system for performing the method. The method includes sending a first setting to configure a radiation device to provide radiation to a substrate undergoing a process operation in a process chamber of the radiation device. The method further includes receiving radiation energy data measured at a plurality of locations of the process chamber and receiving measurement data measured on the substrate during the process operation. The method further includes in response to a variance of the radiation energy data being above a first predetermined threshold and in response to a difference between reference data and the measurement data being above a second predetermined threshold, sending a second setting to configure the radiation device to provide radiation to the substrate.

To provide more test data during the manufacture of non-volatile memories and other integrated circuits, machine learning is used to generate virtual test values. Virtual test results are interpolated for one set of tests for devices on which the test is not performed based on correlations with other sets of tests. In one example, machine learning determines a correlation study between bad block values determined at die sort and photo-limited yield (PLY) values determined inline during processing. The correlation can be applied to interpolate virtual inline PLY data for all of the memory dies, allowing for more rapid feedback on the processing parameters for manufacturing the memory dies and making the manufacturing process more efficient and accurate. In another set of embodiments, the machine learning is used to extrapolate limited metrology (e.g., critical dimension) test data to all of the memory die through interpolated virtual metrology data values.

Techniques and systems for critical dimension metrology are disclosed. Critical parameters can be constrained with at least one floating parameter and one or more weight coefficients. A neural network is trained to use a model that includes a Jacobian matrix. During training, at least one of the weight coefficients is adjusted, a regression is performed on reference spectra, and a root-mean-square error between the critical parameters and the reference spectra is determined. The training may be repeated until the root-mean-square error is less than a convergence threshold.

A source electrode (5), a drain electrode (6) and a T-shaped gate electrode (9) are formed on a GaN-based semiconductor layer (3,4) to form a transistor. An insulating film (10,11) covering the T-shaped gate electrode (9) is formed. A property of the transistor is evaluated to obtain an evaluation result. A film type, a film thickness or a dielectric constant of the insulating film (10,11) is adjusted in accordance with the evaluation result to make a property of the transistor close to a target property.

A method of evaluating metal contamination by measuring the amount of metal contaminants to a silicon wafer in a rapid thermal processing apparatus includes steps of obtaining a Si single crystal grown by the Czochralski method at a pulling rate of 1.0 mm/min or lower, the crystal having oxygen concentration of 1.3×10.sup.18 atoms/cm.sup.3 or less, slicing silicon wafers from the Si single crystal except regions of 40 mm toward the central portion from the head of the single crystal and 40 mm toward the central portion from the tail, heat-treating the silicon wafer with a rapid thermal processing apparatus and transferring contaminants from members in a furnace of the rapid thermal processing apparatus to the silicon wafer, and measuring a lifetime of the silicon wafer to which contaminants are transferred.

An overlay correcting method capable of optimizing correction of an overlay within a scanner correction limit of a scanner of a scanner system, and a photolithography method, a semiconductor device manufacturing method and the scanner system which are based on the overlay correcting method are provided. The overlay correcting method includes collecting overlay data by measuring an overlay of a pattern; calculating correction parameters of the overlay by performing regularized regression using the overlay data, the regularized regression being based on a correction limit of the scanner such that the correction parameters fall within the correction limit of the scanner; and providing the correction parameters to the scanner.

In learning processing performed before sample observation processing (steps S705 to S708), the sample observation device acquires a low-picture quality learning image under a first imaging condition for each defect position indicated by defect position information, determines an imaging count of a plurality of high-picture quality learning images associated with the low-picture quality learning image for each defect position and a plurality of imaging points based on a set value of the imaging count, acquires the plurality of high-picture quality learning images under a second imaging condition (step S702), learns a high-picture quality image estimation model using the low-picture quality learning image and the plurality of high-picture quality learning images (step S703), and adjusts a parameter related to the defect detection in the sample observation processing using the high-picture quality image estimation model (step S704).

The present invention disclosures a critical dimension error analysis method, comprising: S01: performing lithography processes on a wafer, measuring the critical dimension (CD) values of the test points in each of the fields respectively; M and N are integers greater than 1; S02: removing extreme outliers from the critical dimension (CD) values; S03: rebuilding remaining CD values by a reconstruction model fitting method, and obtaining rebuilt critical dimension (CD″) values, according to relative error between CD″ and CD, dividing the rebuilt critical dimension (CD″) values into scenes and the number of the scenes is A; S04: calculating components and corresponding residuals of the test points in each of the scenes under a reference system corresponding to a correction model by parameter estimation; S05: modifying machine parameters and masks by the correction model according to above calculation results. The present invention uses an outer limit to remove extreme outliers, so as to analyze a critical dimension error during a lithography process quickly and accurately.

A substrate inspection device for inspecting a substrate, includes: a setting part configured to define a group according to a basic state that is not dependent on a presence or absence of a defect in a substrate and set the defined group for each inspection target substrate; an inspection part configured to perform a defect inspection based on a captured image of the inspection target substrate and an inspection recipe corresponding to the defined group to which the inspection target substrate belongs and including a reference image; a recipe creation part configured to create the inspection recipe for each group; and a determination part configured to perform a determination as to whether a group-setting target substrate, for which the group is set by the setting part, belongs to the group defined by the setting part.

A machine learning model trained to provide metrology measurements for a substrate is provided. Training data generated for a prior substrate processed according to a prior process is provided to train the model. The training data includes a training input including a subset of historical spectral data extracted from a normalized set of historical spectral data collected for the prior substrate during the prior process. The subset of historical spectral data includes an indication of historical spectral features associated with a particular type of metrology measurement. The training data also includes a training output including a historical metrology measurement obtained for the prior substrate, the historical metrology measurement associated with the particular type of metrology measurement. Spectral data is collected for a current substrate processed according to a current process. A subset of current data extracted from a normalized set of the spectral data for the current substrate is provided as input to the trained model. Metrology measurement data for the current substrate is extracted from one or more outputs of the trained model.