Polishing method and polishing apparatus

Abstract

A substrate polishing method capable of reducing an influence of variation in spectrum of reflected light from a substrate, such as a wafer, and determining an accurate film thickness is disclosed. The method includes: polishing a surface of a substrate by pressing the substrate against a polishing pad on a rotating polishing table; producing a spectrum of reflected light from the surface of the substrate each time the polishing table makes one rotation; creating a three-dimensional data containing a plurality of spectra arranged along polishing time; and determining a film thickness of the substrate based on the three-dimensional data.

What is claimed is: 
     
         1 . A polishing method comprising:
 polishing a surface of a substrate by pressing the substrate against a polishing pad on a rotating polishing table;   producing a spectrum of reflected light from the surface of the substrate each time the polishing table makes one rotation;   creating a three-dimensional data containing a plurality of spectra arranged along polishing time; and   determining a film thickness of the substrate based on the three-dimensional data.   
     
     
         2 . The polishing method according to  claim 1 , wherein determining the film thickness of the substrate comprises:
 inputting the three-dimensional data into a film-thickness calculation model that has been constructed according to an artificial intelligence algorithm; and   outputting the film thickness from the film-thickness calculation model.   
     
     
         3 . The polishing method according to  claim 2 , wherein the film-thickness calculation model is a trained model that has been constructed with use of a training data set containing a combination of a plurality of training three-dimensional data and a plurality of film thicknesses associated with the plurality of training three-dimensional data, respectively, and each of the plurality of training three-dimensional data includes a plurality of reference spectra arranged along polishing time. 
     
     
         4 . The polishing method according to  claim 3 , wherein the film-thickness calculation model is the trained model obtained by:
 creating the plurality of training three-dimensional data each containing a plurality of reference spectra produced when a plurality of reference substrates are polished;   dividing the plurality of training three-dimensional data into a plurality of groups according to a clustering algorithm;   creating test three-dimensional data by arranging a plurality of spectra along polishing time, the plurality of spectra of the test three-dimensional data being produced when a test substrate is polished;   selecting, from the plurality of groups, one group including a training three-dimensional data that best matches the test three-dimensional data; and   constructing the film-thickness calculation model with use of the training data set containing the combination of the plurality of training three-dimensional data belonging to the selected group and the plurality of film thicknesses associated with the plurality of training three-dimensional data, respectively.   
     
     
         5 . A polishing apparatus comprising:
 a polishing table for supporting a polishing pad, the polishing table being rotatable;   a polishing head configured to press a substrate against the polishing pad to polish a surface of the substrate;   a sensor head located in the polishing table, the sensor head being configured to direct light to the surface of the substrate and receive reflected light from the surface of the substrate; and   a processing system having a memory storing therein a program configured to cause the processing system to produce a spectrum of the reflected light, create a three-dimensional data containing a plurality of spectra arranged along polishing time, and determine a film thickness of the substrate based on the three-dimensional data.   
     
     
         6 . The polishing apparatus according to  claim 5 , wherein the memory stores therein a film-thickness calculation model that has been constructed according to an artificial intelligence algorithm, and
 the processing system is configured to input the three-dimensional data into the film-thickness calculation model, and output the film thickness from the film-thickness calculation model.   
     
     
         7 . The polishing apparatus according to  claim 6 , wherein the film-thickness calculation model is a trained model that has been constructed with use of a training data set containing a combination of a plurality of training three-dimensional data and a plurality of film thicknesses associated with the plurality of training three-dimensional data, respectively, and each of the plurality of training three-dimensional data includes a plurality of reference spectra arranged along polishing time. 
     
     
         8 . The polishing apparatus according to  claim 6 , wherein the film-thickness calculation model is the trained model obtained by:
 creating the plurality of training three-dimensional data each containing a plurality of reference spectra produced when a plurality of reference substrates are polished;   dividing the plurality of training three-dimensional data into a plurality of groups according to a clustering algorithm;   creating test three-dimensional data by arranging a plurality of spectra along polishing time, the plurality of spectra of the test three-dimensional data being produced when a test substrate is polished;   selecting, from the plurality of groups, one group including a training three-dimensional data that best matches the test three-dimensional data; and   constructing the film-thickness calculation model with use of the training data set containing the combination of the plurality of training three-dimensional data belonging to the selected group and the plurality of film thicknesses associated with the plurality of training three-dimensional data, respectively.