Polishing apparatus using neural network for monitoring

A polishing system, comprising:

• a carrier to hold a substrate;
• a support for a polishing surface;
• an in-situ monitoring system having a sensor, the in-situ monitoring system configured to generate a plurality of initial values for a plurality of different locations on the layer;
• a motor to generate relative motion between the sensor and the substrate;
• and
• a controller configured to
• input a first multiplicity of initial values from the plurality of initial values into a first plurality of corresponding input nodes of a neural network, the first multiplicity of initial values corresponding to a first multiplicity of locations from the plurality of different locations that are within a first region on the substrate, and wherein initial values from the plurality of initial values that are not input into input nodes of the neural network include a second multiplicity of initial values corresponding to a second multiplicity of locations from the plurality of different locations that are within a different second region on the substrate;
• receive a first multiplicity of modified values from a plurality of corresponding output nodes of the neural network;
• and
• at least one of detect a polishing endpoint or modify a polishing parameter based on the first multiplicity of modified values and the second multiplicity of initial values.

A computer storage medium encoded with instructions that, when executed by one or more computers, cause the one or more computers to perform operations comprising:

• during polishing of a layer on a substrate at a polishing station, receive a plurality of initial values for a plurality of different locations on the layer from an in-situ monitoring system;
• input a first multiplicity of initial values from the plurality of initial values into a first plurality of corresponding input nodes of a neural network, the first multiplicity of initial values corresponding to a first multiplicity of locations from the plurality of different locations that are within a first region on the substrate, and wherein initial values from the plurality of initial values that are not input into input nodes of the neural network include a second multiplicity of initial values corresponding to a second multiplicity of locations from the plurality of different locations that are within a different second region on the substrate;
• receive a first multiplicity of modified values from a plurality of corresponding output nodes of the neural network;
• and
• at least one of detect a polishing endpoint or modify a polishing parameter based on the first multiplicity of modified values and the second multiplicity of initial values.