Machine learning model based controller for rapid thermal processing chamber
Embodiments disclosed herein include a method of developing a reduced order model (ROM) for a model based controller. In an embodiment, the method comprises obtaining a design of a plant, and building a detailed model of the thermal network of the plant from the design of the plant. In an embodiment, the method further comprises obtaining a training input recipe, and running the detailed model using the training input recipe. In an embodiment, the method further comprises generating a plurality of snapshots, wherein each snapshot includes the temperatures of a plurality of components in the detailed model, and utilizing a dynamic mode decomposition with control (DMDc) operation in order to extract the ROM from the plurality of snapshots.
1 . A method of developing a reduced order model (ROM) for a model based controller, comprising:
obtaining a design of a rapid thermal processing (RTP) tool, the RTP tool comprising a plurality of heater zones at a lid of a chamber, and a reflector plate over a bottom of the chamber;
building a detailed model of the thermal network of the RTP tool from the design of the RTP tool;
obtaining a training input recipe;
running the detailed model using the training input recipe;
generating a plurality of snapshots, wherein each snapshot includes the temperatures of a plurality of components in the detailed model;
utilizing a dynamic mode decomposition with control (DMDc) operation in order to extract the ROM from the plurality of snapshots;
using the ROM in a recipe to set a temperature for each zone of the plurality of heater zones and to set a temperature for the reflector plate; and
processing a substrate between the plurality of heater zones and the reflector using the recipe.
2 . The method of claim 1 , further comprising:
calibrating the detailed model with available experimental data.
3 . The method of claim 1 , wherein the DMDc operation includes a non-linear component.
4 . The method of claim 1 , wherein the ROM is in the format of {dot over (x)}=Ax+Bu, wherein A and B are matrices.
5 . The method of claim 1 , wherein the ROM is an approximation of the actual governing equations of thermodynamics for the RTP tool.
6 . The method of claim 1 , wherein an error between an output of the ROM and an output of the detailed model is within 10%.
7 . The method of claim 1 , wherein the design of the RTP tool is a computer aided design (CAD) file.
8 . The method of claim 3 , wherein the ROM is in the format of {dot over (x)}=Ax 4 +Bx+c+Du, wherein A, B, and D are matrices.
9 . A processing tool, comprising:
a chamber;
a plurality of heater zones at a lid of the chamber;
a reflector along a bottom of the chamber;
a substrate support between the plurality of lamps and the reflector; and
a controller coupled to the chamber for controlling a temperature of the substrate, wherein the controller is a model based controller that utilizes a reduced order model (ROM) generated with a dynamic mode decomposition with control (DMDc) process, wherein the controller uses the ROM in a recipe to set a temperature for each zone of the plurality of heater zones and to set a temperature for the reflector plate, and wherein the controller is configured to process a substrate on the substrate support between the plurality of heater zones and the reflector using the recipe.
10 . The processing tool of claim 9 , wherein the processing tool is a rapid thermal processing (RTP) tool.
11 . The processing tool of claim 9 , wherein the ROM is in the format of {dot over (x)}=Ax+Bu, wherein A and B are matrices.
12 . The processing tool of claim 9 , wherein the ROM is in the format of {dot over (x)}=Ax 4 +Bx+c+Du, wherein A, B, and D are matrices.
13 . The processing tool of claim 9 , wherein the ROM is generated from a plurality of snapshots.
14 . The processing tool of claim 9 , wherein the ROM is an approximation of the actual governing equations of thermodynamics for the processing tool.
15 . The processing tool of claim 13 , wherein the ROM is generated before the processing tool is assembled.
16 . A method of developing a reduced order model (ROM) for a model based controller, comprising:
generating a plurality of snapshots, wherein each snapshot includes the temperatures of a plurality of components in a processing tool, the processing tool comprising a plurality of heater zones at a lid of a chamber, and a reflector plate over a bottom of the chamber;
utilizing a dynamic mode decomposition with control (DMDc) operation in order to extract the ROM from the plurality of snapshots;
using the ROM in a recipe to set a temperature for each zone of the plurality of heater zones and to set a temperature for the reflector plate; and
processing a substrate between the plurality of heater zones and the reflector using the recipe.
17 . The method of claim 16 , wherein generating the plurality of snapshots comprises:
obtaining a computer aided design of a rapid thermal processing (RTP) tool;
building a detailed model of the thermal network of the RTP tool from the computer aided design of the RTP tool;
obtaining a training input recipe; and
running the detailed model using the training input recipe.
18 . The method of claim 16 , wherein generating the plurality of snapshots comprises:
running a training recipe on a processing tool; and
recording temperatures of a plurality of components at a plurality of times.