Method and system for a two-step prediction of a quality distribution of semiconductor devices
AssigneeAdvanced Micro Devices, Inc.
InventorRichard Good
Life of this patent
Estimating the term…
Abstract
By performing a two-step approach for predicting a quality distribution during the fabrication of semiconductor devices, enhanced flexibility and efficiency may be accomplished. The two-step approach first models electrical characteristics on the basis of measurement data, such as inline measurement data, and, in a second step, an appropriate distribution for the electrical characteristics may be established, thereby obtaining modeled wafer sort data which may then be used for predicting a quality distribution of the semiconductor devices under consideration.
Claims22
A method, comprising:
- obtaining measurement data in a data processing system from a plurality of selected samples of a group of substrates after performing a first sequence of manufacturing processes in a manufacturing environment configured to produce semiconductor devices on the basis of a predefined overall manufacturing flow, said first sequence comprising at least one production process and at least one metrology process;
- determining a plurality of predicted electrical characteristics of semiconductor devices for said plurality of selected samples using said measurement data, said plurality of electrical characteristics corresponding to a state of said semiconductor devices after passing through said manufacturing flow;
- and
- determining a predicted quality distribution for said group of substrates by using said plurality of predicted electrical characteristics.
A method, comprising:
- creating a model by using a multi-dimensional regression technique, said model relating historical measurement data, obtained from a plurality of substrates comprising semiconductor devices after performing one or more manufacturing processes in a semiconductor manufacturing environment, to final electrical characteristics of said semiconductor devices;
- determining a subset of predicted values of said final electrical characteristics, said subset including said predicted values for selected samples of a group of substrates after performing said one or more manufacturing processes, by using said model and measurement data obtained from said selected samples;
- determining a complete set of predicted values of said final electrical characteristics, said complete set including said predicted values for each semiconductor device of said group;
- and
- determining a quality distribution of said semiconductor devices on the basis of said complete set of predicted values.
A yield prediction system, comprising:
- an interface configured to receive measurement data relating to a sequence of manufacturing processes of an overall manufacturing flow in a manufacturing environment;
- a first prediction unit connected to said interface and configured to perform a data processing to determine final electrical characteristics of sample semiconductor devices on the basis of a multi-dimensional model and measurement data obtained from said sample semiconductor devices;
- and
- a second prediction unit connected to said first prediction unit and configured to determine said electrical characteristics for each semiconductor device associated with said sample semiconductor devices and to determine a yield distribution for said semiconductor devices associated with said sample semiconductor devices.
Dependencies are inferred from the claim text.