Implementation of deep neural networks for testing and quality control in the production of memory devices

1. A method, comprising:

preparing a first plurality of samples of an integrated circuit for cross-sectional imaging;

generating by an electron microscope of cross-sectional images of the first plurality of prepared samples of the integrated circuit;

training a neural network to generate pixel level coordinate locations of features, the pixel level coordinate locations including pixel level coordinates of corners of a bounding box for the features, of the first plurality of prepared samples of the integrated circuit from the cross-sectional images generated by the electron microscope;

applying the trained neural network to one or more additional samples of the integrated circuit to generate pixel level coordinate locations of the features of the one or more additional samples of the integrated circuit;

analyzing results of the applying of the trained neural network to the one or more additional samples of the integrated circuit;

adjusting processing for fabrication of the integrated circuit based upon the analyzing of the results of the application of the trained neural network to the one or more additional samples of the integrated circuit; and

fabricating the integrated circuit by the adjusted processing.

2. The method of claim 1 , wherein preparing a first plurality of samples of an integrated circuit for cross-sectional imaging includes:

fabricating a plurality of integrated circuits comprising a plurality of layers formed over a surface of a substrate; and

slicing the fabricated plurality of integrated circuits perpendicularly to surface of the substrate.

3. The method of claim 1 , wherein preparing a first plurality of samples of an integrated circuit for imaging includes:

fabricating a plurality of integrated circuits each having a plurality of layers formed upon a substrate, wherein the features include voids in the integrated circuits.

4. The method of claim 3 , wherein the plurality of layers include alternating layers of a plurality of metal layers and dielectric layers and the voids are voids in the dielectric layers.

5. The method of claim 1 , wherein preparing a first plurality of samples of an integrated circuit for imaging includes:

fabricating a plurality of integrated circuits each having a plurality of layers formed upon a substrate, wherein the features include shapes of features in the integrated circuits.

6. The method of claim 1 , wherein analyzing results of the applying of the trained neural network to the one or more additional samples of the integrated circuit includes determining an area of features of the one or more additional samples of the integrated circuit.

7. The method of claim 1 , wherein analyzing results of the applying of the trained neural network to the one or more additional samples of the integrated circuit includes performing a binning operation of the features according to type.

8. The method of claim 1 , wherein analyzing results of the applying of the trained neural network to the one or more additional samples of the integrated circuit includes correlating the pixel level coordinate locations of features to elements of the integrated circuits.

9. The method of claim 1 , wherein adjusting processing for fabrication of the integrated circuit includes:

adjusting a temperature.

10. The method of claim 1 , wherein adjusting processing for fabrication of the integrated circuit includes:

adjusting a concentration.

11. The method of claim 1 , wherein adjusting processing for fabrication of the integrated circuit includes:

adjusting a time.

12. A method, comprising:

fabricating a first plurality of examples of an integrated circuit by:

forming alternating plurality of metal layers and dielectric layers over a surface of a substrate;

generating a plurality of cross-sectional images from the first plurality of the examples of the integrated circuit by an electron microscope;

training a neural network to generate pixel level coordinate data of features from the plurality of cross-sectional images generated from the first plurality of the examples of the integrated circuit;

fabricating a second plurality of examples of the integrated circuit by:

forming alternating layers of metal layers and dielectric layers over a surface of a substrate according to a first set of processing parameters;

generating a plurality of cross-sectional images from the second plurality of the examples of the integrated circuit by an electron microscope;

analyzing pixel level coordinate data of features from the plurality of cross-sectional images from the second plurality of the examples of the integrated circuit by the trained neural network;

adjusting the first set of processing parameters based upon the analyzing the plurality of cross-sectional images from the second plurality of the examples of the integrated circuit by the trained neural network; and

fabricating one or more additional examples of the integrated circuit according to the adjusted first set of processing parameters.

13. The method of claim 12 , wherein the integrated circuit is a non-volatile memory circuit having a three dimensional architecture.

14. The method of claim 13 , wherein the metal layers include a plurality of word lines and the dielectric layers include one or more inter-word line dielectric layers.

15. The method of claim 14 , wherein the features include voids in the inter-word line dielectric layers.

16. The method of claim 12 , wherein analyzing pixel level coordinate data of features from the plurality of cross-sectional images from the second plurality of the examples of the integrated circuit by the trained neural network includes:

determining a shape for each of the features.

17. The method of claim 12 , wherein analyzing pixel level coordinate data of features from the plurality of cross-sectional images from the second plurality of the examples of the integrated circuit by the trained neural network includes:

determining a size for each of the features.

18. The method of claim 12 , wherein analyzing pixel level coordinate data of features from the plurality of cross-sectional images from the second plurality of the examples of the integrated circuit by the trained neural network includes:

determining a pixel level location for each of the features.

19. A system, comprising:

one or more processors, the one or more processors configured to:

receive, from one or more electron microscopes, cross-sectional images of a plurality of prepared training samples of an integrated circuit produced at a fabrication facility;

from the plurality of prepared training samples of the integrated circuit, train a neural network to generate pixel level coordinate locations of features, the pixel level coordinate locations including pixel level coordinates of corners of a bounding box for the features, of the plurality of prepared training samples of the integrated circuit from the cross-sectional images from the one or more electron of microscopes;

apply the trained neural network to one or more additional samples of the integrated circuit produced at the fabrication facility according to a set of processing parameters to generate pixel level coordinate locations of the features of the one or more additional samples of the integrated circuit;

analyze results of the applying of the trained neural network to the one or more additional samples of the integrated circuit;

determine adjusted processing parameters for fabrication of the integrated circuit based upon the analyzing of the results of applying the trained neural network to the one or more additional samples of the integrated circuit; and

provide the adjusted processing parameters to the fabrication facility for subsequent fabrication of the integrated circuit.