A method for correcting a mask patter includes: acquiring an initial pattern of a mask, the initial pattern including a scribe line area and die areas which are spaced, and the scribe line area is located between two adjacent die areas, each of the die areas includes at least one die sub-area and at least one first sub-test element group (TEG) area, and the scribe line area includes scribe line sub-areas and second sub-TEG areas, the first sub-TEG area and the second sub-TEG area are adjacent to each other, and the first sub-TEG area and the second sub-TEG area constitute a TEG area; performing an optical proximity correction (OPC) on an area of the initial pattern excluding TEG areas, so as to acquire a final pattern.

A method for correcting a mask patter includes: acquiring an initial pattern of a mask, the initial pattern including a scribe line area and die areas which are spaced, and the scribe line area is located between two adjacent die areas, each of the die areas includes at least one die sub-area and at least one first sub-test element group (TEG) area, and the scribe line area includes scribe line sub-areas and second sub-TEG areas, the first sub-TEG area and the second sub-TEG area are adjacent to each other, and the first sub-TEG area and the second sub-TEG area constitute a TEG area; performing an optical proximity correction (OPC) on an area of the initial pattern excluding TEG areas, so as to acquire a final pattern.

A method for manufacturing a semiconductor device, includes receiving a first layout including patterns for the manufacturing of the semiconductor device, generating a second layout by performing machine learning-based process proximity correction (PPC) based on features of the patterns of the first layout, generating a third layout by performing optical proximity correction (OPC) on the second layout, and performing a multiple patterning process based on the third layout. The multiple patterning process includes patterning first-type patterns, and patterning second-type patterns. The machine learning-based process proximity correction is performed based on features of the first-type patterns and features of the second-type patterns.

A deep trench layout implementation for a semiconductor device is provided. The semiconductor device includes an isolation film with a shallow depth, an active area, and a gate electrode formed in a substrate; a deep trench isolation surrounding the gate electrode and having one or more trench corners; and a gap-fill insulating film formed inside the deep trench isolation. The one or more trench corners is formed in a slanted shape from a top view.

A deep trench layout implementation for a semiconductor device is provided. The semiconductor device includes an isolation film with a shallow depth, an active area, and a gate electrode formed in a substrate; a deep trench isolation surrounding the gate electrode and having one or more trench corners; and a gap-fill insulating film formed inside the deep trench isolation. The one or more trench corners is formed in a slanted shape from a top view.

A pattern decomposition method including following steps is provided. A target pattern is provided, wherein the target pattern includes first patterns and second patterns alternately arranged, and the width of the second pattern is greater than the width of the first pattern. Each of the second patterns is decomposed into a third pattern and a fourth pattern, wherein the third pattern and the fourth pattern have an overlapping portion, and a pattern formed by overlapping the third pattern and the fourth pattern is the same as the second pattern. The third patterns and the first pattern adjacent to the fourth pattern are designated as first photomask patterns of a first photomask. The fourth patterns and the first pattern adjacent to the third pattern are designated as second photomask patterns of a second photomask.

A pattern decomposition method including following steps is provided. A target pattern is provided, wherein the target pattern includes first patterns and second patterns alternately arranged, and the width of the second pattern is greater than the width of the first pattern. Each of the second patterns is decomposed into a third pattern and a fourth pattern, wherein the third pattern and the fourth pattern have an overlapping portion, and a pattern formed by overlapping the third pattern and the fourth pattern is the same as the second pattern. The third patterns and the first pattern adjacent to the fourth pattern are designated as first photomask patterns of a first photomask. The fourth patterns and the first pattern adjacent to the third pattern are designated as second photomask patterns of a second photomask.

A mask correction method, a mask correction device for double patterning, and a training method for a layout machine learning model are provided. The mask correction method for double patterning includes the following steps. A target layout is obtained. The target layout is decomposed into two sub-layouts, which overlap at a stitch region. A size of the stitch region is analyzed by the layout machine learning model according to the target layout. The layout machine learning model is established according to a three-dimensional information after etching. An optical proximity correction (OPC) procedure is performed on the sub-layouts.

A mask correction method, a mask correction device for double patterning, and a training method for a layout machine learning model are provided. The mask correction method for double patterning includes the following steps. A target layout is obtained. The target layout is decomposed into two sub-layouts, which overlap at a stitch region. A size of the stitch region is analyzed by the layout machine learning model according to the target layout. The layout machine learning model is established according to a three-dimensional information after etching. An optical proximity correction (OPC) procedure is performed on the sub-layouts.

A semiconductor device may include a substrate including a first logic cell and a second logic cell, which are adjacent to each other in a first direction and shares a cell border, a first metal layer on the substrate, the first metal layer including a power line, which is disposed on the cell border to extend in a second direction crossing the first direction and has a center line parallel to the second direction, and a second metal layer on the first metal layer. The second metal layer may include a first upper interconnection line and a second upper interconnection line, which are provided on each of the first and second logic cells. The first upper interconnection line may extend along a first interconnection track and the first direction. The second upper interconnection line may extend along a second interconnection track and in the first direction.