It is learned on December 19 from Jinrong QiangXianKan (Financial Preview), WIMI has developed a redundant automatic filling system for chip design based on artificial intelligence to improve the yield rate of chip production.
Chip design is the key and the most challenging part of chip production. The process includes: selecting appropriate materials and structures, determining appropriate design methods and processes, and designing systems with specific functions and performance requirements.
In the process of chip physical design, after the timing and function convergence are completed, it will enter the DFM stage. At this time, the layout, whether it is the density of the metal or the width of the pattern, is uneven; Since the flatness of the wafer surface cannot be guaranteed, redundant filling must be used to improve the flatness of the chip.
Redundant fill is a layout structure that does not affect the logic function of the chip, and helps to improve the consistency of interconnect pattern density and is an important method to optimize the CMP process.
WIMI R&D team developed a redundant automatic filling system based on artificial intelligence chip design. Chip design redundancy filling usually adopts rule-based and pane-based redundancy filling methods:
Based on the rule-based redundant metal filling algorithm, two or more steps are used to perform redundant metal filling. Better image gradients are achieved by filling redundant graphs with larger sizes and buffer distances first and then smaller ones. The density of the overall image is further improved due to the use of multi-step padding and the introduction of redundant graphics with smaller sizes and buffer distances.
Based on the redundant metal filling method of the window format, better pattern density and pattern gradient can be obtained. This method uses a specific algorithm to fill redundant metal according to the surrounding environment of the design layout.
First divide the entire chip into panes (such as 50μm), then calculate the pattern density in each pane, and finally determine the filling amount of redundant metal according to the pattern density of adjacent panes.
The goal of redundant metal filling is to minimize the gradient of the graph, and the filling of the window format uses a circular algorithm to minimize the amount of redundant metal filled, and at the same time, the density gradient difference between adjacent windows is also minimized.
Both filling methods have insufficient constraints. The former filling method does not consider the pattern gradient, and the excessive pattern gradient is an important reason for the hot spots of CMP and etching processes. The latter leads to redundant inconsistencies in metal filling. This affects the uniformity of the etching and CMP processes.
Based on this, the WIMI research and development team uses artificial intelligence technology in the chip redundancy automatic filling system, and uses artificial intelligence algorithm technology to select and calculate the filling density and gradient.
The artificial intelligence chip design redundancy automatic filling system obtains the high-yield rate redundant graphics to fill the template map, corresponds to the chip layout that needs to be filled as the training image, and obtains the filling position label and gradient of the redundant graphics through the calculation of the artificial intelligence model.
Redundant filling in chip design allows the consistency of interconnect pattern density to be maintained during chip production, which greatly improves the production yield of chips.