Creating and decomposing a DRC-compliant design for self-aligned multi-patterning processes is not a trivial matter.
The semiconductor industry began using multi-patterning technology to fabricate ICs at the 20-nm process node. When we got to 7 nm, the use of self-aligned multi-patterning (SAMP) techniques was introduced. In previous articles, we’ve explored and compared the three SAMP techniques, including self-aligned double patterning (SADP), self-aligned quadruple patterning (SAQP), and self- aligned litho-etch litho-etch (SALELE). Because the SADP and SAQP manufacturing processes require both track (metal line) and cut (also known as block) masks, their use also creates new design and verification challenges.
We then detailed the layout decomposition challenges and automation solutions for creating the track (metal line) masks. Now we’re completing our SAMP introductory series by exploring and explaining the process for creating cut masks, which are used to define the line ends of metal target shapes and to provide the proper electrical isolation between metal target shapes.
Cut mask shapes placement and decomposition rules
Generating cut masks for metal layers is a demanding process because the cut shapes are added in the gaps between metal target shapes. However, all design requirements between cut shapes must be observed while adding metal shapes, even though designers can’t yet see where the final cuts will need to be placed. Sounds impossible, doesn’t it? Fortunately, there are automation solutions design teams can use to create legal cut masks that comply with foundry requirements, and then some.
Let’s start with the basics. As shown in Figure 1, cut shapes have their own complex design constraints, including cut width, minimum/maximum cut length, minimum tip-to-tip, side-to-side, corner-to-corner spacing between cut shapes, minimum via enclosure distances, and minimum metal area between cut shapes. All of the associated design rules for these constraints must be taken into account when generating the cut mask.
Figure 1 Cut shapes have their own complex design constraints. Source: Siemens EDA
In addition to these common design rules, foundries often provide additional rules to enhance the printability of the cut mask shapes, such as N-Body rules, which are special rules constraining the spacings in a cluster of three or more cut shapes. For example, a basic design rule may state that the minimum spacing between two cut shapes must be more than 50 nm. However, if the mask contains three or four neighboring cut shapes, the foundry N-Body rule may require the spacing between any two cut shapes forming this cluster to be more than 60 nm, so this specific group of cut shapes must be placed in compliance with the more restrictive N-Body constraint.
Design rule compliance gets even more complex when there are optional—nice to have—spacing requirements. Of course, optional spacing constraints are not required, exactly, but designers know that satisfying them will enhance the manufacturing robustness of cut shapes at these locations, which can improve yield. The challenge is to satisfy as many optional spacing constraints as possible without violating any required constraints. Needless to say, ensuring compliance with all required and optional design constraints is extremely difficult to achieve manually (Figure 2).
Automated cut mask decomposition
Electronic design automation (EDA) multi-patterning tools not only automate the decomposition process, but help engineers more quickly and accurately understand and debug situations where the drawn target shapes do not enable proper cut mask generation. IMEC and Siemens EDA conducted a collaborative project to define and describe the automated decomposition process, using the Calibre Multi-Patterning tool as representative of EDA tool capabilities.
The Calibre Multi-Patterning tool helps designers automatically generate legal cut masks that satisfy all cut width and minimum cut length constraints, as well as all spacing constraints between cuts and between cuts and vias, taking into consideration the minimum metal area between cut shapes. Even when the design contains some minimum metal area violations, the tool attempts to fix them while maintaining all other constraints.
Figure 3 displays cut mask shapes generated for a real-world design. Cuts are placed at locations between metal target shapes, adjusting the placement as needed to ensure that there are no spacing or minimum area violations. For example, a cluster of neighboring cuts is placed in an arrangement that complies with minimum side-to-side, tip-to-tip, and corner-to-corner spacing constraints, as well as N-body constraints.
In the top row, the design target is very short and doesn’t satisfy the minimum area constraints. This error is fixed by moving the cuts away from the target to extend the metal length between cut shapes and satisfy minimum area constraints. In the lower left corner, cut shapes are placed to maintain the minimum area requirements of the dummy metal shape between the cut shapes. In the top middle, cuts are moved away from a target shape to maintain minimum spacing between the cut shape and the via.
Solving spacing violations between cut shapes
The simplest way to add cut mask shapes is to place them at every metal line end to create the required electrical isolation. However, it’s often not feasible in real layouts, as shown in Figure 4. Adding a cut shape at every line end generates cut shapes with illegal widths, illegal shapes, or spacing violations between the generated cut shapes.
To solve these problems, three techniques are used to find a legal solution: cut dropping, cut merging, and cut sliding (Figure 5).
It’s not mandatory to add a cut shape at each line end to create the required electrical isolation. Adding one cut shape between two metal lines could be sufficient. However, adding only one cut shape may result in long line end extensions, which can affect design performance. Designers using the Calibre Multi-Patterning tool can specify a maximum line end extension constraint to ensure the tool only drops cuts that don’t create a violation of this configured maximum line end extension setting.
Merging cuts from different tracks to form longer cuts can overcome spacing constraints between cuts. The maximum length of the generated cut shape can be constrained to meet foundry requirements.
One last way to generate legal cut mask shapes is to slide cuts away from a line end. Sliding is done under two conditions: 1) sliding the cut must not result in exceeding the maximum line end extension as configured in the tool, and 2) sliding the cut must not create spacing violations with other cut shapes.
The main objective of cut sliding is to find a legal placement of cut shapes that satisfies all the constraints.