A common thread woven throughout semiconductor design and manufacturing is the importance of the 3D nature of the process and architecture of the circuit. Until recently, designers have only been able to utilize 2D techniques to help them understand their design, understand their processes, and communicate across groups. Unfortunately, some key information is lost, and sometimes even rendered incorrect, as it is translated from 2D depictions into the reality of the third dimension.
Some common 2D techniques for communication in design and manufacturing are:
- 2D cross-section drawings made by hand in non-optimized office software programs
- Mechanical CAD tools like AutoCAD
- Common GDS layered views
Today, semiconductor manufacturers are facing many critical challenges at the sub-90-nm technology nodes, including increased costs, the need for shorter cycle times, the need for critical yield improvements, and optimization of the usage of expensive production and lab equipment. Development challenges for the semiconductor industry today result from tighter geometries and new materials, as well as the introduction of new transistor architectures. These changes cause an increase in the difficulty of design, debug, and analysis for yield enhancement. New schemes and methodologies involving 3D information are being implemented for fab rampup and production control, lab failure analysis (FA), and circuit editing (CE). Semiconductor customers are signalling that 3D analysis (for example, using focused ion beams and scanning electron microscopes, or SEMs) is becoming essential to increasing yield and for routine FA operations.
Along with the need for more intensive analysis methods, smaller geometries and more complex designs are requiring accurate 3D information to decrease operator guesswork, to allow for planning of experiments, and to increase operator efficiency and decrease time-to-result. For example, for CE, a 3D model interaction enables design of edit (DOE), which allows for successful planning and execution of a device edit with less guesswork and an increased success rate. Without understanding the analysis completely and visualizing the interaction of that region with the surrounding area, there is a much higher chance of failure during device edit or any typical failure analysis.
Having accurate 3D representations has now become an essential communication link between design and manufacturing, for purposes including:
- Communication between engineers, lab, and fab
- Technical and sales publications
- Interactive training
- 3D web and new 3D PDF documents
3D voxel representations
Semiconductor devices and interconnects are very complex, with many details that need to be captured and understood to improve productivity and enhance yield. Solid-modeling CAD software that relies on analytical descriptions of the geometry to capture the details is not suitable because it cannot depict sub-nanometer resolutions and often fails to create an accurate solid model. An example might include mechanical CAD tools based on non-uniform rational-basis spline (NURBS) representations.
However, a different solid-modeling approach is quite capable of supporting advanced-node semiconductor manufacturing. Utilizing 3D modeling software, 3D models can be built from “voxels,” or volumetric pixels, which can be thought of as 3D pixels. Just as SEM images are built from 2D pixels, this new solid-modeling technology builds models from 3D voxels. These 3D models contain an enormous amount of numerical data that is processed by very advanced numerical techniques to enable a superior 3D display and memory management.
The voxel-based approach is well suited to semiconductor manufacturing because it builds models layer by layer. This is important because semiconductor manufacturing distinguishes itself from traditional mechanical engineering in the processes that are used. While mechanical engineers carve away at materials and assemble them, a semiconductor manufacturer uses many complex physical processes and cells to deposit, grow, etch, manipulate, and create materials. Therefore, CAD tools for the semiconductor world need to understand the concepts of mask sets and process sequences. Traditional solid-modeling CAD tools fail under these conditions.
As designs increase in complexity it becomes even more important for manufacturers to be able to communicate clearly and efficiently about their processes and designs. 3D modeling of semiconductor processes can be used across a broad application space that can enable clear communication across numerous boundaries:
- Engineer to engineer (in both lab and fab) for process discussions, as well as transfers of analytic results
- Manufacturer to vendor
- Manufacturing to design
This standardization and improvement in communication can improve the efficiency of actual analysis and can decrease the time spent on expensive in-fab or lab equipment. For example, each day brings many requests for analysis in a semiconductor manufacturing environment. When 3D-visualization tools are used there can be a significant efficiency improvement in time spent detailing and explaining what analysis is to be done, and also in the reporting back of the analysis results. Automated 3D-modeling techniques can greatly reduce the effort and skill required by the personnel involved and can improve the overall quality of the reporting.
Another critical communication area is the manufacturing-to-design linkage that can be created with 3D information. The increased complexity of devices today requires that designers and integration engineers are more aware of their process. Process-aware product design (PAPD) can help ensure that all parts of the manufacturing flow are within the process window and increase understanding of potential problem areas. With the aid of 3D representations, issues in design can be intercepted early before there is any impact on yield. If changes to the process need to be tested, 3D models are being used for DOE to assist with the debug operations.
If done correctly, the 3D voxel representation opens the door to atomistic process modeling. Many physical and chemical models have been developed, but simulations are limited to a very small area of designs because of memory limitations. The 3D voxel technique has proved able to store and retrieve huge amounts of data for such modeling techniques and is therefore very suitable as a 3D-modeling platform.
In all of these areas it is important to convey ideas, visions, intentions, and results. This is true for all semiconductor market segments, whether microprocessor, logic, memory, analog, and related markets such as data storage and MEMS. A high-quality, easy-to-use, customizable silicon accurate 3D representation can help semiconductor manufacturers enhance yields, shorten time to ramp, and reduce development cost by providing measurable cost savings in design and analysis.