Automotive semiconductor products demand high reliability. The current process of performing electrical test after fab-out may not be sufficient for efficient reliability management. This paper proposes an AI solution for improving the reliability of automotive semiconductor products. The solution includes two unique concepts: fab-data augmentation (FDA) to estimate missing values using partially available measurement data during the fabrication process and real-time prediction of reliability using machine learning (ML) models. The ML model is also used to identify and rank critical process steps that impact reliability, and to predict the reliability of wafers in real time. This allows low reliability wafers to be screened out early during the chip fabrication process, improving the overall reliability of the final product.
Continuous scaling of CMOS process technology to 7nm (and below) has introduced new constraints and challenges in determining Design-for-Yield (DFY) solutions. In this work, traditional solutions such as improvements in redundancy and in compensating target designs for low process window margins are extended to meet the additional constraints of complex 7nm design rules. Experiments conducted on 7nm industrial designs demonstrate that the proposed solution achieves 9.1%-41% redundant-via-rate improvements while ensuring all 7nm design rule constraints are met.
As the typical litho hotspot detection runtime continue to increase with sub-10nm technology node due to increasing design and process complexity, many DFM techniques are exploring new methods that can expedite some of their advanced verification processes. The benefit of improved runtimes through simulation can be obtained by reducing the amount of data being sent to simulation. By inserting a pattern matching operation, a system can be designed such that it only simulates in the vicinity of topologies that somewhat resemble hotspots while ignoring all other data. Pattern Matching improved overall runtime significantly. However, pattern matching techniques require a library of accumulated known litho hotspots in allowed accuracy rate. In this paper, we present a fast and accurate litho hotspot detection methodology using specialized machine learning. We built a deep neural network with training from real hotspot candidates. Experimental results demonstrate Machine Learning’s ability to predict hotspots and achieve greater than 90% detection accuracy and coverage, with best achieved accuracy 99.9% while reducing overall runtime compared to full litho simulation.
Achieving lithographic printability at advanced nodes (14nm and beyond) can impose significant restrictions on physical design, including large numbers of complex design rule checks (DRC) and compute-intensive detailed process model checking. Early identifying of yield-limiter hotspots is essential for both foundries and designers to significantly improve process maturity. A real challenge is to scan the design space to identify hotspots, and decide the proper course of action regarding each hotspot. Building a scored pattern library with real candidates for hotspots for both foundries and designers is of great value. Foundries are looking for the most used patterns to optimize their technology for and identify patterns that should be forbidden, while designers are looking for the patterns that are sensitive to their neighboring context to perform lithographic simulation with their context to decide if they are hotspots or not.[1] In this paper we propose a framework to data mine designs to obtain set of representative patterns of each design, our aim is to sample the designs at locations that can be potential yield limiting. Though our aim is to keep the total number of patterns as small as possible to limit the complexity, still the designer is free to generate layouts results in several million of patterns that define the whole design space. In order to handle the large number of patterns that represent the design building block constructs, we need to prioritize the patterns according to their importance. The proposed pattern classification methodology depends on giving scores to each pattern according to the severity of hotspots they cause, the probability of their presence in the design and the likelihood of causing a hotspot. The paper also shows how the scoring scheme helps foundries to optimize their master pattern libraries and priorities their efforts in 14nm technology and beyond. Moreover, the paper demonstrates how the hotspot scoring helps in improving the runtime of lithographic simulation verification by identifying which patterns need to be optimized to correctly describe candidate hotspots, so that only potential problematic patterns are simulated.
KEYWORDS: Design for manufacturing, Silicon, Polishing, Optical proximity correction, Back end of line, Metals, Failure analysis, Yield improvement, Logic, Manufacturing
A set of design for manufacturing (DFM) techniques have been developed and applied to 45nm, 32nm and 28nm logic
process technologies. A noble technology combined a number of potential confliction of DFM techniques into a
comprehensive solution. These techniques work in three phases for design optimization and one phase for silicon
diagnostics. In the DFM prevention phase, foundation IP such as standard cells, IO, and memory and P&R tech file are
optimized. In the DFM solution phase, which happens during ECO step, auto fixing of process weak patterns and
advanced RC extraction are performed. In the DFM polishing phase, post-layout tuning is done to improve
manufacturability. DFM analysis enables prioritization of random and systematic failures. The DFM technique presented
in this paper has been silicon-proven with three successful tape-outs in Samsung 32nm processes; about 5%
improvement in yield was achieved without any notable side effects. Visual inspection of silicon also confirmed the
positive effect of the DFM techniques.
As patterning for advanced processes becomes more challenging, designs must become more process-aware. The
conventional approach of running lithography simulation on designs to detect process hotspots is prohibitive in terms of
runtime for designers, and also requires the release of highly confidential process information. Therefore, a more
practical approach is required to make the In-Design process-aware methodology more affordable in terms of
maintenance, confidentiality, and runtime. In this study, a pattern-based approach is chosen for Process Hotspot Repair
(PHR) because it accurately captures the manufacturability challenges without releasing sensitive process information.
Moreover, the pattern-based approach is fast and well integrated in the design flow. Further, this type of approach is very
easy to maintain and extend. Once a new process weak pattern has been discovered (caused by Chemical Mechanical
Polishing (CMP), etch, lithography, and other process steps), the pattern library can be quickly and easily updated and
released to check and fix subsequent designs.
This paper presents the pattern matching flow and discusses its advantages. It explains how a pattern library is created
from the process weak patterns found on silicon wafers. The paper also discusses the PHR flow that fixes process
hotspots in a design, specifically through the use of pattern matching and routing repair.
KEYWORDS: Semiconducting wafers, Metals, Process modeling, Image processing, Design for manufacturing, Chemical mechanical planarization, Optical proximity correction, 3D modeling, Photovoltaics, Scanning electron microscopy
As a result, low fidelity patterns due to process variations can be detected and eventually corrected by designers as early
in the tape out flow as right after design rule checking (DRC); a step no longer capable to totally account for process
constraints anymore. This flow has proven to provide a more adequate level of accuracy when correlating systematic
defects as seen on wafer with those identified through LFD simulations. However, at the 32nm and below, still distorted
patterns caused by process variation are unavoidable. And, given the current state of the defect inspection metrology
tools, these pattern failures are becoming more challenging to detect. In the framework of this paper, a methodology of
advanced process window simulations with awareness of chip topology is presented. This method identifies the expected
focal range different areas within a design would encounter due to different topology.
KEYWORDS: Chemical mechanical planarization, Metals, Copper, System on a chip, Design for manufacturing, Manufacturing, Dielectrics, Polishing, Device simulation, Logic
Traditionally model based CMP check and hotspot detection are only done at the top level of the design because full
chip assembly is required to capture CMP long range effect. When manufacturing hotspots are found just before tape out
and layout modification is required, this can disrupt the overall schedule by repeating the verification steps with the
changed layout. Hence getting feedback at early design stage is critical to ensure that the design is correct by
construction. In this paper, we present a model-based CMP-DFM methodology which is used at early design phases to
avoid CMP related manufacturing failures. An accurate CMP model has been developed and used to predict surface
topographies for 32nm designs as well as physical hotspots caused by dishing, erosion, and depth of focus. We
demonstrate how to apply a characterized 32nm CMP physical model to run block level simulation with little or no
context information. The block level simulation methodology allows designers to check block robustness against any
possible surrounding environments in which the block may be placed. This approach can be taken for corner case
analysis in CMP-aware RC extraction.
In today's semiconductor industry, prior to wafer fabrication, it has become a desirable practice to scan layout designs
for lithography-induced defects using advanced process window simulations in conjunction with corresponding
manufacturing checks. This methodology has been proven to provide the highest level of accuracy when correlating
systematic defects found on the wafer with those identified through simulation. To date, when directly applying this
methodology at the full chip level, there has been unfavorable expenses incurred that are associated with simulation
which are currently overshadowing its primary benefit of accuracy - namely, long runtimes and the requirement for an
abundance of cpus. Considering the aforementioned, the industry has begun to lean towards a more practical application
for hotspot identification that revolves around topological pattern recognition in an attempt to sidestep the simulation
runtime. This solution can be much less costly when weighing against the negative runtime overhead of simulation. The
apparent benefits of pattern matching are, however, counterbalanced with a fundamental concern regarding detection
accuracy; topological pattern identification can only detect polygonal configurations, or some derivative of a
configuration, which have been previously identified. It is evident that both systems have their strengths and their
weaknesses, and that one system's strength is the other's weakness, and vice-versa.
A novel hotspot detection methodology that utilizes pattern matching combined with lithographic simulation will be
introduced. This system will attempt to minimize the negative aspects of both pattern matching and simulation. The
proposed methodology has a high potential to decrease the amount of processing time spent during simulation, to relax
the high cpu count requirement, and to maximize pattern matching accuracy by incorporating a multi-staged pattern
matching flow prior to performing simulation on a reduced data set. Also brought forth will be an original methodology
for constructing the core pattern set, or candidate hotspot library, in conjunction with establishing hotspot and coldspot
pattern libraries. Lastly, it will be conveyed how this system can automatically improve its potential as more designs are
passed through it.
Automatic layout optimization is becoming an important component of the DfM work flow, as the number of
recommended rules and the increasing complexity of trade-offs between them makes manual optimization increasingly
difficult and time-consuming. Automation is rapidly becoming the best consistent way to get quantifiable DfM
improvements, with their inherent yield and performance benefits for standard cells and memory blocks. Takumi autofixer
optimization of Common Platform layouts resulted in improved parametric tolerance and improved DfM metrics,
while the cell architecture (size and routability) and the electrical characteristics (speed/power) of the layouts remained
intact. Optimization was performed on both GDS-style layouts for standard cells, and on CDBA (Cadence Data Base
Architecture)-style layout for memory blocks. This paper will show how trade-offs between various DfM requirements
(CAA, recommended rules, and litho) were implemented, and how optimization for memories generated by a compiler
was accomplished. Results from this optimization work were verified on 45nm design by model and rule based DfM
checking and by wafer yields.
The advanced process technologies have well known yield loss due to the degradation of pattern fidelity. The
process to compensate for this problem is advanced resolution enhancement techniques (RET) and optical proximity
correction (OPC). By design, the creation of RET/OPC recipes and the calibration of process models are done very early
in the process development cycle with data that are not made of real designs since they are not yet available, but made of
test structures that represent different sizes, distances and topologies. The process of improving the RET/OPC recipes
and models is long and tedious, it is usually a key contributor to quick production ramp-up. It is very coverage limited
by design. The authors will present a proposed system that, by design, is dynamic, and allows the RET/OPC production
system to reach maturity faster through a detailed collection of hotspots identified at the design stage. The goal is to
reduce the lapse of time required to get mature production RET/OPC recipes and models.
This paper describes the feasibility of lOOnm-node lithography using ArF lithography and att-PSM (aUenuated Phase Shift Mask). In the simulation approach, we can find that att-PSM can improve EL window more than 25%compared to BIM (Binary Intensity Mask) in both KrF and ArF lithography. Although the MEF (Mask Error Factor) values of att-PSM and BIM are almost same even in a higher NA region, the total CD variation of aU-PSM is slightly lower than that of BIM because of the increase effect of EL window. Considering the total CD variation, it is necessary to use the ArF lithography machine with higher NA of more than O.7ONA for lOOnm patterning. In the real patterning performance, the ArF lithography and att-PSM can improve EL windows more than 60% in comparison with KrF lithography and att-PSM for sub-l2Onm cell patterns. The case of att-PSM and annular aperture condition, especially small ring width annular condition shows the increasing effect ofprocess windows compared to BIM for lOOnm L/S patterns. For the direct C/H printing below l2Onm feature, we can get about 9% EL window in the case of l2Onm C/H feature. Although we have some technical issues for lOOnm lithography such as the controllability of MEF and EL window extension, the lens quality enhancement for the higher NA and manufacturing defects of att-PSM, etc., there is a sufficient feasibility to obtain lOOnm-node pattern with ArF lithography and att-PSM.
ArF lithography that is expected the candidate for next generation optical lithography and attenuated Phase Shift Mask (att-PSM) will be adapted for 0.12micrometers design-rule and beyond. For the next-generation lithography, the most important requirement for mask process is enough resolution and good pattern fidelity to generate various critical patterns, of which sizes are below 0.5micrometers main pattern including OPC patterns. In this paper we describe in terms of blank mask properties, mask making process and wafer performance of ArF attenuated Phase Shift Mask (att-OSM) using TiN/Si3N4(abbreviated as TiN/SiN) multi-layer for Next Generation Lithography (NGL). In view point of material, we have evaluated for the applicability of TiN/SiN multi-layer to ArF lithography as compared with non- stoichiometric MoSiON-based single-layer structure. In mask making process, we used Chemically Amplified Resist (CAR) process characteristics and Dry etching system for improvement of enough resolution and pattern fidelity. Also we have investigated wafer performance for ArF att-PSM in terms of process windows as compared with BIM (Binary Intensity Mask) in 120nm D/R real cell pattern and 100nm L/S(Line and Space)D/R pattern, respectively.
The patterning potentialities of sub-100nm pattern for ArF lithography was evaluated with conventional alternating PSM (alt-PSM) for dense lines and spaces (L/S) and phase edge PSM (PE-PSM) for isolated lines of memory device. In dense L/S pattern,110nm pattern was defined with relatively small depth of focus(DOF) window(~ 0.2 ?m) due to phase error of mask. As pattern sizes was changed from 130nm to 200nm, critical dimension (CD) difference between two neighboring spaces was varied and it was assumed that micro loading effect was occurred in Qz etching. The linearity was guaranteed to dense L/S of 110nm and isolated line of 90nm, and Iso-Dense bias was controlled within 15nm. The 60nm and 70nm isolated lines of PE-PSM ware defined with good process windows in the case of OA_X size(X-direction size of Cr open area) of 0.5 ?m. The 55nm isolated line was also defined. The pattern shift of isolated lines was occurred with 4~7nm as phase of mask was varies within 190 ~ 200 ° . Though the alt-PSM with high numerical aperture (NA) for ArF lithography was strong candidates for sub-1 OOnm lithography of memory device, the issues of mask fabrication such as tighter phase control and minimizing etch loading effect would be big obstacles. On the contrary, there were many possibilities of sub-100nm patterning in PE-PSM with good process windows, however tighter control of pattern shift due to phase error must be studied intensively.
Optical lithography at resolution limit is a non-linear pattern transfer. One of the important issue is a mask critical dimension control because of nonlinear amplification of mask critical dimension error during image transferring on wafer. This amplification of mask error is called the MEF. This mask error factor has been widely used as an important parameter for indicating tighter CD control for the photomask for low-kl lithography generation.
The defect control of the attenuated-PSM is compared with that of the conventional binary intensity mask (BIM), because the fabrication process for the att-PSM tends to generate more defects than that of the BIM. To repair a defective att. PSM, a similar method used for BIM has been applied. However, this process may cause degradation of pattern fidelity with the repair pattern on the mask are transferred on the wafer, if the transmission and phase of repaired area are not well controlled. In this paper, we have investigated the effect of repairing process on the pattern fidelity to define contact holes using a KrF lithography with an att. PSM. The defects in the various distances form contact hole patterns and of various sizes were repaired. The experimental printability and simulation data from an aerial image model were compared for repaired defects. And the repair tool reliability and the simulation accuracy of the att. PSM was examined using CD-SEM. From the experimental results, repaired defects having larger size than the threshold. One or within a certain range from the pattern induced the pattern deformation. Therefore, the size of defect and the distance between the pattern and defect should be considered in repair process for the Att. PSM fabrication. Based on the experimental and simulation results, the requirements for the repair tool will be proposed.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.