FA4.0

Failure Analysis 4.0 – Key  for reliable electronic devices in smart mobility and industrial production [FA4.0]

FA4.0 introduced digitalization into innovative failure analysis (FA) methods and equipment to ensure that increasingly complex electronic systems operate with high reliability and safety in daily use.

In FA4.0, 4 semiconductor and system suppliers, 11 equipment suppliers supported by 4 RTOs combined their strength to develop innovative FA methods and equipment to  tackle  challenges of  heterogeneous  integration and miniaturization  in microelectronics.  For   the   first  time new  methodologies based on  artificial  intelligence (AI), machine learning (ML) and  digital twin technologies were  applied to improve the  FA  equipment and  methodology landscape  and  provide highly efficient workflows for failure root cause analysis

Achievements and results of the project

The results of  FA4.0 demonstrated impressive improvements in  AI enhanced failure analysis.  It  was demonstrated that  enabling and  implementation of  AI methodologies and   advanced data analysis  including digital twin  technologies have  deep  impact on the efficiency of FA.  Advanced imaging and signal analysis algorithms including neuronal network algorithms and correlative data analysis were developed to push the performance of  FA  tools (see   Figures  1a and b). The  ML based image analysis  (Figure 1b)  achieved a  5  times faster analysis and more than  90% accuracy improvement compared to existing solutions. Several FA equipment have been enhanced by improving hardware components and novel  AI algorithms to achieve higher sensitivities and resolution. Furthermore, standardized data-exchange  and  an  innovative universal  sample holder were implemented to combine  equipment  to highly efficient and more automated workflows.

Background, objectives of the project and challenges

Miniaturization and heterogeneous integration of more functionality in smaller volume drive the development of more  and  more compact and complicated microelectronic systems.  Semiconductor and system companies have an increasing effort to achieve reliability and quality. Effort and cost for FA increase typically by 3-5% per year.  Thus, new methods and tools to improve performance, efficiency and speed are required to tackle this challenge.

Objective of the FA4.0 project was the development of innovative FA   equipment  and   methods in  specific applying new  capabilities of  computer and data science. This includes AI, ML and  digital twin technologies in combination with  automation of failure analysis tools and workflows.

The integration of newly developed AI based algorithms for image and signal analysis allowed the enhancement of tool performance. Failure analysis methodologies, which  used  the   developed universal sample  holder,  allowed  improvement of  through-put by  interlinking  equipment from   different  vendors  for efficient comprehensive workflows.

The  results of  FA4.0 demonstrated that  it  is possible to  establish  an   infrastructure  in  order   to  exchange meta-data and  other  information between equipment of different tool  suppliers. This is a basic requirement for   improving  the   FA  methodology and efficiency.  The   outstanding  important step  achieved in FA4.0 was  the  introduction of integrated workflows (combined equipment), including  a  JSON (Java Script Object Notation) meta-data header and  a universal sample  holder. A  standardisation  initiative has   been started under the umbrella of SEMI.

Technological achievements

The main  technological achievements of the FA4.0 project are in respect to

• Computer and data analysis:

Extended image  and   signal  analysis  techniques  to include ML based approaches by  novel  AI methods for  data handling (see  example Figure 3),  pre-processing,  labelling  and   training  and   implementation into    tool    platforms   for    automated   void    detection (Gimic), automated delamination detection (PVA  TePla), automated TEM  lamella preparation (TESCAN/ORSAY). Impressive enhancement of tool performance was demonstrated e.g.  for  SAM  or AI image data analysis of intermetallic compounds gold  bond  inspection with  a 60x time reduction.

• Improved FA equipment:

— New X-Ray platform (see  Figure 3) with novel X-Ray source (Excillum) improving the imaging resolution to 150 nm (3x improved), new detector concepts by Direct Conversion.

— New tool for thermal surface warpage measurements by CyberTECHNOLOGIES for wider T-range up to 300°C (see Figure 5).

— Enhanced FIB-laser preparation tools and workflows by ORSAY/TESCAN to reduce processing time  (> 50 % time  saving).

— Enhancement of scanning acoustic microscopy tool by PVA TePla with ML based automated defect detection and novel  SAFT (synthetic aperture focusing technique) data analysis, improved data acquisition with a 20-30x time  saving.

• Standardization

Standardized FA  tool interfaces by a universal sample holder and harmonized data interface (JSON-header) which now can be applied to combine different FA  equipment platforms to exchange meta-data and generate digital twins of samples (see Figure 4). In FA4.0 first applications of integrated workflows including tools of different vendors were demonstrated.

• Automated FA workflow

Several automated workflows for FA, enabling for the first time software-based defect navigation with higher accuracy and  repeatability and automated end point detection for preparation were demonstrated.

Market Potential

The  project  opens  new  market  potential  for semiconductor and  system suppliers, FA equipment suppliers, as well as software and design companies.

— Semiconductor/system suppliers are able  to design, develop, and enter  products faster into the market with better reliability and quality:

• Better access to European markets of high reliability and quality products like  autonomous driving, energy etc.

• Better customer satisfaction by faster reaction on field returns.

— Failure analysis (FA) equipment suppliers are able  to develop more efficient FA equipment (see  e.g. Figure 5 on thermal warpage measurement) for faster FA support.

— Software companies like  Matworks and Gimic get access to a new market both for equipment and processes.

Societal & Economic Impact

Societal impact  – new jobs

Three PhD and more than 10 master and bachelor theses were developed out of the  project. Additional new job opportunities  related to computer and data science arise due to the activities of the project (e.g. at Bosch, Infineon).

Environmental impact  – Strengthening reliability which saves material

AI  learning approaches  support higher reliability  and quality of products.

Economic impact  by faster and more efficient failure analysis and standardization

AI enhanced analysis tools enable automated preparation and measurement routines saving time and personal effort. The analysis performance and speed can dramatically be improved (e.g.  20-30x faster SAM defect inspection, 60x reduced time for defect screening at intermetallic contacts).

Standardized FA tool interfaces and data handling e.g. by standardized unified sample holder and   JSON  data header, enable increasing speed of FA workflows and data analysis, reducing failure probability and thus increasing the overall efficiency of the failure analysis process (saving of hours, even days depending on workflow, the real potential is still a topic under investigation).

Patents, Standardisation, Publications

An international standardization initiative on FA tool interfaces started for the unified sample holder and JSON data header under  the umbrella of SEMI. This standardization activity  is  supported by  the  international failure analysis conferences ISTFA and  CAM-Workshop. The  FA4.0  team organized special sessions at  ISTFA 2022   “Standardized and  AI Enhanced FA workflows” and  the CAM-Workshop 2022,  2023,  and  2024  to discuss AI and standardization topics.

A multitude of presentations and publications developed  out of the project including  a  best  poster  award during IPFA 2022  at Singapore and  best attendee paper at Singapore and  a best attendee paper at ISTFA 2023 in Phoenix.

Several  keynote  presentations  came out  of FA4.0: Among  them  were  the  FA conferences IPFA 2022  at Singapore (“FA4.0-Exploring  the   Digital  Future”), the CAM-Workshop     (“Progress  in Failure Analysis –  A  Key for Reliable High-quality European Electronics”), the NordPac Packaging Conferences (“Failure Analysis  4.0”)   as   well   as   the computer  science   conference  AIMSE 2023 (“Impact of AI on Failure Analysis in Microelectronics”).

Future Developments

By the project, a new development trend was generated bringing together data science, AI  and failure analysis experts. In summary “revolutionary” opportunities for failure analysis have been catalyzed in respect to automatization and  performance of  FA techniques and workflows.

FA4.0 is a  door  opener for  future R&D: For  example, the  research in FA4.0 demonstrated that in a next step we need to push knowledge in data management e.g. set-up an ontology in failure analysis (see  FA²IR project).

The  following new  projects, which  are under  preparation, came out of FA4.0:

— The FA²IR project with focuses on more digitalization and AI enhanced FA workflows already started meanwhile. This project also includes ideas to develop generalized AI models for different FA-related use-cases with data from different partners.— A large initiative in the Chips JU is under preparation that  includes a network of FA users, equipment suppliers and an academy.

— A large initiative in the Chips JU is under  preparation that  includes a network of FA users, equipment suppliers and academy.