Interconnecting transistors and other components in the IC, in the package, on the printed circuit board and at the system and global network level, are where the future limitations in performance, power, latency and cost reside.
BY BILL CHEN, ASE US, Sunnyvale, CA; BILL BOTTOMS, 3MT Solutions, Santa Clara, CA, DAVE ARMSTRONG, Advantest, Fort Collins, CO; and ATSUNOBU ISOBAYASHI, Toshiba Kangawa, Japan.
Heterogeneous Integration refers to the integration of separately manufactured components into a higher level assembly that in the aggregate provides enhanced functionality and improved operating characteristics.
In this definition components should be taken to mean any unit whether individual die, MEMS device, passive component and assembled package or sub‐system that are integrated into a single package. The operating characteristics should also be taken in its broadest meaning including characteristics such as system level cost-of-ownership.
The mission of the ITRS Heterogeneous Integration Focus Team is to provide guidance to industry, academia and government to identify key technical challenges with sufficient lead time that they do not become roadblocks preventing the continued progress in electronics that is essential to the future growth of the industry and the realization of the promise of continued positive impact on mankind. The approach is to identify the require- ments for heterogeneous integration in the electronics industry through 2030, determine the difficult challenges that must be overcome to meet these requirements and, where possible, identify potential solutions.
Background
The environment is rapidly changing and will require revolutionary changes after 50 years where the change was largely evolutionary. The major factors driving the need for change are:
- We are approaching the end of Moore’s Law scaling.
- The emergence of 2.5D and 3D integration techniques
- The emerging world of Internet of Everything will cause explosive growth in the need for connectivity.
- Mobile devices such as smartphones and tablets are growing rapidly in number and in data communications requirements, driving explosive growth in capacity of the global communications network.
- Migration of data, logic and applications to the cloud drives demand for reduction in latency while accommodating this network capacity growth.
Satisfying these emerging demands cannot be accomplished with the current electronics technology and these demands are driving a new and different integration approach. The requirements for power, latency, bandwidth/bandwidth density and cost can only be accomplished by a revolutionary change in the global communications network, with all the components in that network and everything attached to it. Ensuring the reliability of this “future network” in an environment where transistors wear out, will also require innovation in how we design and test the network and its components.
The transistors ‘power consumption in today’s network account for less than 10 percent of total power, total latency and total cost. It is the interconnection of these transistors and other components in the IC, in the package, on the printed circuit board and at the system and global network level, where the future limitations in performance, power, latency and cost reside. Overcoming these limitations will require heterogeneous integration of different materials, different devices (logic, memory, sensors, RF, analog, etc.) and different technologies (electronics, photonics, plasmonics, MEMS and sensors). New materials, manufacturing equipment and processes will be required to accomplish this integration and overcome these limitations.
Difficult challenges
The top‐level difficult challenges will be the reduction of power per function, cost per function and latency while continuing the improvements in performance, physical density and reliability. Historically, scaling of transistors has been the primary contributor to meeting required system level improvements. Heterogeneous integration must provide solutions to the non‐transistor infrastructure that replace the shortfall from the historical pace of progress we have enjoyed from scaling CMOS. Packaging and test have found it difficult to scale their performance or cost per function to keep pace with transistors and many difficult challenges must be met to maintain the historical pace of progress.
In order to identify the difficult challenges we have selected seven application areas that will drive critical future requirements to focus our work. These areas are:
- Mobile products
- Big data systems and interconnect
- The cloud
- Biomedical products
- Green technology
- Internet of Things
- Automotive components and systems
An initial list of difficult challenges for Heterogeneous Integration for these application areas is presented in three categories; (1) on‐chip interconnect, (2) assembly and packaging and (3) test. These are analyzed in line with the roadmapping process and will be used to define the top 10 challenges that have the potential to be “show stoppers” for the seven application areas identified above.
On-chip interconnect difficult challenges
The continued decrease in feature size, increase in transistor count and expansion into 3D structures are presenting many difficult challenges. While challenges in continuous scaling are discussed in the “More Moore” section, the difficult challenges of interconnect technology in devices with 3D structures are listed here. Note that this assumes a 3D structure with TSV, optical interconnects and passive devices in interposer substrates.
ESD (Electrostatic Discharge): Plasma damage on transistors by TSV etching especially on via last scheme. Low damage TSV etch process and the layout of protection diodes are the key factors.
CPI (Chip Package Interaction) Reliability [Process]: Low fracture toughness of ULK (Ultra Low‐k) dielectrics cause failures such as delamination. Material development of ULK with higher modulus and hardness are the key factors.
CPI (Chip Package Interaction) Reliability [Design]: A layout optimization is a key for the device using Cu/ULK structure.
Stress management in TSV [Via Last]: Yield and reliability in Mx layers where TSV land is a concern.
Stress management in TSV [Via Middle]: Stress deformation by copper extrusion in TSV and a KOZ (Keep Out Zone) in transistor layout are the issues.
Thermal management [Hot Spot]: Heat dissipation in TSV is an issue. An effective homogenization of hot spot heat either by material or layout optimization are the key factors.
Thermal management [Warpage]: Thermal expansion management of each interconnect layer is necessary in thinner Si substrate with TSV.
Passive Device Integration [Performance]: Higher Q, in other words, thicker metal lines and lower tan dielectrics is a key for achieving lower power and lower noise circuits.
Passive Device Integration [Cost]: Higher film and higher are required for higher density and lower footprint layout.
Implementation of Optical Interconnects: Optical interconnects for signaling, clock distribution, and I/O requires development of a number of optical components such as light sources, photo detectors, modulators, filters and waveguides. On‐chip optical interconnects replacing global inter- connects requires the breakthrough to overcome the cost issue.
Assembly and packaging difficult challenges
Today assembly and packaging are often the limiting factors in performance, size, latency, power and cost. Although much progress has been made with the introduction of new packaging architectures and processes, with innovations in wafer level packaging and system in package for example, a significantly higher rate of progress is required. The complexity of the challenge is increasing due to unique demands of heterogeneous integration. This includes integration of diverse materials and diverse circuit fabric types into a single SiP architecture and the use of the 3rd dimension.
Difficult packaging challenges by circuit fabric
- Logic: Unpredictable hot spot locations, high thermal density, high frequency, unpredictable work load, limited by data bandwidth and data bottle‐necks. High bandwidth data access will require new solutions to physical density of bandwidth.
- Memory: Thermal density depends on memory type and thermal density differences drive changes in package architecture and materials, thinned device fault models, test & redundancy repair techniques. Packaging must support low latency, high bandwidth large (>1Tb) memory in a hierar- chical architecture in a single package and/or SiP).
- MEMS: There is a virtually unlimited set of requirements. Issues to be addressed include hermetic vs. non‐hermetic, variable functional density, plumbing, stress control, and cost effective test solutions.
- Photonics: Extreme sensitivity to thermal changes, O to E and E to O, optical signal connections, new materials, new assembly techniques, new alignment and test techniques.
- Plasmonics: Requirements are yet to be determined, but they will be different from other circuit type. Issues to be addressed include acousto‐ magneto effects and nonlinear plasmonics.
- Microfluidics: Sealing, thermal management and flow control must be incorporated into the package.
Most if not all of these will require new materials and new equipment for assembly and test to meet the 15 year Roadmap requirements.
Difficult packaging challenges by material
Semiconductors: Today the vast majority of semiconductor components are silicon based. In the future both organic and compound semiconductors will be used with a variety of thermal, mechanical and electrical properties; each with unique mechanical, thermal and electrical requirements.
Conductors: Cu has replaced Au and Al in many applications but this is not good enough for future needs. Metal matrix composites and ballistic conductors will be required. Inserting some of these new materials will require new assembly, contacting and joining techniques.
Dielectrics: New high k dielectrics and low k dielectrics will be required. Fracture toughness and interfacial adhesion will be the key parameters. Packaging must provide protection for these fragile materials.
Molding compound: Improved thermal conductivity, thinner layers and lower CTE are key requirements.
Adhesives: Die attach materials, flexible conductors, residue free materials needed o not exist today.
Biocompatible materials: For applications in the healthcare and medical domain (e.g. body patches, implants, smart catheters, electroceuticals), semiconductor‐based devices have to be biocompatible. This involves the integration of new (flexible) materials to comply with specific packaging (form factor) requirements.
Difficult challenges for the testing of heterogeneous devices
The difficulties in testing heterogeneous devices can be broadly separated into three categories: Test Quality Assurance, Test Infrastructure, and Test Design Collaboration.
Test quality assurance needs to comprehend and place achievable quality and reliability metrics for each individual component prior to integration, in order to meet the heterogeneous system quality and reliability targets. Assembly and test flows will become inter- twined and interdependent. They need to be constructed in a manner that maintains a cost effective yield loss versus component cost balance and proper component fault isolation and quantification. The industry will be required to integrate components that cannot guarantee KGD without insurmountable cost penalties and this will require integrator visible and accessible repair mechanisms.
Test infrastructure hardware needs to comprehend multiple configurations of the same device to enable test point insertion at partially assembled and fully assembled states. This includes but is not limited to different component heights, asymmetric component locations, and exposed metal contacts (including ESD challenges). Test infrastructure software needs to enable storing and using volume test data for multiple components that may or may not have been generated within the final integrators data domains but are critical for the final heterogeneous system functionality and quality. It also needs to enable methods for highly granular component tracking for subsequent joint supplier and integrator failure analysis and debug.
Test design collaboration is one of the biggest challenges that the industry will need to overcome. It will be a requirement for heterogeneous highly integrated highly functional systems to have test features co‐designed across component boundaries that have more test coverage and debug capability than simple boundary scans. The challenge
of breaking up what was once the responsibility of a wholly contained design for test team across multiple independent entities each trying to protect IP, is only magnified by the additional requirement that the jointly developed test solutions will need to be standardized across multiple competing heterogeneous integrators. Industry wide collaboration on and adherence to test standards will be required in order to maintain cost and time effective design cycles for highly desired components that traditionally has only been required for cross component boundary communication protocols.
The roadmapping process
The objective of ITRS 2.0 for heterogeneous integration is to focus on a limited number of key challenges (10) that have the greatest potential to be “show stoppers,” while leaving other challenges identified and listed but without focus on detailed technical challenges and potential solutions. In this process collaboration with other Focus Teams and Technical Working Groups will be a critical resource. While we will need collaboration with other groups both inside and outside the ITRS some of the collaborations are critical for HI to address its mission. FIGURE 1 shows the major internal collaborations in three categories.
FIGURE 1. Collaboration priorities.
We expect to review these key challenges and our list of other challenges on a yearly basis and make changes so that our focus keeps up with changes in the key challenges. This will ensure that our efforts remain focused on the pre‐competitive technologies that have the greatest future value to our audience. There are four phases in the process detailed below.
1. Identify challenges for application areas: The process would involve collaboration with other focus teams, technical TWGs and other roadmapping groups casting a wide net to identify all gaps and challenges associated with the seven selected application areas as modified from time to time. This list of challenges will be large (perhaps hundreds) and they will be scored by the HI team by difficulty and criticality.
2. Define potential solutions: Using the scoring in phase (1) a number (30‐40) will be selected to identify potential solutions. The remainder will be archived for the next cycle of this process. This work will be coordinated with the same collabo- ration process defined above. These potential solutions will be scored by probable success and cost.
3. Down select to only the 10 most critical challenges: The potential solutions with the lowest probability of success and highest cost will have the potential to be “show stopping” roadblocks. These will be selected using the scoring above and the focus issues for the HI roadmap. The results of this selection process will be commu- nicated to the relevant collaboration partners for their comments.
4. Develop a roadmap of potential solutions for “show stoppers”: The roadmap developed for the “show stopping” roadblocks shall include analysis of the blocking issue and identification of a number of potential solutions. The collaboration shall include detail work with other units of the ITRS, other roadmapping activity such as the Jisso Roadmap, iNEMI Roadmap, Communications Technology Roadmap from MIT. We are continuing to work with the global technical community: industry, research institutes and academia, including the IEEE CPMT Society.
The blocking issues will be specifically investigated by the leading experts within the ITRS structure, academia, industry, government and research organizations to ensure a broad based understanding. Potential solutions will be identified through a similar collaboration process and evaluated through a series of focused workshops similar to the process used by the ERD iTWG. This process is a workshop where there is one proponents and one critic presenting to the group. This is followed by a discussion and a voting process which may have several iterations to reach a consensus.
The cross Focus Team/TWG collaboration will use a procedure of iteration to converge on an understanding of the challenges and potential solutions that is self‐ consistent across the ITRS structure. An example is illustrated in FIGURE 2.
FIGURE 2. Iterative collaboration process
It is critically important that our time horizon include the full 15 years of the ITRS. The work to anticipate the true roadblocks for heterogeneous integration, define potential solutions and implement a successful solution may require the full 15 years. Among the tables we will include 5 year check points of the major challenges for the key issues of cost, power, latency and bandwidth. In order for this table to be useful we will face the challenge of identifying the specific metric or metrics to be used for each application driver as we prepare the Heterogeneous Integration roadmap chapter for 2015 and beyond.
BILL CHEN is a senior technical advisor for ASE US, Sunnyvale, CA; BILL BOTTOMS is President and CEO of 3MT Solutions, Santa Clara, CA, DAVE ARMSTRONG is director of business development at Advantest, Fort Collins, CO; and ATSUNOBU ISOBAYASHI works in the Toshiba’s Center for Semiconductor Research & Development, Kangawa, Japan.
