3D sequential integration is a way to stack active transistor layers with closely spaced vertical connections, increasing the number of devices that can be integrated within a given footprint. For semiconductor manufacturers, the promise comes with a difficult process constraint: fabricating the upper tier must not degrade the transistors or interconnects already built below it. Research reported by CEA-Leti in 2018 explored how process engineering, material transfer and reliability testing can address that challenge.
Why 3D sequential integration demands a different process window
In conventional fabrication, transistor processing can involve high temperatures. In a sequential stack, however, the lower tier is already in place when the upper tier is made. Excessive heat can affect the bottom-level devices and the intermediate back-end-of-line (BEOL) interconnect structure. The upper-tier process must therefore balance transistor performance with the thermal limits of the completed layers underneath.
CEA-Leti’s CoolCube approach investigates monolithic 3D integration, in which active device layers are fabricated sequentially rather than assembled as separate chips. The work described a 500°C thermal-budget target for top-tier processing in one process-development program. This makes several fabrication steps more demanding, including gate formation, source-and-drain engineering, surface preparation and bonding.
Lower temperatures can create trade-offs. For example, they may increase access resistance or make it harder to achieve the desired channel properties. They can also affect gate-stack quality, where microscopic defects may contribute to charge trapping. Consequently, a successful process needs to be assessed not only for whether it forms a transistor, but also for electrical performance, reliability and compatibility with the already-processed wafer.
Process advances address upper-tier integration barriers
The reported work examined several process steps identified as potential obstacles to manufacturable 3D sequential integration. These included obtaining low-resistance polysilicon gates for top-tier transistors and developing selective silicon epitaxy for source-and-drain regions at low temperature. The research also evaluated the stability of intermediate BEOL structures, including copper and ultra-low-k (ULK) interconnects, during subsequent processing.
Protecting front-end-of-line (FEOL) equipment from contamination is another practical requirement when wafers that contain BEOL materials return for upper-tier device fabrication. The work described a contamination-control sequence involving bevel etching, decontamination and encapsulation. Such process controls are important because integration schemes must fit the operating requirements of fabrication environments, not just produce functional devices in isolation.
Researchers also demonstrated Smart Cut layer transfer above a processed CMOS wafer. This offers an alternative to a scheme based on bonding and etch-back for creating the upper channel layer. Layer-transfer choices matter because they influence how a crystalline silicon layer is introduced while protecting structures already present on the wafer.
Low-temperature epitaxy and surface preparation
Selective epitaxy can build up transistor source-and-drain regions while limiting unwanted growth on surrounding dielectric materials. At reduced temperatures, both the wafer surface condition and the growth chemistry become especially important. Surface preparation aims to remove or control oxide and contamination before epitaxial growth, while maintaining a clean interface as the wafer moves through the process.
The studies compared preparation approaches and investigated silicon and silicon-germanium growth at temperatures in the 500°C to 575°C range. One reported result was that an in-situ surface treatment reduced the hydrogen-bake temperature needed to eliminate an interfacial oxygen peak on silicon. The work also found that a liquid silicon precursor could provide higher growth rates than disilane at a given germanium concentration in the conditions studied. These findings inform process development; they do not remove the need to validate selectivity, material quality and integration compatibility for each specific flow.
Reliability testing puts performance in context
Transistor performance alone is not enough to establish whether a stacked process is viable. The report also described a reliability study of a fully integrated 3D sequential technology processed at 630°C. In that study, researchers examined the effects of upper-tier fabrication on bottom-tier devices, as well as the reliability of the upper-tier transistors themselves.
The reported measurements showed no visible difference in bias-temperature-instability (BTI) reliability between bottom-level transistors exposed to the upper-tier process and planar reference devices. The upper-tier devices were also reported to meet the study’s stated BTI lifetime requirement for digital CMOS applications. A 3D inverter combining a bottom-level NMOS and a top-level PMOS served as an example of a two-tier logic structure.
The 630°C study represents a separate process evaluation from the work targeting a 500°C thermal budget. Taken together, the results illustrate why temperature limits should be considered in relation to the specific process sequence and qualification criteria. They also highlight the value of testing the complete integration flow rather than drawing conclusions from an individual fabrication step.
Implications for semiconductor process development
For process engineers and technology teams, the practical lesson is that 3D sequential integration is a coordinated integration problem. Device formation, wafer handling, surface chemistry, BEOL stability, contamination controls and reliability qualification all affect whether a proposed stack can advance beyond a demonstration.
- Evaluate the full thermal history: assess how each upper-tier step affects lower-tier transistors and interconnects.
- Qualify interfaces and materials: check surface preparation, epitaxial quality, gate-stack behavior and bonding stability.
- Include contamination controls: consider the requirements of returning BEOL-processed wafers to FEOL processing.
- Measure reliability as well as performance: use tests such as BTI evaluation to understand device behavior after integration.
These considerations are relevant to semiconductor manufacturing teams exploring higher-density architectures and new ways to combine device layers. They also reinforce that scaling depends not only on transistor dimensions, but on the ability to control interactions across the entire process flow.
Conclusion
Progress in 3D sequential integration depends on making upper-tier fabrication compatible with the electrical and structural integrity of the lower tiers. Low-temperature gate and epitaxy processes, contamination management, layer-transfer techniques and device-level reliability testing each address part of that challenge. The reported advances provide process-development pathways, while underscoring the need for integrated validation before a stacked technology can meet its intended manufacturing and application requirements.