Online information source for semiconductor professionals

IEEE IITC Conference takes on 3D interconnects

26 March 2010 | By Mark Osborne | News > Wafer Processing

Popular articles

New Product: Applied Materials new EUV reticle etch system provides nanometer-level accuracy - 19 September 2011

Oberai discusses Magma’s move into solar PV yield management space - 29 August 2008

‚??Velocity‚?? the new buzzword in Intel‚??s PQS annual awards - 12 April 2012

Applied Materials adds Jim Rogers to Board of Directors - 29 April 2008

New Product: ASML Brion‚??s Tachyon MB-SRAF enables OPC-like compute times - 19 September 2011

In what the IEEE is praising as the ‘one of the conference’s strongest technical programs in years,’ the 13th annual IEEE International Interconnect Technology Conference (IITC), being held in Burlingame, CA between June 7-9, is have a special focus on 3D circuit interconnections. Mark Bohr, Intel Senior Fellow and the company’s Director of Process Architecture and Integration, will give the keynote presentation addressing the recent shift in the goals for the semiconductor industry, from simply making ever-smaller transistors or chips with higher transistor counts, to integrating more and different systems into small packages for overall increased performance and power efficiency.

“We are extremely gratified with the response to our Call for Papers seeking high-quality work in all areas of the interconnect field, with a special focus on 3D architectures, memories and new materials. These will be presented in addition to papers on core interconnect topics, from the best and the brightest people in the field,” said Michael Armacost, IITC 2010 Publicity Chair and Managing Director at Applied Materials, Inc. “The IITC conference is a premier forum for specialists and suppliers globally to come together to discuss, demonstrate and exhibit advanced interconnect techniques, materials, architectures and processes that shed light on the best ways forward.”

IITC 2010 will feature numerous papers on 3D technology. Three especially interesting papers describe impressive progress in 3D interconnect reliability evaluation, materials and alignment technique. A fourth shows how 3D interconnects can enable the heterogeneous integration of a sensor array and a signal processing CMOS IC.

•    Unperturbed Reliability: “Vias” are vertical holes filled with metal, used to electrically and physically interconnect a chip’s various levels. Researchers are trying to perfect through-silicon vias (TSVs), which pass completely through a chip and connect it with other chips in the stack. But do TSVs embedded in ultra-thin substrates influence the operation and reliability of adjacent transistors? In a first, researchers from IMEC and KU Leuven explored the question. They used a 130-nm CMOS process technology to fabricate copper TSVs with 5.2-µm diameters and 22-µm lengths. They placed them 1.1-µm away from NMOS and PMOS transistors, singly and in groups, on either or both the source and drain sides of the transistors, and in differing layouts. No performance degradation was observed. Further, to determine whether stresses resulting from the thermal mismatch between copper TSVs and silicon substrates had any impact, they performed extensive thermal cycling. It turned out that while there was some residual tensile stress, it had a negligible impact on carrier mobility. (Paper #9.4, “Electrical Evaluation of 130-nm MOSFETs with TSV Proximity in 3D-SIC Structure,” Y. Yang et al, IMEC and KU Leuven)

•    Low-Temperature TSV Barrier: Without a barrier layer, copper from a via would diffuse into the silicon substrate and degrade transistors. Bonded 3D circuits compound the problem because the high processing temperatures needed to build TSV barrier layers can often damage the bond as well as the transistors. Fujitsu researchers experimented with a relatively low-temperature plasma-enhanced chemical vapor deposition (PECVD) technique, in a series of wafer-on-wafer experiments featuring damascene processing and 45-nm design rules. Because a barrier layer’s thickness and density are both key to its performance and reliability, the researchers built silicon-oxy-nitride (SiON) barrier films at various densities and thicknesses, at processing temperatures as low as 150°C. Their results showed no degradation in electrical characteristics or reliability. (Paper #9.6, “Development of Low Temperature Dielectrics Down to 150°C for Multiple TSVs Structure with Wafer-on-Wafer (WOW) Technology,” H. Kitada et al, Fujitsu/University of Tokyo/Dai Nippon Printing/ATTO Co. Ltd.)

•    Practical Wafer Alignment: As circuit density increases, wafer-on-wafer alignment accuracy becomes crucial. But the alignment techniques used in laboratory settings aren’t necessarily practical for volume production. Sematech researchers will describe a practical alignment scheme that achieved submicron alignment accuracy for 300mm wafers. These were bonded face-to-face via thermocompression bonding of the copper interconnects on each wafer. The wafers were aligned using a first-of-its kind aligning/bonding tool which researchers will describe at IITC. They also developed inspection techniques to verify the integrity of the bonds and whether there were any voids. In particular, in order to better observe specific bonding structures, they developed a prototype integrated infrared high-speed focused ion beam (FIB) technique with CAD overlay capabilities. While their work represents a practical approach to wafer alignment for copper/copper thermocompression wafer bonding, they say throughput needs to be improved before it can be declared truly production-worthy. (Paper #12.2, “Recent Advances in Submicron Alignment 300 mm Copper-Copper Thermocompressive Face-to-Face Wafer-to-Wafer Bonding and Integrated Infrared, High-Speed FIB Metrology,” W.H. Teh et al, SEMATECH/EV Group/FEI Co.)

•    3D Interconnects for Biosensor/CMOS IC Integration: Georgia Institute of Technology researchers will describe two novel 3D interconnect technologies useful for the integration of a large biosensor array with a signal-processing CMOS IC. One is a TSV technology that can be used with relatively thick substrates, such as those often needed for sensors. It enables the formation of a seed layer in deep via holes and features a CMP-free planarization technique that allows vias to be fabricated after the sensor is made, which allows maximum material and process freedom for the sensor designers. The other technology is a compliant interconnect architecture that allows low-force and low-resistance temporary electrical contact between the sensor and the signal-processing IC. In this way, should the sensor become contaminated by whatever it is measuring, it can be removed and the CMOS IC can be reused. (Paper #2.2, “A 3D Interconnect System for Large Biosensor Array and CMOS Signal-Processing IC Integration,” H.S. Yang et al, Georgia Institute of Technology)

Making Better Memories

•    Filling Gaps With Glass: Open spaces or gaps can occur during fabrication of the interlayer dielectric (insulating layers) of a chip’s interconnect. If they’re not filled with dielectric material, a contact bridge, or short circuit, between metal lines could result. The problem is worsening as circuits get denser because the gaps tend to have higher aspect ratios (they are taller and narrower) and are thus harder to fill completely. Spin-on glass is used to fill gaps elsewhere on a chip, and attempts have been made in post-curing to improve its gap-fill performance for interconnects, but it hasn’t been effective to date. At IITC, Samsung researchers will describe a technique they used with a 30-nm 1G DRAM memory. First, they determined that the filling behavior of spin-on glass during spin-coating is mainly dependent on the capillary effect. Then, they developed a wettable surface treatment to enhance filling, and they optimized baking temperatures to minimize the viscosity of the glass. They say these techniques nearly eradicated the problem of contact bridges, and that spin-on glass may be suitable for future memories. (Paper #10.1, “Robust Spin-on Glass Gap-fill Process Technology for sub-30-nm Interlayer Dielectrics,” K-M. Byun et al, Samsung)

•    Double-Patterned Interconnect for Flash Memory: Photoresist patterning starts to lose resolution at about 45-nm half-pitch, and the industry is moving toward double-patterning (the use of physical spacers to create masks that can double line densities) as a way to fabricate chips at smaller dimensions. It isn’t restricted to interconnect applications, but is seen as promising for high-density flash memory interconnects. IMEC and Applied Materials researchers will describe a spacer-defined double-patterning approach incorporating 193-nm immersion lithography. They used three masks to produce 20-nm half-pitch test structures, including trenches for copper lines, in a silicon dioxide dielectric. The team then successfully metalized them using an amorphous carbon layer and a TiN metal hard mask. (Paper # 10.3, “Integration of 20-nm Half Pitch Single Damascene Copper Trenches by Spacer- Defined Double Patterning (SDDP) on Metal Hard Mask (MHM),” Y.K. Siew et al, IMEC and Applied Materials)

Advances in Core Interconnect Technology

•    Reduced Electromigration with Better Dielectric Reliability: At 45-nm and smaller geometries, copper electromigration (EM) is a problem. A self-aligned cobalt tungsten phosphide (CoWP) cap deposited on top of the interconnect reduces EM, but unfortunately it also can degrade the reliability of the fragile, porous ultra low-k dielectric which lies between the lines. IBM and Lam Research conducted studies to determine whether this degradation can be prevented. Using a CVD porous ultra-low-k dielectric (k=2.4) and 32-nm process technology on 300-mm wafers, they studied the dielectric’s conduction behavior and breakdown characteristics under a wide range of temperatures and electric fields. They found that adding a CoWP cap doesn’t change the dielectric’s fundamental leakage conduction mechanism and kinetics, provided it is optimized. At IITC, they will describe how they accomplished that, by selectively depositing the CoWP cap on the copper but not on the low-k surface. (Paper # 7.3, “A Comparative Study of ULK Conduction Mechanisms and TDDB Characteristics for Cu Interconnects with and without CoWP Metal Cap at 32-nm Technology,” F. Chen et al, IBM and Lam Research Corp.)

•    Self-Forming Barriers Integrated with Extreme Low-K: The 28-nm technology node and beyond will feature extremely low-k dielectrics. The question is, can the barrier layer that isolates copper lines from the dielectric be built in a self-forming fashion, so that both it and a copper seed layer can be created simultaneously, thereby reducing fabrication complexity and creating very thin layers? While self-forming layers have been investigated, TSMC researchers performed exhaustive analyses of their interfaces and integration with low-k dielectrics and capping layers. They looked at corrosion resistance, breakdown voltage and electromigration lifetimes in various 28-nm, dual-damascene structures with three extreme low-k dielectrics (k=2.2 to 2.8). They found self-formed barriers made from a CuMn alloy can indeed be integrated with extreme low-k dielectrics effectively, but that doing so is complex and relies on the right combination of dielectrics, interfaces and patterning approaches. (Paper # 3.3, “Interface Effect on Mn-Containing Self-Formed Barrier Formation with Extreme Low-k Dielectric Integration,” S-C. Pan et al, Taiwan Semiconductor Manufacturing Co.)

•    Hybrid Ti/TaN Barrier: Another approach is to fabricate hybrid barriers. Thin barriers made from tantalum nitride (TaN) are in use now. However, 45-nm and beyond technology will employ extremely low-k dielectrics, and TaN would degrade their reliability because moisture from the dielectric will oxidize the barrier, according to Panasonic and Renesas researchers. Meanwhile, titanium (Ti) offers favorable interface properties with extreme low-k dielectrics, is compatible with TaN and shows good wettability for copper. The researchers investigated the interactions between Ti, TaN, low-k dielectrics and copper, and found that a thin Ti/TaN multilayer barrier combining the materials’ desirable properties can meet the needs of forthcoming technology generations. (Paper #13.4, “Highly Reliable 45-nm-Half-Pitch Cu Interconnects Incorporating a Ti/TaN Multilayer Barrier,” M. Hamada et al, Panasonic Corp. and Renesas Technology Corp.)

Other Noteworthy Developments

•    Interconnects For Graphene: Carbon-based materials like carbon nanotubes and graphene nanoribbons have been viewed as contenders for the interconnects on future ultra-high-density chips, because they have low resistivity and high current-carrying properties. A team led by Fujitsu developed a multi-layer graphene growth technology, using a photoemission-assisted plasma-enhanced CVD technique, to grow networked nanographite wires on dielectrics without the use of metal catalyst films. Their accomplishment is an initial step toward building multilayer graphene interconnects. (Paper # 11.2, “Networked-Nanographite Wire Grown on SiO2 Dielectric without Catalysts using Metal-Photoemission-Assisted Plasma-Enhanced CVD,” M. Sato et al, Fujitsu/Tohoku University/JASRI/CREST)

What is interconnect and why is it so important?

“Interconnect” is the wiring system that connects transistors and other components on an integrated circuit, or computer chip. It may also refer to chip-to-package or chip-to-chip interconnections. Transistor speed used to be the limiting factor for chip performance, but with today’s multi-million-transistor chips, the interconnect itself has become a limiting factor. This is because the electrical resistance of the wires, or lines, increases as they are made thinner to accommodate more transistors. It also arises because capacitive coupling can occur among adjacent lines spaced very closely together.  Both inhibit the passage of signals.

The interconnect problem threatens to retard the development of chip technology, which in turn threatens the progress of the electronics industry, one of the world’s largest. IITC papers address this problem directly.

Copper lines offer less resistance and higher current-carrying capability than the previously used aluminum lines, but at small geometries the standard silicon dioxide insulator around the lines, called the dielectric, inhibits current flow. The search is on for substitutes. A dielectric’s relative permittivity (i.e. its interaction with current flow) is expressed as “k.” The lower the number, the faster the current flow. (Vacuum, the perfect low permittivity material, has k=1; silicon dioxide has k~4.)

But dielectric materials with very low k-values are porous, fragile and absorb moisture.  They can be easily damaged by typical chip-making processes such as exposure to harsh plasma during photoresist-stripping and to chemical-mechanical polishing, used to planarize each interconnect layer.

The semiconductor industry has introduced 45/40nm and 32nm initial chips into volume production.  Many of these use non-porous, or dense, low-k dielectrics made from carbon-doped oxide, or SiCOH, films (k~2.7-3.0).  Next are chips at the 28nm technology node expected to appear by 2010/2011, followed by 22nm technology chips by 2012, assuming many technical challenges can be solved.  20nm and 15nm options are also being discussed at this time. For these and future chips, porous low-k dielectrics may be necessary, with k-values ranging from 2.5 to 2.0.

At the advanced technology nodes’ smaller dimensions, there is an increased focus on new materials and processes to improve the reliability and manufacturability of copper based interconnects. These include barrier, seed, copper filling, and capping technologies.

3D Interconnects
The electronics industry continually strives to place more functions in the same size or smaller packages. One way to do that is with 3D integrated circuits, where individual chips are thinned (usually), stacked and then interconnected so they function as a single unit. These stacks are challenging to fabricate, and despite much progress in recent years the industry is still trying to determine the best ways to accomplish 3D interconnect architecture.

There are a variety of 3D techniques under consideration, but many of them require the etching and filling of relatively large, deep holes through the backside of the wafer, called through-silicon-vias, or TSVs.  Processes such as deep etching, highly conformal insulator deposition, and high-aspect-ratio metal fill are needed to create these TSV structures. The demand for these types of applications warrants special consideration in interconnect technology, and IITC has included this discussion in its technical sessions.

Related articles

Sematech eyes TSV volume production as positive results are reported - 10 June 2010

300mm fab automation: Towards 7th heaven - 01 June 2005

WiSpry appoints two MEMS veterans to Advisory Board - 30 May 2008

SanDisk’s Eli Harari to receive IEEE Robert N. Noyce Medal - 24 June 2009

EVENT: SEMATECH Hosting Advanced Gate Stack Symposium - 15 September 2005

Reader comments

No comments yet!

Post your comment

Please enter the word you see in the image below: