Mauro J. Kobrinsky

Three-dimensional wafer stacking via Cu-Cu bonding integrated with 65-nm strained-Si/low-k CMOS technology

The authors report the first demonstration of integrating wafer stacking via Cu bonding with strained-Si/low-k 65-nm CMOS technology. Sets of 330 mm wafers with active devices such as 65-nm MOSFETs and 4-MB SRAMs were bonded face-to-face using copper pads with size ranging between 5 /spl mu/m/spl times/5 /spl mu/m and 6 /spl mu/m/spl times/40 /spl mu/m. The top wafers were thinned to different thicknesses in the range 5 to 28 /spl mu/m. Through-silicon-vias (TSVs) and backside metallization were used to enable electrical testing of both wafers in the Cu-stacked configuration. We tested individual transistors in the thinned silicon of bonded wafer pairs where the thinned silicon thickness ranged from 14 to 19 /spl mu/m. All results showed that both n- and p-channel transistors preserved their electrical characteristics after Cu bonding, thinning, and TSV integration. We also demonstrated the functionality of stacked 65-nm 4-MB SRAMs by independently testing the cells in both the thinned wafer and the bottom wafer. For the SRAM, we tested a wider thinned wafer thickness range from 5 to 28 /spl mu/m. On all tested samples, we did not find any impact to the electrical performance of the arrays resulting from the three-dimensional (3-D) integration process. The stacked SRAM is an experimental demonstration of the use of 3-D integration to effectively double transistor packing density for the same planar footprint. The results presented in this letter enable further exploratory work in high-performance 3-D logic, which takes advantage of the improved interconnect delays offered by this Cu-bonding stacking scheme integrated with modern CMOS processes.

Comparisons of conventional, 3-D, optical, and RF interconnects for on-chip clock distribution

This paper analyzes the performance of different interconnect technologies for on-chip clock distribution, including conventional, three-dimensional, optical, and radio frequency interconnects. Skew, power, and area usage were estimated for each of these technologies based on the 2001 International Technology Roadmap for Semiconductors. Our results indicate that most of the skew and power are associated with local clock distribution. Consequently, since the alternative clock distribution approaches that have been proposed focus on global clock distribution, we have not found significant advantages over conventional clock distribution in terms of skew and power. Furthermore, it was found that low skews could be attained with conventional clock distribution schemes if the clock signals are not scaled down.

Design and Fabrication of 3D Microprocessors

Stacking multiple device strata can improve system performance of a microprocessor (μP) by reducing interconnect length. This enables latency improvement, power reduction, and improved memory bandwidth. In this paper we review some of our recent design analysis and process results which quantitatively show the benefits of stacking applied to μPs. We report on two applications for stacking which take advantage of reduced wire length- “logic+logic” stacking and “logic+memory” stacking. In addition to optimizing minimum wire length, we considered carefully the thermal ramifications of the new designs. For the logic+memory application, we considered the case of reducing off-die wiring by stacking a DRAM cache (32 to 64MB) onto a high performance μP. Simulations showed 3x reduced off-die bandwidth, Cycles Per Memory Access (CPMA) reduction of 13%, and a 66% average bus power reduction. For logic+logic applications, we considered a high performance μP where the unit blocks were repartitioned into two strata. For this case, simulations showed that stacking can simultaneously reduce power by 15% while increasing performance by 15% with a minor 14° C increase in peak temperature compared to the planar design. Using voltage scaling, this translates to 34% power reduction and 8% performance improvement with no temperature increase. We found that these results can be further improved by a secondary splitting of the individual blocks. As an example, we split a 32KB first level data cache resulting in 25% power reduction, 10% latency reduction, and 20% area reduction. We also discuss the fabrication of stacked structures with two complimentary process flows. In one case, we developed a 300mm wafer stacking process using Cu-Cu bonding, wafer thinning, and through-silicon vias (TSVs). This technology provides reliable bonding with non-detectable bonding-interface resistance and inter-strata via pitch below 8μm. We investigated the impact of this wafer stacking process to the transistor and interconnect layers built using a 65nm strained-Si/Cu-Low-K process technology and found no impact to either discrete N- and P-MOS devices or to thin 4Mb SRAMs. We verified fully functional SRAMs on thinned wafers with thicknesses down to 5μm. Although wafer stacking leads itself well to tight-pitch same-die-size stacking, die stacking enables integration of different size dies and includes opportunity to improve yield by stacking known good dies. We demonstrated a die stack process flow with 75μm thinned die, TSV, and inter-strata via pitch below 100μm. We also found negligible impact to transistors using this process flow. Multiple stacks of up to seven 75μm thin dies with TSVs were fabricated and tested. Prospects for high volume integration of 3D into μPs are discussed.

Experimental validation of crosstalk simulations for on-chip interconnects using S-parameters

Since the design of advanced microprocessors is based on simulation tools, accurate assessments of the amount of crosstalk noise are of paramount importance to avoid logic failures and less-than-optimal designs. With increasing clock frequencies, inductive effects become more important, and the validity of assumptions commonly used in simulation tools and approaches is unclear. We compared accurate experimental S-parameters with results derived from both magneto-quasi-static and full-wave simulation tools for simple crosstalk structures with various capacitive and inductive couplings, in the presence of parallel and orthogonal conductors. Our validation approach made possible the identification of the strengths and weaknesses of both tools as a function of frequency, which provides useful guidance to designers who have to balance the tradeoffs between accuracy and computation expenses for a large variety of cases

A novel technique for full-wave modeling of large-scale three-dimensional high-speed on/off-chip interconnect structures

This paper presents a novel, rigorous, and fast method for full-wave modeling of high-speed interconnect structures. In this method, the original wave propagation problem is represented into a generalized eigenvalue problem. The resulting eigenvalue representation can comprehend conductor and dielectric losses, arbitrary dielectric and conductor configurations, and arbitrary materials such as dispersive, and anisotropic media. The edge basis function is employed to accurately represent the unknown field, and the triangular element is adopted to flexibly model arbitrary geometry. A mode-matching technique applicable to lossy system is developed to solve large-scale 3D problems by using 2D-like CPU time and memory. A circuit-based extraction technique is developed to obtain S-parameters from the unknown fields. The proposed technique can generate S-parameters, full-wave RLGC, propagation constants, characteristic impedances, voltage, current, and field distributions, and hence yield a comprehensive representation of interconnect structures. Experimental and numerical results demonstrate its accuracy and efficiency.

Experimental validation of crosstalk simulations for on-chip interconnects at high frequencies using S-parameters

Since advanced microprocessors are designed based on simulation tools, accurate assessments of the amount of crosstalk noise are of paramount importance to avoid logic failures and less-than-optimal designs. With increasing clock frequencies, inductive effects become more important, and the validity of assumptions commonly used in simulation tools and approaches is unclear. We compared accurate experimental S-parameters with results derived from both magneto-quasi-static and fullwave simulation tools, for simple crosstalk structures with various capacitive and inductive couplings, in the presence of parallel and orthogonal conductors. Our validation approach made possible the identification of the strengths and weaknesses of both tools as a function of frequency, which provides useful guidance to designers who have to balance the trade-offs between accuracy and computation expenses for a large variety of cases.

Demonstration of a reliable high-performance and yielding Air gap interconnect process

Capacitance coupling in copper low-k interconnects can be further reduced by implementing Air gaps in the intra-layer dielectric. This paper describes the evaluation of an integrated Air gap technology using 32 and 22 nm node technology vehicles. Electrical, reliability, and yield results are presented.

A low power electro-optic polymer clad Mach-Zehnder modulator for high speed optical interconnects

Electro-optic (EO) polymer cladding modulators are an option for low-power high-speed optical interconnects on a silicon platform. EO polymers have inherently high switching speeds and have shown 40 Gb/s operation in EO polymer clad ring resonator modulators (RRM). In EO polymer clad RRM, the modulator’s area is small enough to be treated as a lumped capacitor; the capacitance is sufficiently low that the modulation speed is limited by the bandwidth of the resonator. A high Q resonator is needed for low voltage operation, but this can limit the speed and/or require precise control of the resonator’s wavelength, necessitating power consuming heaters to maintain optimal performance over a large temperature range. Mach Zehnder modulators (MZM), on the other hand, are not as sensitive to temperature fluctuations, but typically are relatively long and must employ power consuming terminated travelling wave electrodes. In this paper, a novel MZM design is presented using an EO polymer clad device. In this device, the electrodes are broken into short parallel segments and the waveguide folds around them. The segments of the electrode length are designed to provide good signal integrity up to 20 GHz without termination. The electrodes are driven by a single drive voltage and provide push-pull modulation. Modulators were designed and fabricated using silicon nitride waveguides on bulk silicon wafers and were demonstrated at high speed (20 GHz). A VπL as low as 1.7 Vcm is measured on initial devices. An optimized device could provide 40 Gb/s performance at 1 V drive voltages, ~100 fF total device capacitance and less than 2 dB optical insertion loss.

Demonstration of a sidewall capacitor to evaluate dielectrics and metal barrier thin films

A sidewall planar capacitor (SW CAP) vehicle is developed to closely simulate processing conditions for metal barrier and dielectric in an integrated structure. For a known tantalum barrier for copper on a low-K dielectric, SW CAP TDDB is similar to those measured on an integrated vehicle. SW CAP results are useful for comparing electrical reliability of different dielectric systems, and effective in determining physical continuity of copper metal barriers.

Demonstration of new planar capacitor (PCAP) vehicles to evaluate dielectrics and metal barrier thin films

Planar capacitors can quickly test material properties of metals and dielectrics for interconnects. A sidewall capacitor device is used to evaluate metal thin-film barriers. Etch stop planar capacitors in turn can test multi-layer etch stops, exposing differences between leaky and good etch stop films. Fillable planar capacitors are also fabricated and results presented for that class of fill materials.