論文

A low-pressure deposited, low-resistivity Ru thin film enables interconnects without adhesive liners for sub-2nm nodes and beyond. The newly developed low-pressure Ru deposition offers (1) a larger grain size of 14 nm, (2) smaller surface roughness of Ra = 0.16 nm, and the highest film density of 12.1 g/cm3, resulting in a record-low resistivity of 22.45 μΩcm at a thickness of 20 nm in the as-deposited state, without any additional thermal annealing. The 10 nm Ru thin film exhibits sufficiently high adhesive strength on SiO2 without the need for adhesive liners. The effective resistance, including liner resistance, is reduced by 15.9% compared to that with a Ta adhesive liner. The Ru thin film without the liner is a strong candidate for future advanced technology nodes.

Abstract By atomistic simulations, we investigated effects of oxidation on resistivity of CuAl 2 , which is expected to be a new interconnect material. Cu/Al/O reactive force field was developed for molecular dynamics (MD) simulations of CuAl 2 oxidation. From atomic positions of oxidized CuAl 2 films derived from MD simulations, resistivity was calculated by quantum transport simulations. Calculated resistivity was well reproduced by Fuchs-Sondheimer model with mean free path of 21 nm and surface specularity parameter of 0. Derived mean free path and surface specularity parameter realize precise modeling of interconnect resistivity composed of CuAl 2 .

We demonstrated Joule-heated Au nanosheet H2S sensors for low-power operation. We confirmed that low temperature regions in the Joule-heated Au nanosheet caused lower response and recovery characteristics than uniformly heated Au nanosheets. By using Pt electrodes, which has lower thermal conductivity than Au, heat dissipation to the electrodes could be suppressed, resulting in lower power consumption and faster recovery characteristics. We then discussed the optimal sensor structure by developing an analytical model of electrical and thermal resistances. We introduced semi-elliptical intermediate electrodes between the channel and pad electrodes to efficiently suppress the heat dissipation, demonstrating that the optimal channel length and thermal conductivity of the intermediate electrode κ int exist depending on the channel width. Finally, we proposed the sensor design strategy of considering the κ int dependences of the electrical and thermal resistances. This strategy is useful for all metal nanosheet sensors because it gives an estimation of their optimal structures.

A 200nm-pitched, superconducting niobium (Nb) interconnects with TiN protecting layer are fabricated on a 300-mm wafer for cryo-CMOS used in Quantum State Controller (QSC) application. The developed Nb interconnects show superior superconductivity with a Tc of 7.8 K and a record-high Jc of 64MA/cm2 at 4.2 K, which is high enough to local interconnects used in a cryo-CMOS circuit. For the first time, the SPICE simulation, using a 65nm-node cryogenic (4K) PDK with a technology file including Nb superconducting interconnects, estimates 40% improvement in signal delay in a typical use case. This result indicates that the power consumption in cryogenic CMOS operation can be reduced by the superconducting interconnects, as the operating voltage can be lowered while maintaining the same operational speed as that with the Cu interconnects.

In this study, density functional tight binding (DFTB) parameter for CoSn were developed toward high-throughput atomistic transport simulation. The developed DFTB parameters based on Bayesian optimization reproduced band structure of hcp Co, α-Sn, β-Sn and CoSn. To validate capability of transport simulation with the developed DFTB parameters, anisotropic bulk resistivity of CoSn was simulated by DFTB based non equilibrium Green's function simulation with frozen phonon approach. Calculated resistivity along a-axis ρa = 40.0 μΩ∙cm and c-axis ρc = 3.2 μΩ∙cm successfully reproduced experimentally reported values.

Electric field sensing can be used for the early prediction of lightning. In this study, we have suggested the method of sensitivity enhancement of graphene electric field sensor by Fowler–Nordheim (FN) injection into SiO2 under the graphene. The electric field sensing mechanism relies on the transfer of electrons between graphene and the interface trap states at the SiO2/graphene interface. The charge transfer alters carrier concentration in the graphene channel, which is reflected as a change in the drain current. The FN injection increases the interface trap states. Although time constant of charge transfer between graphene and the interface trap states was slightly degraded, lower limit of detection and sensor response were improved by the FN injection; the sensor response at 1 kV/m after 30 s of FN injection was 1230% of that before FN injection. The experimental result indicates the proposed method by FN injection is widely applicable for sensitivity improvement of electric field sensor composed of 2-D materials.

Gas sensor responses are considerably affected by gas flow rates, thereby inhibiting the accurate detection of target gas concentrations in variable-flow applications such as breath analyzers. To address this challenge, a flow-adaptive sensing system using a spatially distributed array of uniform Au nanosheet sensors was developed. The array layout was designed via computational fluid dynamics (CFD) simulations to generate spatiotemporal signal patterns from the array output upon gas exposure. These patterns were used to train a deep neural network, which accurately estimated gas flow rates and (Formula presented.) concentrations under various conditions. This system does not require active flow control during operation. The proposed approach employs sensor array design, CFD-guided layout, and artificial-intelligence-based inference to overcome the fundamental limitations of gas sensing. The sensing system can operate reliably under various flow conditions, highlighting the conceptual potential of the array-based sensing method for use in portable and pumpless gas-sensing platforms.

MOS capacitors fabricated on substrates with doping concentrations as high as 1018 cm−3 were characterized at 4.2 K. The highly doped substrate exhibited an intrinsic imaginary component of impedance at 4.2 K. The imaginary component is attributed to the time delay induced by hopping phenomena, leading to a decrease in the gate capacitance. Furthermore, we investigated the time constant associated with dopant ionization under depletion conditions and determined it to be 0.35 μs. An equivalent circuit model of the highly doped substrate at 4.2 K is also shown.

The Pt-Rh thin-film sensors exhibit excellent sensitivity and selectivity for H2 gas detection. Here, we studied the mechanism of highly selective detection of H2 by the Pt-Rh thin-film sensors with ambient-pressure X-ray photoelectron spectroscopy (AP-XPS) measurements at working conditions, which were paralleled with electric resistivity measurements. The elemental composition and chemical state of surface Pt and Rh drastically change depending on the background gas environments, which directly link to the sensor response. It is revealed that surface segregated Pt atoms accelerate dissociative adsorption of H2, resulting in a reduction of the sensor surface and then a decrease of electric resistivity of the film, whereas a thin oxidized Rh layer blocks dissociation of the other reducing agent, that is, NH3. This is supported from the adsorption energetics obtained by the density functional theory (DFT) calculations.

The sensing of gas components in a mixed gas is required for breath-based health monitoring and diagnosis. In this work, we report the simultaneous detection of mixed-gas components using a sensor consisting of [EMIM][BF4]-based ionic gel with four electrodes made of Au, Pt, Rh, and Cr. The voltage between any given pair of electrodes depends on the gas molecules absorbed in the ionic gel and the elements the electrodes are made of. When the voltage signals between all pairs of electrodes were used, H2, NH3, and C2H5OH concentrations were simultaneously estimated by a neural-network-based inference. From molecular dynamics simulations, the origin of the voltage signal was attributed to the catalytically generated adsorbates on the electrodes.

By utilizing stable thiol self-assembled monolayers (SAMs) on gold (Au) nanosheets, we investigated the temperature dependence of the amount of resistivity increases Drad by SAM modification. Drad increased with increasing temperature from 5 to 375 K in all the measured nanosheets for various thicknesses (9-20 nm). We consider from the DFT calculation that the temperature dependence originates from unoccupied states of sulfur atoms above the Fermi level of Au. The available unoccupied states increase at higher temperatures, which enhances scattering rates and thus increases the resistivity. Furthermore, we calculated the temperature dependence of Drad using the Bloch-Grüneisen theory and Fuchs-Sondheimer (FS) model. Based on the results, we conclude that for ultra-thin metal films, the temperature-dependent specularity parameter PT which represents the fraction of electrons scattered specularly should be introduced into the conventional FS model to predict Drad over a wide temperature range.

Changes in surface chemical state of a resistivity-based Pt thin-film sensor for H2 gas were observed under working conditions with direct in situ spectroscopy.

H-2 and NH3 detection with low power consumption was demonstrated by integrated chemiresistive Pt and PtRh nanosheet sensors on glass substrates. The self-heating effects realized low power and local heating of metal nanosheet sensors, enabling the integration of sensors with different operating temperatures. Based on different resistance changes in Pt and PtRh nanosheets toward H-2 and NH3, the concentration of each gas was detected from a gas mixture by consuming around 1-mW power. For decreasing the power consumption and further integration of sensors, sensor scaling and pulsed operations were numerically and experimentally studied. In addition to good connectivity of metal nanosheet sensors to large-scale integration (LSI) circuits, improvements of the power consumption by sensor scaling were proven. The pulsed operations required for integrated sensor arrays maintained a sensor response, or a resistance change, of approximately 60, even when the power consumption was reduced by 20.

A small-size and low-power consumption acetone sensor for the possible integration with LSIs was realized by a low-resistance (NH4) x WO3 nanowire. By utilizing hydrothermal synthesis, a large amount of NH4+ was inserted in the hexagonal tunnel of h-WO3 nanowires. The insertion and the activation of NH4+ were confirmed by FT-IR and XPS. The low nanowire resistance of 1.9kΩ made it possible to realize self-Joule heating effective to activate catalytic reaction with low applied voltage of 0.60 V and low power consumption of 0.11 mW. With the self-Joule heating, selective acetone sensing against ethanol was successfully demonstrated.

Hydrogen is an important biomarker for the human digestive system. However, accurate detection of ppm-level hydrogen in breath is difficult due to the competing detection of high concentration water. We fabricated Pt thin films that respond to hydrogen in air at concentrations as low as 500 ppb. In both dry and humid air, these films have almost identical response to hydrogen, i.e., their resistance decreases linearly with increasing hydrogen concentration regardless of relative humidity. Even at high relative humidity, these Pt thin films can detect ppm-level hydrogen. Furthermore, it was strongly suggested that these films can be applied to low-level hydrogen in the air expired by a healthy human. Based on the chemical kinetics, namely the adsorption and desorption of hydrogen and oxygen, the sensor response is quantitatively described by relating the hydrogen surface coverage to the magnitude of electron scattering at the Pt surface. The proposed model successfully reproduces the effects of hydrogen concentration and time on the sensor response, particularly at hydrogen concentrations below 20 ppm. Based on this model, these Pt thin film sensors have the potential to detect 1 ppm hydrogen in expired air within 30 s.

Band tails in heavily doped semiconductors are one of the important parameters that determine transfer characteristics of tunneling field-effect transistors. In this study, doping concentration and doing profile dependences of band tails in heavily doped Si nanowires were analyzed by a nonequilibrium Green function method. From the calculated band tails, transfer characteristics of nanowire tunnel field-effect transistors were numerically analyzed by Wentzel-Kramer-Brillouin approximation with exponential barriers. The calculated transfer characteristics demonstrate that the band tails induced by dopants degrade the subthreshold slopes of Si nanowires from 5 to 56mV/dec in the worst case. On the other hand, surface doping leads to a high drain current while maintaining a small subthreshold slope.