Program - Symposium B: Heterogeneous Integration Challenges of MEMS, Sensor, and CMOS LSI

Radio frequency RF MEMS systems have many benefits over conventional semiconductor devices for controlling and routing microwave and millimeter-wave signals. RF MEMS switches exhibit very low insertion loss and power consumption, and ultrahigh linearity, which enable high efficiency phase shifters or tunable filters for mobile communication and radar systems. Reliability inhibited deployment of RF MEMS switches in military or commercial systems, mainly due to failure related to electrical charging of the dielectric layer on top of the bottom electrode. Proposed solutions to charging include hermetic packaging, minimizing the electric field across the dielectric, and tailoring the polarity and waveform of bias control signals. These approaches provided improvements in reliability, but did not eliminate RF MEMS switch failure due to fast electrical charging (10s of µsec) and slow discharging (100’s of sec) of the oxide or nitride dielectric layers used in RF MEMS capacitive switches. This presentation shows that a novel ultrananocrystalline (UNCD) diamond film dielectric layer with 2-5 nm grains and 0.4 nm wide grain boundaries exhibit unique fast charging (~ 100 µsec) but also fast discharging (~ 100 µsec) behavior, the latter being 5-6 orders of magnitude faster than the discharging of oxide and nitride dielectric layers. Thus, the UNCD dielectric layer eliminates RF MEMS switch charging-induced failure, due to fast charge motion in and out of nanograins though atomic wide grain boundaries. The switch pulls down charging the dielectric layer to failure, but the latter discharges quickly when the voltage is removed. Thus, the switch always recovers fully from charging before the next switch operation, providing for the first time an operating RF MEMS switch continuously “on” without adverse impact on switch reliability. The results presented here demonstrate a new paradigm in RF MEMS switch operation with no dielectric charging failure. The paper will describe the first demonstration of monolithically integrated RF MEMS switches/CMOS devices with the CMOS devices driving the RF MEMS switch to billion of cycles.

Heterogeneous integration or More Than Moore is rapidly drawing interest as a new research field. This work focuses on applying the technology of the heterogeneous integration to wireless transceiver circuits. The assembling effect of different devices such as CMOS and MEMS (Micro Electro Mechanical Systems) on wireless transceiver circuits is a key issue, and was evaluated in this paper. Then, measurement results were compared with the simulation results. By realizing passive elements such as inductors in wireless transceiver circuit with MEMS process, these passive elements can obtain the tunable characteristic and high Q-factor, which is not possible to realize using only CMOS process. As a means of integrating MEMS and CMOS devices, flip-chip is very attractive because of its small footprint and short bump interconnect. In the flip-chip, However, MEMS and CMOS chip are implemented in a manner where both devices are facing each other. Thus, there is concern about degradation in Q-factor, altering the values of the passive elements and the added noise by capacitive and inductive coupling from various parts of wireless transceiver. In this work, by applying flip-chip bonding to MEMS and CMOS chip, the changes of the inductance and the Q-factor of the inductor built on MEMS was measured before and after the implementation. Also, the effect of the coupled noise from the clock which is used in the control circuit of MEMS was observed. By comparing the measured results with 3D electromagnetic simulation and circuit simulation results, the issues which are missed in the design stage were cleared for the establishment of the integrated design methodology for heterogeneous integration, which is the ultimate goal.

Today’s most advanced electronic devices demand powerful technology to support them. Insulated gate bipolar transistors (IGBTs) are one potential solution to address these needs, owing to their fast switching and high efficiency, producing advantages for complex amplifiers with low pass filters and pulse width modulation. In order to support the development of IGBT devices, however, an appropriate characterization method is needed. While many 2D techniques have been established, these techniques are generally limited to surface examinations. When these devices fail, however, the damage occurs in three dimensions. X-ray microscopy is now an established technique for 3D failure analysis and microstructure characterization, due to the high penetration power of x-rays. Their non-destructive nature enables precise characterization of internal features without physically destroying the sample, enabling repeated studies on one sample to examine, for example, the time evolution of damage propagation. Commercially-available microscopes are now capable of producing resolutions in the tens of nanometers, delivering similar characterization capabilities as electron microscopes, but without sectioning the sample and with no charging effect. In the study presented here, one IGBT device was purposely damaged by an avalanche test between the polysilicon trench gates and prepared with FIB to an appropriate size for imaging with ultra-high resolution x-ray microscopy. The surface defects were first examined by scanning-electron microscopy and subsequently imaged in 3D with a commercial nano-scale x-ray microscopy platform. While the SEM results clearly showed some damage between the polysilicon trench gates, the x-ray results revealed a complex damage structure in 3D. Further analysis revealed the precise nature of the damage, enabling a more comprehensive characterization of the IGBT device than has previously been possible.

The motivating principle behind this research is the development of a small, wearable sensor that would use humidity and temperature measurements as metrics for health monitoring. If it is to be useful as a health monitoring tool, the device needs to respond quickly and predictably to changes in humidity. Also important is keeping the cost of the materials low. In this work, collagen is shown to be a viable humidity sensing material for use in capacitive relative humidity (RH) sensors. As a natural by-product of meat and leather industries, collagen presents itself as an interesting and inexpensive alternative to polyimide dielectric sensing materials. We used gelatin, a partially hydrolyzed form of collagen, to allow for easier thin film deposition. We have successfully fabricated devices by depositing a collagen thin film (15 μm) via spin coating, followed by Au/Pd electrodes (60 nm) via sputter coating. A plastic mask made with a rapid prototyping machine was used during physical vapor deposition (PVD) to pattern electrodes. This simple method eliminates the need to use more complicated photolithography processing. Interdigitated electrodes, rather than parallel plate electrodes, form a 6 mm wide, planar capacitor that has little dependence on dielectric thickness and is not affected by dielectric swelling. Characterization methods include rapid humidity change to evaluate response, steady-state measurements to evaluate consistency and obtain capacitance-humidity relationships, temperature modulation, and scanning electron microscopy (SEM). Initial findings indicate that these devices very closely match the results of the commercial relative humidity sensor (Omega OM-73) used for reference. The capacitance-humidity relationship is shown to be non-linear, with an average change of 3.0 fF for every 1% change in RH around 60% RH, and an average change of 7.0 fF for every 1% change in RH around 80% RH. In this work, we present the fabrication and characterization of these novel collagen-based relative humidity sensors.

The frequency responses of coupled vertically oscillating resonators are simulated and measured to investigate the source of acceleration sensitivity in out-of-plane vibratory MEMS gyroscopes. The acceleration sensitivity in MEMS gyroscopes is one of the important issues, since the gyroscopes are often used in harsh conditions in terms of vibration, especially for automobile application. The tuning fork type gyroscopes (TFG), in which two coupled vibratory gyroscopes are operated anti-phase manner, is the best solution for eliminating acceleration output by differentiating two outputs. However, the unbalanced mass and stiffness of springs from fabrication imperfection are said to be the source of the acceleration output which still exists in TFGs. We have investigated the mechanism how in-phase acceleration causes anti-phase vibration appeared as angular-rate output in in-plane TFG in the previous works and proposed the way to reduce the acceleration output by decoupling the in-phase (fin) and anti-phase (fanti) modes of the sensing vibrations. In this paper, the out-of-plane gyroscopes with out-of-plane resonators for sensing are being investigated. To simplify our analysis, instead of fully operating out-of-plane gyroscope, we designed four types of vertically coupled resonators by changing coupling scheme, coupling spring and coupling frame, and resonator, linear vibration and torsional vibartion. Each type of coupled resonator is designed with decoupling of in- and anti-phase mode (decoupling ratio, DR= (fanti-fin)/(fanti)) values of 0.09, 0.13 and 0.29. We fixed the anti-phase frequency at 16.5kHz and shifted in-phase frequency. In all designs, resonator proof-mass size is same. The designs were fabricated on SOG process of 20µm thick. In experimental evaluation, the devices were placed on piezo-actuator in vacuum chamber for applying vertical mechanical excitation with a frequency sweep in the range of 0-20kHz and observed under scanning laser Doppler Vibrometer. We maintained the vacuum about 20Pa. The results of all devices showed that at anti-phase frequency, two resonators are excited in anti-phase mode and as decoupling ratio increases from 0.09 to 0.29, resonators anti-phase mode excitation amplitude decreases by one-eight.

This paper reports a MEMS based vibrating viscometer suitable for in-process measurement in industrial fluid plants and low-cost disposable uses for medicine. Vibrating body of the proposed sensor is unique dual spiral geometry. The geometry provides a seamless surface for sensing the viscous stress by the simple structure to assemble the parallel plates to create the Couette flow into the MEMS device. The principle of the viscosity measurement based on the proposed dual spiral geometry was successfully verified by the experimental results using the first prototype devices. In the prototype viscosity sensor, the spiral structure is formed by penetrating trenches of 40 micro meters width on Si wafer of 500 micro meters thickness. The trenches were shaped by Deep-RIE. The width of the spiral wall is 80 micro meters. The dual spiral is composed of the two independent spirals. One spiral is named vibrating spiral which is vibrated by a piezo actuator and another spiral is called sensing spiral which is driven by viscous stress. The sensor chip is mounted in a holder included a piezo actuator and a pin to push the center of the vibrating spiral. In experiment, the whole holder is dipped in a liquid. As the piezo actuator is activated, the vibrating spiral is deformed and the viscous stress is developed in the liquid in the gap. The sensing spiral is deformed by the viscous stress. Frequency response curve of sensing spiral is calculated by analyzing the equation of motion. Our theory differs from the popular vibrating viscometer in that the external force which acts on the sensing body is only the viscous force. The viscosity of the liquid is obtained by curve fitting of the frequency response. Because the prototype sensor dose not have integrated strain gauges to measure the displacement of the spirals, a testing system which has a laser displacement sensor was built to examine the measurement principle. Sample liquids are Japanese standard reference liquids. Experimental results of the frequency response are collected. Parameter of the sensor is calibrated using the fitting value of the JS10 one of the reference liquid. By using the fitting value of the others, measurement values of the viscosity are calculated. The measurement results and deviations agree with the reference values less than 10 %. Although the deviations of the values are not sufficiently small, the experiment verified that the proposed dual spiral-shaped structure can be applied to the viscosity measurement.

MEMS-based biosensors are a promising new platform for the delivery of diagnostic services close to the point of care, where issues like reliability, ease of use, and low cost are of primary importance [1]. The intrinsically parallel nature of MEMS fabrication allows for very low unitary cost of elementary MEMS components. As an added benefit, the reuse or modification of standard Complementary MOS (CMOS) technologies allows the coexistence on the same silicon chip of MEMS components and the driving and conditioning circuitry. In the case of biosensors, specific issues related to the bio-activation of the sensor surface and its compatibility with on-chip MEMS and electronics have to be taken into account [2]. While standard bio-functionalization procedures normally involve immersion of the sample in the required solutions, this approach may not be feasible for silicon chips containing mechanically sensitive MEMS components and electronic circuits. Moreover, the bio-coating technique must be compatible with sensor package and wiring. In this work, the use of drop-coating [3] as a substitute to immersion for the creation of bioactive surfaces on MEMS sensors is investigated. The target sensor platform is a CMOS-based resonant sensor based on the microbalance principle [4]. Preliminarily, a test to verify the effectiveness of the functionalization protocol was performed: test silicon dioxide surfaces were cleaned in an ammonia-based hydroxylation solution, and silanized through drop-coating with an aqueous-based APTES (amino-propyl-triethoxysilane) solution as the preliminary step towards the deposition of a bioactive layer [5]. The surfaces were studied by means of conventional and angle resolved x-ray photoelectron spectroscopy. The spectroscopic characterization confirmed that the resulting surface chemical composition was not significantly different upon the two alternative processing approaches: both the atomic percentages values and the outermost layer in-depth distribution of the functionalities are comparable for the two approaches. Subsequently, a sample containing several MEMS resonators [4] underwent a similar procedure. The amino coated resonators were then exposed to a solution containing an oligonucleotide specifically designed to link to a portion of human MGMT (methylguanine-DNA methyltransferase) mRNA [6], and subsequently to its FITC fluorescent labeled complementary target. A comparison between this sample and a reference sample, not exposed to the target, shows a clear fluorescence signal and can interpreted as the occurrence of a specific binding between probe and target. References 1 H.-H. Tsai et al, Sens. Act. B: Chem. 144 (2010) 407–412. 2 A.C.R. Grayson et al., Proc. IEEE 92 (2004), 6-21. 3 E. Espinosa et al., Sens. Act. B: Chem. 144 (2010), 462-466. 4 D. Paci et al., Sens. Act. B: Chem. 129 (2008), 10-17. 5 S. Lenci et al., Appl. Surf. Sci. (in press). 6 L. Tedeschi et al., Biosens. Bioelectron. 20 (2005), 2376-1285.

In this work we present a self-aligned process for the fabrication of round ultra-thin silicon membranes which constitute the sensing element of a 64 element micromechanical capacitive biosensor array. To fabricate the array direct bonding between a Si wafer holding the membranes and a substrate wafer is required. The operation of the sensing elements rely on the changes in membrane surface stress and its subsequent deflection upon binding events between receptors immobilized on its surface and target molecules in its vicinity. Surface stress based biosensors are of great interest as they allow label-free sensing resulting in simplified sample preparation, small size and ability for parallelization into arrays for high throughput analysis [1-2]. The process begins by first forming a 5000Å thick thermal SiO2 layer on the membrane wafer. Circular sensor cavities are then formed in the oxide using optical lithography and CHF3 anisotropic dry etching. Next, a boron ion implantation follows through the openings in the thick oxide. This step creates a highly boron doped region which subsequently forms the flexible membrane electrode of the capacitive sensing element after wet etching. The whole procedure, effectively, aligns the sensor membrane placing it over the sensor cavity in which it is allowed to deflect. The ion implantation and annealing conditions determine the thickness of the final membranes and these conditions are selected based on simulation results and wet etching test experiments. The chosen conditions were 2E16 ions/cm2 at 150 keV, followed by thermal annealing at 1050 C for 1h. Next, the membrane wafer is bonded with the substrate wafer which has gone through phosphor implantation (1E15 ions/cm2, 20keV) followed by thermal annealing (1000 C, 20min) to render it conductive and form the sensor fixed substrate electrode. The wafer stack is then grinded mechanically, from the membrane wafer side, down to about 50µm while the rest of the wafer is etched in EDP solution until the ultra-thin Si membranes are revealed. The process then goes on with Al metallization and low temperature oxide (LTO) to passivate and finalize the device. To test the array performance 15mer synthesized thiol-modified oligonucleotide probes and corresponding targets were used. In order to immobilize the probes on the arrays, they were first functionalized with 3-glycidoxypropyl-tri-methoxy silane (GOPTS). Selective spotting of the probes on each membrane was then performed using the Laser Induced Forward Transfer (LIFT) technique. The hybridization procedure was monitored using a specially designed setup [3]. First experimental results indicate that the sensors are able to detect the hybridization of 10mM synthesized 15mer oligos. [1] J. Fritz, Analyst 133 (2008) 855–863. [2] A. Boisen and T. Thundat, Materials today, 12 (2009) 32-38. [3] V. Tsouti et.al , Biosens. Bioelectron. 26 (2010) 1588-1592.

Brain-machine-interfaces (BMI) use neuronal activities of spike and local field potential (LFP) with a high spatiotemporal resolution, in order to control of artificial machines such as robot-hand/prosthetics. However, the recording of spike/LFP with penetrating-probes into a tissue induces damage to brain/neurons, while the recording of electrocorticogram (ECoG) with planer-electrode onto brain surface has the advantage of the non-invasive measurement. To determine the electrode type in the future BMI, a further analysis of data correlation between the non-invasively recorded ECoG and the invasively recorded spike/LFP is necessary. One possible approach to analyzing the correlation is simultaneous recordings of ECoG and spike/LFP with a high spatial resolution. Here we propose a hybrid integration of ECoG recording- and spike/LFP recording-electrode arrays. We designed and fabricated ECoG and spike detecting hybrid-neuroprobe devices, based on a single-chip integrated IC-processed planer-microelectrodes and vapor-liquid-solid (VLS) grown penetrating-microprobes. The length of the probe was set at 200 µm, making the I-II cell layers of a cortex possible to reach. Diameters of the probe-tip and the planer-electrode were set at 1 μm and 20 μm, respectively, and individual electrodes were spaced 100 µm apart. After packaging the device, each recording site of the array was metalized with a low impedance material of Pt-black by the electroplating. The Pt-black tipped probe resulted in low enough impedance for spike/LFP and ECoG recordings. In order to demonstrate simultaneous recordings of ECoG and spike/LFP, the fabricated hybrid-neuroprobe device was placed on the visual cortex of a Long–Evans rat. A display monitor was placed in front of the rat, and we used computer controlled visual stimuli via the display. We confirmed the ECoG and spike/LFP recording capability of the hybrid neuroprobe device, as visually evoked ECoG and spike/LFP signals were simultaneously obtained via planer- and penetrating-microelectrode arrays. Based on the design of the prototype device, further high density and/or multi-channel ECoG and spike/LFP will be obtained to reveal correlation of ECoG and spike, mechanism of nervous system, as well as to develop a new class of neural recording methodologies for the future BMI technology.

Electrical immunodetection of the avian influenza A (H5N1) hemagglutinin (HA) peptide, the IN peptide, with anti-HA antibody was demonstrated using a field-effect transistor (FET) with an n-type silicon (Si) channel and a nickel (Ni) self-aligned silicided source/drain that was fabricated by a conventional top-down process. Confocal fluorescence microscopy revealed that IN peptide was specifically bound to immobilized antibody on the patterned SiO2 overlaying the Si channel, confirming that the binding capability of the immobilized antibody was retained. Positively charged ions, which modulate the electrostatic field across the SiO2 layer, and which are contained in the PBS solution used in these studies, result in the accumulation of electrons in the n-type Si channel, leading to a rapid increase in the channel current. Upon introducing antigen-containing peptides into the SU-8 reservoir, the accumulated electrons caused by positively charged ions in the PBS solution are then reduced by the specific binding of negatively charged peptides to the immobilized antibody. From the time-dependent I-V measurements, the settling time required to stabilize the electrical signals caused by negatively charged antigens was estimated to be 32 s

Resonant SOI-CMOSFET NEMS gas/bio sensors represent an important fundamental research topic and an emerging technology for many applications, thanks to the high flexibility offered by the surface functionalization methods which can tailor sensing layers for a large variety of gases and biomolecules. Self-assembled monolayers (SAMs) are one of the most relevant functionalization approaches to achieve selectivity in the sensing processes. Although the optimization of these organic films has received considerable attention, most of the functionalization routes have focused on the coating of sensors with relative large areas. The downscaling of the SAMs to small nano-devices, such as Si and poly SiGe NEMS and MEMS-based sensors, requires additional tools, both for their selective deposition and characterization. In this work, two main approaches to selective Si nanowires functionalization are being investigated: i) selective Joule heating ablation of the protective polymer layer on the suspended NW surface, and ii) e-beam lithography to open the resist layer on the NW area. In a following step, a NH2-SAM layer is deposited on the unprotected silicon oxide surface and glutaraldehyde molecules are used as a linker between the SAM amino functionality and biotin. The frequency shift response during CO2 sensing is studied as a function of an increasing number of amino functionalities of the silanization precursors. Finally Kelvin probe atomic force microscopy allows mapping the functionalized area due to the observed shift in surface potential between bare and SAMs-coated SiO2 surface.

As device size of ULSI shrinks, the size of memory capacitors has been shrinking. To maintain the capacitance of memory capacitors, it is necessary that surface area of capacitors is not reduced even though capacitor projected area decreases. Thus, capacitor structure will become more complicated like as a trench features. Thus, ultra-thin and conformal capacitor electrodes will be required. Ru and Pt is one of the major candidates for electrodes of memory capacitor such as dynamic random access memories (DRAMs) or ferroelectric random access memories (FeRAMs) because Ru is a good conductor even though it is oxidized to RuO₂ and Pt has good anti-oxidization property. The property of Ru and Pt as the electrodes was reported in many researches. Recently, embedded DRAMs or FeRAMs (eDRAMs or eFeRAMs) were getting much attention as one of the applications of DRAMs or FeRAMs. The eFeRAMs are put on a chip with the logic circuit in order that the size and the energy consumption of ULSIs is expected to be reduced by employing non-volatile memory. The eDRAMs or eFeRAMs are fabricated after the logic circuit. To prevent the logic circuit from thermal damage, the eDRAMs or eFeRAMs should be fabricated at low temperature. To fabricate conformal films for these devices, we applied supercritical fluid deposition (SCFD) method. In SCFD, supercritical CO₂ (scCO₂) is used as solvent and deposition is performed by chemical reactions of precursors in scCO₂. Supercritical fluids have intermediate property of gas and liquid. The density of supercritical CO₂ can approach or exceed that of liquids, and thus it can be a good solvent for organometallic compounds and their organic decomposition products. Precursor transport occurs in solution and reduction occurs at the solution/solid interface at significantly lower temperatures and higher reagent concentrations than those of vapor-phase techniques such as CVD. Thus, SCFD has the advantage of both gas-phase and liquid phase reaction. In this work, Ruthenium and Platinum thin films were deposited onto planar and trench structure SiO₂/Si substrates by SCFD. To deposit Ru films, we employed H₂ as the reducing agent and acetone as an additive material. Thus, continuous Ru films were obtained. In SCFD, H₂ was often used as reducing agent. However, Pt films could not be deposited by H₂ reduction. Then, we tried to deposit Pt films by reduction using cyclohexane. Cyclohexane may be promising reducing agent because Pt acts as auto-catalysis and promotes decomposition of cyclohexane. Reduction by cyclohexane enables to obtain continuous Pt films with carbon impurities. To remove carbon in Pt films, we employed post deposition annealing (PDA) under O₂ atmosphere and obtained pure Pt films which do not contain any carbon. After PDA with O₂ gas, electrical resistivity of Pt film decreases from 50 μΩ-cm to 15 μΩ-cm. The oxidation chemistry to deposit these noble metals will be also reported.

MEMS devices and technologies have been developing, and these devices are used in various fields and products now. But, in many cases, these MEMS devices are mainly composed of silicon materials, and the kind of materials used in these devices is limited. The authors think that it will be important to use functional materials such as magnetic materials and piezoelectric materials to create functional and high-performance MEMS devices. However, there are obstacles for utilization of these functional materials in these devices. One of the obstacles is difficulty of etching these materials precisely using conventional dry etching methods such as Reactive Ion Etching (RIE). Then, we will suggest the new dry etching method, new facing targets sputter-etching (NFTSE), that easily etch these materials. Sputter-etching methods with magnetic field such as magnetron had been researched widely in semiconductor field once, before RIE technology appeared. The Sputter-etching methods have the merit that they could etch various materials physically using only Ar gas as etching process gas. But the substrates and the materials etched by sputter-etching methods suffer from strong damages by attacks of high-energy particles such as Ar ions. Also, it is difficult to obtain superior etching uniformity on substrate surface across wide area without swing or rotation of magnets that form the magnet fields on substrate surface. Therefore sputter-etching methods are not used in a semiconductor field now because of these demerits. However, these demerits do not become serious problems in NFTSE, in the cases of shaping micron size 3D structures of the functional materials in fundamental research of some advanced MEMS devices. Because, the structure size is over ten-odd micrometers, so the damage from the high-energy particles affects only a restricted surface of the structure. Moreover NFTSE area is uniformity in comparison with magnetron sputter-etching owing to unique design of the magnetic field. Then, in order to confirm the performance of NFTSE method, 1 μm thickness Cu films that deposited on silicon wafer by a magnetron sputtering method were etched by NFTSE method, using 10μm pitch resist mask on the Cu films. As a result, superior perpendicular cross section of Cu films was observed by scanning electron microscope (SEM). Moreover, NFTSE method can obtain less than 10% etching uniformity across 80mm width without swing or rotation of the magnets because of unique magnetic field structure. In addition, NFTSE method has the possibility that the uniform etching area could be spread to 8 inches by optimization of the magnetic field structure. As a result, NFTSE may become one of important manufacturing technology of MEMS devices that consist of the functional materials. For these reasons, NFTSE method will support to create value added MEMS devises in near future.

As the gate pitch of MOSFET scaling into sub-28nm, the strain effect induced by conventional technologies declined in nanometer device’s channel. The new stress-liner material like diamond-like carbon (DLC) with higher stress was thought as one of main techniques to solve this issue. In this paper, we fabricated DLC films with extremely high compressive stress (>8~12GPa) using S-bend filtered cathodic vacuum arc (FCVA) technology and studied the process compatibility of DLC films integration with standard CMOS process. DLC’s compositional, mechanical and thermal properties were characterized using various tools such as multi-wavelength Raman spectroscopy (633nm, 532nm, 325nm), XPS, c-AFM, TEM, and TXRF. We optimized process parameters including substrate bias voltage (carbon ions energy), base pressure and filter coil current to maximize the compressive stress in deposited films with constant thickness (20nm). We firstly presented the correlations between compressive stress and G-peak dispersion extracted by multi-wavelength Raman spectra. It’s concluded that G-peak dispersion is a more insightful and reliable merit compared with ID/IG, G peak position and FWHM of G peak. The evolution of compressive stress and surface microstructure of DLC films with post deposition thermal annealing were studied by XPS and conductive-AFM. Thermal stability and stress evolution satisfy the thermal budget of BEOL. Meanwhile, metal contamination was measured by TXRF to exclude existence of deep impurities ensuring the compatibility of DLC films with CMOS process line. Finally, we studied local stress in silicon induced by DLC films after patterning into strips by UV-Raman spectroscopy and clarified the stress transfer mechanism which can provide insights about the integration of multiple strain techniques in advanced CMOS strain engineering.

In microfluidic devices, dielectrophoresis (DEP) is used to transport particles (e.g., cells, DNA). Conventional DEP devices use top or bottom electrodes which functions as collection points. The DEP device structures are limited because the photolithography is developed to fabricate planer structure. Electrodes are not fabricated on the sidewall of trench in the microfluidic devices. Photolithography for 3D structures permits wide variety of the device design [1,2]. Combining the electrodes fabricated on the sidewall with the branch of the microfluidic trenches, particle separation is achieved [3]. Photolithography for 3D structure requires uniform photoresist coating. Spray coating of photoresist is one of the promising methods for the uniform coating. We are investigating spray coating of photoresist by experimental and numerical approaches [4]. It was revealed that uniformity of the resist film is degraded by the lateral gas flow along the sample surface. Here, a shield plate with an aperture was used to suppress the lateral flow for improving the uniformity of the resist film. Gas flow was numerically analyzed by solving 2D Navier-Stokes equation. Calculation area was 50mmx46mm. Gap between shield plate and sample was 5mm. Aperture was φ9mm. With the shield plate, it was found that gas flow has enhanced vertical velocity component in the aperture area. Because the resist particles are delivered by the gas flow, the resist particle can enter trench structure to a deep level. Spray coating of the photoresist onto a substrate with trench structures was carried out using the shield plate. One direction scanning was repeated to reproduce the 2D flow in the numerical analysis. After the spray coating, line and space pattern were made to be orthogonal to the trench direction. Thickness distribution of photoresist film was investigated by white light interferometry and SEM. Without the shield plate, resist film thicknesses on top and bottom of the trench were 8.28μm and 2.59μm, respectively. With the shield plate, the thicknesses on top and bottom were 3.28μm and 2.21μm, respectively. Thickness ratio of bottom to top increased from 0.31 to 0.67. Deposition area was reduced from φ36mm to φ18mm by the plate. Bump formation which is frequently found around the convex trench corner [4] was suppressed. Bump thickness decreased from 6.80μm to 3.78μm. Uniformity of the photoresist film was improved. With the angled exposure technique [5], device structures can be fabricated on the trench sidewall. This research was supported by a MEXT program for forming strategic research infrastructure for private Universities from 2008 and Scientific Research (B) (20360116). [1] H. Houjou et al., Transducers’05, 3E4.35, p.1437. [2] P. Pham et al., J. Electrostatics 65 (2007) 511. [3] Y. Kim et al., J. Micromech. Microeng. 21 (2011) 015015 [4] S. Kumagai et al., Jpn. J. Appl. Phys. 50 (2011) 106501 [5] M. Sasaki et al., IEEE Optical MEMS Nanophotonics 2009, p.75

Combinatorial sputtering is one of the useful methods that could search for optimal composition of alloy materials or new alloy materials. Combinatorial sputtering provides sample groups (library) having different compositions in a limited place. Then, the properties of the samples can be scanned in an efficient way. It was devised that a Combinatorial Arc Plasma Deposition (CAPD) method which is a new combinatorial deposition method to search for alloy materials especially amorphous alloys. This method can synthesize the samples having wide composition ranges (more than 50 atomic %). However deposition areas (sample size) are too small and film thickness is too thin for evaluation of some properties. The New Facing Targets Sputtering (NFTS) method can be applied to new co-sputtering method for combinatorial sputtering. NFTS enables us to synthesize binary/ternary composition distribution thin film and have possibility to synthesize quaternary/quinary composition distribution thin film. One of the features of this method is that, range of composition ranges could be controlled. Furthermore larger and thicker film samples could be synthesized onto a single substrate compared to CAPD. In this research, binary and ternary composition distribution metal thin films were synthesized by using NFTS onto a single 52 mm×72 mm glass substrate. After deposited, compositions of films were characterized by the energy-dispersive X-ray spectroscopy (EDX). As an example, combinations of Cu, Zr and Ti targets were used to confirm the performance of the NFTS combinatorial sputtering method. As a result, composition distributions of synthesized thin films of Cu-Zr were formed in the range of Cu content between about 40 and about 70 atomic%. The films of Cu-Zr-Ti were formed in the range of Cu content between 25 and 60 atomic%, Zr content between 10 and 25 atomic% and Ti content between 30 and 55 atomic%. NFTS co-sputtering method can easily control composition distributions and their ranges by changing the power or the distance between targets and the substrate. In a presentation, we will report power dependency of the ternary composition distribution.

Recently, vibration energy harvesting is receiving a considerable amount of interest, because the vibration energy exists in many environments, such as mechanical vibration and human motion: it commonly goes unused. Although several methods are useful for obtaining electrical energy from vibration energy, piezoelectric materials are the most suitable because of their capability of converting applied strain energy into useful electric energy directly and the ease with which they can be integrated into a system. The basic structure of the piezoelectric vibration energy harvester is a MEMS cantilever with a seismic mass at the end. A piezoelectric thin film is deposited on the cantilever. The vibration energy is converted to electric energy by the direct transverce piezoelectric effect of the film. Therefore, the effective direct transverce piezoelectric coefficient (e_31,f) is important for the piezoelectric vibration energy harvester. [1] Lead-free ferroelectric BiFeO₃ is one of the promising materials for piezoelectric devices, because BiFeO₃ has a huge spontaneous polarization as large as 100 μC/cm² and high Curie temperature (T_C; 850 °C). Especially, relatively low dielectric permittivity (~50) of BiFeO₃ is suitable for vibration energy harvester, because the electric energy generated by the direct piezoelectric effect is inversely proportional to the dielectric permittivity. However, the direct piezoelectric property of BiFeO₃ films has not been reported. In this study, the direct piezoelectric property of epitaxially grown BiFeO₃ films was evaluated. Furthermore, the dependence of the crystal orientation on the direct piezoelectric property was investigated. (001) and (111) rhombohedral BiFeO₃ epitaxial films were fabricated by pulsed laser deposition, and the (001) and (111) film showed rectangular hysteresis loops with a remanent polarization of 70 and 96 μC/cm², respectively. The e_31,f coefficient of the films was measured by a method based on substrate bending and collecting developed charges. Details of the measurement method were reported elsewhere.[2] The e_31,f coefficients of -3.5, and -1.3 C/m² were observed for the (001) and (111) BiFeO₃ epitaxial films, respectively. It is suggested that the large direct piezoelectric response of the (001) film is caused by the effect of engineered-domain configuration. This study was supported by Industrial Technology Research Program in 2011 from New Energy and Industrial Technology Development Organization (NEDO) of Japan. [1] T. Yoshimura et al., Materials Science and Engineering, 18 092026 (2011) [2] T. Yoshimura et al., Jpn. J. Appl. Phys. 49, 021501 (2010).

Lead zirconate titanate is a popular material for microelectromechanical systems (MEMS) applications, known for its high spontaneous polarization and dielectric constant. These electrical properties depend highly on its crystal orientation as well as the substrate used for deposition. The formation of intermetallic compounds at the interface between the functional material layer (PZT) and the electrode layer (Platinum) is shown to facilitate the growth of a certain orientation of PZT. For this study, lead zirconate titanate with the morphotropic phase boundary composition Pb(Zr0.52Ti0.48)O3 films were deposited on Pt(111)/Ti/SiO2/Si (100) substrates by chemical solution deposition. The reactions between the precursors – lead acetate trihydrate, zirconium propoxide and titanium isopropoxide during sol preparation have been analyzed to understand the factors affecting the stability and life of the prepared sol. As the annealing temperature increases from 500°C to 700°C, a change in crystal structure from amorphous to polycrystalline pervoskite was observed. The films were characterized with X-ray diffraction (XRD), optical microscopy and scanning electron microscopy. Relevant peaks in the XRD patterns were identified revealing that that (101) is the preferred orientation. While only the presence of the intermetallic-Pt3Pb has been earlier reported, in the present study other intermetallic compounds such as PtZr, TiPt8 and Ti6Pb4 have been identified. Different intermetallic compounds at the PZT/Pt interface not only affect the orientation of the overlying PZT film but also its dielectric and polarization constants widely. The conditions for the formation of these intermetallics have been established by varying deposition cycles, annealing temperature and different kinds of substrates.

Recently the importance of point-of-care testing is growing rapidly as the demands of decreasing turn-around-time in bed-side diagnostics and upgrading medical environments in ubiquitous healthcare are growing. Biologically modified field-effect transistor (BioFET) is one of the most attractive approaches because of the on-chip integration of the sensor array, fast response, high reliability and lowcost mass production. However, the BioFETs used to detect macromolecules have been operated only in buffer solution with low salt concentrations because of the Debye screening length of blood or serum. Here we report a novel detection technique for direct label-free immunodetection of cancer markers in human serum using a Si-FET that was fabricated by conventional photolithographic processes. The proposed sensing method shows no dissociation of antigen–antibody binding as in general immunoassays, unlike the previous reports on Si-FET sensors. This method therefore overcomes the Debye length problem of immunodetection in human fluids, such as serum, that are generally encountered by FET based biosensors.1) In addition, in order to provide the microfluidics for test samples transportation to silicon chip and electrical connectivity for quantitative analysis and to use it easily to any non-skilled persons, we developed portable clinical analyzer platform, which consisted of small-size silicon-chip packaging cartridge and portable reader with easy user-interface. We demonstrate that connecting our simple electrical detection method, which does not require pretreatment of serum, with well-established whole blood filter technology will contribute to the development of new point-of-care testing (POCT) sensors. (1)Biosensors and Bioelectronics 25 (2010) 1767–1773

Fabricating micro and nano devices with integrated permanent magnetic components that exhibit appropriate magnetic properties has proven challenging to the micro and nanoelectromechanical systems (MEMS/NEMS) research community. While many research efforts have been pursued to develop processes for integrating hard magnets into devices, the processes significantly affect both the magnetic properties of the material and the entire device design due to constraints that the processes place on the fabrication sequence [1]. For example, despite the excellent hard-magnetic properties of NdFeB and SmCo alloys, methods to produce high-aspect ratio micro and nanostructures with these materials suffer from a lack of repeatability and undesirable oxidation . We have recently investigated several electrochemical deposition strategies, including direct plating, pulse plating, reverse pulse plating, and electroless plating, to produce high performance CoNiReP hard-magnetic alloys with tunable material properties. Patternable films, nanowires, and helical and tubular structures have been created. The processes developed can be economically upscaled and provide relatively few constraints on subsequent process flow. Using direct plating and reverse pulse plating with applied current densities ranging from -10 to -50 mA cm-2, we have attained films with intrinsic coercivities (Hc) as high as 3.5 kOe and energy products up to 6.2 MG Oe [2]. Films up to 3 μm thick that can be patterned by photolithography have been achieved with newly developed electrolytes. Pulse plated CoNiReP nanowires with semi-hard-magnetic properties have also been fabricated using porous anodized aluminum oxide templates. Hc values ranging from 290 to 590 Oe along the nanowire axis were obtained. Helical and tubular structures have been created in which the diacetylenic phospholipid 1,2-bis(10,12-tricosadionyl)-sn-glycero-3-phosphocholine (DC8,9PC) is used as a template. The phospholipidic nanostructures were metalized by electroless CoNiReP plating. The magnetic properties of all of these structures have been fully characterized. In addition, all structures were wirelessly manipulated using a 5-degree-of-freedom electromagnetic manipulation system [3] to investigate their potential as platforms for targeted drug delivery. [1] S. Guan et al., J. Electrochem. Soc. 152 (2005), C190. [2] S. Pané et al., Electrochim. Acta 13 (2011), 8979. [3] M. Kummer et al., IEEE Trans. on Rob. 26 (2010), 1006.

The heterogeneous integration of Si and Ge is attracting more attentions in the fields of microelectronics because Ge with higher carrier mobility of electrons and holes can be used as the transistor channel for future CMOS [1]. Ge has to integrate with silicon to sufficiently make use of the platform from the matured silicon CMOS technology. On the other hand, Ge can play an important role in the combination of silicon and III-V compound semiconductor owing to the matched lattice parameter between Ge and III-V [2]. However, ~4% lattice difference between Ge and silicon results in high defect density during Ge epitaxy on silicon [3]. Moreover, Ge-on-insulator (GeOI), such as silicon-on-insulator (SOI) structure, should be employed to suppress the current leakage caused by narrow bandgap as Ge is used as a transistor channel material. Wafer bonding is enabling technology to achieve heterogeneous integration due to match-free requirement of lattice pattern instead of the direct Ge epitaxy on silicon. However, the process of the heterogeneous wafer bonding has to be performed under the low temperature thanks to the difference of thermal expansion coefficient between both materials. There are a great number of defects in GeOI manufactured by the smart-cut® process [4] but the origin of the defects in GeOI is not being reported to date. The atomic level Ge/SiO₂ wafer bonding was demonstrated by 150°C annealing in the previous study [5] but the inspection for tiny voids at the bonding interface was restricted because of the reflecting infrared imaging. In this paper, we will report the origin of voids at the interface of wafer bonding between Ge and silicon by the detection of scanning acoustic microscope (SAM) with high resolution and the approach to suppress voids. For comparison, Ge/Si and Ge/SiO₂ wafer bonding were performed by 150°C annealing. Based on Si/Si and Si/SiO₂ wafer bonding, the visible voids (observed by infrared transmission imaging) at the Si/Si interface can be viewed but the invisible voids can be only detected by the testing of bonding strength [6]. In particular, it is difficult to examine the invisible voids in the wafer bonding with oxide because the porous oxide may absorb interface voids [7]. Our results indicated that the voids at the Ge/SiO₂ bonding interface can be also generated although the voids are more sensitive in Ge/Si wafer bonding due to the inappropriate wafer bonding approach. As a result, an improved method will be obtained to suppress the interface voids. [1] D. Lin, G. et al. IEEE International Electron Devices Meeting, (2009) 327-330. [2] W.K. Liu, et al., J Cryst Growth, 311 (2009) 1979-1983. [3] V.A. Shah, et al., Solid State Electron, 62 (2011) 189-194. [4] T. Akatsu, et al., Mat Sci Semicon Proc, 9 (2006) 444-448. [5] X.X. Zhang, et al., ECS Transactions, 33(3) (2010) 457-466. [6] X.X. Zhang, et al., Electrochem Solid St, 8 (2005) G268-G270. [7] Q.Y. Tong, et al., J Microelectromech S, 3 (1994) 29-35.

In this presentation, the possible applications of carbon based MEMS and NEMS are presented. Among carbon based structures, carbon nanotubes and graphene were chosen for the main materials of our studies. The three novel device structures, those are a transistor with carbon nanotube channel and suspended graphene based moving gate, an electromechanical non-volatile memory, and a graphene based mechanical resonators will be introduced. Novel field effect transistors with suspended graphene gates are demonstrated. By incorporating mechanical motion of the gate electrode it is possible to improve the switching characteristics. A new concept of non-volatile memory based on the carbon nanotube field effect transistor together with microelectromechanical switch will be introduced. A graphene xylophone structure which is composed of array of graphene resonators is suggested for its RF applications.

Polycrystalline silicon (Poly-Si) has been widely utilized as movable structure and often subjected to cyclic loading at high frequency under harsh environment such as high temperature and humidity in MEMS (Micro Electro Mechanical Systems). Although various researches about mechanical strength of Poly-Si have been reported, its reliability including initial fracture strength and fatigue lifetime has not been revealed completely [1-4]. One of the main reasons is large standard deviation (SD) in strength (2.55GPa in average and 28% in SD) [1], which is caused by the surface roughness of side walls formed by etching process for fabricating specimens. Thus, a new method which can eliminate etching influence for specimen should be developed to evaluate intrinsic material’s reliability. In this paper, we have proposed and demonstrated, for the first time, a novel testing method without applying the stress on the etched surface roughness. The specimen is a square membrane of 250-nm- and 500-nm-thick Poly-Si thin film. The membrane has a cylindrical Si-weight in the center and is supported by Si frame. Load to the membrane is applied by vibrating the Si frame in out-of-plane direction around the resonant frequency. The vertical displacement of Si-weight generated by resonant vibration is measured using a laser displacement sensor and the stress distribution of poly-Si membrane is calculated from the measured displacement using FEM analysis. The initial fracture strength of the film can be evaluated by breaking the membrane, since the fracture origin is on the surface free from etching process damages. We have evaluated the initial fracture strength by gradually increasing the displacement of the shaker. The initial fracture strength of poly-Si was 2.59GPa and 2.28GPa in average and its SD was 0.17GPa (6.6%) and 0.28GPa (12.1%) with 250nm and 500nm thickness of poly-Si membrane, respectively. The difference of average and SD of strength in thickness of membrane is explained by increase of surface roughness. Surface roughness measured by AFM was 1.8nm and 2.6nm for 250nm and 500nm thickness poly-Si, respectively. The average strength was almost same but SD was extremely small compared to the previous research result [1]. It is found that the developed method can decrease the SD and evaluate intrinsic initial fracture strength as mentioned above. As a next step, fatigue lifetime will be also examined using the same setup with long-term constant cyclic loadings. [1]T. Tsuchiya et al., IEEJ Transactions on Sensors and Micromachines, 116, 10 (1996). [2]D.A. Lavan et al., American Society for Testing and Materials, 1413 (2000) [3]W.N. Sharpe Jr et al, Mechanics of Materials, 36 (2004) [4]S. Kamiya et al., Journal of Micromechanics and Microengineering, 18 (2008)

Crystallization structures in the Si thin film affect the performance of MEMS/NEMS. Grain structures degrade mechanical characteristics of the film. Energy losses occur at the grain boundaries during device actuation. High temperature annealing (650-800°C) improves crystallinity of the deposited Si thin film [1,2]. However, random generation of crystallization nuclei limits the improvement. Grain growth is stopped by the collision with neighboring grains. To control the nuclei generation, we used metal-induced lateral crystallization (MILC) using Ni nanoparticles (NPs) synthesized within cage-shaped protein, apoferritin [3]. Because apoferritin molecules (φ12nm) are generated according to gene information, they have atomically the same structure. Using them as templates, uniform φ7nm NPs are synthesized. The size-regulated NPs are ideal “nanoblocks” for constructing functional structures [4]. Amorphous Si film (600nm) was deposited on a Si substrate with thermally oxidized layer (3μm). P ions were implanted into the Si film for electric conduction. Apoferritin molecules with Ni NPs were adsorbed on the amorphous Si film. The protein parts were eliminated by heat treatment (500°C, O2: 0.5slm) to leave Ni NPs on the surface. Positions of the Ni NPs determined the positions of crystallization. The sample was annealed for MILC (680°C, Ar: 0.5slm). The adsorbed Ni NPs reacted with Si to form silicide NiSi2. NiSi2 is generated at lower than 400 deg. Because NiSi2 has similar lattice constant with Si (mismatch: 0.4%), the NiSi2 functions as crystallization nuclei before the thermally generated nuclei induce the random crystallization. Electron BackScatter Diffraction analysis revealed crystallization structure in the Si film. Without MILC, grain was less than 1μm. With MILC, the grain reached 50-60μm. Crystalline orientations were almost the same direction. MEMS resonators were fabricated with the Si film and actuated under atmosphere. MILC shift the resonance frequency from 656 kHz to 676 kHz. This is explained by the enhanced tension by MILC [5]. Tensions measured by MEMS strain gauges were 368MPa and 461MPa, without and with MILC, respectively. Q factor increased from 36.3 to 42.3. 20% increase was observed. The improved crystallinity of the Si film contributed to increase the Q factor. Device operation under vacuum can increase the Q factor. For the advanced MILC, Ni ferritin molecules were patterned [4]. Crystallized area expanded outwards from the Ni ferritin patterned area. Crystallization structures in the MEMS/NEMS devices can be controlled by designing Ni ferritin adsorption positions. This study is supported by CREST/JST and KAKENHI (21710128), JSPS. [1] A. Ferri et al., J. Appl. Phys., 100 (2006) 094311 [2] H. Miura et al., Appl. Phys. Lett., 60 (1992) 2746 [3] H. Kirimura et al., Appl. Phys. Lett., 86 (2005), 262106. [4] I. Yamashita et al., Biochim. Biophys. Acta, 1800 (2011) 846 [5] S. Kuamgai et al., Proc. Transducers ’11, p. 1733

In the medical field, ultrasonic diagnostics is indispensable technology due to many advantages compared to X-ray and CT devices such as non-invasive, on-site investigation, and compact and simple equipments. Recently, further small and high resolution two-dimensional array transducers are required for the three-dimensional imaging intravascular endoscope application. Pb(Zr_xTi_1-x) O₃(PZT) ceramics which shows high piezoelectricity is used to conventional ultrasonic transducers. However, PZT ceramics is not suitable for micro-fabrication process and is difficult to form connecting wires for array transducers by conventional wire bonding technique. Micro-machined Ultrasonic Transducers (MUTs) on Si substrates are essential to realize above applications. MUTs have smaller element size than conventional ceramics transducers and the acoustic wavelength. We have proposed epitaxial PZT thin film which is grown on epitaxial γ-Al₂O₃ as a buffer layer on Si substrate for monolithically integrated with piezoelectric MUTs and Si-LSI. An element of our proposed pMUTs array has a semispherical cavity under piezoelectric PZT thin films formed by surface bulk micromachining technique using XeF₂ gas etching. Advantages of this technique are suitable for miniaturization and high density integration compared with conventional back-side etching technique such as TMAH etching. Moreover, a circular shape attains higher density than a square shape. In this research, we demonstrate high density piezoelectric micro-machined ultrasonic transducers array using epitaxial PZT thin film on γ-Al₂O₃/Si substrate. Fabricated pMUTs array is 100μm in diameter with 56 elements/mm². Semispherical cavity under the 500nm-thick PZT thin film diaphragm was formed by XeF₂ gas etching through the hole about 10μm in diameter from the surface of the transducer. Ultrasonic wave transmitting experiment was carried out using fabricated pMUTs as a transmitter and a commercial PZT hydrophone as a receiver. AC voltage (1.2MHz, 10V_p-p) was applied to the pMUTs under water to generate transmitting ultrasonic waves. Transmitted ultrasonic wave was received by a commercial ultrasonic hydrophone locating 1.5cm far from fabricated pMUTs. As a result, fabricated pMUTs has been successfully generated ultrasonic wave with 1.2MHz which is same frequency of applying voltage. Furthermore, the transmitting power - frequency characteristic under water was measured with changing frequency ranges from 1.0MHz to 8.0MHz. The pMUTs has the first resonance frequency at 1.2MHz, second resonance frequency at 1.6MHz, and third resonance frequency at 2.2MHz. These frequency ranges can be applied to medical ultrasonic diagnosis application.

In this contribution, I will highlight the recent progress on the fabrication of all-oxide piezo-MEMS devices by pulsed laser deposition. In our devices, we use Pb(Zr,Ti)O3 (PZT) as the piezoelectric material. The ferro- and piezoelectric properties are strongly related to the crystal orientation as well as the strain state of the PZT layer. Successful integration of these devices into silicon technology is therefore not only dependent on the ability of epitaxial growth on silicon substrates, but also the control of the crystallographic orientation as well as the strain state. We have fabricated all-oxide piezoMEMS devices and studied the resulting properties, such as piezomechanical actuation and sensor properties. In the contribution, I will further highlight our recent progress on large area deposition by PLD. Currently, we are able to fabricate piezoMEMS devices using a 200 mm technology.