Geir Uri Jensen

Fabrication and characterization of a wideband MEMS energy harvester utilizing nonlinear springs

This paper presents the fabrication, characterization and modeling of a wideband MEMS electrostatic energy harvester utilizing nonlinear springs. The experimental results show that the vibration energy harvester displays a strong softening spring effect. For narrow band vibration, the energy harvester exhibits a widening bandwidth during frequency down-sweeps. For increasing levels of broadband random noise vibration, the energy harvester displays a broadening bandwidth response. Furthermore, the vibration energy harvester with softening springs not only increases the bandwidth, but also harvests more output power than a linear energy harvester at a sufficient level of broadband random vibration. At a broadband random vibration of 7.0 × 10−4 g2 Hz−1, we found that the bandwidth increases by more than 13 times and the average harvesting output power increases by 68% compared to that of a linear vibration energy harvester. Numerical analysis confirmed that the softening springs are responsible for the band broadening.

Monte Carlo simulation of semiconductor devices

This paper gives a review of applications of the Monte Carlo technique and recent literature on Monte Carlo modeling of semiconductor devices. The emphasis of the original research results reported in this paper is on self-consistent ensemble Monte Carlo simulation of GaAs/AlGaAs Heterostructure Field-Effect Transistors (HFETs) and novel HFET structures. We consider both electron and hole transport keeping in mind possible applications for complementary devices and circuits. Hole transport properties in GaAs are treated using corrected expressions for the scattering rates and a very accurate analytical description of the valence bands valid to about 1 eV. We review the corrected rates. Our simulations demonstrate superlinear behavior in the velocity-field relationship at low temperatures, and a temperature maximum is found in the low-field mobility between 40 K and 60 K at a doping density of 1016 cm−3. We describe in detail our two-dimensional self-consistent Monte Carlo simulator and demostrrate the usefulness of Monte Carlo simulations for developing and validating simple device models utilized in computer-aided design tools for VLSI circuits. From this perspective, the unified charge-control model is also briefly outlined. Short-channel effects in self-aligned HFETs are shown to be caused by the injection of charge from the contact regions into the buffer beneath the device channel for gate lengths shorter than about 0.5 μm, given an adequately large aspect ratio. For possible incorporation in charge-control models, the threshold voltage shift as well as the output conductance in saturation are found to be approximately inversely proportional to the gate length under the same conditions. Our simulations show that a p-i-p+ buffer structure improves the carrier confinement to the channel, reducing the output conductance by about 80% in a 0.3 μm device. Two device concepts recently proposed, aimed at increasing carrier velocities, are studied. The Variable Threshold HFET (VTHFET) as well as the Split-Gate HFET (SGHFET) utilize a gate voltage swing that is made to depend on lateral position. This position dependence raises the resistivity and thus the electric field and the velocity near the source contact. For a certain doping configuration, a p-type-like VTHFET is shown to exhibit a 78% increase in the current-gain cutoff frequency fT, a 59% increase in the maximum transconductance, and substantially higher K-factor compared with a conventional 0.5 μm GaAs/AlGaAs HFET. In n-type VTHFETs, the only pronounced improvement is broader and flatter high gm region. The much larger improvement using p-type VTHFETs is shown to be associated withthe lower hole mobility, the absence of satellite valleys, and the smaller velocity overshoot. The n-VTHFET also suffer from a big shift in overall threshold voltage. The remarkable modulation of velocity and energy profiles achieved could in its own right warrant applications in other n-type devices, such as real-space transfer devices. Guidelines for succesful utilization of the VTHFET concept in other materials and for other device geometries are given.

Wideband MEMS Energy Harvester Driven by Colored Noise

We experimentally investigate the usefulness of nonlinear springs in a MEMS electrostatic energy harvester under colored noise vibrations. The experimental characterization of the energy harvester using nonlinear springs is compared with analytical and simulated results for an energy harvester with linear springs. For random vibrations with a bandwidth of 50 Hz and varying center frequencies, we found that the maximum output power of the nonlinear-spring harvester is 1.7 times lower than that of the linear-spring harvester, but the 1-dB bandwidth is two times larger. For vibration center frequencies of 38 Hz and 58 Hz below the resonant frequency, the nonlinear-spring harvester always achieves more power than the linear-spring one regardless of the vibration bandwidth. By varying the bias voltage, we found that the nonlinear-spring harvester obtains an average power of about 1 μW at 180 V, corresponding to an efficiency of 92% for a white noise excitation of 7.29×10-4 g2/Hz and 1.2 μW at 36 V for a frequency down-sweep at 0.15 g. A design variety of the device reached an output power of 7 μW at 120-V bias and 0.36- g acceleration.

Monte Carlo simulation of electron transport in mercury cadmium telluride

We derive expressions for electron scattering rates in mercury cadmium telluride accounting for correct wave functions in narrow‐band‐gap materials. These scattering rates differ slightly from the rates obtained from standard expressions for wide‐band‐gap materials. The difference is related to spin‐flip processes and has a relatively small effect on the transport properties. However, it is very important for spin‐orientation phenomena. Monte Carlo simulations have been performed to investigate the mobility and steady state velocity‐field characteristics of electron transport in mercury cadmium telluride with x=0.205 at 77 K. The simulations include scattering by polar optical phonons, ionized impurities, and alloy scattering. The Pauli exclusion principle as well as the dependence of the screening length on the distribution function have been accounted for. The simulations show that the screening length increases with increasing electric field with this dependence being the strongest for low carrier conc...

Sodium contamination of SiO2 caused by anodic bonding

In this paper we present an investigation of sodium contamination of SiO2 (oxide) during anodic bonding. Sodium contamination can be deleterious to the electrical properties of silicon structures. Silicon wafers with metal–oxide semiconductor (MOS) capacitors were bonded to Corning 7740 (Pyrex) glass wafers. The concentration of mobile ions was measured on capacitors outside and within glass cavities using the triangular voltage sweep method. Using secondary ion mass spectrometry analysis, it was confirmed that the ions were sodium. We found an increase in sodium concentration Nm between 1010 and 1013 cm−2, depending on the oxide location and the geometry of the glass cavity. The gate aluminium of the MOS capacitor was found to partly shield the oxide from contamination, causing a two to five times smaller increase in Nm. Reducing the bonding voltage from 800 to 500 V did not affect the increase in Nm significantly. In contrast, changing the ambient in the bonding chamber from vacuum to 1020 mbar air, reduced the contamination of capacitors situated outside the glass. A plasma-enhanced chemical vapour deposited Si3N4 film was found to be very beneficial in protecting the capacitors. The Si3N4 prevented sodium contamination of the capacitors situated within the glass cavities, and radically reduced the contamination of the capacitors situated outside the glass. The results suggest that the contaminating sodium originated from the bulk glass.

Modeling of Spring Constant and Pull-down Voltage of Non uniform RF MEMS Cantilever

In this paper, we are going to present a model of spring constant and pull down voltage for non uniform RF MEMS cantilever. In order to reduce the pull down voltage, it is usual to use a beam, which is narrower close to anchor and wider at the end or electrode area for a cantilever. Compare to uniform beam, this beam have lower spring constant which reduce the pull down voltage. A comprehensive model for spring constant and pull down voltage of the nonuniform cantilever is developed through basic force deflection mechanism of the suspended beam

Protection of MOS capacitors during anodic bonding

We have investigated the electrical damage by anodic bonding on CMOS-quality gate oxide and methods to prevent this damage. n-type and p-type MOS capacitors were characterized by quasi-static and high-frequency CV-curves before and after anodic bonding. Capacitors that were bonded to a Pyrex wafer with 10 μm deep cavities enclosing the capacitors exhibited increased leakage current and interface trap density after bonding. Two different methods were successful in protecting the capacitors from such damage. Our first approach was to increase the cavity depth from 10 μm to 50 μm, thus reducing the electric field across the gate oxide during bonding from approximately 2 × 105 V cm−1 to 4 × 104 V cm−1. The second protection method was to coat the inside of a 10 μm deep Pyrex glass cavity with aluminium, forming a Faraday cage that removed the electric field across the cavity during anodic bonding. Both methods resulted in capacitors with decreased interface trap density and unchanged leakage current after bonding. No change in effective oxide charge or mobile ion contamination was observed on any of the capacitors in the study.

The microwave dielectric properties of dual-layer PZT/ZrO2 thin films deposited by chemical solution deposition

The dielectric properties of dual-layer PZT/ZrO2 thin films were measured at microwave frequencies in both a metal?insulator?metal (MIM) capacitor and a coplanar waveguide (CPW) up to 50?GHz. Both PZT and ZrO2 films were prepared by the chemical solution deposition method. The measured dielectric loss of the PZT/ZrO2 film was approximately 0.08 at 30?GHz, much lower than that of typical PZT thin films. The dielectric constants obtained using the MIM capacitor with 360?nm PZT/65?nm ZrO2 and using the CPW with 420?nm PZT/280?nm ZrO2 were 47 and 130, respectively, at 50?GHz. Capacitance tunability was ~30% at +25?V and up to 50?GHz. The measured values obtained indicate that PZT/ZrO2 thin films may be suitable for the use of dielectric layers in tunable RF devices and RF MEMS capacitive switches operating at millimetre wave frequencies.

Microfabricated 1-3 composite acoustic matching layers for 15 MHz transducers.

Medical ultrasound transducers require matching layers to couple energy from the piezoelectric ceramic into the tissue. Composites of type 0-3 are often used to obtain the desired acoustic impedances, but they introduce challenges at high frequencies, i.e. non-uniformity, attenuation, and dispersion. This paper presents novel acoustic matching layers made as silicon-polymer 1-3 composites, fabricated by deep reactive ion etch (DRIE). This fabrication method is well-established for high-volume production in the microtechnology industry. First estimates for the acoustic properties were found from the iso-strain theory, while the Finite Element Method (FEM) was employed for more accurate modeling. The composites were used as single matching layers in 15 MHz ultrasound transducers. Acoustic properties of the composite were estimated by fitting the electrical impedance measurements to the Mason model. Five composites were fabricated. All had period 16 μm, while the silicon width was varied to cover silicon volume fractions between 0.17 and 0.28. Silicon-on-Insulator (SOI) wafers were used to get a controlled etch stop against the buried oxide layer at a defined depth, resulting in composites with thickness 83 μm. A slight tapering of the silicon side walls was observed; their widths were 0.9 μm smaller at the bottom than at the top, corresponding to a tapering angle of 0.3°. Acoustic parameters estimated from electrical impedance measurements were lower than predicted from the iso-strain model, but fitted within 5% to FEM simulations. The deviation was explained by dispersion caused by the finite dimensions of the composite and by the tapered walls. Pulse-echo measurements on a transducer with silicon volume fraction 0.17 showed a two-way -6 dB relative bandwidth of 50%. The pulse-echo measurements agreed with predictions from the Mason model when using material parameter values estimated from electrical impedance measurements. The results show the feasibility of the fabrication method and the theoretical description. A next step would be to include these composites as one of several layers in an acoustic matching layer stack.

Design of active phase shifters based on multichannel heterojunction field effect transistors (MCHFET's)

Design criteria of active phase shifters based on GaAs/AlGaAs multichannel (MC) HFET in the frequency range 4-60 GHz are presented. The phase characteristics of MCHFET devices were studied using the computer aided design program TOUCHSTONE. The dependence of transmission phase on various intrinsic elements in the equivalent circuit model as a function of control gate bias was also studied. There are limited gate bias ranges which correspond to the active regions of the two conducting wells for which a quasi-linear continuous phase shift for analog applications was achieved. Continuously varying the gate bias from V/sub gs/=-1.9 V to V/sub gs/=-0.6 V results in a quasilinear phase shift of 10/spl deg/, 15/spl deg/, 21/spl deg/, and 29/spl deg/ at f=12, 20, 30, and 60 GHz, respectively. Similarly, varying the gate bias from V/sub gs/=-0.4 V to V/sub gs/=0.7 V a quasi-linear phase shift of 21/spl deg/, 26/spl deg/, 27/spl deg/, and 23/spl deg/ at f=12, 20, 30, and 60 GHz, respectively, was achieved. The gain variation was less than 3 dB in these bias regions. With digital applications in mind, a maximum differential phase shift of around 50/spl deg/ was obtained by switching the gate bias discretely. The transmission phase of single gate MCHFET mostly depends on variation of gate source capacitance with gate bias rather than on other intrinsic elements. The dependence of phase shift on various geometrical and structural parameters is also presented. To test the practicality of the device, other scattering parameters (e.g., S/sub 11/, S/sub 22/, S/sub 12/) and the noise figure (NF) were finally studied.