Most graphene-based NEMS devices are fabricated in a “one-off” manner using slow, limited scale methods such as mechanical exfoliation, electron beam lithography, or transfer from copper foils which can’t be incorporated into standard micro- and nanofabrication lines. In order to overcome this limitation, a method was developed that can be used to manufacture graphene-based NEMS devices at the wafer scale using conventional microfabrication techniques. In this project graphene is grown directly on thin film copper using chemical vapor deposition. The copper film is then patterned and etched to produce graphene-based NEMS resonators. This research is still in the early stages of development but this manufacturing method offers the potential to increase the throughput, yield, and repeatability of manufacturing graphene resonators while reducing the manufacturing cost and complexity.
You are here
One of the major challenges in producing highly accurate graphene-based sensors and filters is the poor fabrication repeatability of graphene resonators due to small variations in the residual stress and initial tension of the graphene film. This has meant that graphene-based nanoelectromechanical resonators tend to have large variations in natural frequency and quality factor from device to device. This poor repeatability makes it impossible to use these resonators to make accurate, high-precision force and displacement sensors or electromechanical filters. However, by actively controlling the tension on the graphene resonator it is possible both to increase repeatability between devices and to increase the force/mass sensitivity of the nanoelectromechanical resonators produced. Such tension control makes it possible to produce electrometrical filters that can be precisely tuned over a frequency range of up to several orders-of-magnitude. In order to controllably strain the graphene resonator, a microelectromechanical system (MEMS) is be used to apply tension to the graphene. The MEMS device consists of a graphene resonator connected between a set of gold electrodes. Each gold electrode is located on a different MEMS stage. Each stage is connected to a set of flexural bearings which are used to guide the motion of the stage. The displacement stage is actuated using a thermal actuator that allows a uniform and constant tension to be applied to the graphene resonator. The displacement of the actuator and the tension applied to the graphene are measured using a pair of differential capacitive actuators. The resonator is actuated electrostatically using the electrical backgate, and the resonant frequency is measured from the change in conductance of the graphene as it approaches resonance. Using this setup, it is possible to tune the natural frequency of the graphene resonator with high precision and accuracy.
Atomic force microscopy is capable of producing very high resolution (sub-nm-scale) surface topology measurements and is widely utilized in scientific and industrial applications, but has not been implemented in-line with manufacturing systems, primarily because of the large setup time typically required to take an AFM measurement. In order to overcome this limitation, the NDML has developed a single-chip-AFM-based inspection system where a wafer can be precisely and repeatably loaded into the setup and measurements can be taken in under 60 seconds. This inspection system consists of several single-chip AFMs integrated into a positioning stage to make measurements at multiple spots on a wafer at the same time. Each single-chip AFM is a MEMS device that is approximately 2 mm wide by 1 mm tall and is capable of scanning a 10 μm by 10 μm area. Thermal actuators in the MEMS device are used to do the scanning in both the x and y directions as well as to excite the z axis of the AFM so that it can be run in taping mode. Each AFM is attached to a flexure stage in the top plate of the inspection system so that the AFM can be precisely moved to the desired inspection location on the wafer. The flexure plate is coupled to the inspection plate using a kinematic coupling so that the flexure plate can be precisely located with respect to the inspection plate after each loading operation. In order to take a measurement, the flexure plate is removed from the inspection plate, a wafer is loaded into the inspection plate using an exactly constrained, passive alignment system, and the flexure plate is then placed back onto the inspection plate. This brings the AFMs back into contact with the surface that is to be measured and the AFMs can then start taking measurements without any additional alignment operations. The overall measurement procedure takes less than one minute, which is faster than most nanomanufacturing processes. This guarantees that the inspection step will not be the bottleneck in the manufacturing process.
The NDML is developing a new microscale selective laser sintering system (μ-SLS) that uses micromirrors to achieve write speeds on the order of 300 mm/s. In conventional selective laser sintering processes a thin layer of plastic or metal powder is spread over a build platform and a laser is used to sinter together the powder particles in the desired locations. A new layer of powder is then spread over the original layer and the process is repeated to build up a three-dimensional structure. This process can be used to produce high quality parts but the minimum feature size of the SLS processes is typically on the order of a few hundred microns, which is about two orders-of-magnitude larger than the feature sizes required for building true cellular materials. This feature size is typically set by the spot size of the laser beam and the heat diffusion from the laser melt pool. In order to achieve a feature-size resolution of approximately 1 μm in the μ-SLS system, several innovative design features have been implemented. First, the μ-SLS uses ultra-fast nanosecond lasers in order to achieve precise control over the heat-affected zone of the μ-SLS powder bed. The laser is coupled to a fiber optic lens and then directed off a micro-mirror array through a 10x, long working distance objective lens. This allows each 10.8 μm by 10.8 μm pixel in the micromirror array to be focused down to a spot size of approximately 1 μm. Second, the μ-SLS replaces the microscale powders used in conventional SLS processes with a nanoparticle ink. The use of nanoparticles in the μ-SLS system is necessary because in order to build layers that are approximately 1 μm thick, it is necessary to use particles that are at least one order of magnitude smaller than the desired layer thickness. The use of the nanoparticle ink also helps to prevent agglomeration of the nanoparticles during the powder-spreading process. Finally, a one degree-of-freedom nanopositioning system with a resolution of better than 100 nm is integrated with a slot die coating system in order to precisely control the thickness of the powder layer that is spread during the build process. Therefore, through the use of (1) ultra-fast lasers, (2) a micromirror-based optical system, (3) nanoscale powders, and (4) a precision spreader mechanism, the μ-SLS system is capable of achieving build rates of approximately 1 cm3/hr while achieving a feature-size resolution of approximately 1 μm.
In addition to the experimental μ-SLS test bed being developed in the NDML, new molecular scale models are being developed to quantify and certify the μ-SLS build process. Modeling of the μ-SLS process is challenging, because most macroscale models of the SLS process contain assumptions that are no longer valid when the size of the particles that are being sintered is smaller than the wavelength of the laser being used to sinter them. Therefore, in modeling the μ-SLS process we must account for the wave nature of light and can no longer rely on the ray tracing models commonly used to model the SLS process. Also, heat transfer in the μ-SLS process is dominated by near-field radiation due to the diffraction of the light off the nanoparticles in the powder bed and the ultrafast lasers that are used in the μ-SLS system. This means that the assumptions of heat transfer by conduction and far-field radiation in the macroscale SLS systems are no longer valid for the μ-SLS system. Finally, the agglomeration of nanoparticles in the powder bed must be accurately modeled in order to precisely predict the formation of defects in the final parts produced. The figure on the left shows a simulation of the formation of the nanoparticle powder bed, which aligns well with the measured particle distribution in the actual powder bed in the μ-SLS system. Overall, the goal of this modeling effort in the NDML is to be able to predict the quality of a part produced using any given processing conditions, in order to produce parts that are “born certified” and do not need to be tested post fabrication.
In the silicon exfoliation process, nanoelectronic devices are first fabricated on a standard silicon wafer. A nitride passivation layer this then grown on top of the nanoelectronic devices and a nickel layer is electroplated on top of the nitride in order to apply a compressive stress to the silicon. This stress initiates a crack approximately 10 µm below the top surface of the silicon wafer. This crack can then be propagated using a mechanical, wafer scale exfoliation tool developed in the NDML in order to produce a uniform 10 µm thick, extremely flexible, single crystal silicon film. This method is a cost effective and efficient method for creating high performance, flexible nanoelectronic devices. After exfoliation, the thin silicon films can be further processed to expose Through Silicon Vias (TSVs) and divide the film up into 10 µm by 10 µm chiplets for assembly into complex, 3D, hybrid electronic structures and devices.
One of the greatest challenges in the development of nanoscale sensor systems and other nanomaterials-based devices is our inability to precisely, accurately, and repeatably place the desired nanomaterials and nanostructures where we want them, at a high rate and a low cost. For example, carbon nanotubes (CNTs) have long been seen as potentially transformative material due to their outstanding electrical (current density >109 A/cm2) and mechanical (elastic modulus > 1 TPa) properties. However, very few carbon nanotube-based devices exist due to our inability to manufacture CNTs at a large scale with the desired properties on any type of substrate. One potentially promising method for overcoming these is the direct printing of carbon nanotubes onto substrates. To achieve this type of direct printing, a chemical vapor deposition (CVD) furnace with integrated positioning and sensing to controllably grow and place CNTs onto a substrate is needed. This tip-based fabrication method would allow different types of nanotubes (single walled, multi-walled, functionalized, etc.) to be directly and precisely placed on prefabricated structures at a low cost and high speed.
Carbon nanotubes (CNTs) may be used to create nanoscale compliant mechanisms that possess large ranges of motion relative to their device size. Many macroscale compliant mechanisms contain compliant elements that are subjected to fixed-clamped boundary conditions, indicating that they may be of value in nanoscale design. The combination of boundary conditions and large strains yield deformations at the tube ends and strain stiffening along the length of the tube, which are not observed in macroscale analogs. In this project molecular dynamics simulations are used to help model and design a nanoscale linear bearing system. These simulations show that the large-deflection behavior of a fixed-clamped CNT is not well-predicted by macroscale large-deflection beam bending models or truss models. However, pseudo-rigid-body modeling may be adapted to capture the strain stiffening behavior and, thereby, predict a CNT’s fixed-clamped behavior with less than 3% error from molecular simulations. The resulting pseudo-rigid-body model may be used to set initial design parameters for CNT-based compliant mechanisms. This removes the need for iterative, time-intensive molecular simulations during initial design phases.
Molecular dynamics simulations can be used to examine the mechanical properties of carbon nanotubes when both axial and lateral forces are applied to the CNT. However, this project shows that an elastic tube model of a (5,5) carbon nanotube predicts stretching and bending moduli that differ by 19%. This is due to (1) differing energy storage mechanisms in each mode and (2) the inability of the tube model to capture these effects. Conventional tube models assume a common energy storage mechanism in stretching and bending. They show that energy is stored primarily through bond stretching/rotation and bond torsion/van der Waals interactions in stretching and bending, respectively. This knowledge underscores the need to use different moduli to predict stretching, bending, and combined bending and stretching when using the tube model.
In this project a growth model was developed to control the diameter and number of walls in multi-walled carbon nanotubes grown by chemical vapor deposition (CVD). From this model, tunable process parameters may be used to control the geometry or stiffness of multi-walled carbon nanotubes (MWCNTs). This is important to devices that rely on the compliance of MWCNTs in order to achieve specific performance requirements, e.g. deflection or stiffness. Examples of these types of devices include relays, resonators and flexural bearings for small-scale actuators. It is necessary to control the stiffness of these mechanisms because the force, stroke, and device bandwidth depend upon the stiffness of the constituent MWCNTs. For a given length MWCNT, the stiffness is controlled by the MWCNT diameter and the number of walls in the MWCNT. The diameter and number of walls are controlled by adjusting several growth parameters – temperature, catalyst film thickness, and hydrocarbon concentration. Using thermodynamic relations, optimal growth parameters can be determined for CNTs with specific stiffnesses and natural frequencies. Overall, I was able to control the CNT geometry with less than 7 percent error and the stiffness and natural frequency with less than 2 percent error.
As mechanical devices move towards the nanoscale, smaller and more sensitive force and displacement sensors need to be developed. Currently, many biological, materials science, and nanomanufacturing applications could benefit from multi-axis micro- and nanoscale sensors with fine force and displacement resolutions. Unfortunately, such systems do not yet exist due to the limitations of traditional sensing techniques and fabrication procedures. Carbon nanotube-based (CNT) piezoresistive transducers offer the potential to overcome many of these limitations. Previous research has shown the potential for the use of CNTs in high resolution micro- and nanoscale sensing devices due to the high gauge factor and inherent size of CNTs. However, a better understanding of CNT-based piezoresistive sensors is needed in order to be able to design and engineer CNT-based sensor systems to take advantage of this potential. The purpose of this research is to take CNT-based strain sensors from the single element test structures that have been fabricated and turn them into precision sensor systems that can be used in micro- and nanoscale force and displacement transducers. In order to achieve this purpose and engineer high resolution CNT-based sensor systems, the design and manufacturing methods used to create CNT-based piezoresistive sensors were investigated. At the system level, a noise model was developed in order to be able to optimize the design of the sensor system. At the element level, a link was established between the structure of the CNT and its gauge factor using a theoretical model developed from quantum mechanics. This model was confirmed experimentally using CNT-based piezoresistive sensors integrated into a microfabricated test structure. At the device level, noise mitigation techniques including annealing and the use of a protective ceramic coating were investigated in order to reduce the noise in the sensor. From these investigations, best practices for the design and manufacturing of CNT-based piezoresistive sensors were established. Using these best practices, it is possible to increase the performance of CNT-based piezoresistive sensor systems by more than three orders of magnitude. These best practices were implemented in the design and fabrication of a multi-axis force sensor used to measure the adhesion force of an array of cells to the different material’s surfaces for the development of biomedical implants. This force sensor is capable of measuring forces in the z-axis as well as torques in the θx and θy axis. The range and resolution of the force sensor were determined to be 84 μN and 5.6 nN, respectively. This corresponds to a dynamic range of 83 dB, which closely matches the dynamic range predicted by the system noise model used to design the sensor. The accuracy of the force sensor is better than 1% over the device’s full range.
Si–Ti–N nanostructured coatings were synthesized by inertial impaction of nanoparticles using a process called hypersonic plasma particle deposition (HPPD). A study detailing the effect of plasma gases showed that in the case of an Ar+H2 plasma gas mixture, crystalline phases in the coatings consisted of TiN, TiSi2, and Ti5Si3. When the system was switched to a plasma gas mixture of Ar+N2, the only crystalline phase was TiN. Transmission electron microscopy confirmed the presence of TiN crystallites in an amorphous matrix. Warren–Averbach analysis indicated the average size of the TiN crystallites to be 14.9 nm. In separate experiments with the same conditions, aerodynamic lenses were used to deposit particles directly onto TEM grids. We observed agglomerated structures with an Ar+N2 plasma, while with an Ar+H2 plasma we found discrete TiSix and TiNx particles. In-situ particle diagnostics indicated only small changes in the particle size distributions when the plasma gases were changed.
The excellent electrical and mechanical properties of porous nanostructured titania (NST) make it an ideal material for many different applications including dye sensitized solar cells (DSSC). While DSSCs are cheaper to manufacture than silicon-based cells, they are also less efficient. The efficiency of the DSSCs depends on how well the sensitizing dye coats the cell and how well the electron recombination source infiltrates into the porous NST. Consequently, the structure of the NST has a great effect on the efficiency of the cell. Different processing parameters were studied to determine their effects on pore size and surface area. Overall, it was found that increasing the deposition rate and the thickness of the original titanium film increased the pore size of the NST. Also, it was found that increasing the concentration of the hydrogen peroxide used to oxidize the titanium, increased the pore size of the NST. In general, it was also found that as pore size increased, surface area decreased.