Nicholas Morrill and Dr. Robert Davis, Physics & Astronomy
Microelectric mechanical devices (MEMs) are structures on the micron scale that act as sensors and actuators. Devices such as these are used in familiar electronics such as iPhones and Nintendo Wii to sense motion. Current methods for MEMs fabrication are insufficient due to geometric design limitations, limited use of materials other than silicon, and cost. Research conducted at BYU has developed a new fabrication process overcoming these limitations and has already produced working devices1. This process includes patterned growth of Carbon Nanotubes (CNTs) on top of a sacrificial layer, followed by infiltration of appropriate materials and a release process enabling motion of the devices.
This process, however, is not yet perfected. The research done as a result of the ORCA Grant addressed the needs to further improve this new fabrication process regarding surface adhesion. The final step in the process requires a wet chemical etch of a sacrificial layer that releases the appropriate portions of devices from the surface, thus enabling their mechanical functions. Previous to this research, devices frequently came off of the surface during the release etch, rendering them useless, or they were so weakly attached that they broke off during operation. Research enabled through ORCA Grant funding discovered significant improvements that have hopefully solved this problem. Further testing is in progress to confirm this result.
There were three logical explanations for the poor adhesion: 1) stress due to differences in the coefficients of expansion of filler materials as compared to substrate materials, 2) reaction of hydrofluoric acid (HF) with alumina-silicon dioxide interface (interface under structures), and 3) poorly infiltrated devices causing loss of structural integrity where surfaces are anchored to substrate. The respective possible fixes would be: 1) matching the substrate with the filler material, 2) changing the sacrificial layer so that HF is no longer necessary, and 3) improving infiltration of filler material by adding fill-holes to large structures.
The first possible solution is the most difficult response, since the process would necessarily be different for every material; therefore, this possibility was held off unless it became necessary. The second possible fix was seen as the most simple. We tried changing from a PECVD silicon dioxide, which is less dense, to a denser thermal silicon dioxide. Pre-released structures showed a qualitative adhesion improvement, determined by sensitivity to microprobes before breaking, but no quantitative data was gathered since the structures still came off during the HF etch. This suggested that reason 2, the interfacial reaction, was dominant, and we therefore next tried solution number 2. There are limited choices for a new sacrificial layer, since it must support growth of CNTs, be deposited at least 1um thick, and withstand the high temperatures involved in the remaining parts of the fabrication process. One possible material selected through collaboration with various professors was chrome. In trying this, however, we found that the coefficients of expansion between silicon and chrome were too great and that the chrome delaminated from the substrate before thick enough layers were achieved. As a result of this, we looked more closely at solution 3.
With an optical microscope, we viewed the underside of several devices that improperly released from the surface after the release etch was completed. This happened often when probes lightly pushed on the devices and broke free from the surface. The underside view revealed that the inner area of the large structures was broken open. Behind this removed area were CNTs that appeared to have little filler material on them. This strongly suggests that the reason they broke open was because the area that was attached to the surface was largely an empty void of CNTs and no filler material, which is weak as compared to the scale of these devices. This motivated the development of a new lithography mask enabling the addition of fill-holes to the larger portions of structures necessary for anchoring the devices to the substrate.
Our most recent research was a first attempt to release a group of devices for operation, using thermal oxide with the newly designed devices with fill-holes. The first set released smoothly, enabling the operation of devices that could not be used before. The anchor pads held on remarkably better than in previous attempts. More devices are being fabricated at this time to confirm this result. Also, issues with bubbles forming on the surface during the release etch are causing a masking effect, which causes uneven etching. We will be applying a surfactant to the etch chemicals to see if this problem can be solved.
Based on the research, it seems likely that the poor adhesion was a result of all three factors in the hypothesis, as was seen through the effect of switching to thermal oxide, including fill-holes, and the behavior of different filler materials with more extreme differences in coefficient of expansion. We hope, however, that poor filling was the dominating factor and that, when properly controlled, it will override the importance of the remaining factors. In other words, if further testing reveals that proper filling leads to successful release, we will be able to ignore problems with the HF interfacial reaction and stress due to different coefficients of expansion. Future work will confirm the success of the new mask by repeating the experiment, and by analyzing the effect of the addition of surfactants to the release etch.