Devan Anderson and Professor Brian Jensen, Mechanical Engineering
Introduction
Carbon Nanotubes (CNTs) are grown on a silicon wafer that usually has had various metals deposited onto the surface. For many applications of CNTs it is desirable that the CNT arrays remain attached to the silicon wafer, however, thus far the CNT community has been unsuccessful in producing good CNT growth with strong adhesion to the silicon substrate. It is the purpose of these tests to better understand how we might produce good growth and get better adhesion.
There are many different recipes to grow CNTs. The problem of adhesion we thought would be best addressed by trying to deposit different metals to grow on. Titanium has shown to greatly improve the adhesion of CNTs but the growth has always been short and ratty. So the idea was proposed to grow CNTs off of a thin layer of alumina (which has consistently shown good growth) with a thick titanium layer underneath. After the growth process is complete an annealing process would be used to diffuse the titanium up through the alumina. It is hoped that the process would capture the good growth characteristics of alumina grown CNTs and the strong adhesion properties of titanium grown CNTs.
Methodology
Four different wafers were prepared: three wafers were created with a 100 nm coat of titanium followed by alumina which is varied in three thickness (6 nm, 16 nm, 30 nm) with a standard 4 nm top coat of iron, and one wafer with just 30 nm of alumina and 4 nm of iron. The wafers were diced to one cm square samples.
The pull tests simply needed to show whether or not adhesion had been improved. A simple force test would be used to measure the strength of the adhesion. Originally I used hot glue because I believed it would be strong enough and thought that it would not wick into the CNT structure, this did not work and is discussed further in the Results section, JB weld was later used which had its own problems, also discussed in the Results section.
I machined a mask with a hole large enough for the aluminum blank to fit through but not large enough for the sample to fit through. I used the mask to hold the sample down during the force test (The mask and force sensor are shown in Figure 1). I set the data logger to record the measured force every 0.02 seconds for 10 seconds.
Results
This first test (Figure 2) was on 16 nm alumina with titanium. Hot glue was used which, during the force test, separated from the test specimen leaving nanotubes intact. From this test all we know is that the force required to remove the nanotubes is greater than 2.7 N for a .25″ diameter circle.
For the next set of tests JB weld was used as the adhesive. Both a 16 nm and a 30 nm alumina with titanium sample were tested both yielding the exact same results (see Figure 3). The force gauge was maxed out at about 13 N for a 0.25” diameter circle before the glue separated from the wafer. A damaged wafer under the CNTs revealed that the JB weld had in fact seeped into the sample and bonded with the silicon making this data practically useless.
For the final set of tests extra care was used to try to stop the glue from wicking into the sample. Very little pressure was used and the samples were set to dry upside down (with the sample on top). Just as before a 16 nm alumina and titanium sample was tested and yielded similar results to those found in Figure 4. However, a standard 30 nm alumina with no titanium sample was tested, the results are found in Figure 5. The growth was long enough that the glue did not seep down to the substrate providing us accurate data. About 1 N of force was required to remove a 0.20” diameter circle.
For the next set of tests JB weld was used as the adhesive. Both a 16 nm and a 30 nm alumina with titanium sample were tested both yielding the exact same results (see Figure 3). The force gauge was maxed out at about 13 N for a 0.25” diameter circle before the glue separated from the wafer. A damaged wafer under the CNTs revealed that the JB weld had in fact seeped into the sample and bonded with the silicon making this data practically useless.
Discussion and Conclusion
More experimentation could be done with the 30 nm alumina with titanium sample. It exhibits growth similar to the non-titanium samples but does not seem to have improved adhesion; this could be because the titanium is not diffusing through during the annealing process or perhaps the benefit of improved adhesion only comes from growing off a titanium composite. I think doubling the anneal time for this particular sample would be a good thing to try because it would tell us if there is any benefit from the annealing process.
Although we continue to get bad growth on the 16 nm alumina samples adhesion was definitely improved. From the hot glue test (Figure 3) we can see that adhesion is at least 2.5 times better than the adhesion of the normal 30 nm alumina samples (Figure 4). It is likely that during the ramp in portion of the growth cycle that the thin 16 nm alumina layer is diffusing with the titanium so by the time growth occurs there is no longer a pure alumina layer to grow the CNTs on, hence the ratty growth similar to growth on pure titanium. More experimentation is needed to determine the optimum thickness of alumina to get good growth and good adhesion. Until better growth is achieved it is impossible to get good adhesion data for the 16 nm alumina and titanium sample.