Deposition temperature series
With the new heater installed in the AJA sputter tool, the production team decided to run a set of wafers at six different chamber temperatures to re-characterize membrane morphology, mechanical strength and overall quality of material. These films were all deposited with 25 W substrate bias and annealed at 1000 C in the RTP with no susceptor. I included a historic sample that was deposited at ‘450 C’ with the old heater. Below are the micrographs and histograms:
Overall, all six membranes have very similar pore morphologies. The historic sample most closely resembles the deposition at 300 C. Below is a chart plotting the differences in porosity and average pore diameter:
Because the morphologies are so similar, we will need to use burst pressure and membrane yield as a metric for optimization.
Spectroscopic ellipsometry was performed on the RT, 100 C and 200 C amorphous silicon films to verify the index of refraction. As a reference, the index of single crystal silicon is 3.63 @ 1.38 eV [Palik]. The index of the RT, 100 C, 200 C material was 4.04, 4.09, and 4.0 @ 1.38 eV respectively. The graph below shows the index values at all energies:
This confirms that our initial amorphous silicon film is of high quality (density).
It would be useful to know whther the density of the initial amorphous silicon films were the same- if they were, then I am not surprised by this lack of dependence- if they were not, then I am surprised. In any case, this is useful information!
I’ve updated the post and added ellipsometry data for the RT, 100 C and 200 C film. As you suspected, the initial densities were very similar.
I think there is something wrong with the Palik index data – single crystal Si at 1.38eV should be around 3.5. It has no effect on the conclusion, though. We have very high quality material.
I also think that the pore processing program is not capturing some important distinctions in pore morphology. If you look at SC172, there are clearly a collection of larger pores, some with elongated shapes that stand out as being considerably larger than the smaller pores. SC169 does not have as obvious a collection of larger pores, but the histogram looks similar. It would look better to me if SC169 did not have the one larger outlier on the histogram – can you check to see if this outlier is really there? Also can the pore processor give some measure of how circular the pores are? I think this may be important to some of the properties of these membranes, and there are considerable differences in pore shape between several of the temperatures.
Thanks!! This data is great!
Dave, you probably meant 3.63, not 1.63 for the index. But Palik has also data on various types of a-Si and I bet your index is on the high side- which begs the question: if the index of a-Si were lower, this would probably mean the density was lower, so after RTP, would the pore density be higher or the porosity larger or would we have more pinholes?
Oops, I meant to write n=3.63.
I think a roundness index would be instructive and I will look into adding this measure to the pore processor. Perhaps a ratio of the major:minor axis. By eye, it appears that the films deposited at lower temperatures look “nicer”.
I’ve also added a historic sample to the panel as a point of comparison. Our older material circa Dec 2008 had an index of 3.86 @ 1.38 eV and had a slightly tighter distribution. Although the difference in index between the old and new material is significant, I’m not really seeing this translate into a change in morphology. Although so much has changed since Dec that we might not be making a fair comparison…
You can ask Mike B. to add an measure of roundness rather than take the time yourself. Ratio of the major and minor axis is good. Actual circumference compared to circumference for the equivalent circle could also be helpful to detect and analyze the wavy boundaries that occur after media attack on membranes.
Can you remind me what the program is calling “pore diameter”. Is this an average of the major/minor axis or is it the diameter of a perfect circle with the same area, or something else?
Please check on the outlier of SC169 to see if that bar at 35nm should really be there, as it affect the interpretation of cutoff and size distribution. I don’t see a pore that looks like that big of an outlier, so I think the program is getting fooled by some of the bright diffraction spots.
Also, was the old image processed with the same procedure? The graph shows a porosity comparable to SC169, but by eye the porosity looks quite a bit lower in the old one. I wonder if the program is having difficulty with the extra background features in the newer films. Does the older film look less porous to anyone else?
Anyway, if we are going to make decisions based on these numbers, we need to make sure we fully trust the results. I know it’s less pretty, but maybe you could show the image with the counted pores outlined in red? Something like that is generated by the program, right?
Thanks!!
The pore diameter is calculated from the circle of equal area as the pore.
I’ve updated the micrographs to include what the pore processor identifies as pores. I cleaned up some of the diffracting features in Photoshop before the analysis and this seems to have shifted the histograms. It’s important that anyone who does pore processing be careful not to let diffracting features skew pore sizes.
Thanks. The histograms looks more reasonable to me and I see no more anomalies…