I'm afraid I will not be of much help. I looked into the question of dielectric absorption back in 1982 in order to write a report for a class on circuit theory. Over the years I came to realize that the circuit design community did not have a good understanding of the importance of dielectric absorption or how to model it. So I dusted off the old class report and typed it in, with the result being what you read. Other than writing this report, which I wrote after reading a few papers, I have never done any work in this area. My overall conclusion is that the Cole-Cole model is a very good one (very predictive), but that it is difficult to extract the parameters and difficult to create a simulatable model. I tried to address the second issue by writing the fracpole code (available on the website). However, this still leaves open the issue of the parameters. I do not have access to the measurement equipment that would be required to extract the parameters, so I too am constrained to looking on the web. By using Google to search for "Cole-Cole" and "dielectric" I found a large number of references, but nothing on SiO₂ (there was a lot on various body tissues, as well as things you might put in a microwave oven).

However, perhaps I can offer a tiny amount of help. My understanding is the epsilon_infty would normally be the free space dielectic constant, which would be 1 when working with relative dielectric constants, and epsilon₀ would be the accepted value for the dielectric constant for SiO₂, which I don't remember for sure but the value of 11.7 keeps coming to mind. As alpha approaches 1 the value of tau₀ becomes unimportant because the time-constant distribution spread is so wide. In my experience with discrete capacitors, it was possible to choose a value for tau₀ that simply placed it in the middle of the frequency band of interest, then choose alpha to fit the loss tangent data. This assumes that alpha is between 0.75 and 1.0.

Hope that helps.  Good luck.

-Ken