Flipped permittivities?

[benhills]

To start, fantastic work on this one, love the model and your paper.

I could be misinterpreting something here, but I believe that the direction on the anisotropy may be flipped. Lab experiments indicate that wave polarized parallel to the c-axis should have a greater permittivity (not smaller). See Fujita et al. (2000). Is it possible that I am misreading the notation used somewhere?

Are you still developing this model? I am planning to use this and your spectral fabric model for a paper I am writing on Hercules Dome. If the model is still in development, I may submit more PRs as I play around with it more. Also, I would love to chat Herc Dome results with you at some point if you are interested in them. They are confusing... as are all polarimetry measurements, at least to me.

[dlilien]

I read Fujita as saying the opposite. From the start of the second column on the first page: $\varepsilon_\perp$ and $\varepsilon_\parallel$ increase from about 3.15 to 3.19 and from about 3.12 to 3.16, respectively, as ice temperature increases. So, greater permittivity perpendicular to the c axis?

I am pretty sure that this code reproduces Fujita's plots, and we checked against my re-implementation of Fujita, so I tend to think that this model is ok.

Sorry for dropping the ball on looking more at the Herc Dome polarimetry...would still be happy to chat more on that.

[benhills]

Very weird, I agree with your reading of Fujita (2006), but see Fujita (2000) Figure 1. Tom Jordan's work is also written like $\Delta\epsilon = (\epsilon_\parallel - \epsilon_\perp) > 0$ and he is citing either Fujita (2000) or Fujita (2006) in those cases, or both.

I am only more confused now...

Fujita (2000):
https://eprints.lib.hokudai.ac.jp/dspace/bitstream/2115/32469/1/P185-212.pdf

Some Tom Jordan papers with the above notation:
https://ieeexplore.ieee.org/document/8755860
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2022JF006673
https://www.cambridge.org/core/journals/annals-of-glaciology/article/estimation-of-ice-fabric-within-whillans-ice-stream-using-polarimetric-phasesensitive-radar-sounding/F334B5CE907DBED94E7614EFB32E1DAF

[dlilien]

OK, yes, I just checked the experimental data in Matsuoka et al., 1997, and Fujita et al., 1993, and it appears that Fujita et al., 2006, has an error. $\Delta\epsilon=(\epsilon_\parallel-\epsilon_\perp)>0$. I admit I only checked against the 2006 paper, and I would guess Nicholas did the same...

I think @nicholasmr is at IPICs, but we should probably merge this in when he gets back and confirms.

[benhills]

Happy to write some tests or a demo or something to go along with this if the preference would be to do a bigger PR.

[nicholasmr]

Hi Ben,

I asked Fujita about this today at the conference, and it is indeed a typo in his 2006 paper. Since our model behaves similar to other implementations of Fujita's 2006 model (as David notes), you are probably right that eps_c and eps_a are flipped somewhere in the code. I'll look through this in detail maybe next week, else the week after that. This should, however, not affect model results as far as I can see.

I'm also happy to chat about your work/data, just let me know.

I'm glad that you find the model useful. In case you haven't already seen it, there is also a python notebook that demonstrates how specfab can be used in combination with this model.

[benhills]

Oh yeah it shouldn't really matter for any of the theoretical stuff you guys have done or for the fujita paper, just when trying to interpret measurements.

I am probably going to see what Nick Holschuh thinks about this herc dome stuff but then I will share with you two.

[nicholasmr]

I looked at this and could not find other places where these were swapped.
However, it seems not to make a difference when swapping the two permittivities expect for phases (e.g. the phase angle of $C_{HHVV}$).
I think this makes sense since it is the magnitude of the anisotropy, $|\Delta \epsilon|$, that determines the birefringence/nodal-structure in the anomaly plots (which are on a depth scale, not time).
I have pushed a new revision which fixes this issue, but will leave it open for a bit longer, just in case.
c73a2bd

[benhills]

Yes, I agree that this will not change any interpretation in your published work. The issue would come when trying to interpret the orientation of a real girdle, which would be rotated $90\degree$ in the case of flipped permittivities.