Why not publish it and use the peer-review process as part of your dissemination strategy and sourcing unexpected collaborators?

Open Research Software (Ubiquity Press) might be an option.

Hi Ivo ! Actually we have published a paper, a few years ago, where the principle of the code is described - and the code is given with the paper, although you have to find the link in the pdf to find it (sigh). But... Your suggestion is excellent. Our plan is to publish more and more of our codes so that they can be reused. the journal you mentioned seems perfectly suited for that purpose. Thank you very much !!!

Hi Antoine,

As software journals now evolve to fill the niche, a link in a standard paper leading to a code, can now be converted into so much more, all of which fit the "publish or perish" needs of young researchers: 

- code leads to a DOI publication;

- minimum metadata and standards are checked;

- code is archived long-term;

- versioning is possible;

- code evolves wihtout your input;

- reproducibility & collaborations are multiplied.

Keep me posted if you submit, I need success stories for training young researchers to use Open Science principles to support their career objectives.

Regards

Antoine, wow! I'd be very eager to try your code! Before that, couple questions:

(1)  Do you have a version in FORTRAN (say gfortran or fortran-90)? -- or even better, in ratfor90? (I am ashamed to admit it, but I am a dinosaur and inh my old age:) still use gfortran and ratfor90 in my LINUX machine...)

(2) Have you tried your code to look at nonlinear reflection (say for a medium with nonlinear refractive index?   There could be multi-stable and hysteretic solutions there, see e. g.  

 http://psi.ece.jhu.edu/~kaplan/PUBL/AEK.pubs/RUSS/19.pdf

for first theor-prediction,  and a review with experimental results in attachments) as well as multi-histresises  in nonlinear resonators,  http://psi.ece.jhu.edu/~kaplan/PUBL/AEK.pubs/133.pdf

I am curious how your code would handle muti-stability... In my last paper on the subject (133.pdf) I used an "inverse propagation" trick for plain waves to getr multi-hystersises, but would be happy to look into any other  numerical technique , especially if it would manage to handle the case of limited size beams (e. g. diffraction and related stuff)  -- I still don't know how to do it in the presence of nonlinear reflection...

Best, Alex Kaplan,   http://psi.ece.jhu.edu/~kaplan/

BTW, when you said "We are open to suggestions and many other things :)" -- did you include a possible collaboration?

Dear Alexander, thank you very much for your answer !

As long as you're using Linux, you don't seem like a dinosaur to me. Whatever the distribution, you should be able to install Octave on your computer and play with the program - for we don't have a Fortran version right  now, alas. I do program in Fortran, though, but I use Octave (Matlab equivalent) on a regular basis for programs that don't require much computational power. Mastering Octave is usually extremely quick. It's just really easy compared to Fortran (all the codes I used during my PhD were in Fortran, too).

And you got the message. We are open to collaborations :) I am currently generalizing this code to account for nonlocality in plasmonic structures. I will publish the result quite soo (well it is submitted, and will soon be on arxiv, just the time to validate it, I guess).See attached file.

I don't know how easy it would be to take nonlinear effects into account - since it's harmonic and based on a plane wave decomposition. On the other hand, if you know what happens to a plane wave (as your publications suggest), you should be able to see what happens for a beam. I don't feel skilled enough to do it right now alone, but I can definitely help you for that.

Best,

Antoine.

Ivo, I'll keep you updated. I have plenty of ideas of codes to publish that could really be of interest to a quite wide community. This journal is what I was somehow looking for. But each time I had to do really new physics with my codes to be able to publish it. The physics was the main reason the paper got through, each time.

Plus, I really do want to make Science in new ways. Something more collaborative and effective than the regular way we publish our results.

Thanks !

Antoine.

PS: My ReseachGate hack is a success it seems (I'm frustated it's so difficult to share something else than just publications). If you have something to say, just ask a question !

Antoine, thanks! OK, I'll try to learn Octave, or more likely Matlab (all the people around me switched to Matlab, so at least there would be somebody to answer my questions, sigh...) Nonlocality is a large step forward, good you've done it; people need it for many things.

Nonlinearity -- well, of course, it is a huge zoo (I've been doing it for 50+ years, still find exciting things to do...:-), but there is a big sub-field, where you can ignore high-orders harmonics, and look only into a so called nonlinear index of refraction (or intensity-induced index of refraction), whereby the index depends on light intensity (usually due to so called chi^3 effects) and you may neglect generation of 3-rd and other order harmonics. It is a very generic effect, happens in many materials (air including), no phase-matching conditions  required, and it is responsible for many other effects, such as self-focusing and  self-bending of light, soliton formation e. g. in fibers, pulse-shortening, hysteretic effects, ets. Many people would be happy to have an easy adjustable code to use for those effects....

OK, I'll try to learn Matlab, and if I do, will try to see on simple examples whether nonlinearity could be easily incorporated. My problem at this point is that I am heavily engaged in a few concurrent projects, and it is unclear to me when I would be able to put my hands on Matlab (or the other way around:-), although I will need it for one of the other projects too. So, if you have a Fortran version (or, by a tiny yet lucky for me chance a Ratfor90) -- please let me know.  On my side, I can promise you that if I ever use your codes -- I'll automatically go into collaborative mode with you.

Thanks again; and good luck with your research.

Alex Kaplan

Oh, about Linux: see I was first exposed to computers in 1982 (and it was long after start of my research carrier), it was UNIX (at Purdue); and I became a UNIX person (just  a dumb user, mind you...) -- such a beautiful system... and after that the only thing to do was to switch to Linux...

I've been thinking about how to extend what I had done to nonlinear media. If I'm right (but this doesn't mean it's doomed to fail), the index will then depend on the intensity of light locally, right ? Which means that the layers will no longer be homogeneous ?

The code relies on such an assumption. But we may get there, still, because we have other methods (more complicated, I admit) that could allow us to take all of that into account. But I'm not a specialist of nonlinearity, as you may have guessed (more in love with poles and zeros and the complex plane and Green's functions :).

That may be a subject for a (good) intern, though, I guess. We will probably continue our discussion by email... And definitely, writing this code in Fortran would not be an issue, it was actually in my projects some times ago. I could totally do that.

Antoine.

Hi,

source code is quite straight and comprehensible. So I might give a try porting it to python (upd: just found http://ompc.juricap.com). But I have a couple of questions:

How the field distribution of wave/beam is calculated in demo images? I mean is it the direct result of code or it appears as a result of postprocessing (e.g. by multiplication of  planewave solution with the profile of gaussian beam). In short - how amplitudes and phases in demo images were calculated? I can not run code right now and figure it out.

How would you compare it e.g. to Empy (http://lbolla.github.io/EMpy/) in terms of calculating optical properties of pure multilayers (no gratings, rings etc)?

Hi Shavkat !

Empy is very similar, except that transfer matrices are not as stable numerically as scattering matrices. The drawback is that the matrices are more complicated to handle (there is a cascading process instead of just multiplying the matrices). Btw there are actually two kinds of "transfer matrices", I could'nt really determine which one is used in Empy. Both have stability issues in the case of large number of layers or when evanescent waves are involved.

As for the computation of the field, we do compute the field in every layer for every plane wave that composes the incident beam, as you have guessed. Then we sum up the whole thing to get the image in each layer. Maybe it's what you call post-processing...

I do have a RCWA code in matlab, by the way. It's among the codes I'd like to release soon.

Anyway, I'm very glad to see that you consider the code interesting enough to try to translate it into Python - I would be totally unable to do so myself, and I think it's a good idea. Keep me updated !

Best,

Antoine.

Update: the code has evolved and the paper is almost ready to be submitted to JORS. I'll keep you updated (I'll probably ask another fake question at some point on Researchgate :)

Yay ! Our paper is out. It took quite a lot of time, but there it is. I definitely recommend JORS for publishing papers about your code. It really forces you to make things straight. Fake question will follow...

http://openresearchsoftware.metajnl.com/articles/10.5334/jors.100/