how can I get the same curve?

[fanjiaping]

dear all
     I want to reproduce the curve(paper.jpg) from the previously reported literature, but unfortunitily,  my calculated curve(123456789.jpg) is quite different from the one in that literature.
Especially in the region from 0 energy to -2 energy, there is no transimission in the curve from that paper. However, in my calculated curve the transmission is obviously not zero in the range from 0 to -2. I have adjusted the number of electrod lays, but it doesn't work. So I need your suggestion on how I can  adjust other parameters to obtain the identical to the one in that paper.(I almost use the default parameters!)

[zh]

Obviously, use the many parameters mentioned by literature as you can. Otherwise, it is impossible to reproduce the results of other people, because the slightly difference in the geometry structure of two-probe system could lead to different transmission spectrum.

[fanjiaping]

how can I get the similar transimission as mentioned in the literature? The only thing I can do for getting the similar transimission curve is searching for the same configuration? if one obtains a transimission, which can't be reproduced unless they use the same configuration, how can we believe the software and the results?

[zh]

This is not due to the software. Mostly, when some people presented the computational method and model in their paper, they did not give everything of these method and model. If you want to reproduces the results published by other people, the model in your calculations should be very similar to (same as) theirs. But now, you did an reverse strategy, i.e., judge the rationality of your model by comparing your obtained transmission spectrum with the one of others.

To reproduce other people' results, you may do it in the following way:
1). use the same computational method and code; if this can not achieved, the important parameters affecting the accuracy should be same as those used by other people, for example, the exchange-correlation functional.
2). the same procedure to construct the model (i.e., the geometry structure of calculated system). For a two-probe system, the geometry structure of interface part, i.e., the molecule + surface layers, is one of most important factors affecting the obtained results.
3). tune other parameters in the computational methods.

[fanjiaping]

Thank you for kindness !

[Anders Blom]

There is a quite important and serious point here: how to reproduce the results obtained by another group, using the software. This is not a trivial matter.

It is undoubtedly the case that most articles published on simulation results do not contain enough information to allow someone else to in a straight-forward manner replicate the calculation. Most significantly, the exact geometry is typically not specified.

First of all, this is not an issue which is exclusive to ATK, it applies equally to all simulation codes, and in fact also to experimental results. For experimental work, however, there seems to have appeared - for exactly the reason of reproducibility (and cases of fraud, to be honest) - an agreed upon set of specifications, which should allow for another group to repeat the experiment under equivalent circumstances. I think that a similar agreement is needed also for simulations.

Many journals today provide possibility for the authors to publish "supporting material", which would be an excellent place to specify the exact geometry (why not a text file with the coordinates?) and the parameters used in the calculation. In fact, why not the entire Python script?!

Perhaps users of ATK could lead this revolution!

[fanjiaping]

Thank you for your patient guide !