Several people have asked about a plot for relative energies, etc.
Here’s an example for an NEB trajectory.
My idea is that this “Plot Conformers…” dialog would allow you to switch between RMSD or relative energies. It would also offer a “Scale Energies” to convert from Hartree, eV, etc. to kcal/mol as needed. (In this case, the relative energies were in Hartree.)
And you can click on the plot to switch conformers.
Other ideas or suggestions?
And related - what are other properties to track (e.g., forces / gradients, velocities, etc.)
I think that it would be nice to track forces, although I think it would add quite a lot of complexity.
For example, do you plot the total force, the total force in the x, y, and z directions individually, the force on each atom, the force in each direction on each atom, or something else entirely?
I would really enjoy seeing all of those as options, but that adds a lot of complexity right off the bat.
Also, in general I feel like plotting introduces a lot of desire for customization (if you give a mouse a cookie…). For example in NEB plots like the one you have shown, I would really prefer a way to label key points, for example the reactant/product energies and the saddle point energy.
It has been a long time since I’ve used Spartan, but IIRC they had a pretty reasonable plotting interface. One thing I especially liked was that right next to the plot they had a table where you could see all of the energies listed out, and if you were plotting say, dihedral angle on the x-axis then you could see that listed in the table as well. I’m not sure how easy that sort of thing would be to pull off though since I am not sure if Avogadro can retain that sort of information across the individual geometries in an XYZ file.
For conformers it would be nice to be able to plot a Boltzmann distribution, since realistically a scatter plot or line plot doesn’t really contain all that much information.
Some of this is simply the need to add some tutorials.
The plot code already offers the ability to temporarily label:
It’s obviously not perfect - I’d like to add a “click to add manual label” mode for the spectra as well. At the moment these don’t export into a PDF or other image and the needs differ (e.g here you probably just want the y value to a few decimal places.)
In this particular plot, clicking will jump to that frame, so maybe a control-click or shift-click would add a manual label?
I think you want some sort of histogram of energies with Boltzmann weighting or something, which makes sense. Submit a feature request, but it’s definitely “after 2.0 is released.”
One of the changes with the AutoOpt tool is that measurements, etc. will persist across multiple frames. So yes, if you create a measurement of a dihedral angle, that could show up in a table. Certainly even now if you have this, you can display the distances across the frames:
I’ll set this down as a tutorial to add to the website.
As far as forces, it’ll take a bit of time to get some of the “plumbing” in place (e.g., reading it from files for each frame, etc.). But I can easily imagine wanting to plot maximum force or RMS magnitude or something over time. (As you said, I can imagine a lot of things you might want on the plot.)
Another program I use to visualise my calculation shows indeed both the energy and RMS force (which I personally like). They are overlapped in one plot with two different y-axis.
Seeing as it is an NEB you could also consider using a spline to connect the points. It is more of an visual choice but a cubic spline would porbably better present the PES than straight lines.
There’s an argument to be made both for and against spline interpolation for NEB plots.
The density of sampling on the potential energy surface is directly proportional to the probability of a continuous energy profile; for smaller calculations, say 5 points between reactant and product, there is not a guarantee of the PES being free of discontinuity and/or unresolved features.
I for example have been running many calculations of a sort of “PES” in my research, and if my sample number was lower than it is now I would absolutely not have a good picture of the trend. With how high my sampling step is though, I can nearly guarantee that a smooth, interpolated line is indeed the correct way to plot the data.
For right now, I’m going to stick to connecting lines. It’s something worth discussing for post-2.0 (e.g., identifying the type of multiple coordinate sets - NEB, IRC, conformers, trajectory, etc.)
I’ll put in some initial work to handle force and velocity magnitudes, but at the moment, it’s not a high priority. (i.e., just because the window can plot them, they have to be read in from various formats)
Thanks for this feature, really useful! Another improvement may be to set the X-axis value to other properties of the structure, such as a given bond/angle/dihedral (e.g., the one selected on the properties dialog when opening the conformer data dialog). This would unlock the possibility to show relaxed scans (e.g, by opening the allxyz file from an Orca relaxed scan run).
There’s an open feature request for other properties: Add new properties for conformers · Issue #2577 · OpenChemistry/avogadrolibs · GitHub
I’m open to suggestions on the UI. I’m not so sure about a simple selection - I think it would be better to require some sort of “add this to analyze” context menu - much like the constraint (and resulting lock icon) in those tables.
But I’m very flexible on how the UI would work best.
Similarly, I can imagine people might want something like a Ramachandran plot for clustering geometric parameters (dihedral, dihedral or bond, bond, etc.) in 2D
Hi, yes, something parallel to the constraint mechanisms sound a much straightforward way to proceed. I have not so much experience with C++, but I’ve already dug a bit in the code and will take a look to it when I have some time just in case I find a way. Thanks for the reply!
If you’re interested, I can certainly point you to some appropriate sections.
Also, while we’re very cautious with AI tools, they can be very helpful at describing what particular parts of the code do.
Right now, these additions would be on my agenda for after 2.0 is released (e.g., for this summer or fall).
I think the idea would be to track some sort of “analysis list” for properties to plot. I think now is a good time to talk about:
So please continue to post some suggestions and if you’re interested in the code, I can make some further suggestions towards implementation.
I’ve taken some time to compile a working version of 1.103 (the upgrade to Qt6 broke my previous environment), but I finally have it. I will explore the files that I already identified as relevant (mainly in the qtplugins directory), and of course, AI will surely give me a hand.
For the moment, I just made minimal changes to have a proof of concept (and start identifying issues to solve, such as the way the conformer is selected), and this is what I got:
I will be working on this not so fast, but I hope to have a working version ready for next year’s computational lab in my chemistry courses (by mid october).
Thanks again
Great! Let me know how I can help. I can certainly suggest some places (e.g., storing the list of properties to analyze)
Several people have asked to be able to scan a bond length or angle (i.e., picking the geometry and generating the individual conformers) which seems related.
If nothing else, I should have more time over the summer to help with this.
OK, thanks. I’ll reach out once I have something more concrete.
