We developed a new in-browser data analysis tool for educational purposes called tinyplot. You can try it out here: margrafgroup.github.io/tinyplot/

It uses pyscript, so the first time you use it, there's some overhead of importing the underlying libraries.

I'd be really interested in feedback!

A desert

As a rule of thumb: where are the articles you cited published? That's what I would aim for.

Avogadro 2.0 drops April 1st — and no, that's not a joke! 🧪⚗️

Now's the perfect time to help improve translations before release. Every language matters for chemistry worldwide.

#opensource #openscience

hosted.weblate.org/engage/avoga...

Thanks, such a cool conference!

A very good read by @philipcball.bsky.social

Blue abstract picture

Blue

It consistently and confidently suggested wrong or outdated syntax, to the point that it was slowing the process down until I abandoned it. I found the difference between the performance for html/js and pyscript really interesting. Not surprising as such, but unexpected in magnitude.

But then, for some numerical stuff, I wanted to use pyscript. This is a need library that lets you run python in the browser via web assembly. It's fairly new, and the API for interacting between python and js has changed pretty significantly between versions. Here the llm was completely useless 3/n

The experience was interesting but pretty mixed. The thing that worked really well was coding up the front-end of a web app in html and javascript. This ran out of the box from a pretty minimal prompt, and the design choices of the llm were quite nice. 2/n

I did my first bigger coding project with an llm today (copilot w gpt-5). Usually, for the things I do for research, writing the code is not the time limiting factor, and I do it in python, which I know well. This was for a teaching project involving javascript and html, which I barely know. 1/n

Photograph of David Waldeck - in the background is the University of Pittsburgh's Cathedral of Learning

I share the sad news that Prof. David H. Waldeck passed away after a battle with cancer. Dave was an extraordinarily creative scientist, a devoted educator, and to all of us a friend, mentor, and leader.

He will be deeply missed

#chemsky #chemchat

We are looking for two talented people to join us at the Department of Chemistry at the University of Warwick as assistant professors. Permanent positions. 🧪🍎🥽
warwick-careers.tal.net/vx/appcentre...

Local Symmetry Breaking in 3D Hybrid Perovskites with 3-Hydroxyazetidinium Two-dimensional (2D) hybrid perovskite semiconductors with nonprimary ammonium cations (NPACs) have recently attracted interest for spin-optoelectronics owing to the symmetry-breaking distortions in their crystal structures. However, implementing this design strategy in three-dimensional (3D) analogs remains largely unexplored, primarily due to stricter restrictions on cation size. Here, we introduce a family of 3D hybrid perovskites (3DHPs), ASnX3, where A = 3-hydroxyazetidinium (AzOH; +NH2(CH2)2CHOH) and X = Cl, Br, and I. The choice of a bulky NPAC, coupled with the incorporation of a polar group (−OH), targets broken symmetry within the 3D frameworks. While single-crystal X-ray diffraction analysis reveals spatially averaged centrosymmetric cubic (i.e., Pm-3m space group) unit cells, with the largest lattice parameters among existing ASnX3 analogs, first-principles molecular dynamics and electronic structure calculations indicate that the relatively fixed dipoles of AzOH within the SnX6-derived framework introduce local inversion asymmetry and spin polarization. By incorporating a polar cation with limited mobility into the 3D perovskite framework, (AzOH)SnX3 unlocks potential for spin-optoelectronics, photovoltaics, ferroelectrics, and nonlinear optics.

Here is Rayan Chakraborty's masterpiece
doi.org/10.1021/jacs...
showing some new 3D halide perovskites. That's very rare! Again, thanks to Department of Energy, NSF and Duke Climate Initiative. Oh, and a modest 3672 atom DFT simulation probing the local order (along with many other simulations)!

Really cool stuff!

Blame it on the character limit. Wagner is too epic for 300 characters ⚔️

Job advertisement of the University of Bayreuth  - Chair of Physical Chemistry V: Theory and Machine Learning Fixed-Term Research Associate (m/f/d, E13 TV-L, 67%) in the field of Machine Learning for Materials Chemistry

📢 We're hiring a PhD student in ML for Materials Chemistry at Uni Bayreuth! Apply if you're into atomistic simulations and the intersections between physics, chemistry and ML.

We offer a great environment in one of the nicest towns in Germany. Apply here: bit.ly/4aPBgE9.

#compchem #chemsky

Abstract blue and yellow picture with some bubbles on the right side.

Bubbles 2

Yes, I'm pretty happy with the feedback so far. Lots of people I remember from Twitter, some I know in person.

Not even close for me (just JPC).

Thanks, nice example! With the hype around generative ML models, the number of candidates for 'designed' materials of course increased substantially in recent years. My feeling is that the number of candidates is not necessarily the limiting factor, though.

That mix of disappointment and relief when weekend plans get canceled at the last minute. I'm sure there's a German word for that.

(I'm German. There isn't.)

Funnily enough, these are certainly not the most sophisticated calculations or ML models we have used, but the key was to be in exchange with experimentalists from an early stage in the project, to ensure that we are predicting things they actually can and want to make.

Predict before You Precipitate: Learning Templating Effects in Hybrid Antimony and Bismuth Halides Hybrid organic–inorganic (HOI) antimony and bismuth halides exhibit diverse structural features and have been studied intensely for their promising electronic and optical properties. There are well-explored structure–property relations for these materials. However, a thorough understanding of the synthesis routes and templating effects is lacking, turning their targeted synthesis into an open challenge. In this study, we assemble a literature data set of established HOI material candidates and train an explainable machine learning classification model to explore the templating effects in more detail. With a classification accuracy upward of 70%, our model is effective in predicting HOI structure types based on the reactants and points out several structural and electrostatic design features for the organic cation that influence the inorganic substructure most strongly. We further demonstrate the validity of our classifier on 9 newly synthesized members of this materials class and propose incremental learning routes to expand the model in future research.

I would say our own clearest examples of this is this work with @jabla.bsky.social and @lotschgroup.bsky.social: pubs.acs.org/doi/full/10.... and this one with Stefan Hecht: pubs.rsc.org/en/content/a...

Thanks, this is a nice example. I get the impression, that this is still pretty rare. Not because the methods are not accurate enough, but because the connection between computational and experimental groups is missing.

What's your favorite example of the computational discovery/prediction of a new material? #chemsky

Abstract picture that looks like bubbles are leaving a coarse cube downwards.

Bubbles

Hi #chemsky,
I have been asked to come up with chemistry ideas for a new teaching format at german schools that focuses on democracy.
What's lacking are ideas how to use chemistry in this context.

So what could be the chemistry of democracy and how do you present it in a school lesson?

Hello bluesky, I'm a theoretical chemist at University of Bayreuth. My group works on atomistic modeling of energy materials with electronic structure methods and ML. I'm hoping to connect to other #compchem and #materials folks!