Water in Crystals : A Database for ML and a Knowledge Base for Vibrational Prediction

• Hydrate crystals are excellent reference systems to learnaboutaqueous systems. We have created a database of density functionaltheory (DFT)-optimized (optPBE-vdW) structures and vibrational frequenciesfor 101 crystalline hydrate and hydroxide bulk systems and over 300unique oscillators and use it to explore and discuss the tradeoffbetween prediction accuracy and insight. Starting from a machine-learninggeometrical descriptor, we gradually include more physics/chemistryflavor in the descriptor and examine how the frequency predictionpower varies. The most accurate models are the machine-learned model(of modest insight) and a physically motivated model containing theelectric field and field gradient. Furthermore, detailed comparisonswith experimental correlations show that, where available data exists,our DFT results largely overlap with the experiment. A small blind-testshowed that our machine-learned (ML) descriptor model can be usedto predict experimental vibrational frequencies based only on theexperimental structures and our best-regressed model, with encouragingresults.

Ämnesord

TEKNIK OCH TEKNOLOGIER  -- Elektroteknik och elektronik -- Reglerteknik (hsv//swe)

ENGINEERING AND TECHNOLOGY  -- Electrical Engineering, Electronic Engineering, Information Engineering -- Control Engineering (hsv//eng)

Publikations- och innehållstyp

ref (ämneskategori)

art (ämneskategori)