Tom Lehrer’s periodic table song “The Elements” may finally have competition for chemistry’s catchiest tune. Undertaking a playful experiment, chemist Oliver Steinbock of Florida State University used artificial intelligence to generate a playlist of thermodynamics-themed songs to help students remember key equations and concepts in this tricky topic—from a breathy rendition of the equipartition theorem (MP3) to a soulful explanation of the van der Waals equation (MP3) (Nat. Rev. Chem. 2026, DOI: 10.1038/s41570-026-00849-0; Suno playlist).
“There’s something between music and memory. Jingles and tunes are often stuck in your head for a very long time,” Steinbock says. Rhythmic mnemonics are nothing new to education and are indeed a common tool used by language learners to memorize vocabulary and repetitive structures. But the same strategy is rarely applied to the sciences—perhaps because of the perceived seriousness of the subject or a lack of creative confidence among science instructors.
But, by removing the need for musical skill and subsequent performance, AI democratizes this tool for learning, Steinbock argues. “At least for some people, this might then become an easier way to memorize or open up the subject.”
Steinbock began by instructing a large language model (LLM) to generate song lyrics for a particular thermodynamics topic, specifying both the tempo and style of the proposed song. After a few iterations and targeted tweaks, he was happy with the generated lyrics and copied them to AI music-generator Suno. “The AI doesn’t really make many scientific errors, but sometimes it takes too much liberty making it songlike, so you have to adjust it a little bit before generating the music,” he explains. After transferring the content to Suno, he simply toggled the style buttons and, with one click of a mouse, created a chemistry song in less than a minute.
The resulting music generally received positive responses from Steinbock’s colleagues and students—a mixture of curiosity about the educational potential and intrigue at the unusual lyrics. In particular, the LLM effectively parsed dense equations into simple and catchy lines such as, “T is how U answers S / When volume holds its breath,” spelling out the full equation T = (∂U/∂S)V over the course of the chorus.
“I think this is an absolute work of genius,” says Mark Elliott, director of undergraduate studies at Cardiff University. “You want the students to understand it, not just to remember it, and the key advantage of a tune is that you can’t miss a word. So if you’re trying to remember a series, each bit follows the previous in a rhythm, and you’ve built that connection with what came before and what comes next.”
Of course, this approach to learning won’t appeal to everyone, and Steinbock encourages teachers and students to experiment and see what works for them. “I can envisage an assignment where you take a difficult topic and try to make it into a nice song that is engaging, fun, and still conveys the correct science,” he says. “Students could compare and critique each other’s songs and verify the accuracy of the AI. That could be a formal assignment.” This approach would likely work particularly well in secondary schools, Elliott says, especially where students are given the creative freedom to generate their own lyrics first.
Steinbock now plans to test out the new playlist on his next cohort of thermodynamics students. Meanwhile, his novel workflow has set Elliott’s brain cells firing. “It’s so inventive and has inspired me to try something that I didn’t know could be done,” he says. “I can certainly see how this could be useful if you’ve got periodic trends of reactivity, for example. You could also develop a concept-album theme. Could you have a song for each of the first-row transition metals, in the same way that [Gustav] Holst did for the planets?”
