New research from MIT explores fire from a whole series of new perspectives. The research uses deep-learning approaches that extract the vibrational features of flames as flickering objects and renders them into sounds and materials.

The 19th-century physicist Michael Faraday was known for his contributions to electromagnetism but also for his public speaking. His annual Christmas lectures at the Royal Institution evolved into a holiday tradition. His most famous Christmas lecture was about the chemical history of a candle. He showed how a candle inside a lampglass could be used to block out breezes and create a quiet flame.

The flame of that shape is never the same at any one time. You can't see a body of flame like the one you saw rising from the ball. It consists of a lot of different shapes, succeeding each other so quickly that the eye can only notice them all at once.

MIT researchers brought the simple experiment into the 21st century. A single candle flame was sonified by the combined high-resolution image and deep machine learning of Mario and Markus Buehler. They used that single flame as a basic building block and then used it to create music and design novel structures that could be 3D-printed into physical objects. At the American Physical Society meeting in Chicago last week, Buehler described this and other related work.

The dynamics of a flickering candle flame. Researchers use deep learning to first explore what the vibration of a single flame sounds like and then generalize the approach to a larger fire that creates a variety of sounds.
Enlarge / The dynamics of a flickering candle flame. Researchers use deep learning to first explore what the vibration of a single flame sounds like and then generalize the approach to a larger fire that creates a variety of sounds.

Buehler specializes in developing artificial intelligence models to design new proteins. He is known for using sonification to illuminate structural details. The Hierarchical elements of music composition are similar to the Hierarchical elements of the structure of the human body. Music has a limited number of notes and combinations that can be used to create different types of music, similar to how a limited number of building blocks can combine in any number of ways to create novel structures with unique properties. A sound signature is similar to a fingerprints.

Advertisement

Several years ago, Buehler led a team of MIT scientists that mapped the structure of proteins in spider silk threads onto musical theory to produce the sound of silk. The goal was to create a new way to make designer proteins. That work inspired a sonification art exhibit in Paris. MIT engineers collaborated with artist Tomas Saraceno to create an interactive harp-like instrument inspired by the web of a spider, with each strand in the web tuning to a different pitch. You can combine those notes in different patterns on the web.

Buehler's team developed an even more advanced system of making music out of aProtein structure and then converting the music back to create novel proteins not seen in nature. Synthetic spiderwebs and other structures that mimic the spider's process were the aim. The same approach was used in 2020 by Buehler's team to model the spike protein responsible for the novel coronaviruses.

Machine-learning-rendered image of a flame and its 3D-printed fabrication.
Enlarge / Machine-learning-rendered image of a flame and its 3D-printed fabrication.

He wondered if this approach could be expanded enough to study fire. Fire has all the elements of a vibrating string or vibrating molecule but in a dynamic pattern that is interesting. What would they sound like if we could hear them? Can we make fire? Can we push the envelope to create bio-inspired materials that you could actually feel and touch?

The simple experiment involving a single candle flame was similar to the one that happened centuries before. A single flame can be seen as a basic building block of fire, even though a larger fire will have too many perturbations. The researchers lit a candle in a controlled environment with no air movement. They played sounds from a speaker and used a high-speed camera to record how the flame flickered and changed over time.

Advertisement
Assembly simulation of flames into a princess in a fairytale garden.
Assembly simulation of flames into a princess in a fairytale garden.

There are characteristic shapes that are created by this, but they are not the same shapes every time. There are thousands of images for each expectation of the acoustic signal.

They trained a neural network to classify the audio signals that created the flame shape. The frequencies of fire were sonified by the researchers. The audio signal changes dramatically when a flame moves. The flame becomes a kind of musical instrument, which we can play by exposing it to air currents, in order to get the flame to flicker in particular ways.

The definition of music is that of fire being repetitive and vibrating, and with different patterns in fire.

The various shapes of flickering flames are used as building blocks to design novel structures on the computer and then 3D-print those structures.