4 ms·
Inverting the structure–property map of truss metamaterials by deep learning
- ur-whale 5y agoThe should be a scale of 1 to 10 for how obscure a HN post title is (probably automatically computed by some deep NN). This one would likely take the cake, especially if you consider that actually reading the article itself does not help very much.
- atoav 5y agoReally? The paper was more or less precisely what I expected and I don't even have a ML (or engineering) background.
- Isamu 5y agoDefinition: A metamaterial is any material engineered to have a property that is not found in naturally occurring materials. FTA: for applications from wave guiding to artificial bone.
- kragen 5y agoThat Wikipedia definition is not a good definition because it would include things like pure hydrazine, high-speed steel, and cubic boron nitride as metamaterials, which they are not. A better definition is that a metamaterial is a material whose properties come from a highly ordered structure rather than from its composition. Negative refractive indices, negative Poisson's ratio, solids with lower densities than aerogels, stuff like that. Usually things like foams, fiber-matrix composites, and gold nanoparticle pigments are excluded because they're too disordered, and things like wood, limpet teeth, nacre, and bone are often excluded because they're natural, though sometimes they are included. In this case the particular properties they're trying to achieve are found in natural materials, just not the ones they're making the trusses out of.
- wefarrell 5y agoWould graphene nanotubes be considered a metamaterial?
- smaddox 5y agoIsolated carbon nanotubes, probably not. But certain arrangements of a large number of carbon nanotubes, almost certainly. Vantablack should almost certainly be described as a metameterial.
- kragen 5y agoPeople usually don't call things like Vantablack metamaterials because they're too disordered. Also that would commit them to calling ordinary lampblack a metamaterial, and lampblack isn't cool.
- smaddox 5y agoA usefull word for defining metameterials, at least in the field of electromagnetic metamaterials is "mesoscopic". Perhaps we need a generalization of this term to arbitrary length scales. Electromagnetic metameterials are composed of nanometer-scale structures that alter the angstrom-scale (i.e. atom-scale) behavior of pure materials. These are often described as mesoscopic structures---structures that span the length scales between atoms and traditional fabrication scales. Some great examples from nature include Beatle carapaces and the pads of gecko feet.
- kragen 5y agoThose are excellent examples, though there's some debate about whether iridescence and rugate filters and stuff like that are "metamaterials". I think it's partly a question of wanting metamaterials to be "new and hot", which the Beatles are not, however melodic their carapaces. You can get macroscopically-structured electromagnetic metamaterials, though; some are pictured on the Wikipedia page. It's just that they're active at microwave wavelengths or longer instead of visible wavelengths.
- wantsanagent 5y agoTLDR: Imagine you have a bridge with well known stress characteristics, loads, etc. Now you want to run an optimizer to produce a lightweight bridge made out of a material that uses repeating structures throughout. This exists today but takes a long time to run, you can get all sorts of interesting material compositions out of such programs. Instead of doing the above train a deep neural network to predict from a set of known stress properties of shapes made out of different types of materials to a set of the structures of those materials. (Massive over simplification: should I use pyramids or cubes as my basic building block for a material?) After training on millions of known examples, the network is able to take in a desired set of stress characteristics and produce the material which will have those characteristics, very quickly. They seem to be particularly interested in predicting materials to make up bone replacements.
- thunkshift1 5y agoThanks