Chemistry on Computers
One of the most lengthy components of chemistry is all that trial and error to find out exactly how to make the chemical you want, or even just to assess the chemical properties once you have made a chemical. The problem is, there simply isn’t a way to predict how a molecule will behave. Until now. Computer scientists have recently developed an algorithm which can predict certain features of a molecule based off other examples. While this doesn’t give exact answers they found that by using 1000 examples they could calculate the atomization energies of over 6000 other molecules to a precision of 1%. The major problem that had been facing this “designed” approach had lay in the complexity of chemistry equations, most notably the Schrödinger equation which can only be solved efficiently and accurately for 1 electron systems. Instead this Machine Learning approach, which has been used in stock markets for years, gives approximations accurate enough to find what you need to know. As such, research that previously took years may be able to be completed in a fraction of the time.
Image
