Bioinspired nonequilibrium search for novel materials
- PDF / 2,878,348 Bytes
- 10 Pages / 585 x 783 pts Page_size
- 41 Downloads / 237 Views
Introduction Biology is not just a source of inspiration for novel materials, but also a source for novel ways of searching for materials in design space. After all, biology arrives at new functional materials through a process that is fundamentally different from traditional materials design on a computer. The evolutionary process in biology exploits numerous complex and competing effects, from population fluctuations and sexual recombination to temporal variations in the environment. Such effects together allow biology to explore parts of parameter space in a way that is differently biased than most engineering approaches. As a result, biological materials are often more robust, adaptable, and multifunctional than traditional materials design approaches. Further, biological search strategies are not predicated on the target material being near equilibrium or any other such simplifying assumption. In designing and optimizing materials, or in searching for novel materials properties, the idea of sampling a multitude of solutions in parallel and successively refining the most promising ones by letting them compete against a large number of variations offers many advantages, whether such an evolutionary process is applied to biological materials or not.1 When computers are used in this process, so-called genetic algorithms have long embodied some of the central features associated with bioinspired searches, such as finding solutions that are closer to a desired goal through a cycle of repeated recombination.2,3
Due to the specific construction of genetic algorithms, however, their success has been limited to certain classes of problems. This state of affairs changed in the early 2000s with the introduction of a new breed of evolutionary computation strategies of which the covariance matrix adaptation evolution strategy (CMA-ES) is the most prominent one.4 These evolution strategies have been found highly effective in applications ranging from identifying gaits for robot locomotion5 to crystal structure design6,7 to optimizing directed self-assembly of diblock-copolymer films.8,9 They also have been used as a tool to search for, and discover, novel properties in materials both in and far from thermal equilibrium.10,11 However, reaping the full benefits of bioinspired searches in materials design requires some thought. For example, evolutionary algorithms, applied in a black box manner to a materials design problem, may not always provide the best results (the term black box here refers to an application of the algorithm that is indifferent to its inner workings). In fact, biological evolution itself is known to face several strong constraints and does not produce the “fittest” genotypes under many conditions. For example, evolving a new function might require five different parameters to change, each of which lowers fitness individually. Such valley crossings in a rugged landscape can be difficult, unless evolution (e.g., the balance between mutation and selection) is operating in a regime where population fluctuatio
Data Loading...
