The `photometric selection' approach as a high throughput, sample tolerant, low cost and highly automatable method of carrying out serial crystallography is presented. Crystalline samples are loaded and distributed onto a simple transparent substrate and an in-line camera identifies crystals using image recognition algorithms from the computer vision project OpenCV. In contrast to established serial techniques, which generally require that crystal samples be refined with narrow size distributions and defined habits, the sample requirements when using photometric selection are shown to be minimal. We demonstrate how broadly effective photometric selection can be by collecting high-quality datasets from three exemplar systems: a small-molecule organometallic, a small-molecule organic and a metal-organic framework system. In contrast to previously established grid-scanning techniques, data collection using photometric selection can be up to six times faster.
Coulson et al. (Fri,) studied this question.