Bay-functionalized perylene diimides combined with time-series classification for rapid trace detection of illicit drugs

Luminescent sensor arrays combined with machine learning have been demonstrated as a promising strategy for rapidly and selectively detecting chemicals of interest. However, the real-time identification of analytes over a wide concentration/quantity range remains a challenge for the field. Here, we report a family of perylene diimide derivatives bearing different bay substituents, which were selected to modulate their ionization potentials for the detection of illicit drugs. Thin films of the perylene diimides exhibited different changes in their luminescence when exposed to illicit drugs and commonly used medicines. These differences in luminescence change were used in conjunction with a time-series classification algorithm— k -nearest neighbors with dynamic time warping to show that a combination of sensing materials could be selected to allow differentiation between five illicit drugs (methamphetamine•HCl, 3,4-methylenedioxyamphetamine•HCl, cocaine•HCl, fentanyl•HCl, tetrahydrocannabinol) and three non-illicit drugs (aspirin, paracetamol, and caffeine) with a mean accuracy of 94.1% after 17 second exposures to the analytes for amounts ranging from 0.1 μg to 10 μg.