Article Semi-Automatic Elemental Identification of Laser-Induced Breakdown Spectra Using Wavelength Similarity Coefficient I. Rosas-Roma ´n, M.A. Meneses-Nava, O. Barbosa-Garcı ´a, and J.L. Maldonado Abstract This work proposes a method to perform elemental identification on plasmas produced using the laser-induced breakdown spectroscopy (LIBS) technique. The method is based on the preservation of the relative relevance of the spectral line emission intensities, which is lost during the parametric correlation procedure, by the introduction of a similitude coef- ficient called wavelength similarity coefficient. Furthermore, it was shown that for identification purposes, a simplified plasma model is sufficient to predict adequately the relative emission intensities in LIBS plasmas. As a result, it is possible to automatically identify the species with high emission signals, while trace detection is also possible by relaxing search conditions, although manual refinement is still required. Keywords Laser-induced breakdown spectroscopy, LIBS, elemental identification, wavelength similarity coefficient Date received: 29 April 2016; accepted: 13 December 2016 Introduction The elemental identification of materials plays an important role in many areas of science and industry. Although there are many techniques employed for chemical analysis with high precision, including electronic microscopy, mass spec- trometry, chromatography, X–ray spectroscopy, etc., they share some drawbacks: all of them require expensive equip- ment, are time-consuming, require complex sample prep- aration procedures and need to be operated by qualified personnel. In contrast, laser-induced breakdown spectros- copy (LIBS) offers some interesting benefits. Laser-induced breakdown spectroscopy signals can be acquired in just a few seconds; samples do not need special preparation and can be applied on solids, liquids, and gases. Additionally, this kind of analysis requires only a few mg of material, making this technique a semi-destructive test. With this spectroscopic technique, a plasma is generated by focusing a laser pulse on the surface of a sample. Light produced by plasma is collected and then analyzed in a time-resolved dispersive instrument. The spectral wave- length position of emission lines provides a fingerprint for each atomic species contained in the sample making elemental identification possible. However, assigning measured line emission with a par- ticular species is a complex task that involves discrimination of lines separated by a few thousandths of a nanometer or even overlapped ones. Intensity match also represents a challenge mainly due to the lack of complete information of atomic emission parameters on databases. Besides, the emission data available on the literature are obtained from plasmas produced through other techniques than LIBS, where experimental conditions result in different relative amplitude emissions. Several approaches addressing line emission identifica- tion have been reported. Neural networks 1 and fuzzy logic 2 are examples of artificial intelligence based methods. Other similarity tests, i.e., procedures to quantify the like- ness of two entities, are those based on statistics, such as fast Fourier transform (FFT), 3 dot–product, 4 and correl- ation 5 among others. Proper elemental identification with artificial intelligence methods heavily depends on adequate generation of train- ing sets and advanced programming techniques. On the Centro de Investigaciones en Optica A.C., Guanajuato, Mexico Corresponding author: Ignacio Rosas-Roma ´n, Centro de Investigaciones en Optica A.C., Loma del Bosque 115, Colonia Lomas del Campestre. Leo ´n, Guanajuato, Me ´xico. C.P. 37150. Email: irosas@cio.mx Applied Spectroscopy 0(0) 1–7 ! The Author(s) 2017 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/0003702817693236 journals.sagepub.com/home/asp