Chemical Profiling and Chemometric Classification of Magic Pen (Disappearing) Ink Samples Using ATR-FTIR Spectroscopy

Abstract

Ink has always been an important aspect of document examination, which while may appear optically similar to the naked eye, but can vary significantly in terms of their chemical constituents. Majority of studies has been focussed on distinguishing these visibly similar inks. However, disappearing inks, also known as magic inks presents a unique set of challenge to document security given its ability to fade entirely within approximately seconds to days, thus enabling undetectable forgery. The current study attempted a non-destructive chemical profiling and classification of blue disappearing inks available in Indian markets using Attenuated Total Reflectance – Fourier Transform Infrared (ATR-FTIR) spectroscopy in conjunction with chemometric methods
In the present study, a total of ninety different blue magic pen (disappearing) inks, representing the 30 different brands were purchased from e-shopping websites and local markets of Jhansi district of Uttar Pradesh in India. The samples were prepared by drawing circles of 2 cm radius and filling it completely with ink in five subsequent strokes on 80 gsm white A4 copying paper for each pen which was then spectroscopically analyzed by ATR-FTIR within the range of 4000- 600 cm-1. The spectral data was further subjected to chemometric analysis viz. HCA and PC-DA.
FTIR spectroscopy of magic pen inks demonstrated significant similarities in spectra with characteristic peaks for O-H (alcohol/water), C-H (aliphatic), C=C (aromatic) and C-O stretching peaks indicating the presence of thymolphthalein as fundamental constituent to most of the blue disappearing ink formulations available in the markets. Chemometric analysis using Hierarchical Cluster Analysis (HCA) revealed four major clusters within the samples. PC based Discriminant Analysis (DA) model was then used to assess the classification of samples in these respective groups which performed reasonably well achieving an accuracy of 100% in classification and cross-validation. Also, PC based DA model for classification of samples according to manufacturer successfully, linked samples to their respective manufacturer with an original and cross-validated accuracy of 97.8% and 82.2% respectively.