Key Points
- Documentary evidence can be vital to forensic investigations, though tampering techniques make proper authentication a complicated process.
- In a new study, gas chromatography–ion mobility spectrometry (GC–IMS) was combined with machine learning algorithms to investigate the temporal evolution of ink stains.
- The decision tree regression model demonstrated high temporal prediction accuracy (test R²=0.954) through interpretable feature engineering. A stepwise strategy combining classification and regression models was proposed, allowing simultaneous ink characterization and age estimation.
Researchers from China, led by Wenhui Lu of Shangdong University, developed a new approach for studying temporal evolution during ink analysis, combining gas chromatography–ion mobility spectrometry (GC–IMS) with machine learning (ML) algorithms. Their findings were published in the Journal of Chromatography A (1).
Close-Up of a Fountain Pen on Handwritten Document | Image Credit: © villorejo – stock.adobe.com
Authenticating documentary evidence can face various challenges due to the existence of sophisticated tampering techniques, such as content manipulation, page substitution, handwriting forgery, and digital alteration. Vital forensic evidence, disputed documents require rigorous examination, mainly through ink analysis, handwriting verification, and material characterization, to establish legal validity. Statistical analysis of Chinese civil litigation cases (2018–2023) reveals that 73% of document-related disputes involve contracts (57%), private loans (10%), and labor conflicts (6%). There is a critical demand for reliable dating authentication methods; however, evaluating the temporal evolution of ink remains challenging due to the complexity of ink compounds, environmental sensitivity, and technical constraints.
Ion mobility spectrometry (IMS) is an analytical technique that characterizes volatile organic compounds (VOCs) through differential drift times of gas-phase ions under an applied electric field. Compared to gas chromatography (GC), IMS offers advantages like ambient temperature, pressure operation capabilities, and significantly reduced analysis times. When GC is combined with IMS, the scientists think that GC–IMS could be a solution for forensic evidence analysis with the efficient separation ability of GC and the rapid trace-amount detection advantage of IMS. Complex-matrix samples can be initially separated with GC, then introduced into ion-mobility tubes for secondary separation and detection. This technique can skip sample pre-treatment during analysis, which enables non-destructive testing, and it enables more efficient qualitative analysis for identifying characteristic markers in writing ink.
To investigate the abundant GC–IMS data of complex ink volatiles, researchers must further optimize the interpretable analysis of GC–IMS data for the rapid identification of ink volatile markers and summarizing temporal evolution stages. In this study, scientists used GC–IMS with ML algorithms to investigate the temporal evolution stages classification and aging time prediction of gel-pen ink. Ink-specific volatile markers were correlated with aging mechanisms with kinetic modeling and heatmap analysis. Three distinct temporal evolution stages were categorized: rapid evaporation, slow-release, and chemical stabilization through multivariate analysis of volatiles. Further, six tree-based ML algorithms were systematically evaluated. The Categorical Boosting (CatBoost) model achieving superior performance (accuracy = 100%) in classifying five detailed aging stages of gel-pen ink.
In this research, the scientists hoped to address three critical challenges: (i) establishing a GC-IMS-based protocol to decode the aging mechanisms of gel-pen ink through volatile organic compound fingerprinting, (ii) integrating unsupervised and supervised learning to classify temporal evolution stages and predict aging timelines, and (iii) unveiling chemically meaningful aging markers through interpretable machine learning approaches.
The decision tree regression model demonstrated high temporal prediction accuracy (test R²=0.954) through interpretable feature engineering. A stepwise strategy combining classification and regression models was proposed, allowing simultaneous ink characterization and age estimation.
This study provides a new approach for evaluating temporal patterns in gel-pink ink using GC–IMS data-driven interpretable models, establishing theoretical foundations for authenticating disputed documents in forensic applications. The scientists hope this methodology could provide a validated approach for classifying temporal evolution stages and predicting aging time, significantly improving the efficiency of forensic analysis in judicial investigations.
Reference
(1) Lu, W.; Chen, J.; Zhang, L.; Nie, Z. Temporal Evolution Stages Classification and Aging Time Prediction of Gel-Pen Ink Using GC-IMS Combined with Machine Learning for Forensic Science Applications. J. Chromatogr. A 2025, 1755, 466063. DOI: 10.1016/j.chroma.2025.466063