Review: Forensic Analysis of Cigarette Ash – Brand Determination Through Trace-Metal Analysis

Emily C. Lennert




cigarette, ash, trace, metal, inductively coupled laser, ICP, mass spectrometry, MS, principle component analysis, PCA, partial least squares, PLS, discriminant analysis, DA

Article Reviewed

  1. Groth, A. C.; Barnes, J. H.; Lewis, C.; Murray, C. K.; Albahadily, F.; Jourdan, T. H. Forensic analysis of cigarette ash – brand determination through trace-metal analysis. Journal of Forensic Sciences. 2016, 61 (4), 913-921.


The opinions expressed in this review are an interpretation of the research presented in the article. These opinions are those of the summation author and do not necessarily represent the position of the University of Central Florida or of the authors of the original article.


Cigarette butts have long been used in investigations and may provide a source of DNA evidence or indicate a brand. In the absence of cigarette butts, ash may be present. Trace metal concentrations in cigarette ash may be used in brand determination, which may then be used to help link a suspect to a crime. However, cigarette ash analysis has seen little forensic use. The primary application of trace metal analysis of cigarette ash has been in counterfeit product identification, using trace metal concentration to distinguish between genuine and counterfeit cigarettes. In this study, the authors examined cigarette ash from American and international brands to determine trace metal content and evaluate the utility of the information that is obtained by the analysis.

Three sample sets were created. The first set consisted of American brand cigarettes. A total of 84 samples from 14 American brands were obtained; 2 packs from each brand were purchased approximately 2 months apart to allow for samples from separate manufacturing lots, and 3 samples were obtained from each pack. A second set of samples originating from 24 cigarette packs of international origin was obtained, with 3 samples obtained from each pack for a total of 72 samples.

Samples were artificially “smoked” to produce ash samples using an air sampler pump. The cigarette was attached to a tube, which was then attached to the pump and controlled by an on/off switch to allow for controlled “puffing” to mimic smoking. A flow rate of 2 L/min was used, with each puff lasting 2 seconds and a 30 second interval between each puff. The ash samples were prepared for analysis by a digestion in acid. A solution of 4 mL concentrated nitric acid and 1 mL concentrated hydrochloric acid was added to each sample, and the sample was allowed to digest for up to 1 week prior to microwave digestion by a Discover SP-D countertop digestion system. After digestion, samples were transferred into storage vessels and diluted with water, then nitric acid.

Samples were analyzed via inductively coupled plasma – mass spectrometry (ICP-MS). ICP-MS uses plasma, i.e. high temperature ionized gas, to ionize a sample. ICP-MS can be used to analyze metals as well as some non-metals. Calibration curves were prepared for metals of interest to allow for quantitation of metal content in samples. Samples were analyzed for the presence of 63 elements, which can be found in the analysis section of the study. Although included in analysis, sodium, calcium, and magnesium were excluded from results due to high abundance in the environment. Due to the use of hydrochloric acid in the digestion process, chlorine was excluded as well. Following analysis, statistical methods were used to determine the statistically significant elements and to create a classification model for each sample set.

A principal component analysis (PCA) model was created and used to determine the significant variables in the dataset. Upon examination of the model, samples of the same brand had a tendency to cluster together, with overlap occurring between brands in most clusters. This indicates that, although composition within the same brand tends to have little variation, different brands tend to contain similar trace metal compositions. The similarity in compositions between different brands leads to overlapping of clusters within the PCA model. Using the factor loadings, it can be determined which variables contributed most to the variance in each PC. Variations in the significant variables, i.e. trace metals, within samples lead to some of the samples clustering and other samples being separated into another clusters, all of which is observed in the PCA model.

  • In the U.S. model, Marlboro varieties clustered together, although the different Marlboro brands were not differentiated as individual clusters. Upon examination of the loadings, barium, rubidium, and zirconium had strong loadings, indicating these metals as the largest contributors to variance. Cobalt, manganese, chromium, hafnium, and vanadium were moderately loading metals, indicating a moderate affect on variance between samples.
  • A PCA model containing only Marlboro varieties was created. In this model, only Marlboro Reds were differentiated in an individual cluster; whereas, all of the other Marlboro samples clustered together.
  • A PCA model was created with both U.S. and international samples to determine if samples originating in different countries could be differentiated. Although some overlap was observed between the U.S. and international samples, the two groups showed distinct separation. Strong positive loadings for platinum, strontium, and antimony were indicative of international samples. Strong negative loadings for nickel, arsenic, germanium, and boron were indicative of U.S. samples.

Partial least squares – discriminant analysis (PLS-DA) was then used to create a classification model for each sample set. By this method, a classification model is created and a subset of the sample data is tested against the model to determine its accuracy in classification. A sensitivity and specificity score is then determined for each class. Sensitivity scores indicate the degree to which samples within a class were correctly predicted; a score of 1.0 indicates that all samples belonging to the class were correctly predicted to belong within the class. Specificity scores indicate the degree to which samples of another class were predicted to belong to the selected class; a score of 1.0 indicates that no samples belonging to other classes were predicted to belong to the chosen class.

  • In the U.S. model, high sensitivity and specificity values were obtained for some samples, indicating satisfactory classification performance. A table of scores can be seen in table 3 within the study. Overall, few scores over 0.90 were observed. Modification of the model to group all Marlboro brands as one class improved Marlboro scores considerably as compared to the scores of each individual Marlboro brand. Natural American Spirit brands were also combined into one class, and perfect scores of 1.0 were obtained for sensitivity and specificity.
  • A classification model containing only Marlboro varieties was created and tested. Classification results indicated perfect scores for Marlboro Reds, with lower scores for most other Marlboro brands. Scores can be seen in table 4 within the study.
  • A classification model of U.S. and international brand was created and tested as well. Only one U.S. sample was incorrectly predicted as being of international origin, for a classification accuracy of 98.8%. Two international samples were incorrectly predicted as being of U.S. origin, for an accuracy of 97.2%. Scores for each class can be seen in table 5 within the study.
    A third sample set consisted of Marlboro Reds, a U.S. sample, which were “smoked” in different ways to determine the effect of smoking habits on the trace metal concentrations present in the resulting cigarette ash. These samples were tested against the U.S. classification model to determine if the samples would still classify correctly with the Marlboro class in the model. All samples within the set of varying smoking parameters were correctly classified as Marlboro cigarettes.

The results of this study indicate that manufacturers can be differentiated through cigarette ash analysis. Variation between brands was demonstrated to be greater than that within samples of the same brand. The authors suggest that further studies should be conducted to explore the possibility of further differentiation.

Scientific Highlights

  • A classification model was developed for U.S. cigarettes alone and for U.S. and international cigarettes together.
  • In the U.S. model, cigarette brands formed some distinct clusters, with Marlboro brands grouping together as one class, as well as Natural American Spirit brands.
  • U.S. and international cigarette showed some overlap, but were separated into 2 distinct groups by PCA.
  • Marlboro Reds, smoked under varying parameters, classified correctly as Marlboro cigarettes in the U.S. model


In the absence of cigarette butts, which may provide fingerprint of DNA evidence, cigarette ash may be analyzed. Trace metal concentrations in cigarette ask may be indicative of the brand or manufacturer of the cigarette that produced the ash, proving a potential link to a suspect.

Potential Conclusions

  • Based on the results, a classification model for cigarette brand identification based on trace metal concentration in cigarette ash may be developed.
  • Cigarette ash has the potential to provide an evidentiary link, which could be used to place a suspect at the scene.