- Email: [email protected]
Differentiation of red lipsticks using the attenuated total reflection technique supported by two chemometric methods
Forensic Science International 280 (2017) 130–138
Contents lists available at ScienceDirect
Forensic Science International journal homepage: www.elsevier.com/locate/forsciint
Differentiation of red lipsticks using the attenuated total reflection technique supported by two chemometric methods Marta Gładysz, Małgorzata Król* , Paweł Koscielniak Laboratory for Forensic Chemistry, Department of Analytical Chemistry, Faculty of Chemistry, Jagiellonian University in Kraków, Gronostajowa 2, 30-387 Kraków, Poland
A R T I C L E I N F O
Article history: Received 19 June 2017 Received in revised form 2 September 2017 Accepted 25 September 2017 Available online 6 October 2017 Keywords: Red lipsticks Differentiation Attenuated total reflection Principal component analysis Cluster analysis
A B S T R A C T
The main challenge for the identification and differentiation of lipsticks for forensic purposes is the fact that they have a similar chemical composition — in particular, samples of the same hue. The objective of the presented research was to solve this problem using a nondestructive analytical technique — attenuated total reflection spectroscopy (ATR). 38 red lipsticks produced by 20 different manufacturers were examined in optimized experimental conditions. To facilitate discrimination and provide more extensive analyses of the obtained data, two chemometric techniques: principal component analysis and cluster analysis were used. Ultimately, nine groups of investigated lipstick samples with comparable chemical compositions were differentiated. Moreover, lipstick smears on six different interfering surfaces were analyzed and spectral subtraction was performed in order to identify individual samples. The established approach succeeded in identifying the index number and the manufacturer of the samples by using an in-lab built ATR spectra library. The developed method demonstrates a great potential for the differentiation and identification of red lipsticks with a very similar hue. It also seems to have good prospects for future application in forensic science investigations. © 2017 Elsevier B.V. All rights reserved.
1. Introduction Lipstick residues and smears may be significant forensic evidence in investigations, in particular in cases of rape, sexual assault, murder and burglary. They may be found left on a variety of surfaces — for instance, tissue paper, cigarette butts, glasses, drinking cups, clothing etc., and constitute indirect evidence linking a suspect with the victim or crime scene [1–4]. Nevertheless, identification of lipstick samples and differentiation between them can be a major challenge, because the majority of lipsticks have a similar composition. They are composed of three main ingredients: oils (typically 40–70%) — a mixture of castor, vegetable, mineral or lanolin oil; waxes (approximately 15%) – especially beeswax or carnauba; and colouring agents (roughly 8%) – organic dyes such as erythrosine, amaranth, rhodamine, tartrazine, derivatives of fluorescein, and inorganic pigments, particularly titanium dioxide and iron oxides. Other constituents
* Corresponding author at: Laboratory for Forensic Chemistry, Department of Analytical Chemistry, Faculty of Chemistry, Jagiellonian University in Kraków, Gronostajowa 2, 30-387 Kraków, Poland. E-mail address: [email protected] (M. Król). https://doi.org/10.1016/j.forsciint.2017.09.019 0379-0738/© 2017 Elsevier B.V. All rights reserved.
occurring in smaller quantities are antioxidant materials and perfumes [3,4]. There have been many publications regarding lipstick examination, reporting results from simple optical [5] to modern analytical methods (i.e. CE [6], HPLC [7,8], UHPLC–MS [9], GC–MS [10], ICP-OES [11], ICP–MS [12], AAS [11,13], LIBS [14,15], XRF [16]). However, most of them have focused on (aspects of) the negative impact of the lipstick components on humans and the environment. Among other things, the content of azo [6], xanthene [7,17] or other organic colorants [8,9], as well as hazardous nitrosamines [10] and heavy metals (e.g. Pb) [11–16,18–20] in lipsticks have been studied. In the case of forensic lipstick examination, the most common technique is nondestructive Raman spectroscopy [2,3,21,22]. It should be pointed out here, however, that in every mentioned article, Raman spectroscopy has been accompanied by problems associated with fluorescence. An in-situ method of surface enhanced resonance Raman scattering (SERRS) detection of the colorants in lipstick smears on both glass and cotton surfaces was reported [23], but a surfactant was required to obtain SERRS spectra of the dyes and pigments in such waxy samples. Moreover, the technique of neutron activation analysis (NAA) – a nondestructive and highly sensitive but also very time-consuming and expensive method – was employed to study the presence of trace elements in samples of lipsticks [24].
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
Another technique described in the literature used to differentiate lipstick is TLC [1,4,25]. However, although the information obtained by the TLC analytical methods was usually sufficient for lipstick differentiation from the forensic point of view, these methods have the important drawback of destructive sample preparation. One of the techniques, which – to the best of our knowledge – has not yet been applied to lipstick examination is attenuated total reflection spectroscopy (ATR). Two significant advantages of ATR, namely that the sample is not destroyed and no sample preparation is required, have led to this technique being widely employed in forensic science investigations [26–29]. Consequently, the aim of the presented research was to develop the first ATRbased procedure enabling differentiation of red lipsticks. Additionally, chemometric methods such as principal component analysis (PCA) and cluster analysis (CA) were applied in order to facilitate discrimination and provide more extensive analyses of the obtained data. Lastly, the in-lab built ATR spectra library, enabling identification of all investigated samples, was tested by analyzing traces of random lipsticks on six different, potentially interfering surfaces. 2. Experimental The solutions (isopropanol and methanol) used throughout the experiments were supplied by Sigma-Aldrich (Germany). 38 red lipsticks (for 5 randomly selected 5 items of the same index number were bought) of a very similar hue representing 20 different manufacturers were either purchased in local shops or donated by Inglot Sp. z o.o. (cosmetics company). Detailed information about all of them is presented in Table 1. The experiments were carried out using a Thermo Nicolet iS50 FTIR (Thermo Fisher Scientific Co, Waltham, MA, USA) with a Smart Orbit micro-ATR accessory. All spectra were collected from 650 to 4000 cm1 by putting small amounts of a sample (approximately 10 mg) on the ZnSe crystal and carefully pressing it with an ATR pressure tower. After each measurement, the surface of the crystal was cleaned with 50% v/v isopropanol and then with methanol in order to avoid contamination. A new background was collected prior to the analysis of each new sample. Processing of the obtained spectra was carried out using Thermo Electron’s OMNIC 9 software. All statistical analyses were performed using Statistica 12.5 PL software (StatSoft, Tulsa, OK, USA) and OriginPro 2017 software (OriginLab Corporation, Northampton, MA, USA). 3. Applied chemometric/computational methods 3.1. Principal component analysis PCA is a well-known multivariate statistical method enabling reduction of the dimensionality of the original dataset by creating a linear combination of variables called principal components (PCs). However, the determination of the optimal number of components that should be chosen is a crucial step. Inappropriate estimation of PCs could lead to loss of information. The Scree test, Kaiser’s stopping rule and the percentage of cumulative variance are the most frequently used strategies [30], and these were applied in this study. Furthermore, PCs are a very useful tool for demonstrating the relationships between analyzed samples. 3.2. Cluster analysis In CA, samples are assembled in high dimensional space. At the beginning, each sample constitutes its own separate cluster, and then two objects, which are the closest to each other, are
131
combined. This procedure is accomplished repeatedly until all samples are arranged into one cluster. It is worth pointing out that several methods for measuring similarity and combining clusters are available. Their choice depends on the investigated samples and this should be evaluated experimentally [31]. In this study, Euclidean distances were calculated and Ward’s method was employed. Furthermore, in the literature, over 20 different rules for stopping an obtained dendrogram are known. In this article, Mojena’s stopping rule was considered to be the most efficient. According to Mojena’s stopping rule, one should select the number of groups where the following inequality (Eq. (1)) is fulfilled diþ1 > d þ ksd
ð1Þ
where d0,dn, dn1 are distances corresponding to n, n 1, . . . 1 clusters. The terms d, sd, and k are the mean, the standard deviation of the d values, and a constant, respectively [32]. As stated by Milligan and Cooper [33], the value of k should be 1.25. 3.3. Correlation method Correlation coefficients (CC) were computed as Pearson correlation coefficients by means of Thermo Scientific OMNIC software. It seems important to note that the used algorithm additionally included elimination of the effects of baseline variation. 4. Results and discussion Firstly, the parameters of the experiments were optimized. It is known that increasing the number of scans reduces the signal-tonoise level of data, and enables one to distinguish small peaks from noise, and that lower resolution can have an effect on the differentiation of peaks. Therefore, 16, 20 and 60 scans, and resolutions of 4 and 2 cm1 were tested. Despite the fact that only slight differences were observed between registered spectra, 20 scans per second with 4 cm1 spectral resolution for both background and samples were considered to be optimal. The homogeneity (intra-variability) of lipsticks was investigated by analyzing spectra obtained from three different parts of several randomly selected lipstick samples (L3, L10, L14, L21, L31, L32, L34 and L36) during one day. As demonstrated in exemplary Fig. 1a presenting the results obtained for three lipsticks, no visual differences were found on the spectra. Consequently, it was considered that the samples were homogeneous and their composition was stable during one day. Additionally, during three days the inter-variability of lipstick was examined using 5 of the previously selected lipsticks. The similarity of the spectra was calculated as correlation coefficient (L10: 99.90 0.04; L31: 99.91 0.04; L32: 99.94 0.01; L34: 99.83 0.06; L36: 99.92 006). The presented results provided the basis for claiming that the analyzed samples of the lipstick exhibited almost unchanged chemical composition also within a few days. Moreover, both above-mentioned experiments proved the high intra- and inter-repeatability of the proposed ATR-FTIR method. Finally, the variation of chemical composition of lipstick samples within the same series was also investigated during one day by analyzing five items (a–e) of five samples (L10, L31, L32, L34 and L36) from the same manufacturer with the same index number. As it can be seen in exemplary Fig. 1b — the obtained spectra for samples L36 a–e almost completely overlapped. Therefore, it was ascertained that the chemical composition of analyzed lipsticks that were both produced by the same
132
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
Table 1 List of lipsticks examined during the present study. Code of lipstick
Manufacture
Index numbera
Seriesb
L1 L2 L3 L4 L5 L6 L7 L8 L9 L10a L10b L10c L10d L10e L11 L12 L13 L14 L15 L16 L17 L18 L19 L20 L21 L22 L23 L24 L25 L26 L27 L28 L29 L30 L31a L31b L31c L31d L31e L32a L32b L32c L32d L32e L33 L34a L34b L34c L34d L34e L35 L36a L36b L36c L36d L36e L37 L38
INGRID INGRID MISSLYN MISSLYN GOSH p2 BourjoisParis e REVLON L’OREAL Paris L’OREAL Paris L’OREAL Paris L’OREAL Paris L’OREAL Paris MAYBELLINE New York SEPHORA SEPHORA SEPHORA SEPHORA CATRICE CATRICE CATRICE LOVELY WIBO MISS SPORTY EVELINE COSMETICS EVELINE COSMETICS RIMMEL ASTOR MAX FACTOR MANHATTAN INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT MANHATTAN MANHATTAN MANHATTAN MANHATTAN MANHATTAN MANHATTAN MANHATTAN MANHATTAN
289 288 M20.94 M20.87 005 030 13 02 720 330 330 330 330 330 955 B06 R05 N 62 N 19 430 440 140 8 1 301 486 710 510 203 715 550 176 180 429 44 44 44 44 44 21 21 21 21 21 64 12 12 12 12 12 45 M 45 N 45 N 45 N 45 N 45 N 45 F 90 A
Wonder shine Wonder shine Rouge a levres Rouge a levres Velvet touch lipstick Full color lipstick Rouge Edition Longlasting lipstick Super lustrous lipstick Color riche Color riche Color riche Color riche Color riche Color sensational Rouge Baume SPF20 Rouge Cream Rounge Brilliance Color lip last Ultimate color Ultimate color Ultimate stay Creamy color Matte Intense My BFF Aqua Platinum Color edition Moisture renew Perfect stay fabulous Color elixir Soft rouge lipstick Lipstick Lipstick Lipstick matte Slim gel lipstick Slim gel lipstick Slim gel lipstick Slim gel lipstick Slim gel lipstick AMC lip pencil matte AMC lip pencil matte AMC lip pencil matte AMC lip pencil matte AMC lip pencil matte AMC lip paint HD lip tint matte HD lip tint matte HD lip tint matte HD lip tint matte HD lip tint matte Xtreme last & shine Perfect creamy & care Perfect creamy & care Perfect creamy & care Perfect creamy & care Perfect creamy & care Perfect creamy & care Perfect creamy & care
a b
Comment
High passion Cayun pepper Matt, classic red Challenge authority Rouge jet set All you need is red Fire & Ice Matte, cocorico Matte, cocorico Matte, cocorico Matte, cocorico Matte, cocorico Craving coral, matte Soft red Hot Tango Stay Together Pure red-matt Hot’n Spice Hungs-and Hibis-Kisses Behind the red curtain
My cheeky cherry red in love Mayfair red lady Ruby tuesday Watermelon
The index number is a number assigned to each lipstick in a given series. Series name refers to cosmetics collection. The same series name indicates the same collection of lipsticks.
manufacturer and had the same index number were characterized by very good repeatability. 4.1. Visual inspection of spectra As seen in Fig. 2, a typical spectrum of a lipstick sample can be divided into three main ranges. Generally, the first range (between 3740 and 3100 cm1) includes one broad band, which is attributed to water and hydroxyl groups. The second range (from 3050 to 2775 cm1) takes account of symmetric (2850 cm1) and
asymmetric (2916 cm1) C H stretching vibrations and asymmetric CH3 stretching vibrations (2954 cm1). The last range (below 1800 cm1) contains many peaks. C¼O stretching vibrations (1730 cm1), CH2 scissor deformation (1465 cm1), C¼O stretching and CH bending vibrations (1172 cm1), and CH2 rocking mode (712 cm1) are just some of them. All of the mentioned peaks could be attributed to various oils and waxes, especially castor oil and beeswax — two common ingredients in lipsticks [3]. Therefore, it can be supposed that the most informative and distinguishing range is below 1800 cm1.
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
133
Fig. 2. Typical lipstick spectrum, represented by L13.
Fig. 1. Investigation of a) homogeneity of lipsticks L21, L14 and L3; b) variability of lipsticks from the same manufacturer and with the same index number: L36 a–e.
are distinctive and consequently they can be effortlessly distinguished. Concisely, according to PCA analysis, red lipsticks can be divided into the following six groups: G1 {L34}, G2 {L32}, G3 {L31, L33}, G4 {L10}, G5 {L15, L25} and G6, which includes all the remaining samples. In an attempt to find the most efficient grouping of lipsticks, cluster analysis was performed in order to classify lipstick samples in the G6, based on similarities of their ATR spectra in the range from 1800 to 650 cm1 (the fingerprint region). Accordingly, the dendrogram presented in Fig. 4 was created, whereas the results obtained from Mojena’s stopping rule are shown in Table I in the ESM. They indicated that the value of 1.19 should be used as the stopping distance. As a consequence, the dendrogram was divided into the following four groups of lipstick samples: G6a {L1, L8, L20, L27, L28, L29, L30, L35, L36a, L37, L38,}, G6b {L4, L5, L13, L17, L21, L22,
4.2. Differentiation of red lipsticks 4.2.1. Chemometric methods A primary discrimination of lipstick samples was performed using the PCA technique, taking into account all 38 of the collected ATR spectra registered in the range from 4000 to 650 cm1. Fig. I in the Electronic Supplementary Material (ESM) shows a plot for deciding on the number of variables — a scree test. Interpretation of this plot indicates that only the first three components should be retained in the present study (the line drops precipitously), because sufficient useful chemical information is provided by these three components. They also complied with the Kaiser criterion, which only recommends components with eigenvalues greater than unity. Furthermore, approximately 90% of the total variance in the spectra was explained by them: 70.17%, 10.38%, 8.14% for the first, second, and third PC, respectively. Thus, it is reasonable to base inferences solely on these PCs. The PCA score plot of the first three PCs, demonstrating the separation of samples, is presented in Fig. 3. Additionally, in Fig. II in the ESM, two-dimensional plots are presented to make the visual analysis easier. As seen, most of the lipsticks, except L10, L15, L25, L31, L32, L33 and L34, were classified into one multi-element class. Additionally, it was noted that lipstick L15 is very similar to L25 and lipstick L31 is comparable to L33. Moreover, lipsticks L10, L32 and L34 are each completely separated from all other lipsticks. This finding suggests that some of the analyzed lipstick samples
Fig. 3. The PCA score plot of the first three PCs for all investigated lipsticks, taking into account ATR spectra registered in the range from 4000 cm1 to 650 cm1.
134
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
Fig. 4. Cluster analysis of lipsticks from G6, taking into account ATR spectra registered in the range from 1800 cm1 to 650 cm1.
L26}, G6c {L2, L3, L6, L11, L14, L16, L18, L23, L24} and G6d {L7, L9, L12, L19}. Ultimately, as a result of chemometric analysis, nine groups of red lipstick samples with comparable chemical compositions were created: G1–G5 and G6 a–d. It is worth noting that according to chemometric classification, all the investigated lipstick samples manufactured by Manhattan were classified into one group (G6 a). On the other hand, the lipstick samples from Sephora were categorized into various groups (G6 b, c, G5), while some of lipstick samples manufactured by Inglot were placed in group G6 a and some in G1–G3. 4.2.2. Visual classification A visual comparison of the analyzed spectra was also performed. In the first step of differentiation between lipstick samples, one should look at both the relative intensities and the degree of resolution of bands from 3050 to 2775 cm1. As can be seen in Fig. 5, this region enables differentiation between all groups, except between G2 and G6, and between G4 and G5. Further distinction between groups was based on the bands occurring below 1800 cm1. As is evident in Fig. 6a, the most characteristic feature for G5 is a lack of band at 1745 cm1 corresponding to the carbonyl group. Moreover, in G4 one can see a weak doublet at 1375 and 1363 cm1, whereas in G5 there is only a band at 1363 cm1. Both of these bands are due to CH3 deformation [34]. The differentiation of G2 from G6 is presented in Fig. 6b. On the face of it, spectra obtained for these groups are completely unlike each other in the range from 1800 to 650 cm1. The most noteworthy difference is strong absorption near 1000 and 1100 cm1 in G2. It is highly possible that these peaks correspond to phosphates and ethers or alcohol–ethers, respectively, which are found in many plasticizers [34]. Moreover, the differences between G6 a–d are selected in Fig. 6c — only representative spectra of lipstick samples from the groups have been presented. As far as one can see, the region where differences are most clearly visible is from 1400 to 800 cm1. In this region, a little broad band (especially in G6 a–c) at 1150 cm1 characteristic for C O stretch in nonaromatic ethers is observable. Moreover, the existence of a siloxane group is most likely due to a band occurring at 1108 cm1, which is most visible in spectra from
G6 b. Besides, strong absorption in the range from 1100 to 1000 cm1 is characteristic for G6 a, b and, as was mentioned before, these bands are connected with phosphates and ether/ alcohol–ethers [34]. It is noteworthy that there are also three unique bands for G6 d — at approximately 1387, 1376 and 1363 cm1. Summarizing, visual comparison of the obtained spectra led to confirmation of the classification ascertained according to the chemometric results. 4.2.3. Correlation method In order to make the analysis more objective as well as to verify the existence of an unequivocal dependence between the composition of the lipstick samples and the manufacturer (or series), the correlation method was applied. At first, the variability of the same lipstick sample, s, was ascertained as the mean CC (correlation coefficient) value calculated for five items (a-e) of the five samples (L10, L31, L32, L34 and L36) of lipsticks of the same index number, the same series and the same manufacturer. The mentioned samples were representatives of different PCA groups so they characterized by different chemical composition. It is worth pointing out that the CCs value were evaluated by analyzing samples on three successive days and they ranged between 0.04% and 0.33%. On this basis the following assumptions related to ATR spectra of each examined pair of lipstick samples have been made: if CC 100% — 3s, it is highly possible that the spectra correspond to the same lipsticks, if CC < 100% — 3s, the spectra are considered as coming from different lipsticks. In short, the similarity limit in the case of the present study is a value of 99.01%. Comparing the spectra of lipsticks produced by different manufacturers, it was ascertained that they are usually characterized by CCs considerably less than 99.00% (see Table II in ESM), e.g., for L15 (Sephora) and L3 (Misslyn), the CC = 55.81% or for L6 (p2) and L20 (Wibo), the CC = 92.58%. There were a few cases where the compositions of two lipsticks originating from different
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
135
Fig. 5. Differentiation between groups based on the range from 3050 cm1 to 2775 cm1.
manufacturers were very similar, e.g., for L3 (Misslyn) and L22 (Eveline Cosmetics) CC = 99.45%, for L29 (Inglot) and L4 (Misslyn) CC = 99.55%. If the investigated samples were from the same manufacturer but from different series, most of them could also be easily distinguished. It can be seen in Table 2 that CCs between spectra of lipstick samples produced by Inglot, Manhattan and Sephora were not higher than 98.50%. Moreover, it is evident that lipsticks manufactured by Inglot are characterized by the most variable composition (CC from 96% to 24%). The obtained results confirmed the theoretical assumption that lipstick samples from the same manufacturer and from the same series are more similar to each other than lipstick samples from different series (see CC for L28 and L29 from Inglot, or L36 and L37 from Manhattan). On the other hand, some exceptions can be observed: sample L35 from the Xtreme last & shine series is undistinguished (CC between 99.11% and 99.49%) from the samples from the Perfect creamy & care series. Moreover, the data suggest that a comparison of samples with a similar index number (from the same manufacturer and from the same series) can cause problems. In these cases, the CCs between samples were calculated to be greater than 99.01%, and in consequence, they were not differentiable with the use of the presented method. The discrimination of such lipsticks samples can be problematic and
they may differ to a very limited extent, e.g. only slightly in the concentration of colour additives. However, further data are needed in order to confirm this assumption. It has not escaped our notice that there were also lipsticks produced by the same manufacturer in the same series with an index number differing only by one, e.g., L1 — Ingrid no. 289 and L2 — Ingrid no. 288, which revealed, without any doubt, two different spectra (CC = 98.52%). It seems important to reiterate that samples from the same manufacturer, the same series and having the same index number (for instance L36 a–e), as was mentioned before, were considered to be indistinguishable. The CC between spectra was much higher than 99.01%. 4.3. Analysis of lipsticks traces on various surfaces — preliminary examination Due to the fact that lipstick samples can be found left on various surfaces, it is justified to investigate the possibility of identifying lipstick traces in the form of such smears. Firstly, one of the samples (L28) was left as a lipstick mark on various surfaces in order to assess which surfaces caused interfering bands. The surfaces studied included: envelope, paper (80 g/m2), tissue (100% cellulose), cigarette butts, and a transparent plastic cup, as well as the white collar of a shirt. As can be seen in
136
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138 Table 2 Spectra correlation coefficients between samples from the same manufacturer: a) Inglot; b) Manhattan; c) Sephora. Lipsticks samples from the same series are bolded. a) L28 L29 L30 L31 L32 L33 L34
b) L27 L35 L36a L37 L38
c) L12 L13 L14 L15
100.0 99.57 96.44 91.67 90.24 79.95 27.11 L28
100.0 96.70 91.50 90.90 79.48 25.90 L29
100.0 98.46 97.79 98.24 97.93 L27
100.0 94.87 95.95 62.50 L12
100.0 87.83 95.44 74.51 24.67 L30
100.0 83.29 88.50 40.22 L31
100.0 99.11 99.49 99.26 L35
100.0 99.65 99.83 L36a
100.0 96.69 57.46 L13
100.0 65.70 17.65 L32
100.0 53.76 L33
100.0 99.78 L37
100.0 55.17 L14
100.0 L34
100.0 L38
100.0 L15
In striving to find the most efficient approach to identification of lipstick marks, two spectral subtractions were performed: (1) in the whole range of registered spectra, (2) in the fingerprint region. The effectiveness of the proposed approach to the identification of lipsticks marks was verified by analyzing six different traces of red lipstick (L7, L12, L18, L23, L28 and L35) smudges on all the abovementioned surfaces. Overall, the procedure was carried out in a few steps by: measuring the surface spectrum and the spectrum of the lipstick on the same surface, then subtracting one from the other in the whole range of the rejected spectrum and in the fingerprint region (from 1800 to 650 cm1), and subsequently identifying the spectrum by using an in-lab built ATR spectra library. The obtained results indicate that the second approach solely enables one to connect lipstick marks with the manufacturer and index number (see Table 3). It should be highlighted that the described subtraction approach did not yield pure spectra of lipstick samples — additional or negative bands appeared in some cases. Nevertheless, a comparison of the spectral subtraction result with the spectra collected in the in-lab built library enabled – in every case without exception – the successful identification of lipstick samples.
5. Conclusions
Fig. 6. Differentiation of lipsticks according to ATR spectra registered in the range from 1800 cm1 to 650 cm1: a) G4 represented by L25, G5 represented by L10; b) G2 represented by L32, G6 represented by L12; c) G6a represented by L35, G6b represented by L26, G6c represented by L3, and G6d represented by L19.
Fig. III in the ESM, all of the analyzed surfaces, except the transparent plastic cup, provided additional bands to the lipstick spectra. In order to isolate lipstick spectra from surface spectra, Thermo Electron’s OMNIC software was used. This is one of the most difficult issues in the identification of samples in forensic science.
ATR spectroscopy was successfully used for the differentiation of 38 red lipsticks of a very similar hue. Although no unambiguous trends could be identified within manufacturers or series and the differentiation of lipsticks is a complicated issue, combining the two chemometric methods (PCA and CA) enabled the distinguishing of nine groups from investigated lipsticks of similar ATR spectra. Similarities cannot be assumed, in advance, on the basis of manufacturer, series or index number compatibility. However, visual analysis revealed bands that were characteristic for appropriate groups and confirmed the classification obtained by chemometric techniques. Furthermore, the approach supported by the correlation coefficient enabled deeper differentiation of all studied lipstick spectra. And, importantly, the differentiation could be achieved non-destructively in a matter of minutes.
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
137
Table 3 A comparison of the two approaches for the identification of lipsticks in the form of lipstick marks smudged on interfering surfaces. Correct identification of lipsticks is indicated by “+”. Lipstick Code
The range of spectral subtraction 4000–650 cm1
L7 L12 L18 L23 L28 L35
1800–650 cm1
Cigarette butts
Envelope
Paper
Tissue
White collar of shirt
Cigarette butts
Envelope
Paper
Tissue
White collar of shirt
+ + + + + +
+ + + + + +
+ + + + + +
+ + + + + +
+ + + + + +
Table 4 Discriminating power of PCA, CA and CC analysis. Method
Discriminating power (DP)
PCA CA Correlation method
0.29 0.51 0.93
To sum up, on the basis of the performed PCA, CA and CC differentiations, the discriminating power (DP) of the ATR analysis of lipstick samples has been calculated according to the following equation: DP = 1 2 M/(N(N 1)) (where M is number of matching pairs; N is total number of analysed lipsticks). The obtained DP values were listed in Table 4. These results confirmed the above-mentioned conclusion that PCA and CA enables some preliminary classification whereas the correlation coefficient is the best choice for the lipsticks discrimination. Ultimately, by using the developed method and in-lab built ATR spectra library, it was possible to detect and connect some lipsticks left as lipstick marks with their manufacturer and index number, even on interfering surfaces. This finding confirms that the ATR technique is a useful tool in lipstick differentiation and identification. All things considered, it seems reasonable to assume that the proposed method shows promise for future application in forensic science investigations as a good alternative to routinely used methods. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.forsciint.2017. 09.019.
References [1] S. Singh, V. Saran, M. Mishra, A.K. Gupta, Forensic examination of lipstick by the various physio-chemical and instrumental method, Int. J. Soc. Relev. Concern 3 (2015) 1–7. [2] M. López-López, N. Özbek, C. García-Ruiz, Confocal Raman spectroscopy to trace lipstick with their smudges on different surfaces, Talanta 123 (2014) 135–139. [3] F. Salahioglu, M.J. Went, Differentiation of lipsticks by Raman Spectroscopy, Forensic Sci. Int. 223 (2012) 148–152. [4] B. Joshi, K. Verma, J. Singh, A comparison of red pigments in different lipsticks using thin layer chromatography (TLC), J. Anal. Bioanal. Tech. 4 (2013) 1–4. [5] Y. Ehara, Y. Marumo, Identification of lipstick smears by fluorescence observation and purge-and-trap gas chromatography, Forensic Sci. Int. 96 (1998) 1–10. [6] D. Li, Z. Wang, L. Wang, X. Xu, H. Zhang, Ultrasonic extraction coupled with capillary electrophoresis for the determination of azo dyes in lipsticks using ionic liquid as dynamic coating and background electrolyte, Chin. J. Chem. 29 (2011) 147–152.
[7] S. Scalia, S. Simeoni, Assay of xanthene dyes in lipsticks by inverse supercritical fluid extraction and HPLC, Chromatographia 53 (2001) 490–494. [8] S.C. Rastogi, V.J. Barwick, S.V. Carter, Identification of organic colourants in cosmetics by HPLC-diode array detection, Chromatographia 45 (1997) 215– 228. [9] Y. Xian, Y. Wu, X. Guo, Y. Lu, H. Luo, D. Luo, Y. Chen, Simultaneous determination of 11 restricted dyes in cosmetics by ultra-high-performance liquid chromatography/tandem mass spectrometry, Anal. Methods 5 (2013) 1965–1974. [10] Q. Ma, H.-W. Xi, C. Wang, H. Bai, G.-C. Xi, N. Su, L.-Y. Xu, J.-B. Wang, Determination of ten volatile nitrosamines in cosmetics by gas chromatography tandem mass spectrometry, Chinese J. Anal. Chem. 39 (2011) 1201–1207. [11] É. Ferreira Batista, A. dos Santos Augusto, E.R. Pereira-Filho, Chemometric evaluation of Cd, Co Cr, Cu, Ni (inductively coupled plasma optical emission spectrometry) and Pb (graphite furnace atomic absorption spectrometry) concentrations in lipstick samples intended to be used by adults and children, Talanta 150 (2016) 206–212. [12] N.M. Hepp, Determination of total lead in 400 lipsticks on the U.S. market using a validated microwave-assisted digestion, inductively coupled plasmamass spectrometric method, J. Cosmet. Sci. 63 (2012) 159–176. [13] S. Gunduz, S. Akman, Investigation of lead contents in lipsticks by solid sampling high resolution continuum source electrothermal atomic absorption spectrometry, Regul. Toxicol. Pharm. 65 (2013) 34–37. [14] M.A. Gondal, Z.S. Seddigi, M.M. Nasr, B. Gondal, Spectroscopic detection of health hazardous contaminants in lipstick using Laser Induced Breakdown Spectroscopy, J. Hazard. Mater. 175 (2010) 726–732. [15] A. dos Santos Augusto, E. Ferreira Batista, E.R. Pereira-Filho, Direct chemical inspection of eye shadow and lipstick solid samples using laser-induced breakdown spectroscopy (LIBS) and chemometrics: proposition of classification models, Anal. Methods 8 (2016) 5851–5860. [16] E. Da Silva, A. Matthews David, A. Pejovi c-Mili c, The quantification of total lead in lipstick specimens by total reflection X-ray fluorescence spectrometry, XRay Spectrom. 44 (2015) 451–457. [17] Ch.Ch. Wang, A.N. Masi, L. Fernandez, On-line micellar-enhanced spectrofluorimetric determination of rhodamine dye in cosmetics, Talanta 75 (2008) 135–140. [18] I. Al-Saleh, S. Al-Enazi, Trace metals in lipsticks, Toxicol. Environ. Chem. 93 (2011) 1149–1165. [19] I. Al-Saleh, S. Al-Enazi, N. Shinwari, Assessment of lead in cosmetic products, Regul. Toxicol. Pharm. 54 (2009) 105–113. [20] P. Piccinini, M. Piecha, S.F. Torrent, European survey on the content of lead in lip products, J. Pharm. Biomed. Anal. 76 (2013) 225–233. [21] P. Gardner, M.F. Bertino, R. Weimer, E. Hazelrigg, Analysis of lipsticks using Raman spectroscopy, Forensic Sci. Int. 232 (2013) 67–72. [22] F. Salahioglu, M.J. Went, S.J. Gibson, Application of Raman spectroscopy for the diff ; ;erentiation of lipstick traces, Anal. Methods 5 (2013) 5392–5401. [23] C. Rodger, V. Rutherford, D. Broughton, P.C. White, W.E. Smith, The in-situ analysis of lipsticks by surface enhanced resonance Raman scattering, Analyst 123 (1998) 1823–1826. [24] G. Misra, V.K. Mittal, Neutron activation analysis of lipsticks using g-ray spectrometry, J. Appl. Spectrosc. 71 (2004) 270–274. [25] A.F.L. bin Abdullah, Y. Marimuthu, Ch. Kah Haw, N.F. binti Mohamad Said, N.Z. M. Muslim, N.F.N. Hassan, M.H. Yaacob, Y. Chong Hooi, Forensic discrimination of lipsticks by thin layer chromatography and gas chromatography—mass spectrometry, Malays. J. Forensic Sci. 2 (2011) 22–28. [26] J. Manheim, K.C. Doty, G. McLaughlin, I.K. Lednev, Forensic hair differentiation using attenuated total reflection fourier transform infrared (ATR FT-IR) spectroscopy, Appl. Spectrosc. 70 (2016) 1109–1117. [27] M.C.A. Marcelo, K.C. Mariotti, M.F. Ferrao, R.S. Ortiz, Profiling cocaine by ATRFTIR, Forensic Sci. Int. 246 (2015) 65–71. [28] M. Szafarska, M. Wo zniakiewicz, M. Pilch, J. Zie˛ba-Palus, P. Koscielniak, Computer analysis of ATR-FTIR spectra of paint samples for forensic purpose, J. Mol. Struct. 924–926 (924) (2009) 504––. [29] C.M. Orphanou, The detection and discrimination of human body fluids using ATR FT-IR spectroscopy, Forensic Sci. Int. 252 (2015) e10–e16.
138
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
[30] B.R. Kowalski (Ed.), Chemometrics: Mathematics and Statistics in Chemistry, Reidel, Dordrecht, 1984. [31] D.L. Massart, L. Kaufman, The Interpretation of Analytical Chemical Data by the Use of Cluster Analysis, John Wiley & Sons, New York, 1983. [32] R. Mojena, Hierarchical grouping methods and stopping rules: an evaluation, Comput. J. 20 (1977) 359–363.
[33] G.W. Milligan, M.C. Cooper, An examination of procedures for determining the number of clusters in a data set, Psychometrika 50 (1985) 159–179. [34] G. Socrates, Infrared and Raman characteristic group frequencies, Tables and Charts, John Wiley and Sons, Ltd., Chichester, 2001.
Contents lists available at ScienceDirect
Forensic Science International journal homepage: www.elsevier.com/locate/forsciint
Differentiation of red lipsticks using the attenuated total reflection technique supported by two chemometric methods Marta Gładysz, Małgorzata Król* , Paweł Koscielniak Laboratory for Forensic Chemistry, Department of Analytical Chemistry, Faculty of Chemistry, Jagiellonian University in Kraków, Gronostajowa 2, 30-387 Kraków, Poland
A R T I C L E I N F O
Article history: Received 19 June 2017 Received in revised form 2 September 2017 Accepted 25 September 2017 Available online 6 October 2017 Keywords: Red lipsticks Differentiation Attenuated total reflection Principal component analysis Cluster analysis
A B S T R A C T
The main challenge for the identification and differentiation of lipsticks for forensic purposes is the fact that they have a similar chemical composition — in particular, samples of the same hue. The objective of the presented research was to solve this problem using a nondestructive analytical technique — attenuated total reflection spectroscopy (ATR). 38 red lipsticks produced by 20 different manufacturers were examined in optimized experimental conditions. To facilitate discrimination and provide more extensive analyses of the obtained data, two chemometric techniques: principal component analysis and cluster analysis were used. Ultimately, nine groups of investigated lipstick samples with comparable chemical compositions were differentiated. Moreover, lipstick smears on six different interfering surfaces were analyzed and spectral subtraction was performed in order to identify individual samples. The established approach succeeded in identifying the index number and the manufacturer of the samples by using an in-lab built ATR spectra library. The developed method demonstrates a great potential for the differentiation and identification of red lipsticks with a very similar hue. It also seems to have good prospects for future application in forensic science investigations. © 2017 Elsevier B.V. All rights reserved.
1. Introduction Lipstick residues and smears may be significant forensic evidence in investigations, in particular in cases of rape, sexual assault, murder and burglary. They may be found left on a variety of surfaces — for instance, tissue paper, cigarette butts, glasses, drinking cups, clothing etc., and constitute indirect evidence linking a suspect with the victim or crime scene [1–4]. Nevertheless, identification of lipstick samples and differentiation between them can be a major challenge, because the majority of lipsticks have a similar composition. They are composed of three main ingredients: oils (typically 40–70%) — a mixture of castor, vegetable, mineral or lanolin oil; waxes (approximately 15%) – especially beeswax or carnauba; and colouring agents (roughly 8%) – organic dyes such as erythrosine, amaranth, rhodamine, tartrazine, derivatives of fluorescein, and inorganic pigments, particularly titanium dioxide and iron oxides. Other constituents
* Corresponding author at: Laboratory for Forensic Chemistry, Department of Analytical Chemistry, Faculty of Chemistry, Jagiellonian University in Kraków, Gronostajowa 2, 30-387 Kraków, Poland. E-mail address: [email protected] (M. Król). https://doi.org/10.1016/j.forsciint.2017.09.019 0379-0738/© 2017 Elsevier B.V. All rights reserved.
occurring in smaller quantities are antioxidant materials and perfumes [3,4]. There have been many publications regarding lipstick examination, reporting results from simple optical [5] to modern analytical methods (i.e. CE [6], HPLC [7,8], UHPLC–MS [9], GC–MS [10], ICP-OES [11], ICP–MS [12], AAS [11,13], LIBS [14,15], XRF [16]). However, most of them have focused on (aspects of) the negative impact of the lipstick components on humans and the environment. Among other things, the content of azo [6], xanthene [7,17] or other organic colorants [8,9], as well as hazardous nitrosamines [10] and heavy metals (e.g. Pb) [11–16,18–20] in lipsticks have been studied. In the case of forensic lipstick examination, the most common technique is nondestructive Raman spectroscopy [2,3,21,22]. It should be pointed out here, however, that in every mentioned article, Raman spectroscopy has been accompanied by problems associated with fluorescence. An in-situ method of surface enhanced resonance Raman scattering (SERRS) detection of the colorants in lipstick smears on both glass and cotton surfaces was reported [23], but a surfactant was required to obtain SERRS spectra of the dyes and pigments in such waxy samples. Moreover, the technique of neutron activation analysis (NAA) – a nondestructive and highly sensitive but also very time-consuming and expensive method – was employed to study the presence of trace elements in samples of lipsticks [24].
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
Another technique described in the literature used to differentiate lipstick is TLC [1,4,25]. However, although the information obtained by the TLC analytical methods was usually sufficient for lipstick differentiation from the forensic point of view, these methods have the important drawback of destructive sample preparation. One of the techniques, which – to the best of our knowledge – has not yet been applied to lipstick examination is attenuated total reflection spectroscopy (ATR). Two significant advantages of ATR, namely that the sample is not destroyed and no sample preparation is required, have led to this technique being widely employed in forensic science investigations [26–29]. Consequently, the aim of the presented research was to develop the first ATRbased procedure enabling differentiation of red lipsticks. Additionally, chemometric methods such as principal component analysis (PCA) and cluster analysis (CA) were applied in order to facilitate discrimination and provide more extensive analyses of the obtained data. Lastly, the in-lab built ATR spectra library, enabling identification of all investigated samples, was tested by analyzing traces of random lipsticks on six different, potentially interfering surfaces. 2. Experimental The solutions (isopropanol and methanol) used throughout the experiments were supplied by Sigma-Aldrich (Germany). 38 red lipsticks (for 5 randomly selected 5 items of the same index number were bought) of a very similar hue representing 20 different manufacturers were either purchased in local shops or donated by Inglot Sp. z o.o. (cosmetics company). Detailed information about all of them is presented in Table 1. The experiments were carried out using a Thermo Nicolet iS50 FTIR (Thermo Fisher Scientific Co, Waltham, MA, USA) with a Smart Orbit micro-ATR accessory. All spectra were collected from 650 to 4000 cm1 by putting small amounts of a sample (approximately 10 mg) on the ZnSe crystal and carefully pressing it with an ATR pressure tower. After each measurement, the surface of the crystal was cleaned with 50% v/v isopropanol and then with methanol in order to avoid contamination. A new background was collected prior to the analysis of each new sample. Processing of the obtained spectra was carried out using Thermo Electron’s OMNIC 9 software. All statistical analyses were performed using Statistica 12.5 PL software (StatSoft, Tulsa, OK, USA) and OriginPro 2017 software (OriginLab Corporation, Northampton, MA, USA). 3. Applied chemometric/computational methods 3.1. Principal component analysis PCA is a well-known multivariate statistical method enabling reduction of the dimensionality of the original dataset by creating a linear combination of variables called principal components (PCs). However, the determination of the optimal number of components that should be chosen is a crucial step. Inappropriate estimation of PCs could lead to loss of information. The Scree test, Kaiser’s stopping rule and the percentage of cumulative variance are the most frequently used strategies [30], and these were applied in this study. Furthermore, PCs are a very useful tool for demonstrating the relationships between analyzed samples. 3.2. Cluster analysis In CA, samples are assembled in high dimensional space. At the beginning, each sample constitutes its own separate cluster, and then two objects, which are the closest to each other, are
131
combined. This procedure is accomplished repeatedly until all samples are arranged into one cluster. It is worth pointing out that several methods for measuring similarity and combining clusters are available. Their choice depends on the investigated samples and this should be evaluated experimentally [31]. In this study, Euclidean distances were calculated and Ward’s method was employed. Furthermore, in the literature, over 20 different rules for stopping an obtained dendrogram are known. In this article, Mojena’s stopping rule was considered to be the most efficient. According to Mojena’s stopping rule, one should select the number of groups where the following inequality (Eq. (1)) is fulfilled diþ1 > d þ ksd
ð1Þ
where d0,dn, dn1 are distances corresponding to n, n 1, . . . 1 clusters. The terms d, sd, and k are the mean, the standard deviation of the d values, and a constant, respectively [32]. As stated by Milligan and Cooper [33], the value of k should be 1.25. 3.3. Correlation method Correlation coefficients (CC) were computed as Pearson correlation coefficients by means of Thermo Scientific OMNIC software. It seems important to note that the used algorithm additionally included elimination of the effects of baseline variation. 4. Results and discussion Firstly, the parameters of the experiments were optimized. It is known that increasing the number of scans reduces the signal-tonoise level of data, and enables one to distinguish small peaks from noise, and that lower resolution can have an effect on the differentiation of peaks. Therefore, 16, 20 and 60 scans, and resolutions of 4 and 2 cm1 were tested. Despite the fact that only slight differences were observed between registered spectra, 20 scans per second with 4 cm1 spectral resolution for both background and samples were considered to be optimal. The homogeneity (intra-variability) of lipsticks was investigated by analyzing spectra obtained from three different parts of several randomly selected lipstick samples (L3, L10, L14, L21, L31, L32, L34 and L36) during one day. As demonstrated in exemplary Fig. 1a presenting the results obtained for three lipsticks, no visual differences were found on the spectra. Consequently, it was considered that the samples were homogeneous and their composition was stable during one day. Additionally, during three days the inter-variability of lipstick was examined using 5 of the previously selected lipsticks. The similarity of the spectra was calculated as correlation coefficient (L10: 99.90 0.04; L31: 99.91 0.04; L32: 99.94 0.01; L34: 99.83 0.06; L36: 99.92 006). The presented results provided the basis for claiming that the analyzed samples of the lipstick exhibited almost unchanged chemical composition also within a few days. Moreover, both above-mentioned experiments proved the high intra- and inter-repeatability of the proposed ATR-FTIR method. Finally, the variation of chemical composition of lipstick samples within the same series was also investigated during one day by analyzing five items (a–e) of five samples (L10, L31, L32, L34 and L36) from the same manufacturer with the same index number. As it can be seen in exemplary Fig. 1b — the obtained spectra for samples L36 a–e almost completely overlapped. Therefore, it was ascertained that the chemical composition of analyzed lipsticks that were both produced by the same
132
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
Table 1 List of lipsticks examined during the present study. Code of lipstick
Manufacture
Index numbera
Seriesb
L1 L2 L3 L4 L5 L6 L7 L8 L9 L10a L10b L10c L10d L10e L11 L12 L13 L14 L15 L16 L17 L18 L19 L20 L21 L22 L23 L24 L25 L26 L27 L28 L29 L30 L31a L31b L31c L31d L31e L32a L32b L32c L32d L32e L33 L34a L34b L34c L34d L34e L35 L36a L36b L36c L36d L36e L37 L38
INGRID INGRID MISSLYN MISSLYN GOSH p2 BourjoisParis e REVLON L’OREAL Paris L’OREAL Paris L’OREAL Paris L’OREAL Paris L’OREAL Paris MAYBELLINE New York SEPHORA SEPHORA SEPHORA SEPHORA CATRICE CATRICE CATRICE LOVELY WIBO MISS SPORTY EVELINE COSMETICS EVELINE COSMETICS RIMMEL ASTOR MAX FACTOR MANHATTAN INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT INGLOT MANHATTAN MANHATTAN MANHATTAN MANHATTAN MANHATTAN MANHATTAN MANHATTAN MANHATTAN
289 288 M20.94 M20.87 005 030 13 02 720 330 330 330 330 330 955 B06 R05 N 62 N 19 430 440 140 8 1 301 486 710 510 203 715 550 176 180 429 44 44 44 44 44 21 21 21 21 21 64 12 12 12 12 12 45 M 45 N 45 N 45 N 45 N 45 N 45 F 90 A
Wonder shine Wonder shine Rouge a levres Rouge a levres Velvet touch lipstick Full color lipstick Rouge Edition Longlasting lipstick Super lustrous lipstick Color riche Color riche Color riche Color riche Color riche Color sensational Rouge Baume SPF20 Rouge Cream Rounge Brilliance Color lip last Ultimate color Ultimate color Ultimate stay Creamy color Matte Intense My BFF Aqua Platinum Color edition Moisture renew Perfect stay fabulous Color elixir Soft rouge lipstick Lipstick Lipstick Lipstick matte Slim gel lipstick Slim gel lipstick Slim gel lipstick Slim gel lipstick Slim gel lipstick AMC lip pencil matte AMC lip pencil matte AMC lip pencil matte AMC lip pencil matte AMC lip pencil matte AMC lip paint HD lip tint matte HD lip tint matte HD lip tint matte HD lip tint matte HD lip tint matte Xtreme last & shine Perfect creamy & care Perfect creamy & care Perfect creamy & care Perfect creamy & care Perfect creamy & care Perfect creamy & care Perfect creamy & care
a b
Comment
High passion Cayun pepper Matt, classic red Challenge authority Rouge jet set All you need is red Fire & Ice Matte, cocorico Matte, cocorico Matte, cocorico Matte, cocorico Matte, cocorico Craving coral, matte Soft red Hot Tango Stay Together Pure red-matt Hot’n Spice Hungs-and Hibis-Kisses Behind the red curtain
My cheeky cherry red in love Mayfair red lady Ruby tuesday Watermelon
The index number is a number assigned to each lipstick in a given series. Series name refers to cosmetics collection. The same series name indicates the same collection of lipsticks.
manufacturer and had the same index number were characterized by very good repeatability. 4.1. Visual inspection of spectra As seen in Fig. 2, a typical spectrum of a lipstick sample can be divided into three main ranges. Generally, the first range (between 3740 and 3100 cm1) includes one broad band, which is attributed to water and hydroxyl groups. The second range (from 3050 to 2775 cm1) takes account of symmetric (2850 cm1) and
asymmetric (2916 cm1) C H stretching vibrations and asymmetric CH3 stretching vibrations (2954 cm1). The last range (below 1800 cm1) contains many peaks. C¼O stretching vibrations (1730 cm1), CH2 scissor deformation (1465 cm1), C¼O stretching and CH bending vibrations (1172 cm1), and CH2 rocking mode (712 cm1) are just some of them. All of the mentioned peaks could be attributed to various oils and waxes, especially castor oil and beeswax — two common ingredients in lipsticks [3]. Therefore, it can be supposed that the most informative and distinguishing range is below 1800 cm1.
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
133
Fig. 2. Typical lipstick spectrum, represented by L13.
Fig. 1. Investigation of a) homogeneity of lipsticks L21, L14 and L3; b) variability of lipsticks from the same manufacturer and with the same index number: L36 a–e.
are distinctive and consequently they can be effortlessly distinguished. Concisely, according to PCA analysis, red lipsticks can be divided into the following six groups: G1 {L34}, G2 {L32}, G3 {L31, L33}, G4 {L10}, G5 {L15, L25} and G6, which includes all the remaining samples. In an attempt to find the most efficient grouping of lipsticks, cluster analysis was performed in order to classify lipstick samples in the G6, based on similarities of their ATR spectra in the range from 1800 to 650 cm1 (the fingerprint region). Accordingly, the dendrogram presented in Fig. 4 was created, whereas the results obtained from Mojena’s stopping rule are shown in Table I in the ESM. They indicated that the value of 1.19 should be used as the stopping distance. As a consequence, the dendrogram was divided into the following four groups of lipstick samples: G6a {L1, L8, L20, L27, L28, L29, L30, L35, L36a, L37, L38,}, G6b {L4, L5, L13, L17, L21, L22,
4.2. Differentiation of red lipsticks 4.2.1. Chemometric methods A primary discrimination of lipstick samples was performed using the PCA technique, taking into account all 38 of the collected ATR spectra registered in the range from 4000 to 650 cm1. Fig. I in the Electronic Supplementary Material (ESM) shows a plot for deciding on the number of variables — a scree test. Interpretation of this plot indicates that only the first three components should be retained in the present study (the line drops precipitously), because sufficient useful chemical information is provided by these three components. They also complied with the Kaiser criterion, which only recommends components with eigenvalues greater than unity. Furthermore, approximately 90% of the total variance in the spectra was explained by them: 70.17%, 10.38%, 8.14% for the first, second, and third PC, respectively. Thus, it is reasonable to base inferences solely on these PCs. The PCA score plot of the first three PCs, demonstrating the separation of samples, is presented in Fig. 3. Additionally, in Fig. II in the ESM, two-dimensional plots are presented to make the visual analysis easier. As seen, most of the lipsticks, except L10, L15, L25, L31, L32, L33 and L34, were classified into one multi-element class. Additionally, it was noted that lipstick L15 is very similar to L25 and lipstick L31 is comparable to L33. Moreover, lipsticks L10, L32 and L34 are each completely separated from all other lipsticks. This finding suggests that some of the analyzed lipstick samples
Fig. 3. The PCA score plot of the first three PCs for all investigated lipsticks, taking into account ATR spectra registered in the range from 4000 cm1 to 650 cm1.
134
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
Fig. 4. Cluster analysis of lipsticks from G6, taking into account ATR spectra registered in the range from 1800 cm1 to 650 cm1.
L26}, G6c {L2, L3, L6, L11, L14, L16, L18, L23, L24} and G6d {L7, L9, L12, L19}. Ultimately, as a result of chemometric analysis, nine groups of red lipstick samples with comparable chemical compositions were created: G1–G5 and G6 a–d. It is worth noting that according to chemometric classification, all the investigated lipstick samples manufactured by Manhattan were classified into one group (G6 a). On the other hand, the lipstick samples from Sephora were categorized into various groups (G6 b, c, G5), while some of lipstick samples manufactured by Inglot were placed in group G6 a and some in G1–G3. 4.2.2. Visual classification A visual comparison of the analyzed spectra was also performed. In the first step of differentiation between lipstick samples, one should look at both the relative intensities and the degree of resolution of bands from 3050 to 2775 cm1. As can be seen in Fig. 5, this region enables differentiation between all groups, except between G2 and G6, and between G4 and G5. Further distinction between groups was based on the bands occurring below 1800 cm1. As is evident in Fig. 6a, the most characteristic feature for G5 is a lack of band at 1745 cm1 corresponding to the carbonyl group. Moreover, in G4 one can see a weak doublet at 1375 and 1363 cm1, whereas in G5 there is only a band at 1363 cm1. Both of these bands are due to CH3 deformation [34]. The differentiation of G2 from G6 is presented in Fig. 6b. On the face of it, spectra obtained for these groups are completely unlike each other in the range from 1800 to 650 cm1. The most noteworthy difference is strong absorption near 1000 and 1100 cm1 in G2. It is highly possible that these peaks correspond to phosphates and ethers or alcohol–ethers, respectively, which are found in many plasticizers [34]. Moreover, the differences between G6 a–d are selected in Fig. 6c — only representative spectra of lipstick samples from the groups have been presented. As far as one can see, the region where differences are most clearly visible is from 1400 to 800 cm1. In this region, a little broad band (especially in G6 a–c) at 1150 cm1 characteristic for C O stretch in nonaromatic ethers is observable. Moreover, the existence of a siloxane group is most likely due to a band occurring at 1108 cm1, which is most visible in spectra from
G6 b. Besides, strong absorption in the range from 1100 to 1000 cm1 is characteristic for G6 a, b and, as was mentioned before, these bands are connected with phosphates and ether/ alcohol–ethers [34]. It is noteworthy that there are also three unique bands for G6 d — at approximately 1387, 1376 and 1363 cm1. Summarizing, visual comparison of the obtained spectra led to confirmation of the classification ascertained according to the chemometric results. 4.2.3. Correlation method In order to make the analysis more objective as well as to verify the existence of an unequivocal dependence between the composition of the lipstick samples and the manufacturer (or series), the correlation method was applied. At first, the variability of the same lipstick sample, s, was ascertained as the mean CC (correlation coefficient) value calculated for five items (a-e) of the five samples (L10, L31, L32, L34 and L36) of lipsticks of the same index number, the same series and the same manufacturer. The mentioned samples were representatives of different PCA groups so they characterized by different chemical composition. It is worth pointing out that the CCs value were evaluated by analyzing samples on three successive days and they ranged between 0.04% and 0.33%. On this basis the following assumptions related to ATR spectra of each examined pair of lipstick samples have been made: if CC 100% — 3s, it is highly possible that the spectra correspond to the same lipsticks, if CC < 100% — 3s, the spectra are considered as coming from different lipsticks. In short, the similarity limit in the case of the present study is a value of 99.01%. Comparing the spectra of lipsticks produced by different manufacturers, it was ascertained that they are usually characterized by CCs considerably less than 99.00% (see Table II in ESM), e.g., for L15 (Sephora) and L3 (Misslyn), the CC = 55.81% or for L6 (p2) and L20 (Wibo), the CC = 92.58%. There were a few cases where the compositions of two lipsticks originating from different
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
135
Fig. 5. Differentiation between groups based on the range from 3050 cm1 to 2775 cm1.
manufacturers were very similar, e.g., for L3 (Misslyn) and L22 (Eveline Cosmetics) CC = 99.45%, for L29 (Inglot) and L4 (Misslyn) CC = 99.55%. If the investigated samples were from the same manufacturer but from different series, most of them could also be easily distinguished. It can be seen in Table 2 that CCs between spectra of lipstick samples produced by Inglot, Manhattan and Sephora were not higher than 98.50%. Moreover, it is evident that lipsticks manufactured by Inglot are characterized by the most variable composition (CC from 96% to 24%). The obtained results confirmed the theoretical assumption that lipstick samples from the same manufacturer and from the same series are more similar to each other than lipstick samples from different series (see CC for L28 and L29 from Inglot, or L36 and L37 from Manhattan). On the other hand, some exceptions can be observed: sample L35 from the Xtreme last & shine series is undistinguished (CC between 99.11% and 99.49%) from the samples from the Perfect creamy & care series. Moreover, the data suggest that a comparison of samples with a similar index number (from the same manufacturer and from the same series) can cause problems. In these cases, the CCs between samples were calculated to be greater than 99.01%, and in consequence, they were not differentiable with the use of the presented method. The discrimination of such lipsticks samples can be problematic and
they may differ to a very limited extent, e.g. only slightly in the concentration of colour additives. However, further data are needed in order to confirm this assumption. It has not escaped our notice that there were also lipsticks produced by the same manufacturer in the same series with an index number differing only by one, e.g., L1 — Ingrid no. 289 and L2 — Ingrid no. 288, which revealed, without any doubt, two different spectra (CC = 98.52%). It seems important to reiterate that samples from the same manufacturer, the same series and having the same index number (for instance L36 a–e), as was mentioned before, were considered to be indistinguishable. The CC between spectra was much higher than 99.01%. 4.3. Analysis of lipsticks traces on various surfaces — preliminary examination Due to the fact that lipstick samples can be found left on various surfaces, it is justified to investigate the possibility of identifying lipstick traces in the form of such smears. Firstly, one of the samples (L28) was left as a lipstick mark on various surfaces in order to assess which surfaces caused interfering bands. The surfaces studied included: envelope, paper (80 g/m2), tissue (100% cellulose), cigarette butts, and a transparent plastic cup, as well as the white collar of a shirt. As can be seen in
136
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138 Table 2 Spectra correlation coefficients between samples from the same manufacturer: a) Inglot; b) Manhattan; c) Sephora. Lipsticks samples from the same series are bolded. a) L28 L29 L30 L31 L32 L33 L34
b) L27 L35 L36a L37 L38
c) L12 L13 L14 L15
100.0 99.57 96.44 91.67 90.24 79.95 27.11 L28
100.0 96.70 91.50 90.90 79.48 25.90 L29
100.0 98.46 97.79 98.24 97.93 L27
100.0 94.87 95.95 62.50 L12
100.0 87.83 95.44 74.51 24.67 L30
100.0 83.29 88.50 40.22 L31
100.0 99.11 99.49 99.26 L35
100.0 99.65 99.83 L36a
100.0 96.69 57.46 L13
100.0 65.70 17.65 L32
100.0 53.76 L33
100.0 99.78 L37
100.0 55.17 L14
100.0 L34
100.0 L38
100.0 L15
In striving to find the most efficient approach to identification of lipstick marks, two spectral subtractions were performed: (1) in the whole range of registered spectra, (2) in the fingerprint region. The effectiveness of the proposed approach to the identification of lipsticks marks was verified by analyzing six different traces of red lipstick (L7, L12, L18, L23, L28 and L35) smudges on all the abovementioned surfaces. Overall, the procedure was carried out in a few steps by: measuring the surface spectrum and the spectrum of the lipstick on the same surface, then subtracting one from the other in the whole range of the rejected spectrum and in the fingerprint region (from 1800 to 650 cm1), and subsequently identifying the spectrum by using an in-lab built ATR spectra library. The obtained results indicate that the second approach solely enables one to connect lipstick marks with the manufacturer and index number (see Table 3). It should be highlighted that the described subtraction approach did not yield pure spectra of lipstick samples — additional or negative bands appeared in some cases. Nevertheless, a comparison of the spectral subtraction result with the spectra collected in the in-lab built library enabled – in every case without exception – the successful identification of lipstick samples.
5. Conclusions
Fig. 6. Differentiation of lipsticks according to ATR spectra registered in the range from 1800 cm1 to 650 cm1: a) G4 represented by L25, G5 represented by L10; b) G2 represented by L32, G6 represented by L12; c) G6a represented by L35, G6b represented by L26, G6c represented by L3, and G6d represented by L19.
Fig. III in the ESM, all of the analyzed surfaces, except the transparent plastic cup, provided additional bands to the lipstick spectra. In order to isolate lipstick spectra from surface spectra, Thermo Electron’s OMNIC software was used. This is one of the most difficult issues in the identification of samples in forensic science.
ATR spectroscopy was successfully used for the differentiation of 38 red lipsticks of a very similar hue. Although no unambiguous trends could be identified within manufacturers or series and the differentiation of lipsticks is a complicated issue, combining the two chemometric methods (PCA and CA) enabled the distinguishing of nine groups from investigated lipsticks of similar ATR spectra. Similarities cannot be assumed, in advance, on the basis of manufacturer, series or index number compatibility. However, visual analysis revealed bands that were characteristic for appropriate groups and confirmed the classification obtained by chemometric techniques. Furthermore, the approach supported by the correlation coefficient enabled deeper differentiation of all studied lipstick spectra. And, importantly, the differentiation could be achieved non-destructively in a matter of minutes.
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
137
Table 3 A comparison of the two approaches for the identification of lipsticks in the form of lipstick marks smudged on interfering surfaces. Correct identification of lipsticks is indicated by “+”. Lipstick Code
The range of spectral subtraction 4000–650 cm1
L7 L12 L18 L23 L28 L35
1800–650 cm1
Cigarette butts
Envelope
Paper
Tissue
White collar of shirt
Cigarette butts
Envelope
Paper
Tissue
White collar of shirt
+ + + + + +
+ + + + + +
+ + + + + +
+ + + + + +
+ + + + + +
Table 4 Discriminating power of PCA, CA and CC analysis. Method
Discriminating power (DP)
PCA CA Correlation method
0.29 0.51 0.93
To sum up, on the basis of the performed PCA, CA and CC differentiations, the discriminating power (DP) of the ATR analysis of lipstick samples has been calculated according to the following equation: DP = 1 2 M/(N(N 1)) (where M is number of matching pairs; N is total number of analysed lipsticks). The obtained DP values were listed in Table 4. These results confirmed the above-mentioned conclusion that PCA and CA enables some preliminary classification whereas the correlation coefficient is the best choice for the lipsticks discrimination. Ultimately, by using the developed method and in-lab built ATR spectra library, it was possible to detect and connect some lipsticks left as lipstick marks with their manufacturer and index number, even on interfering surfaces. This finding confirms that the ATR technique is a useful tool in lipstick differentiation and identification. All things considered, it seems reasonable to assume that the proposed method shows promise for future application in forensic science investigations as a good alternative to routinely used methods. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.forsciint.2017. 09.019.
References [1] S. Singh, V. Saran, M. Mishra, A.K. Gupta, Forensic examination of lipstick by the various physio-chemical and instrumental method, Int. J. Soc. Relev. Concern 3 (2015) 1–7. [2] M. López-López, N. Özbek, C. García-Ruiz, Confocal Raman spectroscopy to trace lipstick with their smudges on different surfaces, Talanta 123 (2014) 135–139. [3] F. Salahioglu, M.J. Went, Differentiation of lipsticks by Raman Spectroscopy, Forensic Sci. Int. 223 (2012) 148–152. [4] B. Joshi, K. Verma, J. Singh, A comparison of red pigments in different lipsticks using thin layer chromatography (TLC), J. Anal. Bioanal. Tech. 4 (2013) 1–4. [5] Y. Ehara, Y. Marumo, Identification of lipstick smears by fluorescence observation and purge-and-trap gas chromatography, Forensic Sci. Int. 96 (1998) 1–10. [6] D. Li, Z. Wang, L. Wang, X. Xu, H. Zhang, Ultrasonic extraction coupled with capillary electrophoresis for the determination of azo dyes in lipsticks using ionic liquid as dynamic coating and background electrolyte, Chin. J. Chem. 29 (2011) 147–152.
[7] S. Scalia, S. Simeoni, Assay of xanthene dyes in lipsticks by inverse supercritical fluid extraction and HPLC, Chromatographia 53 (2001) 490–494. [8] S.C. Rastogi, V.J. Barwick, S.V. Carter, Identification of organic colourants in cosmetics by HPLC-diode array detection, Chromatographia 45 (1997) 215– 228. [9] Y. Xian, Y. Wu, X. Guo, Y. Lu, H. Luo, D. Luo, Y. Chen, Simultaneous determination of 11 restricted dyes in cosmetics by ultra-high-performance liquid chromatography/tandem mass spectrometry, Anal. Methods 5 (2013) 1965–1974. [10] Q. Ma, H.-W. Xi, C. Wang, H. Bai, G.-C. Xi, N. Su, L.-Y. Xu, J.-B. Wang, Determination of ten volatile nitrosamines in cosmetics by gas chromatography tandem mass spectrometry, Chinese J. Anal. Chem. 39 (2011) 1201–1207. [11] É. Ferreira Batista, A. dos Santos Augusto, E.R. Pereira-Filho, Chemometric evaluation of Cd, Co Cr, Cu, Ni (inductively coupled plasma optical emission spectrometry) and Pb (graphite furnace atomic absorption spectrometry) concentrations in lipstick samples intended to be used by adults and children, Talanta 150 (2016) 206–212. [12] N.M. Hepp, Determination of total lead in 400 lipsticks on the U.S. market using a validated microwave-assisted digestion, inductively coupled plasmamass spectrometric method, J. Cosmet. Sci. 63 (2012) 159–176. [13] S. Gunduz, S. Akman, Investigation of lead contents in lipsticks by solid sampling high resolution continuum source electrothermal atomic absorption spectrometry, Regul. Toxicol. Pharm. 65 (2013) 34–37. [14] M.A. Gondal, Z.S. Seddigi, M.M. Nasr, B. Gondal, Spectroscopic detection of health hazardous contaminants in lipstick using Laser Induced Breakdown Spectroscopy, J. Hazard. Mater. 175 (2010) 726–732. [15] A. dos Santos Augusto, E. Ferreira Batista, E.R. Pereira-Filho, Direct chemical inspection of eye shadow and lipstick solid samples using laser-induced breakdown spectroscopy (LIBS) and chemometrics: proposition of classification models, Anal. Methods 8 (2016) 5851–5860. [16] E. Da Silva, A. Matthews David, A. Pejovi c-Mili c, The quantification of total lead in lipstick specimens by total reflection X-ray fluorescence spectrometry, XRay Spectrom. 44 (2015) 451–457. [17] Ch.Ch. Wang, A.N. Masi, L. Fernandez, On-line micellar-enhanced spectrofluorimetric determination of rhodamine dye in cosmetics, Talanta 75 (2008) 135–140. [18] I. Al-Saleh, S. Al-Enazi, Trace metals in lipsticks, Toxicol. Environ. Chem. 93 (2011) 1149–1165. [19] I. Al-Saleh, S. Al-Enazi, N. Shinwari, Assessment of lead in cosmetic products, Regul. Toxicol. Pharm. 54 (2009) 105–113. [20] P. Piccinini, M. Piecha, S.F. Torrent, European survey on the content of lead in lip products, J. Pharm. Biomed. Anal. 76 (2013) 225–233. [21] P. Gardner, M.F. Bertino, R. Weimer, E. Hazelrigg, Analysis of lipsticks using Raman spectroscopy, Forensic Sci. Int. 232 (2013) 67–72. [22] F. Salahioglu, M.J. Went, S.J. Gibson, Application of Raman spectroscopy for the diff ; ;erentiation of lipstick traces, Anal. Methods 5 (2013) 5392–5401. [23] C. Rodger, V. Rutherford, D. Broughton, P.C. White, W.E. Smith, The in-situ analysis of lipsticks by surface enhanced resonance Raman scattering, Analyst 123 (1998) 1823–1826. [24] G. Misra, V.K. Mittal, Neutron activation analysis of lipsticks using g-ray spectrometry, J. Appl. Spectrosc. 71 (2004) 270–274. [25] A.F.L. bin Abdullah, Y. Marimuthu, Ch. Kah Haw, N.F. binti Mohamad Said, N.Z. M. Muslim, N.F.N. Hassan, M.H. Yaacob, Y. Chong Hooi, Forensic discrimination of lipsticks by thin layer chromatography and gas chromatography—mass spectrometry, Malays. J. Forensic Sci. 2 (2011) 22–28. [26] J. Manheim, K.C. Doty, G. McLaughlin, I.K. Lednev, Forensic hair differentiation using attenuated total reflection fourier transform infrared (ATR FT-IR) spectroscopy, Appl. Spectrosc. 70 (2016) 1109–1117. [27] M.C.A. Marcelo, K.C. Mariotti, M.F. Ferrao, R.S. Ortiz, Profiling cocaine by ATRFTIR, Forensic Sci. Int. 246 (2015) 65–71. [28] M. Szafarska, M. Wo zniakiewicz, M. Pilch, J. Zie˛ba-Palus, P. Koscielniak, Computer analysis of ATR-FTIR spectra of paint samples for forensic purpose, J. Mol. Struct. 924–926 (924) (2009) 504––. [29] C.M. Orphanou, The detection and discrimination of human body fluids using ATR FT-IR spectroscopy, Forensic Sci. Int. 252 (2015) e10–e16.
138
M. Gładysz et al. / Forensic Science International 280 (2017) 130–138
[30] B.R. Kowalski (Ed.), Chemometrics: Mathematics and Statistics in Chemistry, Reidel, Dordrecht, 1984. [31] D.L. Massart, L. Kaufman, The Interpretation of Analytical Chemical Data by the Use of Cluster Analysis, John Wiley & Sons, New York, 1983. [32] R. Mojena, Hierarchical grouping methods and stopping rules: an evaluation, Comput. J. 20 (1977) 359–363.
[33] G.W. Milligan, M.C. Cooper, An examination of procedures for determining the number of clusters in a data set, Psychometrika 50 (1985) 159–179. [34] G. Socrates, Infrared and Raman characteristic group frequencies, Tables and Charts, John Wiley and Sons, Ltd., Chichester, 2001.