Organic and inorganic discrimination of ballpoint pen inks by ToF-SIMS and multivariate statistics

Organic and inorganic discrimination of ballpoint pen inks by ToF-SIMS andmultivariate statistics

John A. Denman a, William M. Skinner a, K. Paul Kirkbride b, Ivan M. Kempson a,c,*a Ian Wark Research Institute, University of South Australia, Mawson Lakes, SA 5095, Australiab Australian Federal Police, Forensic and Data Centres, GPO Box 401, Canberra, ACT, Australiac Institute of Physics, Academia Sinica, 128 Academia Road, Section 2, Nankang, Taipei 115, Taiwan

Applied Surface Science 256 (2010) 2155–2163

A R T I C L E I N F O

Article history:

Received 19 June 2009

Received in revised form 17 September 2009

Accepted 17 September 2009

Available online 25 September 2009

Keywords:

Document analysis

Trace analysis

Mass spectrometry

Ink markings

Principal component analysis

Surface analysis

A B S T R A C T

Surface analysis by ToF-SIMS analysis of ballpoint pen ink markings was performed for discrimination.

ToF-SIMS provided non-destructive analysis of ink’s organic and inorganic components directly off paper

with no interference from the paper substrate. Organic and inorganic information were collected

simultaneously and processed with PCA, discriminating 41 out of 45 pairs (91%) of pens analysed.

Minimal sample preparation and analysis time, the simultaneous acquisition of organics and metals, and

ability to analyse trace amounts gives this technique advantages over others currently utilised in the

forensic field. Simultaneous acquisition of organics and inorganics has not been presented before for the

characterisation of these materials. It was indicated that pens from the same manufacturer, but discrete

batches, can be significantly different.

� 2009 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Applied Surface Science

journa l homepage: www.e lsev ier .com/ locate /apsusc

1. Introduction

The analysis of questioned documents is an important area offorensic science investigating the authenticity of a document basedon marks from a variety of instruments including pencils, ballpointor fountain pens, inkjet inks and photocopy/printer toners [1]. It isdifficult however to analyse ink markings with a non-destructivemethod that does not suffer from spectral contamination by theunderlying substrate. There are however deficiencies in standardpractices and discriminating power [2]. Many analytical methodshave been applied to discriminating inks, each offering particularadvantages. For maximum discrimination a range of techniques islikely to be required [3].

Examination of ink may be important for comparing two ormore ink entries on one or more documents to determine if theywere written with the same instrument [4]. This can give insightinto whether entries have been added or altered. In the literature,the general consensus is that while it is difficult to determinewhether an individual pen was used to write a document, it ispossible to identify an ink type or manufacturer [5,6]. Despite the

* Corresponding author at: Ian Wark Research Institute, University of South

Australia, Mawson Lakes Blvd, Mawson Lakes, SA, 5095, Australia.

Tel.: +61 8 8302 3495; fax: +61 8 8302 3683.

E-mail address: [email protected] (I.M. Kempson).

0169-4332/$ – see front matter � 2009 Elsevier B.V. All rights reserved.

doi:10.1016/j.apsusc.2009.09.066

prevalence of inkjet and laser printers and copiers in homes andoffices, it is reported that around 80% of questioned documentsrequiring analysis contain ballpoint pen ink [7].

Most analysis has been on dyes and other organic componentspresent in ballpoint pen inks. It was the aim of this work to exploreTime-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS) forcharacterising and discriminating such organic components, whilealso extending this knowledge to that of inorganic components. Itis hypothesised that an ink may contain an ‘elemental profile’ thatmay be used for the purposes of differentiation, either comple-mentary to or discrete from organic data. ToF-SIMS has previouslybeen found to be a suitable method for the analysis of pencil [8]directly off paper, in a non-destructive manner. ToF-SIMS cansimultaneously collect organic and inorganic information and thusmay be useful for characterising and discriminating ballpoint peninks. Advance statistical analysis can provide additional informa-tion for fundamental analysis and discrimination.

1.1. Ink composition

Ballpoint pen ink composition can be complex depending on anumber of factors such as the colour, tackiness, drying time and thetype of writing instrument. Essentially, a ballpoint pen ink consistsof synthetic dyes in a glycol-based solvent. The dyes (acidic and/orbasic) and pigments (organic and/or inorganic) make up about 25%of the formulation, while the solvent makes up about 50% by

Table 1Summary of analysed dyes: chemical name; common name(s); structure; peak type; and mass monitored. ‘M’ denotes molecular ion, ‘A’ denotes anion.

Dye Other name(s) Structure Peak m/z

Basic Violet 1 Methyl Violet (2B) [M�Cl]+ 358

Basic Violet 3 Crystal Violet, Methyl Violet 10B [M�Cl]+ 372

Basic Yellow 2 Auramine O [M�Cl]+ 268

Basic Blue 26 Victoria Blue (B) [M�Cl]+ 470

Basic Green 1 Brilliant Green 1 [M�Cl]+ 385

Basic Green 4 Malachite Green [M�Cl]+ 329

Basic Orange 14 Acridine Orange [M�Cl]+ 266

Basic Blue 9 Methylene Blue [M�Cl]+ 284

Basic Blue 7 Victoria Blue BO [M�Cl]+ 478

Solvent Blue 2 Neptun Blue [M�Cl]+ 484

Basic Violet 14 Fuchsin, Rosanilin [M+H]+ 302

J.A. Denman et al. / Applied Surface Science 256 (2010) 2155–21632156

Table 1 (Continued )

Dye Other name(s) Structure Peak m/z

Solvent Blue 23 Spirit Blue [M�A]+ 516

Solvent Orange 3 Chrysodine Y Base [M+H]+ 213

Ditolyl guanidine [5_TD$DIFF]– [M+H]+ 240

Diphenyl guanidine [5_TD$DIFF]– [M+H]+ 212

J.A. Denman et al. / Applied Surface Science 256 (2010) 2155–2163 2157

weight [9]. The remainder is a range of additives that may includeresins, viscosity adjusters, antioxidants, surfactants, softeners forink flow ability and fatty acids for ball lubrication.

Most dyes are hydrophobic in nature, composed of severalderivatives with similar structures [10]. Basic dyes (most common)based on triarylmethane and rhodamine, or acid dyes derived fromdiazo compounds or phthalocyanine are used [4]. Dye chemistrydetermines colour, hence different dyes will be found whencomparing blue, black and red inks. Basic dyes have a characteristicstructure containing an iminium group and benzene andnaphthalene rings together with an extended conjugated system.The structures of some common dyes, shown in freebase form, areillustrated in Table 1.

Around 80% of blue and black inks contain the polymethylatedBasic Violet 3 and its homologues [4]. Blue inks may also containBasic Blue 26, Basic Blue 7, Solvent Blue 2 and Solvent Blue 23 dyes.Black inks are additionally characterised by the presence of thedyes Basic Red 1:1, Basic Blue 26, Solvent Black 7 and Acid Yellow36.

The dyes are non-volatile due to their cationic part and remainon the paper long after the solvent and other components havevolatilised. Evaporation of the solvent components is initiallyintensive, and decreases quickly in the first hour after the ink hasbeen applied to the paper [11]. Hence, while other additives in anink formulation may not be easily detectable, analysis of thecommon dyes usually provides satisfactory discriminationbetween different formulations [5].

The solvents (commonly phenoxyethanol, phenoxyethoxyetha-nol, dipropylene glycol, benzylalcohol, butylene glycol, phthalicanhydride or 2-pyrrolidone) in ink act to dilute the colourants andfacilitate its application to the paper. Oleic acid can be used as alubricant to allow the ballpoint to rotate easily. Another additivecan be the aryl guanidine group of chemicals, with structuresshown in Table 1. These are added to form salts with the acidic dyesand raise the pH of the ink [4]. Softening agents, such as phthalates;corrosion inhibitors, such as organophosphates; antioxidants, suchas butylated phenols; and surfactants, such as alkylamines, areother common additives [9].

1.2. Document analysis

Non-destructive techniques are the first to be carried out andused in preference to destructive chemical methods. Visualexamination scrutinizes colour and the morphology of the line[12,13]. Optical techniques utilising examination with visible,ultraviolet (UV) and infrared (IR) light are common [1,14].

IR examination has been extensively researched, particularlyFourier transform, Raman and microscopical reflection–absorptionIR methods [6,15–19]. Rather than specifically identifying indivi-dual components in inks, these methods provide a spectral profileto compare inks for grouping or differentiation. Wang et al. [6]developed a method for classifying blue ballpoint pen inks usingFourier transform IR, that allowed a total of 108 samples to bedivided into 35 subgroups using a pattern recognition system. Thismethod was found to be fast and reliable for bulk ink analysis, buthad yet to be applied to inks on paper.

IR experiments utilising synchrotron sources have also beeninvestigated [16,17]. The high lateral spatial resolution coupledwith the sensitivity of the technique improves spectra, andanalysis can be performed non-destructively off the paper. Despitethe complexity of these spectra due to the overlap of manyvibrational bands, Wilkinson et al. [16] used synchrotronreflectance IR spectroscopy to distinguish a number of inks.Similarly, Perry et al. [17] were able to use synchrotron IRmicrospectroscopic techniques to measure the ageing profile indrying ink by analysing both the gross and subtle changes in thevibrational spectra of the inks components.

Diffuse reflectance infrared spectrometry (DRIFT) has also beenused, but has its limitations. When analysing ink on paper, thetechnique experiences interference from the paper substrate andwhile extracting the ink is possible, this requires a large sampleand is somewhat destructive [4].

Due to the nature of forensic evidence, there is a preferencetowards non-destructive techniques, providing chemical analysiswhile maintaining sample integrity. Laser-desorption massspectrometry (LD-MS) is a method that is minimally destructiveand provides molecular level information concerning an ink’scomposition, including dyes, which is often not possible due to thevolatility requirement of similar techniques such as GC–MS [20–23]. GC–MS can be used in a non-destructive way to analysesolvents with use of head space solid phase micro-extraction [24].Weyermann et al. utilised laser-desorption ionization massspectrometry (LDI-MS) for the analysis of dyes from inks, andcompared its ability to differentiate blue ballpoint pen inks to thestandard method of high-performance TLC (HPTLC) [25]. From 31ink markings on paper, 26 classes could be classified by LDI-MScompared to 18 for HPTLC. Desorption electrospray ionizationmass spectrometry can analyse ink directly off paper with imagingcapability [26]. Analysis of inks by luminescence spectroscopydirectly off paper has also been shown to have potential [27].Coupling with principal component analysis demonstrated theusefulness of advanced statistical techniques.

J.A. Denman et al. / Applied Surface Science 256 (2010) 2155–21632158

Once non-destructive techniques have been exhausted,destructive techniques are often used if further discrimination isrequired such as thin layer chromatography (TLC) and high-performance liquid chromatography (HPLC) [5,7,18,28,29]. TLC hasbeen used with higher discriminating power than other standardforensic protocols, such as microspectrophotometry and filteredlight examinations [30] and new image analysis software mayfurther refine the discrimination [31]. HPLC coupled with aphotodiode array system has been used to differentiate blue andblack ballpoint pen inks [5]. A number of statistical classificationmethods, including principal component analysis (PCA), were usedthat allowed successful discrimination of some, but not allmanufacturers.

Other techniques include capillary electrophoresis (CE) [10,32],inductively coupled plasma mass spectrometry (ICP-MS) [33,34],laser ablation ICP-MS [35], field desorption mass spectrometry(FD-MS) [14] and ion-electrospray ionization mass spectrometry(ESI-MS) [4]. FD-MS, ESI-MS, ICP-MS, and to a slightly lesser extentLA-ICP-MS, remain semi-destructive techniques as they requirethe dissolution or removal of the ink from the paper in order toperform the analysis.

Time-of-flight secondary ion mass spectrometry (ToF-SIMS) is atechnique with particular advantages in analysing inks forpurposes of questioned document analysis. Its ability has beenshown for the analysis of pencils previously [8]. Advantages of thistechnique were identified to be rapid, non-destructive analysiswith minimal sample preparation; simultaneous detection oforganic and inorganic species; surface sensitivity allowing nointerference from underlying matrices; and sensitivity such thatanalysis could be performed on very small areas down to tens ofsquare microns.

While analysis of dyes and other organic components inballpoint pen inks has been well investigated, little researchhas been performed in the area of inorganic analysis. Inorganicelements may be present in a number of components, includingpigments, viscosity adjusters, ball lubricants, antioxidants andsurfactants. It is hypothesised that the amounts and types ofsuch additives would vary between manufacturers and evenbetween different models of pen from the same manufacturer,and hence be used as a basis of discrimination. Vogt et al. usedthe non-destructive technique of particle induced X-ray emis-sion (PIXE) to analyse ballpoint pen inks directly off paper [10].In most inks, Cu, Ni, Zn and Pb were found and sometimesalso Ti, Fe, Cr and Co. It was concluded that a reliabledifferentiation between a large number of samples was notpossible, although the Cu/Zn ratio allowed most samples to begenerally distinguishable. A comparison of the absoluteamounts of the trace elements present was reported to be‘useless’ due to the differences in line width, pressure appliedduring writing and penetration depth of the ink into the paper[10].

Table 224 pens were analysed from a different manufacturers and countries.

Pen

Pens 1–4 Bic Cristal Medium

Pen 5* Bic Cristal Medium

Pens 6–10 Papermate Kilometrico Medium

Pen 11* Papermate Kilometrico Medium

Pens 12–13 Bic Clic 2000 Medium

Pen 14* Bic Clic 2000 Medium

Pens 15–16 Pilot Begreen Rexgrip Medium

Pens 17–18 Papermate Flex Grip Elite Mediu

Pens 19–21 Zebra Tapli Clip Medium

Pens 22–24 Staedtler Stick 430 Medium

* These pens were purchased 7 months earlier than the others of the same make an

Elemental composition of inks has also been analysed by X-ray fluorescence (XRF) [19,36]. One study showed the presenceof S, Cu, Si and P in all samples, as well as Zn, Cl, Br and Ca insome. The technique was used to complement IR and Ramanallowing 90% of the 70 samples to be distinguished [19].

In summary, many techniques have been applied to theanalysis of inks, both on extracted samples as well as directly offa paper substrate. Each has strengths and weaknesses in regardto the information they provide, the level of differentiationpossible and the degree of destructiveness. Mass spectrometricanalysis appears to be an emerging method which can providequalitative data about components in a manner that is leastdestructive to the questioned document and does not sufferfrom interference from the substrate. Incorporation of statisticalmethods such as principal component analysis adds additionalscrutiny [27].

2. Experimental

2.1. Samples

A total of 24 blue ballpoint pens (including replicates)were purchased at a supermarket and news agency (Table 2)which represented seven models of pen from five differentmanufacturers. Duplicates of 3 of those models were purchased7 months earlier for the purpose of exploring inter-batchvariations. Samples were prepared by drawing a 3 cm line onregular 100gsm Reflex white paper (PaperlinX Ltd.).

2.2. ToF-SIMS

ToF-SIMS experiments were performed using a PhysicalElectronics Inc. Model 2100 PHI TRIFT II instrument equippedwith a pulsed liquid metal 69Ga+ primary ion gun. This wasoperated in positive SIMS mode at 15 kV energy, 2 nA beamcurrent with a pulse length of 20 ns. Charge compensation withan electron flood gun acted to neutralise surface charging. Eachspectral analysis was performed over a 100 mm � 100 mm rasterarea, with 1 min acquisitions performed for a total of 10 analysesper sample. Development of this analysis protocol has beendiscussed previously [8]. In this instance, 30 spectra wereinitially acquired from the Staedtler pen. Individual peaks werenormalised to the sum of all peaks selected and a running 95%confidence interval was calculated for the average normalisedpeak intensity as each spectrum was taken into account. Tenanalyses were chosen to be a balance between acquisition timeand minimising errors, hence a total of 240 spectra wereanalysed for the 24 pens. Table 3 summarises the organiccomponents for which spectra were assessed. For inorganicinformation, the spectra were analysed for any prominent

Manufacturer

Bic Australia, origin unknown

Bic Australia, origin unknown

Sanford Australia, ‘made in Malaysia’

Sanford Australia, ‘made in Malaysia’

Bic Australia, ‘made in New Zealand’

Bic Australia, ‘made in New Zealand’

Pilot, ‘made in Japan’

m Sanford USA, ‘assembled in Mexico’

Beautone Australia, ‘made in Japan’

Staedtler Pacific, ‘assembled in Aust’

d model.

Table 3Organic components identified in ballpoint pen inks from ToF-SIMS spectra.

Pen BY2 BV1 BV3 BV14 BB7 BB26 AGs

1–4 U U U

5 U U U

6–10 U U U

11 U U U

12–13 U U U

14 U U U U U

15–16 U U U

17–18 U U U

19–21 U U U

22–24 U U U

J.A. Denman et al. / Applied Surface Science 256 (2010) 2155–2163 2159

elemental peak detected in the inks (Na, Mg, Al, Si, K, Ca, Ti, V, Cr,Mn, Fe, Co, Ni, Cu, Zn and Pb).

2.3. Data analysis

ToF-SIMS spectra were calibrated using the Wincadencesoftware (Physical Electronics Inc.) based on prominent knownpeak identification (CH3, C2H5 and C3H7). Peaks of interest wereselected and integrated using ToFPak plug-in (Physical Electro-nics Inc.) for Matlab (Math-works Inc.). Integrated peak values(i.e. number of counts) of the selected ions were normalised tothe total selected secondary ion intensities to correct fordifferences in total ion yield between analyses and samples.These normalised values were averaged over the 10 spectraacquired for each pen. Data were then processed withSTATISTICA 7 software (StatSoft Inc.) utilising the principalcomponents and classification analysis module. The number ofpossible pairs for discriminatory purposes was equated to n!/2(n � 2)!, hence when comparing any two pens out of the ten(comprising the 7 models and 3 older duplicates), 45 possiblepair combinations existed.

Fig. 1. An example of a positive ion ToF-SIMS spectra from the Bic Clic 2000 Medium ‘old

Basic Violet 1 (BV1) and Basic Violet 3 (BV3) are clearly identifiable.

3. Results and discussion

3.1. Homogeneity and reproducibility

An example of a ToF-SIMS positive ion spectrum from a BicClic 2000 Medium sample is given in Fig. 1. As a test ofhomogeneity and to ascertain a reasonable number of analysesto perform per ink sample, analysis was originally performed on30 different positions on a Staedtler sample. Cumulative meansof peak intensities and 95% confidence intervals of the meanwere calculated (Fig. 2). Only the first 20 points are shown as thefinal ten were redundant. Although 15–20 analyses are neededto minimise the confidence interval to a point where it isessentially constant, approximately 85–90% of the variationoccurred in the first 10 analyses. The errors associated with themeasurements are a combination of the instrumental error andthe heterogeneity inherent in the sample. As a compromisebetween data quality and analysis time, it was decided that datacollection would consist of 10 analyses at random positionsalong each ink line. While instrumental error cannot becompletely removed, this methodology aimed to average outand minimise any errors.

3.2. Analysis of organic components

Dyes are most often used as the component for discriminationbetween different ballpoint pen inks. While the solvents and otherorganic components volatilise, the dyes are non-volatile andremain on the paper. Detection of many of the organic additives isnot possible due to their volatile nature and the ultra-high vacuumconditions of ToF-SIMS. The following summary for each sample isbased upon visual inspection of the spectra and identification ofdominant parent and fragmentation ions indicative of the commondyes that were easily observable. The results are summarised inTable 3. Nomenclature of ‘old’ and ‘new’ samples refers to those

’ sample. Peaks for the Aryl Guanidines, Basic Yellow (BY2), Basic Violet 14 (BV14),

Fig. 2. 95% confidence interval minimisation; results shown for a number of

inorganic species as well as organic dyes Basic Violet 1 (BV1), Basic Violet 3 (BV3)

and Basic Blue 26 (BB26).

J.A. Denman et al. / Applied Surface Science 256 (2010) 2155–21632160

where replicates were bought 7 months earlier for the purposes ofinvestigating different batches.

3.2.1. Bic Clic 2000 Medium ‘old’

A peak at 240 Da was identified to be an aryl guanidinederivative (ditolyl guanidine), with a smaller peak at 212 Daprobably present due to another form of aryl guanidine (diphenylguanidine). A strong signal at 268 Da corresponded to Basic Yellow2, a common blue ink dye. Basic Violet 14 was also present,indicated by the peak at 302 Da. A peak at 372 Da indicated thepresence of Basic Violet 3. This is confirmed by the presence ofpeaks at 256 and 358 Da (Basic Violet 1), which are analogues ofthe dye.

Basic Violet 1 is usually a mixture of hexa- (Basic Violet 3),penta- and tetra-methylated triarylmethane. Hence, in terms ofidentification of components from the spectra, it is difficult toascertain whether in fact either Basic Violet 1 or Basic Violet 3 orboth dyes are present in the original ink formulation.

A peak at 330 Da corresponds to trimethylated arylmethane,which as seen in Table 1, is a demethylated product of Basic Violet3. The peak at 456 Da is often indicative of Basic Blue 26 with a lossof a methyl group, but the lack of a parent molecular ion at 470 Danegates this as an option, so it remains unidentified.

3.2.2. Bic Clic 2000 Medium ‘new’

The major peaks identified corresponded to Basic Yellow 2,Basic Violet 3, and its demethylation product at 330 Da. Peaks at302 and 240 Da, were tentatively identified as Basic Violet 14 andaryl guanidine, which was the case for the ‘old’ sample.

3.2.3. Bic Cristal Medium ‘old’ and ‘new’

Significant peaks corresponded to Basic Yellow 2, Basic Violet 3,and its demethylation product. Peaks at 221 and 281 Da are alsopresent, but not identified. Spectra for both ‘new’ and ‘old’ samplesappear analogous.

3.2.4. Papermate Flex Grip Elite Medium

A peak corresponding to Basic Violet 3 and Basic Violet 1 waspresent along with an associated fragmentation analogue at344 Da corresponding to the demethylation of Basic Violet 1 toform a tetra-methylated analogue. A peak for Basic Yellow 2 wasnoticeably absent, but a strong signal at 478 Da corresponded toBasic Blue 7. The presence of Basic Blue 7 was supported by peaksat 448 and 434 Da, corresponding to the successive loss of C2H6 andCH2, respectively.

3.2.5. Papermate Kilometrico Medium ‘old’ and ‘new’

‘Old’ and ‘new’ pen samples were analogous, with peaksindicative of Basic Violet 3, Basic Violet 1 and its fragmentedanalogue. A peak at 470 Da corresponded with Basic Blue 26,with an associated peak at 456 Da due to the loss of a methylgroup.

3.2.6. Pilot Begreen Rexgrip Medium

The spectra for this sample were dominated by a large peakcorresponding to Basic Blue 7. Fragments at m/z = 462, 448 and434, correspond to the loss of CH4, C2H6 and C3H8 groups,respectively. All other peaks were minor, peaks indicated a minoramount of Basic Violet 1 and 3.

3.2.7. Staedtler Stick 430 Medium

Spectra for the Staedtler ink were relatively uncomplicatedshowing Basic Blue 26 and a series of peaks corresponding to BasicViolet 1 and 3.

3.2.8. Zebra Tapli Clip Medium

The spectra were unusual for this sample as they weredominated by peaks corresponding to aryl guanidines. Ng et al.reported that the aryl guanidines are not dyes in themselves, butrather are added to ink formulations in order to form salts withacid dyes or to raise the pH of an ink [4]. Peaks at 212 and 240 Dacorresponded to diphenyl and ditolyl guanidines, respectively. Asmall peak at 268 Da suggested Basic Yellow 2, but due to thepresence of the other aryl guanidines, is more likely to be dixylolguanidine. The only dyes that can positively be confirmed to bepresent in the Zebra ink sample were Basic Violet 1 and 3.

From Table 3 it is possible to make several inferences withrespect to differentiation between samples and manufacturers.Firstly, in the case where samples were bought 7 months apart,there appears to be little difference between the ink samples. Itwas not possible to differentiate between the ‘new’ and ‘old’examples of Papermate Kilometrico and Bic Crystal inkmarkings, while some differences did exist for the Bic Clic2000 samples. The main difference was the presence of an arylguanidine peak in the ‘old’ sample, which could indicate achange in formulation between batches. These compounds areadded to form salts with the acid dyes or raise the pH of the inkand hence could be a component that is highly variable betweenbatches. On the other hand, it would be reasonable tohypothesise that components such as dyes would remain fairlyconsistent between batches, so that quality and ink colour couldbe steadily sustained. That said, the Bic Clic ‘old’ spectraindicates Basic Violet 14 (302 Da) is present, which is not thecase for the ‘new’ sample, so this would suggest a change in dyeformulation between batches is possible.

Most of the ink samples can be differentiated from each other,with a few exceptions, based upon observing the spectra.Papermate Kilometrico and Staedtler Stick 430 were indistinguish-able, both containing Basic Violet 1/3 mixtures and Basic Blue 26dyes as their major peaks. A few smaller peaks vary betweensamples, but nothing that can confidently be used as a means ofdiscrimination. Bic Clic and Bic Cristal were very similar, whichmay be expected, as both inks come from the same manufacturer.In fact Bic Clic ‘new’ is essentially indistinguishable from BicCristal, it is possible that the same ink is used in both ‘models’ ofpen from this manufacturer. The Basic Violet 14 peak (302 Da) andaryl guanidine peak (240 Da) in the Bic Clic ‘old’ spectra make itdistinguishable from Bic Cristal. Although Papermate Flex GripElite and Pilot Begreen Rexgrip have the same dyes identified ineach spectrum, the Basic Violet peaks for the Pilot sample are muchsmaller than those from the Papermate sample, this could be usedas a point of differentiation between the two, but is debatable.

Fig. 3. PCA plot of cases on the factor-plane (1st vs 2nd PC) for the organic components. The ‘new’ and ‘old’ (indicated) Bic Clic discriminate different batches of ink for the same

product.

J.A. Denman et al. / Applied Surface Science 256 (2010) 2155–2163 2161

While inferences can be made based upon the inspection of themass spectral data, it is difficult to add weight or confidence to theresults. Peaks are not normalised with respect to the total ioncount, hence comparing peak intensities between samples maygive skewed results and interpretation of what are ‘major’ and‘minor’ peaks may differ. In effect, while major differencesbetween samples can be noted, other differences that may allowfurther differentiation may go unnoticed. Based upon 10 differentgroups of manufacturers or models, out of a possible 45 pairs ofsamples, 38 can be confidently differentiated using the method ofvisual spectral interpretation. In an attempt to improve on thisfigure, multivariate analysis was employed.

3.3. Principal component analysis

The relative intensities of the components listed in Table 1 wereselected for the PCA analysis, performed using a correlation matrix.The output from the PCA resulted in 5 significant principalcomponents, representing 94.8% of the variance in the data set. Aplot of the first two principal components on the factor-plane isgiven in Fig. 3. This statistical analysis included the minor peaksthat were not readily identified as important for differentiating theinks by visual inspection.

These results confirm a number of suspicions made earlier fromthe visual inspection of spectra. Firstly, clustering together of ‘new’and ‘old’ samples is observed for both Papermate Kilometrico andBic Cristal samples, indicating no significant difference betweenthe separate batches. On the other hand, while two of the Bic Clic‘new’ samples are clustered, the Bic Clic ‘old’ sample is wellseparated and not correlated with any other sample. A plot of the1st vs 5th PC (not shown), confirms that Basic Violet 14 isresponsible for the differentiation of this sample.

Fig. 3 indicates that Papermate Kilometrico and Staedtler Sticksamples are clustered closely, as are Bic Crystal and Bic Clic (‘new’).Similarly, Papermate Flex Grip and Pilot Begreen Rexgrip cannot bedistinguished, as initially observed. With reference to the unitcircle plot, it is clear which ink components are responsible for theclustering or separation of samples.

Papermate Kilometrico and Staedtler Stick samples werenegatively correlated to the first PC, as were the Basic Violet 1,Basic Violet 3, Solvent Blue 2 and Basic Blue 26 components. It isalso easily observed that the Zebra Tapli samples are stronglycorrelated to the guanidine components, while the Bic Clic and BicCristal samples are correlated to Basic Orange 14 and Basic Yellow.Using the first two principal components, which represent�64% ofthe variation, it is possible to differentiate between 38 of 45possible pairs of samples. This gives the same degree of

differentiation as the visual inspection method, but with greaterconfidence due to its statistical methods. It also adds confidence inspeculations that some of the different pens could be manufac-tured with the same ink.

3.4. Analysis of inorganic components

The PCA model used here included 16 inorganic elements: Na,Mg, Al, Si, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Co, Cu, Zn and Pb. These werechosen from an inspection of the spectra, as well as includingelements expected to be present due to known ink additives.Aluminium silicates and various metallic carbonates are some-times used as ‘extenders’, cheap inorganic pigments that functionto reduce colour strength and also to promote gloss and flow [37].Some dyes form metal complexes, with chromium and coppercommon examples.

Four significant PCs were obtained, that in total represented85.8% of the variance in the data set. A plot of the first two principalcomponents on the factor-plane is illustrated in Fig. 4. It should benoted that one of the Pilot Rexgrip samples is not included in theplot. From the initial PCA it was observed that this singular samplewas so different from all other samples (including the otherRexgrip) that it skewed the plot, and hence was noted to be easilydistinguished, and not included in subsequent analysis. A numberof things are evident from the plot. Similar clustering andseparation of samples is observed to that of the organic data,although some samples are not clustered as tightly.

Once again, Papermate Kilometrico and Staedtler Sticksamples are negatively correlated to the first PC and are onlyslightly distinguishable from each other. Zebra Tapli samples arealso correlated in a similar manner to the organic data, anddistinguishable from all other samples. While closely clustered inthe organic data, the Papermate Flex Grip samples are wellseparated in the inorganic plot, suggesting differences in theinorganic components present. Similarly, Bic Cristal sampleswere more spread out in this instance compared to the organicdata. The two left-most Bic Cristal samples correspond to ‘old’samples, so could indicate a difference between batches in regardto inorganic concentrations or components. In fact, withreference to Fig. 4, and by observation of the original spectrait was established that the two ‘old’ samples had higherconcentrations of Co, Zn and Pb, which were responsible forthe separation. Similarly, the correlations of these elements tothe Bic Clic samples allow differentiation from the remainder ofthe Bic Cristal samples. This was not possible from the organicdata, where these samples were indistinguishable based uponorganic dye components.

Fig. 4. PCA plot of cases on the factor-plane (1st vs 2nd PC) for the inorganic components.

J.A. Denman et al. / Applied Surface Science 256 (2010) 2155–21632162

Pairs of Pilot Begreen Rexgrip and Papermate Flex Grip sampleswere clustered in the organic data, but this is not the case in theinorganic data. It was established that this was due to a highcontribution from manganese. One of the Papermate Flex Gripsamples is also quite different from the other sample in its pair. Theunit circle and reference to the original data revealed a differencein Cr and Cu. Hence, in both cases, the pairs of Pilot Rexgrip andPapermate Flex Grip samples are markedly different from eachother in terms of inorganic components, despite been purchasedtogether in the same package. This is in contrast to all otherpurchased samples, where those pens that came in the samepackage were indistinguishable from each other. Hence for thePilot Rexgrip and Papermate Flex Grip samples, a number ofpossibilities are apparent. One possible explanation is that themanufacturing process of these pens is more variable compared toother brands, and there is in fact intra-batch variation. A morelikely explanation is that the pairs in fact come from differentbatches, where inorganic components have varied, but have endedup being packaged together for sale.

In relation to variation between the different manufacturersand ‘models’, some interesting observations can be made. Firstly, itwas the case that for both the organic and inorganic data, thatPapermate Kilometrico and Staedtler Stick samples were indis-tinguishable from each other, which indicates that they could bethe same ink. Despite the fact that they originate from differentcountries, it is still possible that the ink itself is from the samesource. The Kilometrico sample was ‘made in Malaysia’, while theStaedtler Stick explicitly states that the pen is ‘assembled inAustralia from local and imported materials’. Therefore it ispossible that the ink is in fact from Malaysia, or even vice-versa.What is also evident is that different pen ‘models’ from the samemanufacturer use different inks. The two Papermate pens, Kilo-metrico and Flex Grip are easily differentiated, which may notcome as a surprise as the former model was ‘made in Malaysia’,while the latter was ‘assembled in Mexico’. Bic Clic and Bic Cristalcannot be clearly differentiated by their organic components, butcan be when considering the inorganic components. This may bean example of the manufacturer using a similar ‘recipe’ of dyes toachieve a consistent colour across its range, but varying theinorganic components to suit the individual model. In this case, a‘click action’ pen may contain different amounts or types ofadditives compared to the regular ballpoint model.

For the inorganic data, using the first two principal componentswhich account for �58% of the variation in the data, 38 out of 45possible pairs of samples were differentiated. This compares to theorganic data where the same degree of differentiation waspossible, but due to slightly different clustering of samples. Usinga combination of the organic and inorganic data, 41 out of 45

possible pairs of samples were distinguished (91%). The indis-tinguishable pens included the other ‘old’ and ‘new’ samples. Basedon the organic analysis, it could have been concluded that somemakes of pen used the same ink source, however slight differencesor additional ingredients have introduced variations revealed bythe inorganic analysis, which led to greater differentiation.

4. Conclusions

In terms of the organic components in the blue ballpoint pensstudied, results indicated that most pen manufacturers and modelscould be distinguished. Out of a possible 45 pairs, 38 could beconfidently distinguished based upon the organic components. Ifthe inorganic information was also included, 41 out of 45 pairs(91%) were differentiated.

While some ‘new’ and ‘old’ pairs of samples for somemanufacturers where indistinguishable, another pair wheredistinguishable, indicating that pens from the same manufacturer,but discrete batches, can be significantly different. Additionally,while most pens that were purchased in the same packaging wereindistinguishable from each other, this was not the case for somebrands. From the standpoint of trying to incriminate a particularpen marking, these intra-batch or even inter-batch variations arepreferable. While class characteristics such as colour and mainbulk components may be able to classify a sample into a particulargroup of manufacturer or model, small batch variations mayprovide an additional or even individual level of classification.

In summary, ToF-SIMS can provide the forensic questioneddocument examiner with a technique with a variety of benefits.The surface sensitive nature of analyses, coupled with the non-destructive mode of operation, means that the integrity of theevidence remains intact. Unlike current methods that eitherrequire the removal or dissolution of the ink sample from thepaper, ToF-SIMS has been shown to allow sensitive characterisa-tion and differentiation of ink markings directly off the substrate.The analytical approach offers promise for other fundamentalstudies into inks, such as aging and identifying time sequences oflines.

The studies have identified a number of advantages in theutilisation of the ToF-SIMS for purposes of forensic ink analysis:

Minimal sample preparation and non-destructive analysisallow for the preservation of the integrity of the document;The sensitivity of the technique and small analysis area allowtrace amounts of ink to be analysed;Both organic and inorganic information about ink markings canbe collected simultaneously;

J.A. Denman et al. / Applied Surface Science 256 (2010) 2155–2163 2163

Rapid analysis time is achieved after initial instrumental setup,with analysis time totalling roughly �10 min per sample; andSurface sensitivity of the technique allows the analysis of inkmarkings without interference from the underlying papersubstrate.

References

[1] J.S. Kelly, B.S. Lindblom (Eds.), Scientific Examination of Questioned Documents,CRC Press, London, 2006.

[2] C. Neumann, P. Margot, Forensic Sci. Int. 185 (2009) 29–37.[3] V. Causin, R. Casamassima, C. Marega, P. Maida, S. Schiavone, A. Marigo, A. Villari, J.

Forensic Sci. 53 (2008) 1468–1473.[4] L. Ng, P. Lafontaine, L. Brazeau, J. Forensic Sci. 47 (2002) 1238–1247.[5] A. Kher, E.V. Green, M.I. Mulholland, J. Forensic Sci. 46 (2001) 878–883.[6] J. Wang, G. Luo, S. Sun, Z. Wang, Y. Wang, J. Forensic Sci. 46 (2001) 1093–1097.[7] J. Andrasko, J. Forensic Sci. 46 (2001) 21–30.[8] J.A. Denman, I.M. Kempson, W.M. Skinner, K.P. Kirkbride, Forensic Sci. Int. 175

(2008) 123–129.[9] J.H. Bugler, H. Buchner, A. Dallmayer, J. Forensic Sci. 50 (2005) 1–6.

[10] C. Vogt, A. Becker, J. Vogt, J. Forensic Sci. 44 (1999) 819–831.[11] Y. Xu, J. Wang, L. Yao, Forensic Sci. Int. 162 (2006) 140–143.[12] O. Hilton (Ed.), Scientific Examination of Documents, Callaghan & Co, Chicago,

1956.[13] W.R. Harrison, Suspect Documents, Sweet & Maxwell Ltd., London, 1966.[14] M. Sakayanagi, J. Komuro, Y. Konda, K. Watanabe, Y. Harigaya, J. Forensic Sci. 44

(1999) 1204–1214.[15] R.A. Merrill, E.G. Bartick, J. Hollis Taylor III, Anal. Bioanal. Chem. 376 (2003) 1272–

1278.

[16] T.J. Wilkinson, D.L. Perry, M.C. Martin, W.R. McKinney, A.A. Cantu, Appl. Spectrosc.56 (2002) 800–803.

[17] D.L. Perry, T.J. Wilkinson, M.C. Martin, W.R. McKinney, Synchrotron InfraredMicrospectroscopic Studies of the Changes of the Chemistry of Ink, WashingtonState Convention Center, Seattle, WA, 2001.

[18] A. Kher, M. Mulholland, E. Green, B. Reedy, Vib. Spectrosc. 40 (2006) 270–277.[19] J. Zieba-Palus, M. Kunicki, Forensic Sci. Int. 158 (2006) 164–172.[20] J. Dunn, J. Siegel, J. Allison, J. Forensic Sci. 48 (2003) 652–657.[21] D. Grim, J. Siegel, J. Allison, J. Forensic Sci. 46 (2001) 1411–1420.[22] D. Grim, J. Siegel, J. Allison, J. Forensic Sci. 47 (2002) 1294–1297.[23] D. Grim, J. Siegel, J. Allison, J. Forensic Sci. 47 (2002) 1265–1273.[24] L. Brazeau, M. Gaudreau, J. Forensic Sci. 52 (2007) 209–215.[25] C. Weyermann, R. Marquis, W. Mazella, B. Spengler, J. Forensic Sci. 52 (2007) 216–

220.[26] D.R. Ifa, L.M. Gumaelius, L.S. Eberlin, N.E. Manicke, R.G. Cooks, Analyst 132 (2007)

461–467.[27] C.D. Adam, Forensic Sci. Int. 182 (2008) 27–34.[28] J. Andrasko, J. Forensic Sci. 47 (2002) 324–327.[29] V.F. Samanidou, K.I. Nikolaidou, I.N. Papadoyannis, J. Liq. Chromatogr. R.T. 27

(2004) 215–235.[30] C. Roux, M. Novotny, I. Evans, C. Lennard, Forensic Sci. Int. 101 (1999) 167–176.[31] D. Djozan, T. Baheri, G. Karimian, M. Shahidi, Forensic Sci. Int. 179 (2008) 199–

205.[32] J.A. Zlotnick, F.P. Smith, Forensic Sci. Int. 92 (1998) 269–280.[33] B. Wagner, E. Bulska, T. Meiselb, W. Wegscheiderb, J. Anal. Atom. Spectrom. 16

(2000) 417–420.[34] B. Wagner, E. Bulska, J. Anal. Atom. Spectrom. 19 (2004) 1325–1329.[35] A. Koenig, Direct Elemental Analysis of Ink on Paper by Laser Ablation ICP-MS,

CETAC Technologies, 2003.[36] B. Kanngießer, O. Hahn, M. Wilke, B. Nekat, W. Malzer, A. Erko, Spectrochim. Acta

B 59 (2004) 1511–1516.[37] T. Sharp, What’s in an Ink? Kerley Ink, 1994.