Reconstruction of Obliterated Characters in Polycarbonate through

Article Cite This: Anal. Chem. XXXX, XXX, XXX-XXX

pubs.acs.org/ac

Reconstruction of Obliterated Characters in Polycarbonate through Spectral Imaging Cédric Parisien,† Gitanjali Kolhatkar,† Frank Crispino,‡ André Lajeunesse,‡ and Andreas Ruediger*,† Nanophotonics and Nanoelectronics Group, Énergie-Matériaux-Télécommunication Research Center, Institut National de la Recherche Scientifique, 1650 Boulevard Lionel-Boulet, Varennes, Québec J3X1S2, Canada ‡ Laboratoire de Recherche en Criminalistique, Université du Québec à Trois-Rivières, 3351 Boulevard des Forges, Trois-Rivières, Québec G9A5H7, Canada †

ABSTRACT: In forensic sciences, there is an increasing demand for nondestructive and reliable methods to retrieve obliterated information in polymers. This study demonstrates a case study for the potential of Raman spectroscopy to reconstruct abraded serial numbers. Residual strain and local variations in the structural arrangement are nondestructively imaged through peak shifts and variations of the full width at half-maximum of specific Raman lines, respectively. We qualitatively validate our approach by successfully recovering an obliterated letter stamped with a pressure of ∼170 MPa in a polycarbonate sample, with a subsequent quantitative statistical analysis. The detection threshold is estimated from the propagation depth of plastic deformations to a value of ∼750−800 μm, substantially larger than typical obliteration depths, 200 μm in our case for an initial profile depth of 120 μm.

P

state2,3,6,8,9 around the relief. In polymers, the stamping process induces residual mechanical strain and local changes in the item’s microstructure.2 These deformations propagate relatively deep into the material and remain even when the serial number is superficially abraded. This local strain and structural distortion pattern thus extends over the same region as the relief so that their reconstruction reveals the abraded information. Raman spectroscopy is a vibrational spectroscopy technique based on the inelastic scattering of photons and optical phonons, subject to selection rules of the bond polarizability.9−13 This well-established analytical technique is primarily known for the wealth of chemical information on the samples under study, often providing a “spectral fingerprint”.10 In addition, changes of a Raman peak position reveal strain,10 while the peak width indicates the phonon lifetime and thus the local ordering of the material.11 In particular, Raman spectroscopy is a nondestructive method without requirement for sample treatment prior to the measurements,12,13 an important criterion in forensic science. In this work, Raman spectroscopy is demonstrated to retrieve an obliterated letter in polycarbonate through pixel by pixel imaging of residual mechanical strain and local structural changes. We focus on polycarbonate, an industry favored polymer for the fabrication of many different items, such as automobile parts or bulletproof windows, due to its toughness and tunable optical properties.14,15 By monitoring the shift of

olymers are omnipresent thanks to their versatile properties including color, refractive index, elasticity, flexibility, or even surface roughness that can easily be modulated to meet the requirements of various applications.1 Due to their high tunability, polymers are increasingly favored by the industry and in particular replace metals as the main components in several items such as automobile parts or firearms.1 For safety purposes and traceability, many of these items are marked with a serial number after their fabrication. In the course of criminal investigations, scientists are often confronted with situations where a serial number has been made fully or partially unreadable through physical abrasion.2−4 Many techniques have been developed to enable forensic experts to recover these obliterated serial numbers in metals. Conventional techniques such as chemical etching2,3 or the magneto-optical method5 are restricted to metals only as they either rely on acids or require magnetic ordering.2 To treat polymer samples, heat treatment, chemical swelling, or relief polishing have been suggested,2−4,6 however, these techniques are barely reliable and show poor reproducibility due to a lack of control over the parameters affecting the recovery process.1,2,5,7−9 To develop a nondestructive and reliable method of recovery, the effect of the marking process on the material needs to be understood. In order to write a permanent serial number, a relief is introduced providing visual contrast. While the most commonly used technique is stamping, which consists in hammering the number in the material with a cold or warm punch,2,3 other methods such as laser ablation2,3 or engraving2,3 are commonplace as well. In metals, the stamping process generates an additional local change in the crystallographic © XXXX American Chemical Society

Received: August 1, 2017 Accepted: October 18, 2017 Published: October 18, 2017 A

DOI: 10.1021/acs.analchem.7b03069 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry the Raman peaks and their full width at half-maximum (fwhm) as a function of position, we reconstruct an obliterated letter. Using cross section maps, we determine the propagation depth of the deformation, and thus the detection threshold of the technique.

■

EXPERIMENTAL SECTION Sample Fabrication. The polymer samples were made from transparent polycarbonate. A 0.5 × 0.6 cm2 “H” was cold stamped with a depth of 120 μm. Following the depth profiling of the mark, the right half of the letter was obliterated using an old Swiss Varnamo mechanical milling machine. The depth of the obliteration was controlled with a micrometer adjustment screw to a value of 200 μm. The obliteration depth was constant over the entire area. For the cross-section measurements, the sample was cut using a mechanical slow-saw, to avoid thermal effects, in the nonobliterated section of the letter. Raman Measurements. The Raman measurements were conducted at room temperature using a customized Olympus BX-41 confocal microscope with a 10× objective with an NA of 0.25, yielding a focal spot diameter of 2.3 μm. The linearly polarized Cobolt Blues 25 mW diode-pumped solid state (DPSS) laser operates at 473 nm. The measurements were performed using a laser intensity of 8.4 mW to prevent thermal degradation of the sample and local thermal expansion. The microscope was connected by a multimode optical fiber to an iHR-320 spectrometer from Horiba Scientific operating with a grating of 2400 lines per millimeter. Under these conditions, the system has a hardware resolution of approximately 1 cm−1 per pixel. The detector was a back illuminated deep depleted charge-coupled device Synapse camera (BIDD-CCD) from Horiba Scientific. The microscope was equipped with a Märzhäuser motorized stage and an autofocus module for imaging. The hardware-defined resolution was enhanced by a numerical fitting algorithm under Origin software and a homemade hyperspectral image processing software, proving identical results.

Figure 1. (a) Raman spectrum of polycarbonate acquired in a stressfree region and (b) chemical structure of its monomer unit.

“H”. This analysis reveals that not every bond was affected in the same way, as summarized in Table 1. Peaks 4, 5, and 10 show a shift to lower wavenumbers, demonstrating a residual tensile strain.19−21 In response, peak 2, corresponding to the C−C cycles, shows a shift toward higher wavenumbers, indicating a residual compressive strain.18−20 The other peaks did not show detectable shifts. This is correlated to the polarizability and dissociation energy of the chemical bonds.22 The stiffer bonds, having a higher dissociation energy, are less susceptible to changes of the phonon potential and thus spectral shift,11 whereas the bonds most susceptible to stress provide better sensitivity.11 In this particular case, peak 10 is the most affected by mechanical stress. This peak corresponds to the C−O covalent bond18 that connects the different monomer units in the polymer, a chemical bond highly susceptible to break under stress.14 Its Raman shift over regions 1 and 2 is depicted in Figure 2(b, c), respectively. This peak is located at ∼1237 cm−1 in the absence of strain14 and shifts to 1235.5 cm−1 over the stressed area in both the visibly marked and obliterated regions.23 The similar trend in both regions demonstrates that it is possible to detect residual mechanical strain from the stamping process even in the abraded part of a polycarbonate sample. In addition, the fwhm of peak 10 depicts a variation over region 1 (Figure 2(d)), from 29 cm−1 over the unstressed area to 31 cm−1 over the marked region, suggesting an alteration in the local structural arrangemen.t24 Indeed, the fwhm is related to the lifetime of the optical phonons and therefore to the quality of the local ordering.11 A broadening of the Raman peak is indicative of a more amorphous or disorganized area.25 This analysis thus confirms that the marking process induces strain and local structural changes into the polymer. Reconstruction of the Obliterated Letter. To recover the obliterated information, the shift and the fwhm of the C−O Raman peak were mapped over the complete area presented in Figure 2(a), as illustrated in Figure 3. In the left part of the images, the letter remained unaltered while in the right part, the letter was completely obliterated as described in the Experimental Section. These experiments were performed at

■

RESULTS AND DISCUSSION Interpretation of the Polycarbonate Raman Spectrum. A typical Raman spectrum acquired in a strain-free region of the polycarbonate sample is presented in Figure 1(a). Several peaks can be observed, and those labeled are assigned to a specific chemical bond of the polycarbonate monomeric unit (Figure 1 (b)).16 Indeed, polycarbonate is mainly composed of covalently bonded carbon (C), hydrogen (H) and oxygen (O) atoms.17 Its monomer unit consists in a carbonate function attached to two carbon cycles, corresponding to the bisphenol A molecule.1,14 Because this thermoplastic polymer is amorphous, the chains do not show particular order or any specific conformation.15 Details regarding the peak position and their assignment can be found in Table 1.18 Effect of a Mechanical Stress on the Raman Peaks. The polycarbonate sample marked with the letter “H” is illustrated in Figure 2(a). To identify the bonds affected by the stress applied during the stamping, the Raman shift of peaks 1− 10 was monitored across the visible (white dashed line #1) and obliterated (white dashed line #2) regions of the mark, as described in Figure 2(a). Region 2, which corresponds to an area where the letter was visible prior to the abrasion, was determined by taking into account the geometry of the letter B


Article

Analytical Chemistry Table 1. Identification of the Peaks Composing the Polycarbonate Raman Spectrum Raman peak

identification16

chemical bond energy (kJ/mol)

Raman shift in the absence of stress (cm−1)

1 2 3 4 5 6 7 8 9 10

coplanar deformation C−C (cycle) bending C−C (cycle) stretch C−H (aromatic) symmetric vibration C−H (out of plan) stretching C−CH3 vibration C−H (out of plan) stretching C−O (cycle) stretching C−O (cycle) stretching C−O (cycle) stretching C−O

614 348 413 413 348 413 358 358 358 338

641.0 708.8 737.6 830.1 890.3 921.8 1005.0 1112.3 1179.6 1237.5

strain observed compressive tensile tensile

tensile

the presence of strain gradients. In Figure 3(c), the population centered at ∼1235.7 cm−1 (blue curve) corresponds to the strained pixels, while the population centered at ∼1236.7 cm−1 (green curve) corresponds to the nonstrained pixels. These two populations are separated by over 11.5σ, confirming that they are significantly separated, as 99.7% of each population is confined within 3σ.26 Similarly, for the fwhm, (Figure 3(d)), the population centered at 29 cm−1 (blue curve) corresponds to the nonaltered regions, while the population centered at 30 cm−1 (green curve) corresponds to the regions where structural changes were observed. Both distributions are asymmetric and not as clearly defined as for the shift, due to a gradient in the broadening of the Raman peak. Nevertheless, these two populations are separated by 2.53σ, indicating that there is less than 1% chance that the values are from the same population.26 Propagation Depth of the Strain and Detection Threshold. Often, the depth by which letters are obliterated is a compromise between the need of rendering the serial number unreadable and a sustained mechanical integrity. Obliteration thus often remains superficial. The depth up to which the residual deformation is detectable is therefore relevant. Yet, this is linked to the stamping pressure (H), which can be approximated using the indentation depth (h) and the material’s properties27 according to the following:16

Figure 2. (a) Picture of the marked polycarbonate sample showing the visible part on the left and the obliterated part on the right. The white dashed lines indicate the regions where line-scans were performed. (b) Raman shift in region 1, (c) in region 2, and (d) fwhm in region 1 of the C−O peak as a function of the position on the polycarbonate sample.

⎛h ⎞ H = ⎜ 0⎟ H0 ⎝ h ⎠

1/2

different locations of the same sample and on different samples with comparable results. The acquisition time necessary to obtain the image presented in Figure 3(a, b) was ∼5 h, to ensure a high enough signal-to-noise ratio to extract accurately the peak shift and fwhm values. Mapping the peak position over the sample surface (Figure 3(a)) exposes a strained region over an H-shaped area (blue region), indicated by a spectral shift of ∼1 cm−1, thereby revealing the letter H’s missing half. This peak displacement is consistent with the value obtained in Figure 2(c). The map of the fwhm is presented in Figure 3(b). The turquoise H shaped region indicates a broadening of the C−O peak due to a degradation of the local ordering. Both the peak shift as well as the change of the peak width are independently suited to reconstruct abraded serial numbers. To determine the statistical significance of the image contrast in Figure 3(a, b), we created histograms with the populations of pixels as shown in Figure 3(c, d). On the basis of the hypothesis of two distinct populations of unstrained/undistorted and strained/distorted areas, we numerically fitted each histogram with two Gaussian distributions of respectively identical σ assuming a constant systematic error across each picture.25 The use of Gaussians is a rough approximation and might require further attention in the future, given that our images indicate

+1

(1)

where H0 is hardness (166 MPa for polycarbonate), and h0 is a constant defined by the tool. Using eq 1, the pressure applied to create a ∼120 μm deep mark in polycarbonate was estimated to be about 169 MPa. Furthermore, to evaluate the propagation depth of the strain associated with a stamping pressure of 169 MPa, and thus determine the detection threshold to successfully reconstruct the obliterated information, the cross section of the nonobliterated “H” segment was studied. Figure 4 shows the map of the C−O peak position and fwhm over the depth of the sample. The interface between the brown/black and blue sections at the top indicates the sample surface, while the marked area is represented by the V-shaped indent on the surface. Figure 4(a) reveals that, following the stamping, the deformation propagated deep into the polycarbonate sample, down to 750 μm, as suggested by the shift of the C−O Raman peak toward lower wavenumbers (blue).19 The strain remains perceivable between 750 and 800 μm (light blue-green), however less pronounced in this range. This is further confirmed by the map of the fwhm (Figure 4(b)) that shows structural changes down to ∼750 μm. Therefore, the detection C


Article

Analytical Chemistry

Figure 3. Spectral images of (a) the shift and (b) the fwhm of the C−O Raman peak of polycarbonate showing the successful recovery of the obliterated right part of the letter. The histograms in (c) and (d) illustrate the populations of unaffected and affected pixels for peak shift and fwhm, respectively. Solid blue and green lines are numerical fits for two Gaussian distributions assuming identical σ in each respective case.

in polymers through the monitoring of residual mechanical strain and local structural changes. We apply this efficient and nondestructive method to a polycarbonate sample bearing an obliterated 120 μm deep stamped letter manually punched with a force of ∼169 MPa. In the marked region, the C−O Raman peak displays a shift of ∼1 cm−1 toward lower wavenumbers and significant broadening. This indicates an elongation of the material upon deformation due to a tensile stress, accompanied by a change in the local arrangement. By mapping the shift and the fwhm of the C−O Raman peak, the obliterated letter is reconstructed. The detection threshold of the strain propagation depth of our method was determined to 750−800 μm from cross-section measurements. Avoiding any sample damage, Raman spectroscopy provides control over all the restoration parameters, which makes it a promising approach compared to conventional methods for the recovery of obliterated elements in polymers. Also, the ability to detect strain in a nondestructive way in these materials will certainly find several industrial applications such as quality control analysis and security, as polycarbonate could be used for more

Figure 4. Spectral images of (a) the shift and (b) the fwhm of the C− O Raman peak, acquired on the cross-section of a letter marked on the polycarbonate sample.

threshold for a letter marked with 169 MPa is in the 750−800 μm range, much larger than a necessary obliteration depth (80 μm below the mark).

■

CONCLUSIONS In summary, we demonstrate the potential of Raman spectroscopy for the reconstruction of obliterated information D


Article

Analytical Chemistry

(24) Li, Z. J.; Danilewsky, A. N.; Helfen, L.; Mikulik, P.; Haenschke, D.; Wittge, J.; Allen, D.; McNally, P.; Baumbach, T. J. Synchrotron Radiat. 2015, 22 (4), 1083−1090. (25) Bradley, M. Curve Fitting in Raman and IR Spectroscopy: Basic Theory of Line Shapes and Applications; 2007. (26) de Smith, M. J. Statistical Analysis Handbook: A Comprehensive Handbook of Statistical Concepts, Techniques and Software Tools; 2010. (27) Bucaille, J. L.; Felder, E.; Hochstetter, G. J. Mater. Sci. 2002, 37 (18), 3999−4011.

items that need to bear a safety mark. While this technique was demonstrated on polycarbonate, it is now tested on other (colored) polymers, and a variety of other materials such as ceramics, which are becoming more widely used.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected] (A.R.). ORCID

Cédric Parisien: 0000-0001-9345-3970 Gitanjali Kolhatkar: 0000-0003-0848-4751 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS

■

REFERENCES

We are grateful to J. Plathier for his hyperspectral imaging software. Fruitful discussion with C. Muehlethaler (UQTR) is gratefully acknowledged. A.R. holds an NSERC discovery grant, and G.K. is thankful for an FRQNT postdoctoral scholarship.

(1) Askeland, D.; Fulay, P.; Wright, W. The Science and Engineering of Materials, 6th ed.; Cengage Learning: Stamford, 2006. (2) Strauffer, E.; Bonfanti, M. In Forensic Investigation of Stolenrecovered and Other Crime-Related Vehicles; Academic Press: Burlington, 2006; pp 177−207. (3) Houck, M. M. In Encyclopedia of Forensic Sciences; Academic Press: London, 2013; Vol. 3, pp 43−47. (4) Katterwe, H. J. Forensic Sci. Soc. 1994, 34 (1), 11−16. (5) Dawson, J. National Institute of Justice 2015, 275. (6) Heard, B. J. In Handbook of Firearms and Ballistics: Examining and Interpreting Forensic Evidence; John Wiley and Sons, Ltd.: Oxford, 2008. (7) Conlan, X. a.; Baker, M. J.; Krieg, R.; Lockyer, N. P.; Vickerman, J. C.; Barnett, N. W.; Lim, K. F. Surf. Interface Anal. 2011, 43 (1−2), 625−627. (8) Abdul Wahab, M. F.; Mohamad Ghani, N. I.; Kuppuswamy, R. Forensic Sci. Int. 2012, 223 (1−3), 53−63. (9) Kuppuswamy, R. In Aluminium Alloys, Theory and Applications; InTech: Rijeka, 2010. (10) Chalmers, J. M.; Edwards, H. G. M.; Hargreaves, M. D. Infrared and Raman Spectroscopy in Forensic Science; 2012. (11) Atkins, P.; De Paula, J. Physical Chemistry, 8th ed..; Oxford University Press: New York, 2006. (12) Das, R. S.; Agrawal, Y. K. Vib. Spectrosc. 2011, 57 (2), 163−176. (13) Kudelski, A. Talanta 2008, 76, 1−8. (14) Lee, S. N.; Stolarski, V.; Letton, A.; Laane, J. J. Mol. Struct. 2000, 521 (1−3), 19−23. (15) Wright, S. C.; Huang, Y.; Fleck, N. A. Mech. Mater. 1992, 13, 277−284. (16) Akita, S.; Nishijima, H.; Kishida, T.; Nakayama, Y. Japan J. Appl. Phys,z. 2000, 228−229. (17) Dybal, J.; Schmidt, P.; Baldrian, J.; Kratochvíl, J. Macromolecules 1998, 31 (19), 6611−6619. (18) Debus, J.; Dunker, D. Raman studies on a heavily distorted polycarbonate sample http://arxiv.org/abs/1203.0064. (19) Colomban, P. Adv. Eng. Mater. 2002, 4 (8), 535−542. (20) Nielsen, a. S.; Pyrz, R. Compos. Sci. Technol. 2002, 62, 2219− 2227. (21) Siesler, H. W. Int. J. Polym. Anal. Charact. 2011, 16 (8), 519− 541. (22) Lide, D. R. Handbook of Chemistry and Physics, 84th ed.; CRC Press: Boca Raton, 2003. (23) Sato, H.; Ozaki, Y.; Jiang, J.; Yu, R.-Q.; Shinzawa, H. Raman, Infrared, Near-Infrared Chem. Imaging 2011, 261−281. E


Reconstruction of Obliterated Characters in Polycarbonate through

Recommend Documents