Wood identification by a portable low-cost polymer-based electronic nose

Abstract

The rapid and reliable identification of woods is a difficult task, yet extremely necessary for the control of illegal logging or trade of protected species. In this paper we describe a portable low-cost electronic nose based on conductive polymers, capable of distinguishing pairs of similar-looking woods, aiming to help wood traders to comply with the determinations of the United Nations (UN) Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES). Examples of woods with CITES restrictions are (i) mahogany (Swietenia macrophylla) and Spanish-cedar (Cedrela fissilis), and (ii) Brazilian walnut (Ocotea porosa) and black-cinnamon (Ocotea catharinensis). The e-nose consists of an array of four different gas sensors fabricated by the deposition of thin doped conductive polymer films onto the surface of interdigitated metallic electrodes. The electrical conductance responses of the sensors upon exposure to the volatile compounds emitted by the wood specimens, after scratching their surface, were evaluated by principal component analysis (PCA). The data can be processed on small computers or even on smart phones. The analysis time is only 10–15 min. Leave-one-out analysis gave a classification hit rate of 100% for group (i) and 94% for group (ii). Hence, this is an extremely viable method for standard control use in remote places such as roads in the Amazon forest, which are sometimes hundreds of miles away from urban centers, where conventional analyses could be performed.

---

Introduction

The protection of some endangered tree species, especially in tropical forests, currently represents a major environmental concern worldwide. For that purpose The United Nations (UN) established in 1972 an international treaty controlling the trade and exploitation of endangered species of animals and plants (www.cites.org). Mahogany (Swietenia macrophylla), one of the most valuable tropical wood species, known for its attractive properties^1,2 such as resistance, hardness and shiny reddish colour, has been classified as an endangered species since 1996. It has been harvested and traded³ for hundreds of years mostly for luxury furniture manufacturing.^4,5 In Brazil, although mahogany exploitation is forbidden since 2002, the illegal cutting of this tree still occurs very frequently due to its similar appearance to cedar⁶ (Cedrela fissilis) and other reddish Amazonian woods of which exploitation is allowed (e.g. Cedrelinga catenaeformis, Cabralea canjerana). Inaccurate wood identification may occur in loco by loggers or supervisors by mistake, but also might be a fraud intended to avoid legal restrictions to exploitation and trade.

Another example of tropical woods that can be easily mistaken due to their great similarity is Brazilian walnut (Ocotea porosa) and black-cinnamon (Ocotea catharinensis). These species belong to the same genus and are considered vulnerable according to the IUCN (International Union for Conservation of Nature) Red List of Threatened Species.⁷

Nowadays, wood identification is carried out based on macro and microscopic characteristics of the species⁸ by wood anatomy experts, requiring specific apparatus for analytical techniques such as MS (mass spectrometry),⁹ IR (infrared) spectroscopy^10,11 and NIRS (near infrared spectroscopy).⁶

Electronic noses are analytical instruments that mimic the human nose. They are formed by arrays of gas sensors with partial specificity attached to a pattern recognition system capable of recognizing scents or odours.¹² The most popular gas sensors are chemiresistors based on metal oxide semiconductors (MOS) or conductive polymers (CP).¹³ The latter offers attractive features, such as operation at ambient temperature, high sensitivity to a wide range of volatile organic compounds (VOCs), large possibilities of structural variations, and are low-cost.

In the last few years, we have described gas sensors and electronic noses, based on conducting polymers, for sensing low-molecular weight alcohols,¹⁴ for quantitative determination of methanol in Brazilian sugar-cane spirit (“cachaça”),¹⁵ for the determination of the ethanol content in automotive fuel¹⁶ and for the detection of post-harvest fungal attack in oranges.¹⁷ In this paper, we describe the development of a portable, low-cost, CP-based electronic nose and its application in the identification of pairs of woods that are similar in appearance and, therefore, can be easily mistaken by chance or as part of illicit logging of endangered tree species. The chosen pairs of woods for this study are: (i) mahogany and Spanish-cedar and (ii) black-cinnamon and Brazilian walnut.

Experimental section

Wood samples

All the samples were selected in the xylarium of the Departamento de Botânica (SPFw), Instituto de Biociências, Universidade de São Paulo (IB-USP), Brazil. Spare samples are available for future investigations or rebuttal.

The approximately 125 cm³ (5 cm × 5 cm × 5 cm) samples of wood were stored separately in hermetically sealed bottles. Before every measurement, the wood samples were sandpapered in order to increase the release of their volatile compounds.

Conductive polymers

We chose poly(9,9-n-dioctyl-2,7-fluorenylene ethylene) (PDO27FE), poly(2,5-biphenylene ethylene) (PPPX), poly(4′-hexyloxy-2,5-biphenylene ethylene) (PHBPE) and poly(2-bromo-5-hexyloxy-p-phenylene vinylene) (BHPPV) (Fig. 1) by screening among other conductive polymers to compose the electronic nose. They were synthesized as previously described in the literature,^18–21 respectively.

Fig. 1 Structures of the polymers used to fabricate the sensors.

Interdigitated electrodes and sensors

Chromium interdigitated electrodes of 1 cm² area and 24 μm spacing between fingers (Fig. 2) were produced on commercial microscope glass slides by conventional UV-photolithography. The slides underwent a RCA cleaning (a common silicon wafer cleaning process), followed by a chromium deposition by sputtering in an argon atmosphere with 2 mTorr of chamber pressure and 100 W of rf power. The total deposition time was 15 minutes to achieve thickness of ∼400 nm. The chromium was patterned by photolithography and etched using a ceric ammonium nitrate solution.

Fig. 2 The interdigitated electrode.

Solutions of the above polymers containing 2.5 mg of the polymer, 0.7 mg of dodecylbenzenesulfonic acid (DBSA) and 0.75 mL of chloroform were prepared. Then 20 μL of each solution was deposited onto the interdigitated area of the electrodes by spin coating (600 rpm, 10 s), forming films of approximately 1 μm thickness.

Measuring set-up

A pneumatic device for dynamic sampling was assembled as shown in Fig. 3 and 4. A Boyu S-1000A oil-free pump at a flow rate of 70 mL min⁻¹ provided a continuous air supply. The sample was placed inside a thermostated chamber (25 °C). During the exposure period (20 s) solenoid valves v₁ and v₂ were opened while v₃ was closed so that the air stream passed through the samples chamber carrying its headspace to the sensors. During the recovery period (120 s) v₁ and v₂ were closed while v₃ was opened, so a stream of sample-free air cleaned and reset the sensors. The exposure/recovery cycle was repeated 40 times for each wood sample.

Fig. 3 Scheme of the dynamic sampling device.

Fig. 4 Photograph of a prototype of the dynamic sampling device.

The conductance between the contact pairs of each electrode was continuously monitored (20 times per s) by accurate conductivity meters, operating with a 80 mV peak-to-peak 2 kHz triangle wave AC voltage and connected via a 10 bit analog-to-digital converter to a personal computer, in which a software enabled collecting the data and plotting conductance versus time graphs.

Results and discussion

Conductive polymers

The polymers were doped with 4-dodecylbenzenesulfonic acid (DBSA) in order to obtain conductive films. DBSA was chosen as p-dopant because of its organic Lewis acid property and good miscibility with the polymers. In BHPPV, a poly(p-phenylene vinylene) derivative, the electrical conductivity occurs mostly through the extent of the conjugated backbone, whilst in PDO27FE, PPPX and PHBPE, all poly(p-xylylene) derivatives, the conductivity occurs mainly inside the biaryl units in face of the absence of conjugation along the polymer chains. It is worth mentioning that these biphenyl units present six conjugated π bonds, which was shown to be sufficient to ensure modest electrical conductivity after being doped with a Lewis acid.²²

One can note the presence of side-chains in the structures of PDO27FE, PHBPE and BHPPV. This is an important characteristic to achieve solubility of polymers in organic solvents and therefore to enable film processing. Although PPPX does not contain such long side-chains, but phenyl groups, the polymer presented fair solubility in chloroform and could be processed.

Sensors response

For this study two pairs of woods were selected: (a) mahogany (Swietenia macrophylla) and Spanish-cedar (Cedrela fissilis); (b) Brazilian walnut (Ocotea porosa) and black-cinnamon (Ocotea catharinensis). Woods of each pair grow in the same geographical region (South America), are similar in appearance and, therefore, can be easily mistaken.

Fig. 5 shows a typical response of the sensors towards a particular wood sample (mahogany). The first four exposure/recovery cycles out of 40 are represented. The reversibility of the responses is noticeable since the initial conductance values were recovered quickly during the cleaning periods. Each sensor showed a particular response pattern, which is important for the discrimination capability of the e-nose. The exact response mechanism of conductive polymers on exposure to volatile compounds is still not fully understood. It may involve swelling of the polymers caused by absorption of the analyte molecules changing the extrinsic conductivity²³ that may also influence intrachain mobility of free charge carriers due to a solvation-induced alteration of the molecular conformation.¹⁴ Polarization-Modulation Infrared Reflection Absorption Spectroscopy (PMIRRAS)²⁴ could demonstrate the latter for PPPX. Thus, reversible changes in the angle formed between the planes of the biphenyl rings in PPPX may affect the coplanarity of the overlapping orbitals changing the degree of conjugation and the electronic gap (E_gap) of the system causing electrical conductivity to change accordingly.

Fig. 5 Response of the sensors (conductance versus time) upon four cycles of exposure/recovery to the volatile compounds of mahogany (Swietenia macrophylla) at 25 °C. Sensor 1: PDO27FE; sensor 2: PPPX; sensor 3: PHBPE, and sensor 4: BHPPV.

Relative responses (R_a) were calculated for all the cycles of the woods tested according to eqn (1):

R_a = (G₂ − G₁)/G₁ (1)

where G₁ and G₂ are the conductance values at the beginning and the end of the exposure period, respectively.

Principal component analysis

The calculated relative responses (R_a) were used as input data for principal component analyses (PCA). Fig. 6 and 7 show PCA bidimensional scatter plots for both pairs of woods. A clear clustering of the data points can be seen on both graphs and the total variance was over 96%. It is interesting to note that the two clusters corresponding to Spanish-cedar and mahogany, which belong to different genera (Cedrela and Swietenia), are better separated than the clusters of black-cinnamon and Brazilian walnut, which belong to the same genus (Ocotea, family Lauraceae). This confirms the traditional botanical statement that woods belonging to the same taxonomic classification present chemical similarities, such as similar volatile organic compounds. Leave-one-out validation analyses²⁵ gave a hit rate of 100% for Spanish-cedar/mahogany and 94% for black-cinnamon/Brazilian walnut data, showing the high reliability of this e-nose.

Fig. 6 PCA graph for mahogany and Spanish-cedar.

Fig. 7 PCA graph for black-cinnamon and Brazilian walnut.

Conclusions

An electronic nose was assembled based on four gas sensors made of doped conductive polymers with the purpose of wood species classification. The e-nose is portable and has low power consumption. The sensors are robust and have been tested for over one year. In case of replacement, the cost is only US$ 2.00 per sensor. The data can be processed on small computers or even on smart phones. The analysis time is only 10–15 min. Hence, this is an extremely viable method for standard control of woods in remote places such as logging areas and roads in the Amazon forest, which are sometimes hundreds of miles away from urban centers, where conventional analyses could be performed.

The system is a very reliable, light, and innovative tool to improve wood identification, and hence has the potential of being an important weapon against illegal wood logging/trading and can support tropical forests conservation.

Acknowledgements

The authors would like to thank Conselho Nacional para o Desenvolvimento Científico e Tecnológico (CNPq) (Proc. no. 307915/2013-1, 307041/2014-0), Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) (Proc. no. 2011/51249-3) for their financial support, and to M.Sc. Rebecca Ann Povilus for revising the manuscript.

References

1. P. Langbour, J. Gerard, J. M. Roda, P. A. Fauzi and D. Guibal, J. Trop. Forest Sci., 2011, 23, 252–259.
2. E. Schwab, Holz Roh- Werkst., 1992, 50, 57–61.
3. A. Verissimo, P. Barreto, R. Tarifa and C. Uhl, For. Ecol. Manage., 1995, 72, 39–60.
4. P. Kjellberg, Connaissance Arts, 1982, 368, 109–114.
5. B. Bacon, Connoisseur, 1980, 204, 58–60.
6. T. C. M. Pastore, J. W. B. Braga, V. T. R. Coradin, W. L. E. Magalhaes, E. Y. A. Okino, J. A. A. Camargos, G. I. B. de Muniz, O. A. Bressan and F. Davrieux, Holzforschung, 2011, 65, 73–80.
7. www.iucnredlist.org, accessed in July, 2016.
8. IAWA Committee.
9. E. C. Cabral, R. C. Simas, V. G. Santos, C. L. Queiroga, V. S. da Cunha, G. F. de Sa, R. J. Darodad and M. N. Eberlin, J. Mass Spectrom., 2012, 47, 1–6.
10. A. Carballo-Melián, A. M. Goodman, M. G. Baron and J. Gonzales-Rodriguez, Cellulose, 2016, 23, 901–913.
11. P. Zhao and J. Cao, J. For. Res., 2016, 27, 219–224.
12. J. W. Gardner and P. N. Bartlett, Electronic Noses Principles and Applications, Oxford University Press, New York, 1999.
13. E. A. Baldwin, J. Bai, A. Plotto and S. Dea, Sensors, 2011, 11, 4744–4766.
14. M. S. Meruvia, I. A. Hümmelgen, C. S. Gonçalves, A. R. V. Benvenho and J. Gruber, Sens. Lett., 2007, 5, 625–628.
15. L. O. Peres, R. W. C. Li, E. Y. Yamauchi, R. Lippi and J. Gruber, Food Chem., 2012, 130, 1105–1107.
16. A. R. V. Benvenho, R. W. C. Li and J. Gruber, Sens. Actuators, B, 2009, 136, 173–176.
17. J. Gruber, H. M. Nascimento, E. Y. Yamauchi, R. W. C. Li, C. H. A. Esteves, G. P. Rehder, C. C. Gaylarde and M. A. Shirakawa, Mater. Sci. Eng., C, 2013, 33, 2766–2769.
18. J. Gruber, R. W. C. Li, L. H. J. M. C. Aguiar, T. L. Garcia, H. P. M. de Oliveira, T. D. Z. Atvars and A. F. Nogueira, Synth. Met., 2006, 156, 104–109.
19. R. W. C. Li, L. R. F. Carvalho, L. Ventura and J. Gruber, Mater. Sci. Eng., C, 2009, 29, 426–429.
20. R. W. C. Li, L. Ventura, J. Gruber, Y. Kawano and L. R. F. Carvalho, Sens. Actuators, B, 2008, 131, 646–651.
21. (a) A. R. V. Benvenho, R. Lessman, I. A. Hümmelgen, R. M. Q. Mello, R. W. C. Li, F. F. C. Bazito and J. Gruber, Mater. Chem. Phys., 2006, 95, 176–182; (b) R. M. Rosa, R. L. Szulc, R. W. C. Li and J. Gruber, Macromol. Symp., 2005, 229, 138–142.
22. J. H. P. Utley, Y. Gao, J. Gruber, Y. Zhang and A. Munoz-Escalona, J. Mater. Chem., 1995, 5, 1837–1845.
23. J. M. Slater, E. J. Watt, N. J. Freeman, I. P. May and D. J. Weir, Analyst, 1992, 117, 1265–1270.
24. L. Caseli, J. Gruber, R. W. C. Li and L. O. Péres, Langmuir, 2013, 29, 2640–2645.
25. F. M. Lopes, D. C. Martins and R. M. Cesar, BMC Bioinf., 2008, 9, 451.

---