Determination of the Weight Percentage Gain and of the Acetyl

This was expected because the acetyl group content accounts for the acetyl groups already present in wood, which depends on the species, and was found...
0 downloads 0 Views 199KB Size
Download PDF
Anal. Chem. 2008, 80, 1272-1279

Determination of the Weight Percentage Gain and of the Acetyl Group Content of Acetylated Wood by Means of Different Infrared Spectroscopic Methods Barbara Stefke,†,⊥ Elisabeth Windeisen,‡ Manfred Schwanninger,*,§ and Barbara Hinterstoisser†,|

Competence Centre for Wood Composites and Wood Chemistry (Wood K plus), St. Peter Strasse 25, A- 4021 Linz, Austria, Wood Research Munich, Technical University of Munich, Winzererstrasse 45, 80797 Munich, Germany, Department of Chemistry, BOKUsUniversity of Natural Resources and Applied Life Sciences Vienna, Muthgasse 18, A-1190 Vienna, Austria, and Department of Material Sciences and Process Engineering, BOKUsUniversity of Natural Resources and Applied Life Sciences, Vienna, Peter Jordan Strasse, A-1180 Vienna, Austria

The weight percentage gain (WPG) and the acetyl group content of wood due to acetylation with acetic anhydride have been analyzed by means of Fourier transform infrared spectroscopy (FTIR) and near-infrared spectroscopy (NIR). Band height ratios (BHR) (1240/1030 (1230/ 1030) and 1745/1030 (1740/1030)) of the bands at 1745 (1740), 1240 (1230), and 1030 cm-1 were calculated from FTIR-KBr and FTIR-ATR (attenuated total reflection) spectra. The good linear correlation with a coefficient of determination of about 0.94 over a range from 0 to 27% WPG existing between BHRs and WPG and acetyl group content, respectively, requires only a few samples to calibrate FTIR. Partial least-squares regression models based on second derivatives of the NIR spectra in the wavenumber range from 6080 to 5760 cm-1 resulted in a R2 value of 0.99, number of PLS components (rank) between 3 and 5, root-mean-square error of cross-validation between 0.6 and 0.79%, and a residual prediction deviation up to 10. Although a wide range of input parameters (i.e., various wood species and different procedures of acetylation) was used, highly satisfactory results were obtained. Both FTIR and NIR spectroscopic means fulfill the need for determining the WPG and the acetyl content of acetylated wood. By reason of its additional potential for on-line process control, the NIR method may even outperform the FTIR method. Wood is an important industrially applicable resource, and due to its high specific strength-to-weight ratio it is preferred as a building and engineering material.1 Even though nature provides a very broad spectrum of material properties due to the natural variability of wood species, there are some limitations such as photosensitivity and combustibility. Moreover, a lowering or * Corresponding author. E-mail: [email protected]. Tel: +431360066523. Fax: +431360066059. † Competence Centre for Wood Composites and Wood Chemistry. ‡ Technical University of Munich. § Department of Chemistry, BOKU. | Department of Material Sciences and Process Engineering, BOKU. ⊥ Dynea Austria GmbH, Hafenstraβe 77, A-3500 Krems. (1) Sun, R. C.; Sun, X. F. Ind. Crops Prod. 2002, 16, 225-235.

1272 Analytical Chemistry, Vol. 80, No. 4, February 15, 2008

harmonization of wood properties is also often desired, especially when wood is exposed to moisture, such as swelling and shrinking and biodegradability. Methods that enable a decrease of the degradation rate by modifying the basic chemistry of the cell wall polymers could help to enhance wood’s utility. Esterification of wood hydroxyl groups in hemicelluloses, lignin, and parts of cellulose by acetic anhydride leads to a substitution of hydroxyl groups by acetyl groups, whereby acetic acid is formed as a byproduct.2-13 The dimensional stability,3,7,14-19 decay resistance to fungi,8,20-22 resistance to photodegradation by UV radiation,23-25 (2) Ramsden, M. J.; Blake, F. S. R. Wood Sci. Technol. 1997, 31, 45-50. (3) Ramsden, M. J.; Blake, F. S. R.; Fey, N. J. Wood Sci. Technol. 1997, 31, 97-104. (4) Rosenqvist, M. Holzforschung 2001, 55, 270-275. (5) Rowell, R. M. Mol. Cryst. Liq. Cryst. 2004, 418, 881-892. (6) Rowell, R. M. For. Prod. J. 2006, 56, 4-12. (7) Sander, C.; Beckers, E. P. J.; Militz, H.; van Veenendaal, W. Wood Sci. Technol. 2003, 37, 39-46. (8) Hill, C. A. S.; Forster, S. C.; Farahani, M. R. M.; Hale, M. D. C.; Ormondroyd, G. A.; Williams, G. R. Int. Biodeterior. Biodegrad. 2005, 55, 69-76. (9) Tserki, V.; Matzinos, P.; Kokkou, S.; Panayiotou, C. Composites, Part A 2005, 36, 965-974. (10) Tserki, V.; Zafeiropoulos, N. E.; Simon, F.; Panayiotou, C. Composites, Part A 2005, 36, 1110-1118. (11) Hill, C. A. S.; Jones, D.; Strickland, G.; Cetin, N. S. Holzforschung 1998, 52, 623-629. (12) Hill, C. A. S.; Khalil, H. P. S. A.; Hale, M. D. Ind. Crops Prod. 1998, 8, 53-63. (13) Kumar, S. Wood Fiber Sci. 1994, 26, 270-280. (14) Obataya, E. Wood Sci. Technol. 2005, 39, 472-483. (15) van Houts, J. H.; Winistorfer, P. M.; Wang, S. Q. For. Prod. J. 2003, 53, 82-88. (16) Beckers, E. P. J.; de Meijer, M.; Militz, H.; Stevens, M. J. Coat. Technol. 1998, 70, 59-67. (17) Epmeier, H.; Johansson, M.; Kliger, R.; Westin, M. Holzforschung 2007, 61, 34-42. (18) Minato, K.; Takazawa, R.; Ogura, K. J. Wood Sci. 2003, 49, 519-524. (19) Rafidah, K. S.; Hill, C. A. S.; Ormondroyd, G. A. J. Trop. For. Sci. 2006, 18, 261-268. (20) Brelid, P. L.; Simonson, R.; Bergman, O.; Nilsson, T. Holz Roh- Werkst. 2000, 58, 331-337. (21) Ohkoshi, M.; Kato, A.; Suzuki, K.; Hayashi, N.; Ishihara, M. J. Wood Sci. 1999, 45, 69-75. (22) Peterson, M. D.; Thomas, R. J. Wood Fiber 1979, 10, 149-163. (23) Hill, C. A. S.; Cetin, N. S.; Quinney, R. F.; Derbyshire, H.; Ewen, R. J. Polym. Degrad. Stabil. 2001, 72, 133-139. (24) Kosikova, B.; Sasinkova, V.; Tolvaj, L.; Papp, G.; Bohus, J.; Szatmari, S. Drev. Vysk. 2002, 47, 11-18. 10.1021/ac7020823 CCC: $40.75

© 2008 American Chemical Society Published on Web 01/16/2008

mechanical properties,3,13 and hydrophilicity are improved by acetylation.9 Fourier transform infrared spectroscopy (FTIR) is a well-known powerful analytical tool to detect functional groups by measuring fundamental molecular vibrations.26 Especially carbonyl groups (ester groups) that have a high molar absorptivity can easily be seen, even at low concentrations. Therefore, many researchers characterizing and following the acetylation of wood have used this tool.1,10,27-31 However, mid-infrared (MIR) instruments are normally not used for on-line process control under conditions found in an industrial environment. Alternatively, near-infrared (NIR) spectroscopy can be used. Absorption bands in the NIR region arise from overtones and combination bands caused by vibrations of C-O, O-H, C-H, and N-H groups, which have their fundamental molecular vibrations in the MIR region.26,32 The absorption signals of various constituents are connatural and highly overlapping, and therefore in many cases no distinct band can be observed that is directly related to the chemical abundance of a single wood constituent.33 However, it was shown that changes due to the acetylation of wood could clearly be seen in the NIR spectra.27 NIR spectroscopy is a rapid and nondestructive method used for process and quality control in many areas such as food and beverage production, agriculture, biotechnology, petrochemistry, and pharmaceutical production34 as well as for research purposes in wood, pulp, and paper science for more than 20 years.33,35,36 As in various cases, the chemical information is hardly selective within the typically broad and extensively overlapping bands of NIR spectra. Multivariate analysis techniques have to be used to model relevant data for the classification and prediction of wood properties.36-44 No results have been published with regards to (25) Hon, D. N.-S. Weathering and Photochemistry of Wood, 2nd ed., revised and expanded; Marcel Dekker, Inc.: New York, 2001. (26) Socrates, G. Infrared and Raman Characteristic Group Frequencies. Tables and Charts, 3rd ed.; John Wiley & Sons, Ltd.: Chichester, U.K., 2001. (27) Schwanninger, M.; Hinterstoisser, B. In Modifiziertes Holz, Eigenschaften und Ma ¨ rkte. Modified Wood, Properties and Markets; Alfred Teischinger, R. S., Ed.; Institut fu ¨ r Holzforschung gemeinsam mit dem Verband Holzwirte O ¨ sterreichs: Vienna, 2002; Vol. 3, pp 149-169. (28) Adebajo, M. O.; Frost, R. L. Spectrochim. Acta, Part A. 2004, 60, 23152321. (29) O ¨ zmen, N.; C¸ etin, N. S.; Tingaut, P.; Se`be, G. Eur. Polym. J. 2006, 42, 1617-1624. (30) Sereshti, H.; Mohammadi-Rovshandeh, J. Iran Polym. J. 2003, 12, 15-20. (31) Tjeerdsma, B. F.; Militz, H. HolzRoh- Werkst. 2005, 63, 102-111. (32) Williams, P.; Norris, K. Near-Infrared Technology in the Agricultural and Food Industries, 2nd ed.; American Association of Cereal Chemists, Inc.: St. Paul, MN, 2004. (33) Shenk, J. S.; Workman, J. J.; Westerhaus, M. O. In Handbook of Near-Infrared Analysis; Burns, D. A., Ciurczak, E. W., Eds.; Marcel Dekker, Inc.: New York, 2001; pp 419-474. (34) Workman, J. J. Appl. Spectrosc. Rev. 1999, 34, 1-89. (35) Workman, J. J. Appl. Spectrosc. Rev. 2001, 36, 139-168. (36) Tsuchikawa, S. Appl. Spectrosc. Rev. 2007, 42, 43-71. (37) Gierlinger, N.; Jacques, D.; Schwanninger, M.; Wimmer, R.; Hinterstoisser, B.; Paˆques, L. E. Can. J. For. Res.-Rev. Can. Rech. For. 2003, 33, 17271736. (38) Gierlinger, N.; Schwanninger, M.; Hinterstoisser, B.; Wimmer, R. J. Near Infrared Spectrosc. 2002, 10, 203-214. (39) Fackler, K.; Schwanninger, M.; Gradinger, C.; Hinterstoisser, B.; Messner, K. FEMS Microbiol. Lett. 2007, 271, 162-169. (40) Fackler, K.; Schwanninger, M.; Gradinger, C.; Srebotnik, E.; Hinterstoisser, B.; Messner, K. Holzforschung 2007, 61, 680-687. (41) Schimleck, L. R.; Evans, R.; Jones, P. D.; Daniels, R. F.; Peter, G. F.; Clark, A. IAWA J. 2005, 26, 175-187. (42) Schimleck, L. R.; Evans, R.; Matheson, A. C. J. Wood Sci. 2002, 48, 132137.

quantification of acetyl groups or the weight percentage gain (WPG) of wood due to acetylation. Acetylation was performed on solid wood6,45,46 as well as on wood particles of different size and various applications, such as particle boards,47 hardboards,48 flake boards,49 oriented strand boards,50 and fiber boards.51 It was the aim to develop (a) methods for FTIR and NIR to determine the WPG due to acetylation from MIR and NIR spectra and (b) the acetyl group content from MIR spectra, by using uniand multivariate statistical methods. In the case of NIR, this would open up a new on-line measuring method. Therefore, different wood species and mixtures thereof as well as various sample geometries chemically modified with and without catalysts were used to simulate conditions close to those of industrial applications. EXPERIMENTAL DETAILS Materials and Acetylation. Acetylation of Wood Particless Sample-Set A. The material with particle size between 0.5 and 5 mm was a mixture of 60-70% softwood (mainly spruce) and 3040% hardwood (beech, oak, alder, poplar, willow, and others). The material was dried at 80 °C for 5 h with a layer thickness between 15 and 20 mm. Then the material was allowed to cool to room temperature in a desiccator over silica gel. About 5 g was then weighed into 100 mL round-bottom flasks equipped with a stirrer. Each flask was set under vacuum for 15 min at 21 ( 2 hPa at room temperature. Hot acetic anhydride (Aldrich, p.A. grade) (temperature above 100 °C) was introduced in order to obtain ambient pressure in the flasks. Residual anhydride was drained off so that the remaining impregnated particles were just covered with anhydride (to obtain good heat conductivity). The flasks were immediately transferred to an oil bath (110 °C) that was placed on a magnetic stirrer with heating and loosely closed with a glass stopper. The acetylation procedure was conducted for 2.5, 5, 10, 15, 30, and 45 min and for 1, 1.5, 2, 3, 4, and 5 h under continuous stirring of the acetic anhydride-soaked particles. Immediately after the desired reaction time has expired, the flasks were taken out of the oil bath and about 80-100 mL of ice-cold water was added and thoroughly shaken to stop the reaction. The water was decanted, and the particles were then transferred to a 250 mL Erlenmeyer vessel and therein washed three times with warm water (40 °C) under shaking until the pH of the water was (roughly measured with pH paper strips) was about neutral. After washing, the modified particles were dried at 80 °C for 20 h. The material was allowed to cool to room temperature in a (43) Gindl, W.; Teischinger, A. Wood Fiber Sci. 2002, 34, 651-656. (44) Schimleck, L. R.; Tyson, J. A.; Jones, P. D.; Peter, G. F.; Daniels, R. F.; Clark, A., III. J. Near Infrared Spectrosc. 2007, 15, 261-268. (45) Brelid, P. L.; Simonson, R. Holz Roh- Werkst. 1999, 57, 383-389. (46) Stefke, B.; Hinterstoisser, B. In Modifiziertes Holz, Eigenschaften und Ma ¨rkte. Modified Wood, Properties and Markets; Alfred Teischinger, R. S., Ed.; Institut fu ¨ r Holzforschung gemeinsam mit dem Verband Holzwirte O ¨ sterreichs: Vienna, 2002; Vol. 3, pp 25-55. (47) Arora, M.; Rajawat, M. S.; Gupta, R. C. Holzforsch. Holzverw. 1981, 33, 8-10. (48) Chow, P.; Bao, Z. Z.; Youngquist, J. A.; Rowell, R. M.; Muehl, J. H.; Krzysik, A. M. Wood Fiber Sci. 1996, 28, 252-258. (49) Hadi, Y. S.; Darma, I. G. K. T.; Febrianto, F.; Herliyana, E. N. For. Prod. J. 1995, 45, 64-66. (50) Papadopoulos, A. N.; Traboulay, E. Holz Roh- Werkst. 2002, 60, 84-87. (51) Gomez-Bueso, J.; Westin, M.; Torgilsson, R.; Olesen, P. O.; Simonson, R. Holz Roh- Werkst. 1999, 57, 433-438.

Analytical Chemistry, Vol. 80, No. 4, February 15, 2008

1273

desiccator over silica gel and then re-weighed on a four-figure balance. The weight gain was determined according to eq 1

WPG )

(mac - mo.d.) × 100 (%) mo.d.

(1)

where WPG is the percentage of weight gain, mo.d. the mass of untreated oven-dried particles, and mac the mass of acetylated particles. The water content of the oven-dried material that went into the acetylation procedure was determined by gravimetrical means, using a 105 °C drying temperature until the mass of material reached a constant weight. The acetylated material, 50 samples, was used for NIR measurements as is and after milling with a RETSCH Ultra Centrifugal Mill ZM 1000 device with a fixed ring sieve with 250 µm mesh size. For FTIR, 10 samples were milled with a fixed ring sieve with 120 µm mesh size. Acetylation of Spruce Wood ShavingssSample-Set B. Six spruce wood (Picea abies (L.) Karst) shavings (5 × 5 × 25-30 mm) were oven-dried at 50 °C for 21 h, liquid-phase acetylated with a mixture of acetic anhydride and pyridine 1:1 (v/v) for 5 and 10 h, respectively, vacuum-dried at 21 ( 2 hPa at 55 °C until the smell of the reagents and the catalyst was removed, and finally dried at 105 °C for 37 h. The WPG was determined as described above. For FTIR, the samples were milled with a RETSCH Ultra Centrifugal Mill ZM 1000 device with a fixed ring sieve with 250 µm mesh size. The pooled samples have a WPG of 25.9% (5 h) and 26.7% (10 h). Acetylation of Spruce Wood VeneerssSample-Set C. Spruce wood samples (2 × 10 × 20 mm) were dried at 105 °C for 14 h, liquidphase acetylated with acetic anhydride and mixtures of acetic anhydride and pyridine (0.5, 1.0, 5.0, and 10% (v/v) pyridine) for 15, 30, 45, 60, 75, 90, 105, 120, 150, 180, 210, 240, and 270 min, then extracted with acetone using a Soxhlet and dried at 105 °C for 5 h. After each drying step, the samples were stored in a desiccator overnight over silica gel and weight. The WPG was determined as described above. For FTIR, the samples with reaction times between 15 and 90 min (6 × 3 ) 18 samples) and the ones between 105 and 270 min (7 × 3 ) 21 samples) were pooled and milled with a RETSCH Ultra Centrifugal Mill ZM 1000 device with a fixed ring sieve with 120 µm mesh size and dried at 50 °C. Spruce Wood Veneers, Pine Sapwood, and Pine Heartwood VeneerssSample-Set D. Veneers, 1 × 20 × 95 mm, of spruce (Picea abies (L.) Karst) and pine (Pinus sylvestris) sapwood and heartwood were dried at 105 °C for 2 h, stored in a desiccator over silica gel under reduced pressure (300 mbar) for 41 h at room temperature (22 °C), then vacuum-treated at 11 mbar for 30 min. Venting to ambient pressure was performed by introducing liquid acetic anhydride at room temperature (22 °C). Liquid-phase acetylation was carried out on the submerged veneer samples at 110 °C under reflux for 0, 80, 95, and 240 min on pine sapwood, for 0, 80, 120, and 270 min on pine heartwood, and for 0, 15, 105, and 240 min on spruce. Samples, after being taken out of the reaction mixture, were dried at 35-40 °C for 30 min with a blow dryer. To remove unreacted acetic anhydride 1274

Analytical Chemistry, Vol. 80, No. 4, February 15, 2008

and acetic acid by product, samples were roughly washed 4 times with 1 L of distilled water under shaking, and afterward 20 samples per 1.3 L of distilled water were extracted under continuous shaking for 20 h at room temperature. Finally the samples were blow-dried at 35-40 °C before oven drying at 80 °C for 12 h. Before being re-weighed, the samples were allowed to cool to room temperature in a desiccator over silica gel. After NIR spectra collection, the veneers of each species were pooled according to the treatment time and milled with a RETSCH Ultra Centrifugal Mill ZM 1000 device with a fixed ring sieve with 250 µm mesh size and dried at 50 °C. Thereafter, NIR spectra were collected from the milled samples. Infrared Measurements. FT-NIR reflectance spectra were recorded at ambient temperature using a fiber probe connected to a Bruker (www.bruker.de) FTIR spectrometer (Equinox 55; germanium diode detector). One-hundred scans per measuring area (9-10 mm2) were collected at a spectral resolution of 8 cm-1, and a zero filling of 1 was applied.52 Six spectra from each solid wood sample (spruce wood veneers and pine sapwood and pine heartwood veneers) at random positions on the front and back side of the veneers were collected and averaged. Two replicate spectra of wood particles and milled samples were recorded and averaged. FTIR-KBr spectra (two replicates) of 17 milled samples (KBr pellets, ≈1.5 mg sample and 200 mg of KBr were prepared under vacuum in a standard device by applying a pressure of 75 kN‚cm-2 for 3 min) were recorded on the same spectrometer equipped with a DLATGS detector (32 scans per sample, spectral resolution, 4 cm-1; wavenumber range, 4000-600 cm-1). The collected spectra were ratioed against air.52,53 Furthermore, FTIR-ATR spectra from 22 samples (32 scans per sample, spectral resolution, 4 cm-1; wavenumber range, 4000600 cm-1 using a single reflection attenuated total reflectance (ATR) device (MIRacle, Pike Technologies, www.piketech.com) and a DLATGS detector) were recorded with a Bruker FT-IR spectrometer (Vertex 70). All samples used for infrared measurements were stored in a drying oven for several days at 50 °C. Thereafter, they were stored in a desiccator overnight over silica gel before measurement. Data Processing. Postspectroscopic Spectra Manipulation. Postspectroscopic manipulation was kept to a minimum. The FTIR and FTIR ATR wood spectra were baseline-corrected using the rubber-band method and normalized to the highest band only for Figure 1 (OPUS 6 software from Bruker). Second-derivative spectra were obtained by applying the Savitzky-Golay54 algorithm. No absorbance scale is given in Figures 1 and 5, because the spectra are shifted parallel to the wavenumber axis and/or were normalized. Calculation of Band Heights and Band Height Ratios (BHR). The height of the bands of the baseline-corrected FTIR (FTIRATR) at the maxima in the ranges between 1810 and 1710 cm-1, 1290 and 1190 cm-1, and 1045 and 1025 cm-1 was measured. The previous ones were divided by the latter ones giving BHR 1745/ 1030 (1740/1030) and 1240/1030 (1230/1030), respectively. (52) Schwanninger, M.; Hinterstoisser, B.; Gradinger, C.; Messner, K.; Fackler, K. J. Near Infrared Spectrosc. 2004, 12, 397-409. (53) Schwanninger, M.; Rodrigues, J.; Pereira, H.; Hinterstoisser, B. Vib. Spectrosc. 2004, 36, 23-40. (54) Savitzky, A.; Golay, M. J. E. Anal. Chem. 1964, 36, 1627-1639.

Figure 1. (a) FTIR-KBr and (b) FTIR-ATR spectra of acetylated wood with increasing degree of acetylation (WPG %) from bottom to top.

Figure 2. Correlations between the WPG and theBHR 1240/1030 (1230/1030) and 1745/1030 (1740/1030) respectively calculated from (a) FTIR-KBr and (b) FTIR-ATR spectra of milled wood (from sample-sets A and C).

Partial Least-Squares Regression (PLS-R) Modeling. OPUS Quant 2 software was used for data preprocessing (second derivatives), for the calculation of the PLS-R models, and for the prediction of the evaluation samples. Spectra were processed (smoothed and derived) by means of a 17-point smoothing filter and a second-order polynomial to obtain second derivatives, respectively,54 with OPUS software (version 6, www.brukeroptics.de). For calibration (cross-validation and test-set validation), the infrared data sets were regressed against the WPG. The numbers of spectra subjected to PLS-R are indicated in each paragraph of the results and discussion section. Calibration Models and Validation of the Models. In a first step, the preprocessed (second derivatives) NIR data were regressed against the calibration component, and by full cross-validation with one sample omitted, a significant number of PLS components (rank) was obtained.38 In a second step (test set validation), the calibration data set was divided into two groups (1 and 2). After sorting the whole data set according to the WPG, the samples were chosen alternately for the calibration set and for the test set. Each group was used for both cross-validation (CV) and test set validation (TS). First, group 1 was used for CV and group 2 for TS and then in reverse order, to evaluate if the model statistics were identical or at least very similar, leading to the same rank. All models were calculated to a maximum rank of 10, and the results of the

calibration (R2 coefficients of determination and RMSEE ) rootmean-square error of estimation), the cross-validation (R2 and RMSECV ) root-mean-square-error of cross-validation), and the test set validation (R2 and RMSEP ) root-mean-square-error of prediction) were compared. Therefore, test set validation was performed using the calibration with optimal rank in the crossvalidation (as usual in an external validation), and also an optimal rank was defined through test set validation. The comparison of the ranks gives a first indication of the predictive ability of the model, because models with large differences between the ranks determined by CV and TS are never satisfactory.38 The residual prediction deviation (RPD) or ratio of performance to deviation was introduced by Williams32 several years ago and is calculated as the ratio of 2 standard deviations; the standard deviation of the reference data for the validation set and the standard error of prediction (from cross-validation or test-set validation). Determination of Acetyl Groups. The acetyl groups content was determined for 11 samples according to Månsson and Samuelsson (1981)55 by means of an aminolysis with pyrrolidine and subsequent GC analysis on a GC (Carlo Erba HRGC). RESULTS AND DISCUSSION Acetylation of wood results in a decrease of the number of hydroxyl groups and in an increase in the acetyl groups content. (55) Månsson, P.; Samuelsson, B. Svensk Papperstidning 1981, 84, R15-R24.

Analytical Chemistry, Vol. 80, No. 4, February 15, 2008

1275

Figure 3. Correlations between acetyl groups (%) and the BHR 1240/1030 (1230/1030) (open diamonds) and 1745/1030 (1740/1030) (filled diamonds) respectively calculated from (a) FTIR-KBr and (b) FTIR-ATR spectra of milled wood (from sample-sets A and C).

The increase of the number of acetyl groups can be followed in the infrared spectra. Spectra in the Mid-Infrared Region (MIR FTIR). Spectra of acetylated wood have been recorded using KBr-pellets (Figure 1a) and an ATR unit (Figure 1b). A continuous increase of the band heights due to an increase of the WPG and the increasing number of acetyl groups, respectively, due to the acetylation of wood mainly at 1745 (1740) cm-1, 1374 (1370) cm-1, and at 1240 (1230) cm-1 is shown. These bands are assigned to the acetyl groups incorporating the valence vibration of CdO (1745 cm-1) and the corresponding C-O vibration (1240 cm-1) as well as the symmetric C-H bending vibration of the methoxy group (1374 cm-1)56,57 The absorption band at 1030 cm-1 is due to C-O valence vibrations from cellulose,58 hemicelluloses, and lignin.57,59 The correlation between the BHR 1745/1030 (1740/1030) or 1240/1030 (1230/1030) and the WPG allows the determination of the WPG from FTIR-KBr (Figure 2a) and FTIR-ATR (Figure 2b) spectra. The BHR (1745/1030) has successfully been used to determine the WPG of cellulose in a small range (0-5.5% WPG).28 Eleven of the samples shown in Figure 2 covering the WPG range from 0.4 to 17% were used to determine the acetyl group contents to verify the expected good correlation between them and the WPG. A R2 value of 0.95 (acetyl groups (%) ) [0.728 × WPG + 3.9 ] (%)) was found between the acetyl group content (3.96-16.1%) and the WPG. Moreover, the correlation between the BHRs and acetyl group content was investigated (Figure 3). A better correlation with a higher coefficient of determination (R2) was found between BHR and acetyl group content compared with that of BHR and WPG. This was expected because the acetyl group content accounts for the acetyl groups already present in wood, which depends on the species, and was found to be higher in hardwood (up to 5%) than (56) Collier, W. E.; Kalasinsky, V. F.; Schultz, T. P. Holzforschung 1997, 51, 167-168. (57) Faix, O. Holzforschung 1991, 45 (Supplement), 21-27. (58) Fengel, D.; Ludwig, M. Das Papier 1991, 45, 45-51. (59) Collier, W. E.; Schultz, T. P.; Kalasinsky, V. F. Holzforschung 1992, 46, 523-528.

1276

Analytical Chemistry, Vol. 80, No. 4, February 15, 2008

Figure 4. NIR spectra of acetylated wood with increasing degree of acetylation (WPG %) from bottom to top (from sample-sets A and C).

in softwood (about 1%).60 These results confirm that the BHR can be used to determine the WPG and therefore indirectly the acetyl group content in acetylated wood from infrared spectra. Although a slightly higher R2 value was found using the band at 1240 cm-1, it is suggested to use the band at 1745 cm-1, because bands derived from other functional groups do not overlap the latter one. Spectra in the Near-Infrared Region (NIR). The feasibility of NIR, used in many fields for on-line and process control, was investigated for its potential to determine the WPG of various acetylated input materials with regard to wood species and sample forms. (60) Fengel, D.; Wegener, G. Wood: Chemistry, Ultrastructure, Reactions; Walter de Gruyter & Co.: Berlin, 1989.

Figure 5. PLS-R validation results of the acetylated veneer samples (a) CV1, (b) TS2, (c) CV2, (d) TS1, and (e) CVall. (CV, cross-validation; TS, test-set validation, CVall, cross-validation with all samples; spruce, filled diamonds; pine sapwood, open circles; pine heartwood, filled circles). Table (f) shows the validation statistics wher RMSE is the RMSECV for CV and the RMSEP for TS. RPD is the residual prediction deviation.

The bands increasing as a consequence of acetylation can be seen in the spectra shown in Figure 4. The most obvious increase appears in the range from 6100 to 5700 cm-1, where the bands can be assigned to the first overtone (OT) of

C-H stretching, and to a lower extent in the range from 8700 to 8400 cm-1, which can be assigned to the second overtone of C-H stretching.26,32 Due to acetylation that reduces the number of hydroxyl groups a decrease of the band at about Analytical Chemistry, Vol. 80, No. 4, February 15, 2008

1277

Figure 6. Cross-validation results (predicted WPG versus measured WPG) for (a) unmilled (wood particles) and (B) milled wood (48 samples from sample-set A).

7000 cm-1 assigned to the first overtone of O-H stretching is expected. Unfortunately, in the NIR region there is no such intensive band like the CdO band in MIR. The first overtone of the CdO stretching vibration is in the MIR region (3600-3330 cm-1), and the second overtone (5300-5100 cm-1) gives a weak signal which is overlapped by the much stronger water band (5400-5200 cm-1).26,32 As shown in Figure 4, the range of the first overtone of C-H stretching vibrations gives the most intensive acetylationrelated signals, which are overlapped by those derived from wood (Figure 4, 0% WPG). To solve the problem of overlapping bands, multivariate statistical methods have successfully been used.37-40 Several data preprocessing methods such as vector normalization, first and second derivative, multiplicative scatter correction, and straight-line subtraction as well as combinations of these methods were applied to the spectra. The second-derivative method was found to be the most appropriate one for all samples (particles, shavings, milled wood, and veneers) investigated. Overall, the wavenumber range of 6080-5760 cm-1 gave the best results (PLS-R models), although a slightly better model was obtained for the wood particles by adding the range from 7550 to 6800 cm-1. Therefore, 141 spectra of the acetylated veneers (sample-set D) were divided in data sets for cross-validation and test-set validation as described in detail in the Experimental Section. PLS-R models were calculated from the second derivatives of the spectra using the range from 6080 to 5760 cm-1. The low root-mean-square error (RMSE) and the high R2 values of the PLS-R models are almost identical for CV and TS (Figure 5), and the high RPD obtained allow one to draw the conclusion that the model is applicable for process control. From the analytical point of view in accordance with AACC Method 39-00,61 the RPD should be in the following range: g2.5 screening in breeding programs; g5 acceptable for quality control; and g8 good for process control, development, and applied (61) AACC; American Association of Cereal Chemists (AACC), 1999; Vol. AACC Method 39-00, pp 15.

1278

Analytical Chemistry, Vol. 80, No. 4, February 15, 2008

research.61 It has to be kept in mind that the RPD is only correct and comparable when the data are normally distributed. PLS-R models for wood particles (48 samples from sample-set A, non-acetylated and acetylated) before and after milling the samples were calculated (second derivatives, 6080-5760 cm-1), and the cross-validation results are shown in Figure 6. The model statistics (wood particles, rank ) 3, R2 ) 0.98, RMSECV ) 0.78%, RPD ) 7.2; milled wood, rank ) 4, R2 ) 0.99, RMSECV ) 0.65%, RPD ) 8.6) show that by using the milled samples, a better model can be obtained. This was expected, because the measured area is small compared with the particle size. In processes, a larger area will be measured from a distance and therefore results similar to the ones obtained for milled wood particles are expected. Samples covering a wide range of (a) input materialswood from different speciessand (b) wood acetylated with different proceduresswith and without catalyst, varying moisture contents, and different sample geometriesswere selected to investigate if this large variation could be handled with one PLS-R model. A PLS-R model for milled acetylated and non-acetylated wood (62 samples from the sample-sets A, B, and C) was calculated (second derivatives, 6080-5760 cm-1), and the cross-validation result is shown in Figure 7a. The model statistics (rank ) 5, R2 ) 0.99, RMSECV ) 0.72%, RPD ) 9.6) are comparable to those of previous results (Figure 6b). Taking into account that the acetyl group content of spruce is less than half of that found in, e.g., beech, a small increase in the rank and the RMSECV was expected. A further model including additional spruce and pine wood samples (11 samples from sample-set D) gave the same rank and only a slight increase of the RMSECV (Figure 7b: rank ) 5, R2 ) 0.99, RMSECV ) 0.79%, RPD ) 8.6). The R2 value and the RPD are high and the RMSECV is low, making this model qualified for the prediction of the WPG of samples covering a wide range. CONCLUSION On the basis of the knowledge acquired during this study that a good linear correlation exists, ranging from 0% to 27% WPG

Figure 7. Cross-validation results (predicted WPG versus measured WPG) for the milled samples (a) (50 samples from sample-set A, open circles; 10 samples from sample-set C, filled triangles; 2 samples from sample-set B, filled diamonds) and (b) 11 additional milled acetylated veneer samples from sample-set D (filled rectangles).

between BHRs and WPG and acetyl group content, respectively, only a few samples are necessary to calibrate FTIR for KBr pellets and/or ATR. This means that a method for the determination of WPG can be obtained quickly at low cost. NIR was shown to be a good method for the determination of WPG. PLS-R models based on second derivatives of the spectra in the wavenumber range from 6080 to 5760 cm-1 resulted in a R2 value of 0.99, ranks between 3 and 5, RMSECV between 0.6 and 0.79%, and a RPD up to 10. Although various wood species, sample geometries/dimensions, and different acetylation procedures were used to cover a wide range of input material and parameters, good results were obtained. The only drawback of

NIR is the higher number of samples necessary for calibration. While FTIR is a good laboratory method, requiring some time for sample preparation, NIR is preferred for on-line measurements and process control, because it is faster and no sample preparation is needed. The potential of NIR for the on-line quality control of acetylated wood covering a wide range of input material, which is common for industrial processes, could be realized. Received for review October 9, 2007. Accepted November 29, 2007. AC7020823

Analytical Chemistry, Vol. 80, No. 4, February 15, 2008

1279