SciELO - Scientific Electronic Library Online

Home Pagelista alfabética de revistas  

Servicios Personalizados




Links relacionados


Journal of the Chilean Chemical Society

versión On-line ISSN 0717-9707

J. Chil. Chem. Soc. v.50 n.3 Concepción sep. 2005 


J. Chil. Chem. Soc., 50, N° 3 (2005), págs: 565-568





Laboratorio de Recursos Renovables, 1Facultad de Ciencias Químicas, 2Facultad de Ciencias Forestales, Universidad de Concepción, Casilla 160-C, Concepción Chile. E-mail:


The application of near-infrared (NIR) spectroscopy as a tool for the rapid determination of pulp yield and basic density of Eucalyptus globulus wood is described. Previously, we had examined these properties using mid-infrared (MIR) spectroscopy. More precise results were obtained based on NIR spectra where the calibrations models were more robust and better predictors than obtained for the MIR spectra. Twenty-seven samples were used for establishing the correlations (calibration) and three for making predictions based on the calibration (validation). Quant+ chemometrics software was used to obtain the calibration and prediction through principal components regression (PCR). A PCR model based on NIR spectra gave fits with a standard error of prediction (SEP) of 14.3 kg m-3 and 1.57% for basic density and pulping yield, respectively. Using PCR-MIR, the SEP values obtained were 21.44 kg m-3 and 1.73%. The method readily predicts yield (correlation coefficient: R2= 0.51), and density (correlation coefficient: R2 = 0.94) from spectral NIR measurement. Both NIR and MIR performed better in the analysis of the density property than for the pulp yield of the hardwood samples.

Keywords: Near-infrared spectroscopy, Quant+, E.globulus


We have reported that mid-infrared spectroscopy (MIR) was a useful method to predict wood's physical and chemical characteristics (1). A principal component regression (PCR) analysis produced calibration graphs for density and yield with sufficient precision. When the wavelengths as the explanatory variables were selected suitably for the calibration equations, we could predict both basic density and pulp yield for E. globulus. The statistical analysis was based on the Diffuse Reflectance Fourier Transformed Infrared (DRIFT) spectra using the MIR region with low output power. Therefore, the information from such optical method was confined to something for the subsurface layer of a sample. Near infrared spectroscopy (NIR), although yielding spectra that are less readily interpretable, has the advantage of having more available energy and very low noise detectors, and the measurable sample thickness is larger than for MIR . Moreover, good quality spectra in the NIR region can be obtained from solid samples using DRIFT.

Spectra in NIR consist entirely of overtones and combinations of primary bands within the MIR region. For wood or complex mixtures, the excessive overlapping of bands produces a diffuse continuum absorption with few characteristic features, making unequivocal band assignment difficult. Through multivariate analysis, spectral data can be related to any sample feature that is reflected in its spectrum (2). Even so, a major asset of NIR analysis is the ease with which reproducible spectra can be obtained for wood without complicated sample preparation. NIR spectroscopy has proved to be an excellent method for wood and fiber property evaluation. It has been used to evaluate properties such as density, pulp yield, pulp properties, chemical and biological degradation, hardwood-to-softwood ratio in wood chip mixtures, and resin content in wood chips (3-7). The combination of NIR spectroscopy and multivariate data analysis leads to improved models. For example, NIR spectroscopy and principal component analysis (PCA) have been used to classify wood samples of different origins (8,9). Partial least squares regression (PLS) has been applied to determine the chemical components content in Eucalyptus (E. globulus and E. nitens) wood samples from NIR spectra (10). Kraft pulp yields are also modeled in the same work in terms of the wood's glucan and xylan contents through the use of PLS regression.

In this study, the same 30 E. globulus samples previously analyzed by MIR spectroscopy (1) were studied using DRIFT-NIR to estimate density and pulping yield. Similarly, both sets of data were analyzed by PCR.


2.1 Wood preparation for NIR spectroscopy

Wood chips (approximately 2.5x1.5x0.2 cm) were obtained from E. globulus planted in the VIII Region of Chile. Wet chips were oven-dried at 60°C, and stored under dry conditions until used. Chips were milled in a knife mill (RetschMuhle) to pass through a 0.5 mm screen. For FT-NIR analysis, milled samples were sieved to pass through a 350 mm screen. The sieved samples were dried at 60°C overnight and stored under silica gel.

2.2 Pulping and properties determinations

Wood sample pulping was carried out in a M/K reactor system of 10 L with liquor recirculation automatically controlled. The operation conditions were:

Liquor- wood ratio: 4/1
Maximum temperature: 165 C
Time to maximum temperature: 140 min.
Alkali charge (Na2O): 16%
Sulfidity: 30%

The kappa number was determined using the TAPPI standard method 236 om-85(11). The basic density was determined using TAPPI standard method 258 om-94(11). The pulp yield was determined after pulp samples were washed, screened, and air-dried.

2.3 Measurement of NIR spectra and spectral manipulation

Milled wood samples (fraction passing through a 350 mm screen) were previously conditioned in the equipment room for 1 h (50% relative humidity and 23°C), then placed in the sample device of the Perkin Elmer Identicheck reflectance accessory, and NIR spectra were recorded between 1000 and 2500 nm in a Identicheck FT-NIR Perkin Elmer spectrometer using 50 scans. Raw reflectance data were converted to Kubelka-Munk units (K-M) and the spectra baselines were corrected to the regions near 1320 nm, 1870 nm, 2220 nm and 2410 nm. Two spectra were recorded from each wood sample. Duplicate spectra were averaged using the Perkin Elmer software facilities.

2.4 Data analysis

Spectral and chemical composition data were analyzed with the software package QUANT+ available with the Identicheck FT-NIR Perkin Elmer spectrometer. Spectral data were analyzed in digitized form at 2 nm intervals from 1000 to 2500 nm. All spectra were treated by a Kubelka-Munk transformation and a standardized normalization procedure (6,12). The first treatment was applied to the original NIR spectra to correct for any non-linearity of diffuse reflectance data. Utilizing the Kubelka-Munk theory, the reflectance (R) can be related to the absorption coefficient (K) and scattering coefficient (S) by the following equation (13):

In general, a standardized normalization procedure was applied to remove any systematic variation, reducing the effect of a baseline drift in the original NIR spectra. NIR spectra were converted before the PCR analysis to the second derivative mode using the SPECTRUMR software(13).

PCR models were based on a principal component analysis performed on spectral information followed by multiple linear regression between each wood or pulp property and chosen principal components. Minimal values for standard error of prediction (SEP) in full-cross validation procedures were used to define the number of principal components (PC) to be used in each PCR model (14).


The NIR spectra (1000-2500 nm) of five of the E. globulus wood samples are shown in Figure 1. Differences between the bands in the NIR spectra of different wood samples are difficult to discern: many of the maximum showing only as broad humps. The second derivative forms of the spectra are shown in Figure 2. The bands are indeed sharper and better resolved but with inverted peaks. Bands in this mode of the NIR spectra are less resolved than in their corresponding MIR spectra.

Figure 1. NIR spectra for five E. globulus wood samples

Figure 2. Second derivative NIR spectra for five E. globulus wood samples

The basic density and pulp yield of the E. globulus samples measured by the traditional procedures are shown in Table 1.

Data from 27 of the samples were used to establish the calibration data and validate the model. The model predicted the properties of samples 1, 15 and 30.

The individual NIR and MIR calibration plots for density and yield are shown in Fig. 3. NIR data provided good calibration fits (density, R2 = 0.94; yield, R2 = 0.51), which were better than those obtained from MIR (density, R2 = 0.84; yield, R2 = 0.40).

Figure 3. Predicted density and pulp yield by PCR vs values determined by traditional procedures. a) NIR, b) MIR

The statistical data, prediction values, and the PCs used for all calibration procedures are provided in Table 2. In general, NIR gave a smaller SEP with less deviation in the calibration than MIR. It is interesting to compare the deviation of the wood properties obtained above from the NIR spectra with those previously obtained using the PCR method with MIR spectra. NIR had a much faster, simpler sample preparation and shorter scan time with spectra obtained directly from the powdered wood without an intermediate dilution step involving several weighing steps. Additionally, it generally produced better calibration fit. A critical factor in the success of this work was the accuracy of the pulp yield and basic density data: both NIR and MIR gave better results for the density property than for pulp yield.


A direct comparison of the two procedures shows that NIR had a much faster, simpler sample preparation, shorter scan time, and in general gave a better calibration fit than produced by MIR. Although the NIR spectra generally have broad bands overlapping some group frequencies, they may simultaneously contain multiple sample compositional characteristics. In this work, we also showed that the NIR spectra reflected sample compositional characteristics. Our strategies are to propose a classification of eucalyptus wood (E. globulus from E. nitens) coupled with chemical information based on their compositional differences. The method will rely not only on a chemometrical approach, but also on chemical interpretation of the NIR spectra density and yield as well as for glucan, total lignin, hemicelullose, carbohydrate, xylan, and other wood properties and components.


The authors would like to thank FONDECYT (Grant No. 1020161) and Dirección de Investigación, Universidad de Concepción (DIUC 201.023-028-1.4).



1. J. Freer, J. Ruiz, M. Peredo, J. Rodriguez, and J. Baeza, Estimating the density and pulping yield of E. globulus wood by DRIFT-MIR spectroscopy and principal component regression (PCR), J.Chil. Chem. Soc. 48 (3): 19-22(2003).         [ Links ]

2. H. Bailleres, F. Davrieux and F. Ham-Pichavant, Near infrared analysis as a tool for rapid screening of some major wood characteristics in a eucalyptus breeding program, Ann. For. Sci. 59:479-490(2002).         [ Links ]

3. A. Michell, Pulpwood quality estimation by near-infrared spectroscopy measurements on eucalyptus woods, Appita 48(6):425-428(1995).         [ Links ]

4. L. Schimleck an J. French, Application of NIR spectroscopy to clonal Eucalyptus globulus samples covering a narrow range of pulp yield, Appita 55(2):149-154(2002).         [ Links ]

5. D. Easty, S. Berben, F. De Thomas and J. Brimer, Near-infrared spectroscopy for the analysis of wood pulp: quantifying harwood-softwood mixtures and estimating lignin content, Tappi J. 10:257-261(1990).         [ Links ]

6. A. Ferraz, J. Baeza, J. Rodriguez and J. Freer, Estimating the chemical composition of biodegraded pine and eucalyptus wood by DRIFT spectroscopy and multivariate analysis, Biores. Technol. 74:201-212(2000).         [ Links ]

7. A. Michell, Vibrational spectrometry-a rapid means of estimating plantation pulp wood quality, Appita 47:29-32(1994).         [ Links ]

8. L. Schimleck, A. Michell and P. Vinden, Eucalyptus wood classification by NIR spectroscopy and component analysis, Appitta 49(5):319-324(1966).         [ Links ]

9. H. Anti, D. Alexandersson, M. Sjöström and L. Wallbacks, Detection of kappa number distributions in kraft pulps using NIR spectroscopy and multivariate calibration,Tappi J. 3:102-108(2000).         [ Links ]

10. L. Schimleck, P. Wright, A. Michell and A. Wallis, Near infrared spectra and chemical composition of E. globulus and E. nitens plantation woods. Appita 50(1):40-46(1997).         [ Links ]

11. TAPPI Test Methods. Technical Association of the Pulp and Paper Industry. Atlanta. Ga.         [ Links ]

12. K. Beebe and K. Kowalski, An introduction to multivariate calibration and analysis, Anal. Chem. 59:1007A-1017A(1987).         [ Links ]

13. Spectrum for windows guide book, Perkin Elmer 1998.         [ Links ]

14. Quant+ guide book, Perkin Elmer 1998.         [ Links ]


Creative Commons License Todo el contenido de esta revista, excepto dónde está identificado, está bajo una Licencia Creative Commons