In the last 30 to 40 years, various new types of carbon materials have been engineered for multiple industrial uses. It is now well-known that the Raman spectrum is sensitive to the structure, even though the spectrum is rather uncomplicated. Because Raman spectroscopy now has a reputation for providing good information, potential users of Raman equipment can request information on the quality of their sample. However, they are often not able to define clearly what they mean by “quality.” If they are growing diamond films, they may or may not want interstitial sp

 carbon to glue polycrystalline diamond together. If they are growing hard diamond-like carbon (DLC) films, they may want to correlate the spectral characteristics with physical characteristics of the film. In this column, I explain how the Raman characteristics can aid in characterization of carbon materials.

There are essentially two ways that elemental carbon bonds to itself—sp

. The atoms crystallize in a tetrahedral lattice, and the zone center phonon falls at 1332 cm

. In perfect graphite, all the carbons are arranged in infinitely extended aromatic planes that stack in the A-B-A-B arrangement. The zone center phonon falls near 1580 cm

, and has been called the G band. However, graphite rarely appears without a second band in the mid-1300 cm

 region. In 1970, Koenig reported that he could correlate the relative intensity of this band to the crystallite size of the graphite sample being measured, as determined by X-ray diffraction (XRD) (1); the intensity of this band that came to be known as the D (“disorder”) band was related inversely to long range order. It was believed that the D band simply represented the disorder in graphite. However, in 1978, Vidano and Fischbach and, in 1990, Wang, Alsmeyer, and McCreery found that its frequency depended on the laser excitation wavelength. By exciting graphite with varying laser wavelengths they observed that the D band occurred near 1390 cm

 when excited in the UV, and 1270 when excited in the NIR (2,3). They also documented the presence of a band in the mid-2700 cm

 region that fell close to precisely twice the frequency of the D band. In the discussion by Wang and associates, they surmised that the dependence of the D band frequency on the excitation wavelength followed from each wavelength exciting different subpopulations of the materials. However, in 2004, Reich and Thomsen produced a theory based on double resonance that explained the “excitation-energy dependence (of the D band), the overtone spectrum, and the difference between Stokes and anti-Stokes scattering” (4). Although the theory is an elegant solid-state physics exposition of the energy band structure and its effects creating double resonance Raman conditions, it is probable that many readers of this article would not be in a position to become competent in this area. The point is to be cognizant of the best understanding of the origins of the Raman spectra of carbon-based materials so that the spectra can be used to derive empirically physical and chemical information on manufactured materials.

Before the double resonant theory was developed, the D nomenclature was further used to denote “diamond-like,” which was probably a misnomer. The term came from the development of hard carbon films for the computer hard disc industry where the conclusion was that the hardness of the films was related to the D or “diamond-like” band, whose frequency was not far from that of diamond. In fact, the spectra of these films were fit to D and G bands and the characteristics of the fits (peak positions, peak widths, peak amplitudes, % Gaussian in the fitting function, and integrated intensity in the fits) were used to correlate to the physical characteristics of the films. Because of the wavelength dependence of the D band intensity, for empirical correlations to be effective, measurements must be made in a consistent manner—same excitation wavelength, same spectral dispersion, and so forth. Irmer and Dorner-Reisel explained the spectral behavior related to the size of graphitic nanoclusters, hydrogenation, and hydrogen-free tetrahedral amorphous carbon (ta-C) with sp

-bonded content, and how these properties are correlated with hardness and wear properties (5).

In this column, I show examples of measurements of all these materials, so that someone interested in using the spectra for empirical correlations has a roadmap of where to start.

, as seen in Figure 1. A clean diamond has a flat baseline, with some evidence for the second order band near 2600 cm

. Figure 1 shows the Raman spectrum with a picture of the small chip from which the spectrum was acquired. Figure 1 also shows the spectrum plotted in wavelength units, rather than Raman units, and shows the photoluminescence (PL) that sometimes appears in diamond because of impurities. These PL spectral features have been used to characterize diamond, both natural and man-made (made under pressure or by chemical vapor deposition [CVD]) (6).

Figure 2 shows the results of a mapping measurement of a polycrystalline diamond film. The two colored maps were made using the Raman frequency of the two peaks fitted by the software. This material is clearly not high quality, single crystal diamond, but if the small crystals are grown to be used for polishing, the material would work well. If, however, one wanted a diamond window, this sample would clearly not be useful. In addition, it is useful to point out that the peak shifts in this data set are understood to be because of stress.

Figure 3 shows the spectrum of crystalline graphite measured with a 600 and 1800 g/mm grating on the same instrument. Note that the relative intensities do not exactly overlay because of differences in line widths relative to the instrument’s dispersion with the two gratings. Because of this phenomenon, one must make measurements under consistent conditions if meaningful correlations from the spectra want to be derived. Note that the D band appears in this spectrum because truly single crystal natural graphite is rare.

Figure 4a shows a Raman map of graphene. Graphene is technically a single layer of graphite, but the spectra of several-layer graphene have been used to determine the number of layers. In the figure, the sets of red, green, and blue cursors were used to calculate the integrated intensities of the D, 2D, and G bands. Clearly, the D band is quite low in intensity in this sample, which is expected for a near perfect lattice. A mark of true single layer graphene is a high I(2D)/I(G) ratio, which can be seen in the light teal map on the lower right. However, the intensity maps do not show the entire picture. Figure 4b shows a few spectra extracted from the map, indicating how the spectra change in this data set from point to point. A good reference for the Raman spectra of graphene is an article in 

 (7) which discusses many interesting phenomena that can be monitored with Raman spectra.

There is also the possibility of monitoring single-walled and multi-walled carbon nanotubes (SWCNTs and MWCNTs). In addition to the D and G bands (and their derivatives discussed above), the SWCNTs have bands in the low frequency region(<300 cm

), which are diagnostic of the chirality and size of the tube. However, the intensity of the spectra are extremely dependent on the type of tube, which changes the resonance conditions. That means, if you are looking for a particular type of tube, you may or may not see it with the laser wavelength that you are using. There are many good references on the solid states physics of these materials, but in another 

 article, there is a figure that clearly shows the relationship between the excitation wavelength and the type of tube measured (Figure 12 in reference [8]).

We have also measured the spectra of various carbon fibers that were shared with us by Herman Noether, retired from Hoechst Celanese. Thornel was manufactured from pitch, and its structure was known to exhibit radial graphite planes. Other fibers manufactured from polyacrylonitrile (PAN) exhibit turbostratic graphite planes that wrap around the fiber axis. Figure 5 shows the spectra of these three samples. Both Thornel and PAN GY70 were fairly crystalline as determined by their Raman spectra, both of which have sharp G and D bands. The PAN fiber called Celion 3000 exhibits very broad D and G bands and very weak features in the 2D region. I have band-fit all three spectra. Thornel and GY70 both exhibit a weak but sharp band between 1610 and
1620 cm

, and this band has been seen in intercalation studies. To fit the Celion spectrum, two extra bands are required to flatten the residual. If you will be using these spectra to characterize or compare samples, again the samples have to be oriented in the same way, and all instrumental conditions have to be the same (laser wavelength, spectral dispersion, and polarization). Then, you can export the fitted parameters to an Excel file to determine quantitatively how the samples compare. In fact, you can correlate physical chemical properties with these parameters, even if you do not understand what they mean at the nanoscopic level. Tables Ia–Ic show the fitted parameters from these spectra.

An example of how these spectra can be useful is something I came across on the internet while assembling this article. An article that goes back to 1979 discusses the uptake of certain metal chlorides that are used to treat carbon fibers to raise their electrical resistance. This study was sponsored by NASA because of there being several industrial accidents that were a consequence of electrical sparking in manufacturing facilities (9). What was found was that the chemical uptake was much higher in the pitch fibers than in the PAN fibers. Looking at the cartoon figures that I made for Figure 5, it is easy to understand why there is more penetration in the pitch fibers than in the PAN fibers. Polarized Raman spectra would enable differentiating PAN from pitch fibers because of the orientation of the graphite planes in the fiber.

There is also a material called glassy carbon, a spectrum of which I show in Figure 6. Note that it also has D and G bands, wide enough to overlap a bit, but sharper than the bands of DLC, which you will see in a moment. In addition, the D band is more intense than the G band which is rather unusual. According to Professor William White at Penn State, the adjective “glassy” for this material is actually a misnomer. It is highly crystalline, but the crystals are quite small. Professor White told me this more than 30 years ago (10), and the current theories of the structures of sp

 carbons might affect this description, but if you do a Google search on the Raman spectrum of glassy carbon, most spectra are similar to the one in Figure 6.

The Raman spectra of these films have been used extensively for physical characterization. Figure 7 shows the spectra of two such films, with the band-fitting, and Table II shows the band-fit parameters extracted and imported into Excel for further manipulation that will enable correlation with physical properties. There is an excellent review by Whitley that describes how thistechnology is used in the computer disk industry (11). The figure shows two spectra, both of which were successfully fit to a D and G band. I say “successfully fit” because the residual shows only noise, but no indication of the necessity for another band. Comparison of the D bands in the two spectra indicate significant differences in band width. More interesting, however, is the I(D)/I(G) ratio when using the highest count rate in the region of the two bands vs. the amplitude (height) of the fitted peaks vs. the integrated intensities of the fitted peaks. Table III shows the results of these simple calculations for the two samples. Reading across the table from left to right for each sample, it is clear that the calculated ratios are quite different. People tell me that they want to know the D/G ratio as a quality index. But which ratio do they really want? The problem is that the non-spectroscopist is not aware of this complexity. It appears that calculating the D/G ratio should be a simple calculation, but, clearly, the answer is different depending on how the calculation is done. Which one will produce the best correlation with the desired properties? Even if you do not understand what is the origin of these spectral characteristics, you can see that what you choose to calculate will determine what number you will get. Maybe the thing to do is to correlate separately all of these ratios to the physical characteristics of the sample to see which one actually predicts the data the best.

We showed representative spectra of many classes of carbon allotropes, with an indication of the origins of the spectra. We tried to indicate what about the spectra can be important for some obvious applications, providing literature references in order for the reader to start planning their work.

(3) Y. Wang, D.C. Alsmeyer, and R.L. McCreery, 

(6) www.gia.edu/gems-gemology/spring-2016-photoluminescence-spectroscopy-diamond-applications-gemology

(7) L.M. Malard, M.A. Pimenta, G. Dresselhaus, and M.S. Dresselhaus, 

(8) M.S. Dresselhaus, G. Dresselhaus, R. Saito, and A. Jorio, 

(9) T.E. Thompson, NASA-CR-166201 Carbon Fiber Modification NASA Contract NAS 2-10091 Nov 1979 SRI Project 7976

(10) Private Communication, Professor William White, Penn State University. 

(11) A. Whitley, “The Use of Raman Spectroscopy to Monitor the Quality of Carbon Overcoats in the Disk Drive Industry,“ in 

Handbook of Raman Spectroscopy from the Research Laboratory to the Process Line

, I.R. Lewis and H.G.M. Edwards, Eds. (Marcel Dekker, Inc. New York, NY, 2001), pp. 975–998.

Articles

Column: Molecular Spectroscopy Workbench

Here, we explain how Raman spectroscopy is valuable for characterizing industrially important carbon materials.

Use of Raman Spectroscopy to Qualify Carbon Materials

In this second part of this four-part series on spectroscopy instrument components, we look into optical components or subassemblies used for vibrational spectroscopy instruments. Our “under the hood” look continues as we survey the most typical instrument optical design component materials. As we continue publishing this survey series, we note that several tutorial articles and 

The Spectroscopy Instrument Components Terminology Guide

, the latter of which was published digitally in February 2022, are available to our readers.

In Part 2 of our four-part spectroscopy components survey article, we take a closer look at a variety of aspects associated with spectroscopy optical materials and designs to include infrared (IR), Raman, and near-IR (NIR) optics, UV optics, visible optics, charge-coupled devices (CCDs), monochromators, interferometers, diode arrays, digital light processing (DLP) designs, microscopes, and hyperspectral imaging (HSI) systems. These are shown in Table III, with the table containing four columns of information content—the component name, a text description, basic specifications, and references and links are given in the table (1–5). Part I of this series is published in the March 2022 issue. Optical filters (high-pass, low-pass, interference, acousto-optic tunable filters [AOTF], tilting, broadband, narrowband, neutral density, custom, and so forth) is a rather large subject and beyond the scope of this article series. 

 covers the general filter terms (1), and we note the subject of optical filters would be an excellent topic for a future separate article. For the table, all components or optical assemblies are listed in alphabetical order.

The Spectroscopy Instrument Components Terminology

The Concise Handbook of Analytical Spectroscopy: Physical Foundations, Techniques, Instrumentation and Data Analysis, in five volumes, first edition, UV, Vis, NIR, IR, and Raman

 (World Scientific Publishing-Imperial College Press, Hackensack, NJ and Singapore, 2016). ISBN-13: 978-9814508056.

 (John Wiley & Sons, New York, NY, 1st ed., 2002). ISBN: 978-0-471-98847-2

(4) ASTM (American Society for Testing and Materials) ASTM Volume 03.06, “Molecular Spectroscopy and Separation Science; Surface Analysis” (ASTM International, West Conshohocken, PA, 2017).

(5) N.B. Colthup, L.H. Daly, and S.E. Wiberley, 

Introduction to Infrared and Raman Spectroscopy

 (Academic Press/ Elsevier, Boston, MA, 3rd ed., 1990). ISBN: 9780121825546.

Featured Article

In this second part of this four-part series on spectroscopy instrument components, we take a closer look at optical components or subassemblies used for vibrational spectroscopy instruments.

A Survey of Basic Instrument Components Used in Spectroscopy, Part 2: Optical Materials and Designs

The feasibility of using a portable near-infrared (NIR) spectrometer combined with partial least squares for discrimination analysis (PLS-DA) to identify five similar 

 wood species was investigated. To improve model reliability and identification accuracy, the effects of three main spectra preprocessing methods and their combinations were examined. Then, the model performance created by spectra collected before and after specimen surface sanding were compared. In the PLS-DA model based on spectra preprocessed by standard normal variate (SNV) and first derivative combined, the identification accuracy of the five species was above 95%, and the compared results demonstrated the surface natural change influences the NIR model performance. It was shown that a portable NIR device combined with PLS-DA can be used to rapidly and accurately identify five similar Cinnamomum wood species.

 is a genus belonging to the Lauraceae family that is composed of approximately 250 known species mainly distributed in warm temperature and tropical regions (1,2). There are about 46 species in China, and they are mainly distributed south of the Qinling mountains, especially in the southwestern, central, and southern China provinces. The genus, 

, is one of the most popular and important species because its byproducts are used in food flavoring, cosmetics, and medicines (3). The major aromatic product, camphor, is obtained from Cinnamomum camphora, and the timber of this species is considered a good material for high-value furniture, carvings, and structural uses (4).

 has been conducted (4,5,6). The results have shown that the wood structure was homogeneous among genera (4,7). As a result, it is difficult to determine wood species immediately on-site even for a skilled inspector. Although some advancement has occurred in wood species identification auxiliary technology, a device that could identify wood species rapidly and on-site at the species level has long been desired. Near-infrared (NIR) spectroscopy has been investigated as a reliable technique for identifying some wood species, but it has not been tested for the Cinnamomum species. NIR spectroscopy is a nondestructive, fast, and cost-effective analytical technology (8), and an NIR instrument combined with multivariate analysis is a useful tool not only for the quantification but also for the qualification of wood species (9).

Wood species identification by NIR spectroscopy has been investigated extensively. Adedipe and others correctly classified the red and white oak species by using NIR spectroscopy and soft independent modeling of class analogies (SIMCA) (10). NIR spectroscopy, with aid of partial least squares for discrimination analysis (PLS-DA), identified mahogany (

) from three similar-looking woods, crabwood (

) at two conditions of laboratory-processed powder (11) and solid wood block (12). Based on NIR spectroscopy, eight rosewoods were clearly classified into eight categories by principal component analysis (PCA) (13). Horikawa and others demonstrated that two anatomically similar wood species, 

, could be identified by employing NIR spectra collected from heartwood samples (14). These studies demonstrated that NIR spectroscopy is an effective technique for wood species identification in laboratory conditions, but further studies are needed.

The development of hardware and high-performance accessories has promoted the advance of miniature NIR spectrometers (15). Some investigations involving wood species identification and provenance detection by portable NIR devices has been conducted. The origin of true mahogany wood from five countries was detected by portable NIR spectroscopy devices and multivariate data analysis (16). The potential of portable NIR (spectral range: 1595–2396 nm) technology combined with SIMCA and PLS-DA to identify seven high-value 

 wood species was investigated (17). The results showed that visually confusable species of 

 can be identified at species level using portable NIR technology. Therefore, it is promising to identify wood species accurately and rapidly on-site by using a miniature NIR spectrometer.

The main advantages of portable NIR devices include nondestructive and on-site detection; however, the disadvantages include the spot size, limited wavelength range, and limited resolution (18). Because of these drawbacks, it makes it difficult to develop an all-purpose NIR spectra database. The samples from previous research studies are mainly from sawmills, lumber boards, and so on. Wood herbarium, with abundant varieties of wood specimens, is an invaluable resource for establishing a teaching model database. If the NIR spectroscopy model database established using wood specimens from the wood herbarium can be used to identify external wood species, it would promote the development of wood species filed and identified by portable NIR spectroscopy technology dramatically.

NIR investigations involving large-scale wood specimens from wood herbariums are few in number. For this study, specimens were stored in the wood herbarium under natural conditions for many years; as a result, the smell and surface color of the woods has changed. The NIR spectra mainly contain information pertaining to the overtones and combinations of fundamental vibrational transitions, including those of the C-H, O-H, and N-H functional groups (19). The study of Horikawa and others showed that when the aging wood samples were used, the NIR model was ineffective in identifying two similar wood species, 

, when sampling from traditional architecture (14). As a result, researching the feasibility of building an NIR model by wood specimens from wood herbarium is desirable.

To verify if the removal of oxidized film on the surface of the wood specimen interfered in the discrimination analysis of wood species, 29 samples of mahogany were cut from the end of boards that were stored to be exported from the market in the research of Braga and others (12). The spectra were obtained before and after sanding the sample surface. The result showed no significant difference in the performance of two spectra type models, but the influence of the natural change of wood specimens stored in wood herbarium to the portable NIR model was indeterminate.

Cinnamomum porrectum [Roxb.] Kosterm., Cinnamomum tenuipilum Kosterm., Cinnamomum camphora [L.] Presl., Cinnamomum glanduliferum [Wall.] Nees

) from herbarium wood were selected to test the effectiveness of the portable NIR model. There are two principal aims in this research: one was to demonstrate the possibility of the identification of five Cinnamomum species by a portable NIR spectrometer, and the other one was to verify the influence of the natural surface change of the wood specimens to the species identification using portable NIR technology.

Sample Preparation and NIR Spectra Measurement

Cinnamomum porrectum, Cinnamomum tenuipilum, Cinnamomum camphora, Cinnamomum glanduliferum

) were harvested from the wood herbarium of the Southwest Forestry University (Table I). For this study, 30 specimens with no defect were selected for each species. The specimens used were cut from different parts of the tree. The average size of the specimen is around 75 mm × 10 mm × 45 mm, and the total number of the specimens was 150. Prior to spectral measurement, two cross-sections were sanded with grit for each specimen. In addition, three species (

Cinnamomum glanduliferum, Cinnamomum camphora, 

) were selected, and the spectra were obtained before and after sanding the sample surface.

The NIR spectra were collected with the MicroNIR on-site spectrometer (Viavi Solutions, Inc.) over a range of 908–1650 nm. For each of the 150 specimens, five spectra in different points on the cross-section were obtained, totaling 1500 raw spectra. In the present work, the average process was carried out for five raw spectra from the same cross-section. Therefore, there are 60 spectra for each species, and the training set included 40 spectra, with the test set composed of 20 spectra for each species.

Spectra preprocessing is often indispensable in multivariate analysis, and it is a very important step in NIR quantitative and qualitative analysis. The appropriate NIR spectral preprocessing method can effectively improve the applicability of the model. The noise information in the NIR spectra can be filtered, and the effective information can be retained by a reasonable preprocessing method, thereby reducing the complexity of the NIR model and improving the robustness of the model. The commonly used NIR spectral pretreatment methods mainly include smoothing, derivatives, standard normal variate (SNV), and so on. To create high dependability for the discriminate model, the three different pretreatment methods (smoothing, first derivative, and SNV) and combinations of these three methods were carried out in this work.

Smoothing reduces the noise in the data without reducing the number of variables. In smoothing, 

 values are averaged over one segment symmetrically surrounding one data point. The raw value on this point is replaced by the average over the segment, thus creating a smoothing effect. The Savitzky-Golay smoothing was selected and accomplished using the Unscrambler X 10.4 software (CAMO) for this experiment.

The Savitzky-Golay algorithm fits a polynomial function to each curve segment, thus replacing the original values with more regular variations. The length of the segment (right and left of each data point), and the order of the polynomial can be selected. The polynomial order was three, the number of right and left side point was 10, and the number of smoothing points was 21 in this study.

Derivatives are applied to correct for baseline effects in spectra to remove non-chemical effects and create robust calibration models. Derivatives also resolve overlapped bands to provide a better understanding of the data and emphasize small spectra variations not evident in the raw data. The first derivative preprocess based on Savitzky-Golay derivative was carried out in this work.

The Savitzky-Golay derivative can be used to compute first, second, third, and fourth order derivatives. The Savitzky-Golay algorithm is based on performing a least squares linear regression fit of a polynomial around each point in the spectrum to smooth the data. The derivative is then the derivative of the fitted polynomial at each point. The algorithm includes a smoothing factor that determines how many adjacent data point variables will be used to estimate the polynomial approximation of the curve segment. The first derivative of a spectrum is a measure of the slope of the spectral curve at every point. The slope of the curve is not affected by purely additive baseline offsets in the spectrum, and thus the first derivative is an effective method for removing such offsets.

SNV is a row-oriented transformation which removes scatter effects from spectra by centering and scaling individual spectra. Each value 

Partial Least Squares (PLS) for Discrimination Analysis (PLS-DA)

To better obtain the information of NIR spectra, PLS-DA was performed using the Unscrambler X 10.4 software (CAMO). PLS-DA involves developing a conventional partial least squares regression model, in which the 

 vector is composed of a binary variable (0 and 1). The data matrix 

 is formed by the NIR spectra of the sample. If the variable takes the value of 1, the samples belong to that group and if a variable takes the value of 0, the sample is not a member of that group (17). The square of the correlation coefficient for calibration (validation) (

), standard error of calibration (SEC), standard error of cross validation (SECV), the number of correct classifications, and the accuracy of classification models were used to evaluate the models.

The characteristics of raw spectra and after they are preprocessed by different methods is shown in Figure 1. Compared with raw spectra, the spectra barely changed after the smoothing processing (1b), and it shows that there is a serious baseline drift. Other preprocessed methods allowed more specific identification of small and latent absorption peaks, which displayed the specificity of absorption peaks for certain bonding groups. Most of the different absorption of bands are a result of the differences in the chemical composition between the species, but it is clear that overlap between the bands in the same figure is significant. A visual inspection of the NIR spectra is not helpful for wood species identification, and it is very difficult to infer the chemical compounds responsible for the difference bands (peaks) and attribute these band variations to specific species. Therefore, it is necessary to use multivariate analysis for spectral analysis.

Table II shows the optimal number of factors for each model suggested by the software Unscrambler v10.4. (Note the 

 superscripts denote calibration and prediction, respectively.) The results of calibration and prediction models for discriminating five species based on raw spectra were quite poor: the 

. This result demonstrated the performance of the model based on raw spectra may not be enough for five species identification.

The results of calibration and prediction models for discriminating five species under three independent preprocess methods are shown in Table III. First, in contrast to the model created by raw spectra the 

 value of five species were decreased in the model created by the smooth preprocessed spectra.

 were increased in the model created by the first derivative preprocessed spectra compared to the mode created by raw spectra, but the 

 value of another three species decreased. Third, the 

 increased in the model created by the SNV preprocessed spectra, compared to the model created by raw spectra, but the 

 value of another two species decreased. After a single method was processed among five species, the 

 value of two or three species increased, whereas the 

 value of residual species decreased compare to the results of raw spectra. The holistic parameter value may be increased if the different preprocessing methods combined are used to treat spectra. As a result, the three preprocessing methods (smooth, first derivative, and SNV) combined to treat spectra were carried out to obtain a high performance PLS-DA model.

And the results of calibration and prediction models for discriminating five species under four combine preprocessing methods are presented in Table IV. Considering the parameter value of each species, and compared to Tables II and III, it is clear that the result of the model created by SNV and first derivative combined preprocessing were best, exhibiting a relative high 

 value (0.804 to 0.852), along with a relatively low standard error of calibration (SEC) value (ranged between 0.142 and 0.163) and standard error of validation (SECV) value (ranged between 0.155 and 0.178).

Then, the PLS-DA identification models based on raw spectra and spectra preprocessed using seven different methods were established, aiming to test the ability and accuracy of the models. Based on PLS-DA, the results of the unknown samples of five species predicted by identification models are presented in Table V. It is obvious that in the model created by the spectra preprocessed by SNV and first derivative combined, only one sample of 

 was misclassificated into another species. The accuracy of the model was above 95%, the highest accuracy was obtained among eight PLS-DA models. This result indicated that the PLS-DA models had the ability to quickly predict and classify five similar Cinnamomum species.

Scaled loading plots of first three factors of the spectra preprocessed by SNV and first derivative combine (factor 1 has a proportion of variance 83%; factor 2 has a proportion of variance 9%; and factor 3 has a proportion of variance 5%) are shown in Figure 2, which is a plot of X-loading weights for all the components compared to the variable number. If a variable has a large positive or negative loading weight, this means that the variable is important for the component concerned. If a variable has the same sign for all the important components, it is most likely to be an important variable. The highest accuracy was obtained in the model created by spectra preprocessed using SNV and first derivative (Table V). As a result, the loading plots of the first three factors (had a greater contribution to model) of the spectra preprocessed by SNV and first derivative combined was analyzed (Figure 2). From the point of view of factor 1, there existed significant peaks at approximately 982–1106 nm. For factor 2, the main absorptions were at 1025, 1267, and 1434 nm, and the important peaks were approximately 1186 and 1484 nm for factor 3.

) and their NIR mean spectra obtained before and after sanding based on SNV and first derivative combined preprocess were shown in Figure 3. The differences between the same specimen absorbance before and after sanding are displayed in Figure 3. For example, the change in absorbance of the 

 bands located at 950–1000 nm, 1350 nm, and 1410 nm.

The results of calibration and prediction models for discriminating three species, sanded and not-sanded samples, under two combined preprocess methods (SNV and first derivative, smoothing and SNV) are shown in Table VI. It can be seen that the ing and SNV) are shown in Table VI. It can be seen that the 

 value of sanded samples were above 0.895, and the 

 value of not-sanded samples were above 0.833 for the model based on spectra preprocessed by SNV and first derivative combined. The 

 value of the sanded samples were above 0.890, and the 

 value of not-sanded samples were above 0.789 for the model based on spectra preprocessed by smoothing and SNV combined; the sanded sample provided better results than the unsanded (raw) sample. The identification results of unknown samples from three different species, sanded samples, and not-sanded samples using two types PLS-DA models were presented in Table VII. For the sanded samples, all samples were identified correctly for the two models. 100% accuracy was attached for all sanded samples. For the not-sanded samples, one sample of 

 was misclassified into other species for the model created by the spectra preprocessed by SNV and first derivative combined, and one sample of 

 were misclassified into other species for the model created by the spectra preprocessed after smoothing and SNV combined. The accuracy was 95% for not-sanded samples. Although two types of models are created from the spectra obtained from sanded and not-sanded specimens; samples that were sanded produced higher accuracy models in wood species identification using portable NIR technology.

The NIR spectra were collected from the cross-sections of wood specimens in this research. The cross-sections of hardwoods consisted of vessels cells, which have a longitudinal axis that is parallel to the direction of the NIR incident radiation. Therefore, the NIR radiation can travel further into the wood (20). Meanwhile, the sanding process reduces the roughness of cross-sections and exposes more vessels, so more information of the chemistry and structure of wood species can be obtained from the spectra of sanded wood samples. The model created by the spectra from the sanded samples was better than the model created by the spectra from the not-sanded ones.

This result was not consistent with the result Braga and others reported (12). Comparing the samples of their experiment, it can be seen that the main differences are the sample origin and storage conditions. The samples were cut from the ends of boards that were stored and then exported in the research of Braga and others.

The samples of this work are from the wood herbarium and stored here for many years under natural conditions, so natural surface changes can be seen in the sample surfaces. It is demonstrated that natural changes of a sample surface influence the performance of the NIR model for wood species identification using portable NIR technology.

 species from wood herbarium were correctly identified at species-level based on the portable NIR spectra preprocessed by SNV and first derivative combined. Evaluation of the models created by raw spectra and the spectra processed by a single method demonstrated the different preprocessing methods combined are useful for creating a portable NIR wood species identification model, and portable NIR spectroscopy combined with PLS-DA model can be used to rapidly and accurately identify five similar 

 species. From the practical point of view, further advancement of the method would examine the feasibility of identifying external wood samples based on models created by wood specimens from the wood herbarium. However, this study was limited in the number of wood specimens and the external origin of wood samples. The performance comparison of the three species sanded spectral model and the not- sanded spectral model demonstrated that creating a discriminant model using wood specimens from the wood herbarium was feasible. And the study suggests that specimen surface natural changes influence the NIR model performance. The natural change surface of specimens should be removed (samples sanded) to create NIR models with improved identification accuracy.

The study was supported by China National Natural Science Fund (Grant Numbers 31770766 and Grant Numbers 31370711). The authors thank the wood herbarium of Southwest Forestry University for the wood specimens supported.

(1) G.K. Jayaprakasha, L. Jagan Mohan Rao, and K.K. Sakariah, 

(2) S.S. Cheng, C.Y. Lin, C.K. Yang, et al, 

(3) E.J. Doh, J.H. Kim, S. eun Oh, et al, 

(4) M.K. Singh, M. Sharma, and C.L. Sharma, 

(6) I.W. Andianto, T.K. Waluyo, R. Dungani, et al, 

(10) O.E. Adedipe, B. Dawson-Andoh, J. Slahor, et al, 

(11) T.C.M. Pastore, J.W.B. Braga, V.T.R. Coradin, et al, 

(12) J.W.B. Braga, T.C.M. Pastore, V.T.R. Coradin, et al, 

(14) Y. Horikawa, S. Mizuno-Tazuru, and J. Sugiyama, 

(15) C.A. Teixeira dos Santos, R.N. Páscoa, M. Lopo, et al, 

Encyclopedia of Analytical Chemistry: Applications, Theory, and Instrumentation

 (John Wiley & Sons, Hoboken, NJ, 2006), pp. 1–27.

(16) D.C. Silva, T.C.M. Pastore, L.F. Soares, et al, 

(17) F.A. Snel, J.W.B. Braga, D. da Silva, et al, 

(19) Z. Yang, Y.N. Liu, X.Y. Pang, et al, 

(20) B. Leblon, O. Adedipe, G. Hans, et al, 

 are with the Research Institute of Wood Industry at the Chinese Academy of Forestry, in Beijing, China. 

 is with the College of Materials Science and Engineering at Southwest Forestry University, in Yunnan, China. Direct correspondence to 

Peer-Reviewed Research

Portable NIR spectroscopy, combined with discrimination analysis (PLS-DA), can be used to rapidly and accurately identify five very similar wood species of the Cinnamomum genus.

Identification of Five Similar Cinnamomum Wood Species Using Portable Near-Infrared Spectroscopy

When evaluating fruit, such as the Nanguo pear, ensuring high quality of the product is a top priority. The quality of a Nanguo pear will directly affect its appearance, taste, and commodity value. Currently, most of the quality detection methods of Nanguo pears are destructive. In this work, nondestructive testing methods for the internal and external quality of the Nanguo pear were studied. First, the spectral curves of the Nanguo pear were collected to establish the internal comprehensive quality indicators. Second, the successive projection algorithm was used to extract the spectral characteristics of samples. Based on the gray level co-occurrence matrix, texture features of the sample image were extracted. Finally, the model was established by fusing spectral and image features together, and good modeling results were achieved. The internal and external quality of the Nanguo pear were detected simultaneously. The accuracy of the external quality assessment was 97.14%. For the internal quality prediction model, the correction set 

 and RMSEC are 0.959 and 0.483, respectively, and the prediction set 

 and RMSEP are 0.920 and 0.615, respectively. It provides a theoretical basis for the development of an online nondestructive testing system for fruit.

Nanguo pears are known as the “king of pears.” Nanguo pears can be made into fruit juices, preserved fruits, pear pastes, dried fruits, and other byproducts that not only improve the added value of other products, but also plays an important role in promoting the development of the Nanguo pear industry. However, in the blind pursuit of generating a higher yield, the quality of the Nanguo pear fruit has declined, a trend that has resulted in a sharp decrease in its commodity value and quality. This decline of Nanguo pear quality has affected the development of the Nanguo pear industry and the related economic chain.

Currently, fruit quality detection and classification relies on the look, smell, and touch of the fruit under observation. However, that approach to quality detection method requires a sizable labor force and remains an inefficient process. In addition to these disadvantages, quality cannot be accurately judged by visual inspection. At this time, most of the internal quality research needs to be conducted by chemical detection, which causes damage to the fruit itself. Therefore, nondestructive testing and research of the internal quality based on hyperspectral imaging can effectively overcome these shortcomings to achieve rapid and nondestructive testing, which has become the focus of researchers. Currently, the application of hyperspectral imaging technology is used to detect the internal and external qualities (11) of several fruits and vegetables, including apples (2,5), bananas (7), peaches, pears, dates, navel oranges, blueberries (4), Hami melons, oranges, potatoes, spinach, and tomatoes (14). Hyperspectral imaging technology has been widely used in various fields. It has applications in product safety (1,3,12), pest identification, especially in vegetables, fruits (6,10), meat (13), grain (15), and agricultural products (18).

Looking at the literature, we can see that there is increased interest in studying the Nanguo pear. However, most of the studies are about trace elements in a Nanguo pear, the effects of lipids on the aroma and coloring, the effects of enzymes on the decay and storage time of Nanguo pear, and the effects of different regions, climates, and environments on the Nanguo pear. There are few studies on the nondestructive detection and recognition of the internal quality of Nanguo pears, especially on assessing quality using hyperspectral imaging technology.

Hyperspectral imaging can simultaneously detect the spectral and image information of the measured object. We conducted long-term research on the Nanguo pear previously. In that work, the external quality identification of the Nanguo pear was achieved (16), and a nondestructive method for testing the internal quality of Nanguo pear was studied (17).

It can be seen that hyperspectral imaging can detect the internal and external quality of the Nanguo pear. However, using hyperspectral imaging for simultaneous external and internal quality detection has been rarely reported.

Following our past research on the Nanguo pear, the purpose of the current study was to use hyperspectral imaging technology to explore the development of a comprehensive detection method to evaluate the internal and external quality of the Nanguo pear in a nondestructive manner. Our study offers a theoretical basis for the follow-up development of an online quality detection and grading system based on hyperspectral imaging technology. It also provides a theoretical basis and technical support for the development of real-time, online, and nondestructive detection systems for the fruit industry.

The experimental materials were purchased from the Datun fruit farm in Haicheng City. In total, 420 fresh Nanguo pears were collected, including 200 external defect samples (50 heteromorphic fruits, 50 pest damaged fruits, 50 rust fruits, and 50 brown rot fruits) and 220 intact fruit samples. There were 280 correction sets (30 abnormal fruits, 30 pest fruits, 30 rust fruits, 30 brown rot fruits, and 160 intact fruits) and 140 prediction sets (20 abnormal fruits, 20 pest fruits, 20 rust fruits, 20 brown rot fruits, and 60 intact fruits).

Hyperspectral Image Acquisition and Correction

The hyperspectral image was collected by the system. First, the parameters of the hyperspectral image acquisition system were adjusted. Next, the distance between the sample and the lens was set at 272 mm, the platform moving speed was set at 1.1 mm/s, and the camera exposure time was set at 79 ms. Then, according to the above method, the collected hyperspectral images were corrected by the blackboard and whiteboard. Finally, the sample images of Nanguo pears were collected. The black and white correction formula was as follows:

 is the relative reflectivity of the black-and-white corrected hyperspectral image; 

 is the hyperspectral image calibrated by the blackboard; and 

 is the hyperspectral image calibrated by the whiteboard.

According to the agricultural industry standard of the People’s Republic of China (NYT 1076-2006), the external and internal quality indexes of the Nanguo pear were selected. The detection index of the external quality of Nanguo pear samples examined whether there were surface defects. The types of surface defects selected in this paper are common defects, namely abnormal fruit, insect-damaged fruit, rust fruit, and brown rot fruit. Soluble solid content (SSC), titratable acid content (TAC), fruit firmness (FF), and vitamin C (VC) were selected as the evaluation indexes to characterize the internal quality of the Nanguo pear. Based on the study of the four internal quality indicators, the internal comprehensive quality indicators, which can represent the four internal quality indicators, were determined and used for subsequent internal and external quality research and analysis.

Determination of Physical and Chemical Indexes

For the determination of pulp hardness, refer to “NY/T 2009-2011 determination of fruit hardness” (22). For the determination of the content of soluble solids, please refer to “GB/T 12295 determination of soluble solids in fruit and vegetable products refractometer method” (23). For the determination of titratable acid content, refer to “GB/T 12293 determination of titratable acid in fruit and vegetable products” (24). For the determination of vitamin C, refer to “GB/T 6195-86 determination of vitamin C in fruits and vegetables (2,6-dichloroindo-phenol titration)” (25).

Determination of an Internal Comprehensive Quality Index

First, the correlation analysis of the internal quality indicators was examined. Second, the Kaiser-Meyer-Olkin (KMO) (20) and Bartlett (21) tests were used to determine whether these variables were suitable for factor analysis. Finally, Statistical Package for the Social Sciences (SPSS) software was used to analyze the original variables of internal quality indicators, and the principal component eigenvalues and contribution rates were extracted. The score of the coefficient matrix of components was calculated by the regression method to comprehensively evaluate several quality indexes of the Nanguo pear. The rotation score coefficient matrix was used to establish and analyze the subsequent principal factor expression.

In this study, successful projection algorithms (SPA) (9) were used to extract feature variables. In this paper, Gray level co-occurrence matrix (GLCM) (8), was used to extract and analyze the image texture features. The best wavelength extraction method was selected, and the image with the best wavelength was taken as the feature image. Based on the feature image, five texture features, namely contrast, homogeneity, correlation, energy, and entropy, were extracted for subsequent modeling analysis.

The least squares–support vector machine (LS-SVM) (19) has clear advantages in processing nonlinear, small sample, and high-dimensional data. It has been widely used in image recognition, electricity, food inspection, communication, and other fields. Based on the LS-SVM algorithm, the internal quality model of the Nanguo pear was established using the spectral and image information as the input of the model and the internal comprehensive quality index as the output. By taking the spectral and image information as the input of the model and the external quality index as the output, the external quality model of Nanguo pear was established. Nanguo pear samples with five different external quality indexes (heteromorphic fruit, insect damaged fruit, rust fruit, brown rot fruit, and normal fruit) were assigned 1, 2, 3, 4, and 5, respectively. If the difference between the predicted value of Nanguo pear sample and the hypothetical category was within ± 0.5, the sample discrimination was correct. For example, if the prediction result of the Nanguo pear sample was between 0.5 and 1.5, the discrimination was correct for heteromorphic fruit. If the prediction result of the Nanguo pear sample was between 1.5 and 2.5, the discrimination was correct for insect damaged fruit. If the prediction result of the Nanguo pear sample was between 2.5–3.5, the discrimination was correct for rust fruit. If the prediction result of Nanguo pear sample of brown rot fruit was between 3.5–4.5, the discrimination was correct. If the prediction result of Nanguo pear sample of normal fruit was between 4.5–5.5, the discrimination was correct.

Measurement and Correlation Analysis of Internal Quality Index 

After the hyperspectral images of Nanguo pear samples were collected and the spectral curve data were obtained, the internal quality indexes of the Nanguo pear were detected, including SSC, TAC, FF, and VC.

The correlation analysis of the SSC, TAC, FF, and VC physical and chemical indexes were made. The results appear in Table I. It can be seen from Table I that there was significant correlation among SSC, TAC, FF, and VC. SSC is highly correlated with TAC and VC, and FF is also highly correlated with SSC, TAC, and VC.

Selection of Internal Quality Comprehensive Indicators

It can be seen that there was significant correlation between the four internal quality indicators. If any of them were used to establish the internal quality model of the Nanguo pear, the model would be affected by the other three quality indicators. Therefore, it was necessary to conduct factor analysis on these four indicators to find the comprehensive quality indicators that can represent the four internal quality indicators of Nanguo pears.

KMO and Bartlett tests were conducted on the four indicators of internal quality to determine whether these variables were suitable for factor analysis. The KMO test coefficient was 0.692, which was greater than 0.5. The significance of the Bartlett test was 0.02—less than 0.05. It showed that the partial correlation between the variables was strong and independent, making it suitable for factor analysis.

SPSS software was used to analyze the original variables of the internal quality index of the Nanguo pear samples, and principal component analysis (PCA) was used to extract the eigenvalues. The main component analysis method was used to determine the common factor variance statistics for the internal quality index variables.

Variance between the common factors of the four indicators of internal quality was above 0.8, which indicated that the common factors, after being extracted by factor analysis, retained most of the original information and had good interpretation ability for each indicator variable. In the process of factor analysis, PCA was used to extract the characteristic value and the principal component contribution rate, and the maximum variance method was used to rotate. The extracted characteristic values and contribution rates of principal components are shown in Table II.

It can be seen from Table II that the first two factors of the Nanguo pear samples have covered the main information of internal quality indicators.

The regression method was used to calculate the score coefficient matrix of components. It can be seen from Table III that in the composition matrix of the original variable of quality index, the score effect of TAC and hardness was not good, but in the composition matrix after rotation, the score effect of both was significantly improved, and the overall effect of the rotation composition matrix was good. Therefore, the rotation score coefficient matrix was used to establish and analyze the subsequent main factor expression.

From Table III, we can get two main factor scores of the component matrix as follows (for equations 2–4, 

 is the first main factor score expression for the quality indicators‘ rotated score matrix; 

 is the second main factor score expression for the quality indicators‘ rotated score matrix; and 

 is the internal comprehensive quality index of Nanguo pear):

Based on the contribution rate and factor score of the two main factors, the internal comprehensive quality index 

 of Nanguo pears was established. The expression is as follows:

Using the internal comprehensive quality index formula, the 

-value of each sample of Nanguo pears was calculated.

The average value of the comprehensive quality index of external defect samples is lower than that of intact fruit samples, and the average value of the comprehensive quality index of external defect and intact fruit samples is lower than that of intact fruit samples. However, the standard deviation of the comprehensive quality index of external defect and intact fruit samples is greater than that of intact fruit samples, which is because of rust fruit disease that reduces the SSC and VC values. The SSC, FF, and VC values of brown rot fruit decreased more than the SSC, TAC, and VC values of pest fruit, but both decreased.

The hyperspectral image acquisition system was used to collect the samples of Nanguo pears, extract the region of interest (ROI) (100 pix * 100 pix) of the five types of samples, and obtain the average reflectance spectrum curve of the heteromorphic area, insect damaged area, rust area, brown rot area, and normal fruit area, as shown in Figure 1. The spectral curve range was 400–1000 nm.

It can be seen from Figure 1 that the absorption peak of relative reflectance curve of intact fruit sample appears at 675 nm, which is mainly caused by the absorption of chlorophyll in the pulp and epidermis of Nanguo pears. The shape of the relative reflectance curve of the heteromorphic fruit is most similar to that of the intact fruit, but the relative reflectance of the intact fruit is higher. The absorption peak of the rusty fruit decreased with the decrease of chlorophyll. After being attacked by insects, Nanguo pears formed a standing hole on their surface, where the chlorophyll was missing, so there was no obvious absorption peak. There was also no obvious absorption peak in the brown rot fruit.

After preprocessing the spectral data, SPA was used to select the characteristic wavelengths. The minimum effective wavelength was set to four and the maximum effective wavelength was set to 18. The wavelength selection process is shown in Figure 2.

It can be seen from Figure 2 that when the number of characteristic wavelengths is 6, there is a turning point in RMSE, and the RMSE value is 0.25317. Therefore, the number of characteristic wavelengths is 6, which are 527 nm, 606 nm, 750 nm, 796 nm, 825 nm, and 902 nm, respectively.

Texture Feature Extraction Results and Analysis of Image

Here, 750 nm was selected as one of the characteristic wavelengths by SPA, and it can be seen from Figure 1 that the spectral curves of the five samples have the largest difference at the wavelength of 750 nm. Therefore, the images at 750 nm were selected as the feature images. The characteristic images of heteromorphic fruit, insect damaged fruit, rust fruit, brown rot fruit, and intact fruit are shown in Figure 3.

GLCM was used to extract the gray level co-occurrence matrix in the direction of 0° from the feature images of each sample, and then extract five texture feature parameters (contrast, correlation, entropy, homogeneity, energy) from each gray level co-occurrence matrix.

Internal Comprehensive Quality Inspection

Characteristic wavelengths of internal quality comprehensive index extracted by SPA are selected as the spectral characteristic. The texture features extracted by GLCM are taken as image features. The LS-SVM model are established based on spectral features, image features, the fusion information of spectral features, and image texture features to predict the internal comprehensive quality of the samples.

The modeling results show that the LS-SVM model based on spectral characteristics has a good effect. The corrected set 

 and RMSEC are 0.905 and 0.798, respectively, and the predicted set 

 and RMSEP are 0.882 and 0.937, respectively. The discriminant result of the model based on the texture feature is slightly lower than that of the model based on spectral feature. The corrected set 

 and RMSEC are 0.895 and 0.992, respectively, and the predicted set 

 and RMSEP are 0.878 and 1.285, respectively. The accuracy of the LS-SVM model based on the fusion information of spectral features and texture features has been significantly improved compared with the model of single spectral and texture feature. The prediction results of fusion information are shown in Figure 4.

We propose a theoretical basis using vis-NIR spectroscopy for the development of an online nondestructive testing system for the quality of Nanguo pear fruit.

June 2022

Use of Raman Spectroscopy to Qualify Carbon Materials

A Survey of Basic Instrument Components Used in Spectroscopy, Part 2: Optical Materials and Designs

Identification of Five Similar Cinnamomum Wood Species Using Portable Near-Infrared Spectroscopy

Study of Nondestructive Testing of Nanguo Pear Quality Using Vis-NIR Spectroscopy

A Novel High Selectivity Fluorimetric and Colorimetric Probe Based on Rhodamine B Hydrazide Derivatives for Detecting Hg2+ in Aqueous Media