Background removal in soil analysis using laserinduced breakdown spectroscopy combined with standard addition method R. X. Yi,1,2 L. B. Guo,1,2 X. H. Zou,1 J. M. Li,1 Z. Q. Hao,1 X. Y. Yang,1 X. Y. Li,1,* X. Y. Zeng,1 and Y. F. Lu 1

Wuhan National Laboratory for Optoelectronics (WNLO), Huazhong University of Science and Technology, Wuhan, Hubei 430074, China 2 These authors contributed equally to this work * [email protected]

Abstract: The matrix effect of powder samples, especially for soil samples, is significant in laser-induced breakdown spectroscopy (LIBS), which affects the prediction accuracy of the element concentration. In order to reduce this effect of the soil samples in LIBS, the standard addition method (SAM) based on background removal by wavelet transform algorithm was investigated in this work. Five different kinds of certified reference soil samples (lead (Pb) concentrations were 110, 283, 552, 675, and 1141 ppm, respectively) were used to examine the accuracy of this method. The root mean square error of prediction (RMSEP) was more than 303 ppm by using the conventional calibration method. After adoption of SAM with background removal by wavelet transform algorithm, the RMSEP was reduced to 25.7 ppm. Therefore, the accuracy of the Pb element was improved significantly. The mechanism of background removal by wavelet transform algorithm based on SAM is discussed. Further study demonstrated that this method can also improve the predicted accuracy of the Cd element. © 2016 Optical Society of America OCIS codes: (300.6365) Spectroscopy, laser induced breakdown; (350.5400) Plasmas.

References and links 1. 2.

M. H. Zenk, “Heavy metal detoxification in higher plants--a review,” Gene 179(1), 21–30 (1996). D. C. Adriano, Trace Elements in Terrestrial Environments: Biogeochemistry, Bioavailability, and Risks of Metals (Springer-Verlag, 2001). 3. D. W. Hahn and N. Omenetto, “Laser-induced breakdown spectroscopy (LIBS), part I: review of basic diagnostics and plasma-particle interactions: still-challenging issues within the analytical plasma community,” Appl. Spectrosc. 64(12), 335–366 (2010). 4. J. P. Singh and S. N. Thakur, Laser-Induced Breakdown Spectroscopy (Elsevier, 2007). 5. S. Rosenwasser, G. Asimellis, B. Bromley, R. Hazletta, J. Martinb, T. Pearceb, and A. Ziglera, “Development of a method for automated quantitative analysis of ores using LIBS,” Spectrochim. Acta B At. Spectrosc. 56(6), 707–714 (2001). 6. R. Noll, Laser-Induced Breakdown Spectroscopy (Springer Berlin, 2012). 7. M. H. Ebinger, M. L. Norfleet, D. D. Breshears, D. A. Cremers, M. J. Ferris, P. J. Unkefer, M. S. Lamb, K. L. Goddard, and C. W. Meyer, “Extending the applicability of laser-induced breakdown spectroscopy for total soil carbon measurement,” Soil Sci. Soc. Am. J. 67(5), 1616–1619 (2003). 8. M. Z. Martin, N. Labbe, N. O. Andre, S. D. Wullschleger, R. D. Harris, and M. H. Ebinger, “Novel multivariate analysis for soil carbon measurements using laser-induced breakdown spectroscopy,” Soil Sci. Soc. Am. J. 74(1), 87–93 (2010). 9. R. Barbini, F. Colao, R. Fantoni, A. Palucci, and F. Capitelli, “Application of laser-induced breakdown spectroscopy to the analysis of metals in soils,” Appl. Phys., A Mater. Sci. Process. 69(1), S175–S178 (1999). 10. S. H. Chen, X. H. Ma, H. F. Zhao, and H. Lv, “Research of laser induced breakdown spectroscopy for detection of trace Cd in polluted soil,” in 22nd International Conference on Optical Fiber Sensor (ISOP, 2012), paper 8421AL.

#254197 © 2016 OSA

Received 18 Nov 2015; revised 15 Jan 2016; accepted 27 Jan 2016; published 1 Feb 2016 8 Feb 2016 | Vol. 24, No. 3 | DOI:10.1364/OE.24.002607 | OPTICS EXPRESS 2607

11. V. K. Unnikrishnan, R. Nayak, K. Aithal, V. B. Kartha, C. Santhosh, G. P. Gupta, and B. M. Suri, “Analysis of trace elements in complex matrices (soil) by laser induced breakdown spectroscopy (LIBS),” Anal. Methods 5(5), 1294–1300 (2013). 12. F. H. Kortenbruck, R. Noll, P. Wintjens, H. Falk, and C. Becker, “Analysis of heavy metals in soils using laser induced breakdown spectrometry combined with laser-induced fluorescence,” Spectrochim. Acta B At. Spectrosc. 56(6), 933–945 (2001). 13. M. J. C. Pontes, J. Cortez, R. K. H. Galvão, C. Pasquini, M. C. Araújo, R. M. Coelho, M. K. Chiba, M. F. de Abreu, and B. E. Madari, “Classification of Brazilian soils by using LIBS and variable selection in the wavelet domain,” Anal. Chim. Acta 642(1-2), 12–18 (2009). 14. M. E. Essington, G. V. Melnichenko, M. A. Stewart, and R. A. Hull, “Soil metals analysis using laser-induced breakdown spectroscopy (LIBS),” Soil Sci. Soc. Am. J. 73(5), 1469–1478 (2009). 15. A. S. Eppler, D. A. Cremers, D. D. Hickmott, M. J. Ferris, and A. C. Koskelo, “Matrix effects in the detection of Pb and Ba in soils using laser-induced breakdown spectroscopy,” Appl. Spectrosc. 50(9), 1175–1181 (1996). 16. B. C. Windom and D. W. Hahn, “Laser ablation-laser induced breakdown spectroscopy (LA-LIBS): A means for overcoming matrix effects leading to improved analyte response,” J. Anal. At. Spectrom. 24(12), 1665–1675 (2009). 17. J. Pareja, S. López, D. Jaramillo, D. W. Hahn, and A. Molina, “Laser ablation-laser induced breakdown spectroscopy for the measurement of total elemental concentration in soils,” Appl. Opt. 52(11), 2470–2477 (2013). 18. B. Bousquet, J. B. Sirven, and L. Canioni, “Towards quantitative laser-induced breakdown spectroscopy analysis of soil samples,” Spectrochim. Acta B At. Spectrosc. 62(12), 1582–1589 (2007). 19. J. E. Haddad, M. Villot-Kadri, A. Ismaël, G. Gallou, K. Michel, D. Bruyère, V. Laperche, L. Canioni, and B. Bousquet, “Artificial neural network for on-site quantitative analysis of soils using laser induced breakdown spectroscopy,” Spectrochim. Acta B At. Spectrosc. 79, 51–57 (2013). 20. E. C. Ferreira, D. M. Milori, E. J. Ferreira, R. M. Da Silva, and L. Martin-Neto, “Artificial neural network for Cu quantitative determination in soil using a portable laser induced breakdown spectroscopy system,” Spectrochim. Acta B At. Spectrosc. 63(10), 1216–1220 (2008). 21. J.-B. Sirven, B. Bousquet, L. Canioni, L. Sarger, S. Tellier, M. Potin-Gautier, and I. L. Hecho, “Qualitative and quantitative investigation of chromium-polluted soils by laser-induced breakdown spectroscopy combined with neural networks analysis,” Anal. Bioanal. Chem. 385(2), 256–262 (2006). 22. Z. Wang, T. B. Yuan, Z. Y. Hou, W. D. Zhou, J. D. Lu, H. B. Ding, and X. Y. Zeng, “Laser-induced breakdown spectroscopy in China,” Front. Phys. 9(4), 419–438 (2014). 23. X. H. Zou, L. B. Guo, M. Shen, X. Y. Li, Z. Q. Hao, Q. D. Zeng, Y. F. Lu, Z. M. Wang, and X. Y. Zeng, “Accuracy improvement of quantitative analysis in laser-induced breakdown spectroscopy using modified wavelet transform,” Opt. Express 22(9), 10233–10238 (2014). 24. W. Hyk and Z. Stojek, “Quantifying uncertainty of determination by standard additions and serial dilutions methods taking into account standard uncertainties in both axes,” Anal. Chem. 85(12), 5933–5939 (2013). 25. C. Ma and X. Shao, “Continuous wavelet transform applied to removing the fluctuating background in nearinfrared spectra,” J. Chem. Inf. Comput. Sci. 44(3), 907–911 (2004). 26. X. G. Shao, A. K. M. Leung, and F. T. Chau, “Wavelet: a new trend in chemistry,” Acc. Chem. Res. 36(4), 276– 283 (2003). 27. C. M. Galloway, E. C. Le Ru, and P. G. Etchegoin, “An iterative algorithm for background removal in spectroscopy by wavelet transforms,” Appl. Spectrosc. 63(12), 1370–1376 (2009). 28. X. G. Ma and Z. X. Zhang, “Application of wavelet transform to background correction in inductively coupled plasma atomic emission spectrometry,” Anal. Chim. Acta 485(2), 233–239 (2003).

1. Introduction In recent years, many human activities, such as industrial production and the application of pesticides and fertilizers, have resulted in heavy metal environmental pollution. Heavy metals, such as lead (Pb), cadmium (Cd), and chromium (Cr) are harmful to plants, animals, and humans [1, 2]. The residue of heavy metals in the soil is absorbed by plants. When the plants are consumed, these heavy metals are accumulated in human body, since humans are at the top of the food chains, serious diseases appear, such as nerve injury and cancers. Thus, it is necessary to monitor the accumulation of these heavy metals in the soil. There are many wellestablished analytical techniques for detecting heavy metals in soil, including inductively coupled plasma optical emission spectroscopy (ICP-OES), inductively coupled plasma mass spectroscopy (ICP-MS), atomic absorption spectroscopy (AAS). For ICP-OES, ICP-MS, and AAS, the high costs for instrument maintenance, and the complex operational procedures (strong oxidants and corrosive are used to process samples, which take long time and is not safe for experimenters) are the main drawbacks. For XRF, the main drawbacks are inability to

#254197 © 2016 OSA

Received 18 Nov 2015; revised 15 Jan 2016; accepted 27 Jan 2016; published 1 Feb 2016 8 Feb 2016 | Vol. 24, No. 3 | DOI:10.1364/OE.24.002607 | OPTICS EXPRESS 2608

detect light elements, and safety management of radioactive sources. The disadvantages restrict the wide application of these techniques. As a useful analytical tool, laser-induced breakdown spectroscopy (LIBS) [3–6] can overcome the disadvantages described above due to its characteristics of low invasiveness, relatively simple process for sample preparation and ease of use. In the last two decades, many groups have paid more attention to the LIBS technique, it has been applied in the detection of carbon [7, 8], or heavy metals in soil. For instance, R. Barbini et al. [9] studied the influence of plasma temperature on the quantitative LIBS measurement, and a quantitative determination of several metals from Antarctic sea-bottom sediments has been achieved on a wide range of concentrations. S.H. Chen et al. [10] used spectral data processing approaches (i.e. Radial Basis Function (RBF) neural network and Lorentz function) to optimize the LIBS spectral data for the purpose of improving the limits of detection (LOD) in polluted soil. Unnikrishnan et al. [11] used LIBS to detect copper (Cu), zinc (Zn), and calcium (Ca) elements in soil, which proved the feasibility of detecting heavy metals in soil by LIBS. Kortenbruck et al. [12] combined LIBS with laser-induced fluorescence (LIF) for the analysis in soil, 22 different elements were detected. Especially, the limits of detection were Cd (0.3 ppm) and Tl (0.5 ppm). These studies focused on the possibility of heavy metal detection in soil using LIBS but did not investigate the matrix effect, which is the key to the accuracy of quantitative analysis using LIBS. In short, the matrix effect is a phenomenon that ablated mass, element distribution, and laser absorptivity change with the matrix of the samples, which ultimately results in a significantly different signal response at the same concentration level in two different samples [3]. This phenomenon reduces the accuracy of LIBS. The matrix of soil is complex due to the matrix changes associated with its geographical distribution, which creates an unavoidable problem in the field of soil component analysis. Some groups have studied the matrix effect in soil. Pontes et al. [13] studied an analytical methodology for soil classification based on the use of LIBS and chemometric techniques. The methodology was validated in a case study involving the classification of 149 Brazilian soil samples into three different orders. Essington et al. [14] detected different elements with different concentration ranges in soil. The research indicated that the correlation between LIBS response and elemental content was poor (r < 0.98), and the relative errors (REs) of prediction for the LIBS-detected elements were less than acceptable for an analytical technique (> b/k, the parameter b will not affect C0. If C0 is not large enough, the background (parameter b) will seriously influence the results.

Fig. 1. Schematic diagram of the standard addition method.

#254197 © 2016 OSA

Received 18 Nov 2015; revised 15 Jan 2016; accepted 27 Jan 2016; published 1 Feb 2016 8 Feb 2016 | Vol. 24, No. 3 | DOI:10.1364/OE.24.002607 | OPTICS EXPRESS 2610

Background removal by wavelet transform algorithm Wavelet transform (WT) [25, 26] is a new signal processing technique which has undergone rapid development in the past two decades. Compared with Fourier transform, WT has the characteristics of time-frequency localization and multiresolution analyses of signals. The basic principle of background removal by discrete wavelet transform (DWT) is described as follows [27]. First, the spectral signal is decomposed into several levels. Second, the approximation coefficients of the highest level are set to zero, which is considered to only contain the low-frequency background information. Finally, the spectrum is reconstructed using the modified approximation coefficients (low-frequency part) of the highest level and the original detail coefficients (high-frequency part) of each level, and then the backgroundcorrected spectrum is obtained. As the signal peaks and the background partially overlap in the low-frequency domain of the highest level of signal decomposition, an accurate background fitting will inevitably occur. Hence, the iterative WT [27] is adopted for background correction. The process containing decomposition (decompose the signal to a level where all of the background is only just contained), coefficient modification, and reconstruction of the spectral signal will be repeated until the fit converges, so as to make the fitted background approximate the real value (the information of the elements concentrations in the processes of background removal will not be used). 2. Experiment 2.1 Experiment setup The LIBS experimental apparatus used in this study is schematically shown in Fig. 2(a). A Qswitched Nd: YAG laser (Beamtech, Nimma 400, wavelength: 532 nm; repetition rate: 3 Hz; pulse width: 8 ns; pulse energy: 40 mJ) was focused 2 mm under the surface of the soil sample using a 150 mm focal length lens to create plasma. The plasma emission was coupled into an optical fiber by a light collector and then collected by a Czerny-Turner spectrometer (Andor Tech., Shamrock 500i, grating: 1800 l/mm, slit width: 10 μm), the narrow slit width was chosen to reduce spectral interference. A digital delay generator (SRS DG535) was adopted to trigger the laser pulses and control the gate delays and widths of the intensified charge-coupled device (ICCD) (Andor Tech., iStar 320T: DH320T-18F-E3-26mm). The plasma emission spectra were collected with a delay time of 4 μs and gate width of 6 μs to obtain high spectral intensity and signal-to-background ratio (SBR) for Pb element and with a delay time of 2 μs and gate width of 1.5 μs for Cd element. To reduce the effect of the laser energy density fluctuation on the spectral intensity, a laser displacement sensor and a stepper motor in z axis were used to monitor and control the distance between lens and the surface of the sample. Each spectrum accumulated 30 pulses, and 10 spectra were taken for each sample and averaged. To avoid over ablation, the sample was mounted on a stepper motor stage, and the laser beam scanned the sample in a straight line.

Fig. 2. LIBS experiment setup (a) and pellet soil samples (b).

#254197 © 2016 OSA

Received 18 Nov 2015; revised 15 Jan 2016; accepted 27 Jan 2016; published 1 Feb 2016 8 Feb 2016 | Vol. 24, No. 3 | DOI:10.1364/OE.24.002607 | OPTICS EXPRESS 2611

2.2 Sample preparation Soil samples were certified reference materials approved by the State Administration of Quality Supervision, Inspection, and Quarantine of China. The number and concentration of Pb element are shown in Table 1. The main matrixes of these soil samples are different from each other, No.1 is basalt from southern China, No.2 is carbonate from northwest china, No.3 and No. 4 are skarn from Hubei province and Hunan province, respectively, No.5 is limestone from northeast China and No.6 is granite from southwest China. As the concentration of Pb in sample No.1 is too low (14 ppm) to be detected (the spectra are so weak that the predicted result is unreliable. To avoid confusion, we have not added the data of sample No.1), sample No.2, No.3, No.4, No.5, and No.6 were used to test the accuracy of LIBS in this work. The soil samples analyzed were in pressed pellet form, which are shown in Fig. 2(b). It should be mentioned that, the original samples did not contain Cd element, the solution of CdCl2 was added in each sample to produce concentration gradient. Table 1. Sample number and concentration. Sample No. Pb concentration (ppm)

1

2

3

4

5

6

14

110

283

552

675

1141

Cd concentration (ppm)

0

50

100

200

300

400

3. Results and discussion 3.1 Spectra measurement The time-integrated optical emission spectrometry (OES) spectra of laser-induced soil plasmas from soil samples are obtained. The emission spectra of the Pb element are obtained in the spectral range of 403-408 nm, the line peak of Pb I: 405.78 nm is taken as an analytical signal. The diameter spot size of the focused laser beam is about 0.3 mm with a laser fluence of 56.6 J/cm2. The line of Pb (Pb I: 405.78 nm) is chosen to predict the Pb concentration in soil. 3.2 Conventional calibration method

Fig. 3. Comparison of original spectrum of sample No. 5(red) and its corrected spectra, which is corrected by two-point background subtraction method (blue) and wavelet background removal (black), respectively.

To study the effect of background removal method on the spectra, the original data were processed with wavelet background removal method and two-point background subtraction method, respectively. Spectra of sample No. 5 were taken as an example in Fig. 3. The red line is the original spectrum, the blue line is the spectrum processed with two-point background subtraction method, the two points of 405.65 and 405.85 nm were selected as the points to correct the background (the two points were the minimal value points at the edge of

#254197 © 2016 OSA

Received 18 Nov 2015; revised 15 Jan 2016; accepted 27 Jan 2016; published 1 Feb 2016 8 Feb 2016 | Vol. 24, No. 3 | DOI:10.1364/OE.24.002607 | OPTICS EXPRESS 2612

the peak, and the averaged spectral intensities of the two points were regarded as the background), and the black line is the spectrum processed with the wavelet function, the wavelet function is Daubechies (db) 7 and the wavelet transformation was performed at a decomposition level of 7, which have been optimized (60 spectra of these soil samples were used to optimize the parameters of the algorithm. The spectral region from 408.83 nm to 409.12 nm is very smooth and away from the peak, and the value of this spectral region processed with the parameter db 7 and the level 7 was closest to zero, so the parameter db 7 and the level 7 were chosen as optimized parameters). Figure 4 shows the calibration curves for the Pb element in sample No.2 with conventional calibration methods, the solid dots represent sample No.2 (predicted point) and the hollow dots represent other five samples (calibration points). According to the standard concentrations in soil samples No.1, No.3, No.4, No.5 and No.6, the relative intensity ratios of line Pb I 405.78nm /Fe I 406.3nm are plotted as a function of the Pb concentration in a linear scale in Fig. 4. It is calculated that the determination coefficients (R2) are 0.8461, 0.8786 and 0.8836 for the three kinds of data, respectively. The results show that the two kinds of background removal methods could not make the R2 more than 0.9 when using conventional calibration method.

Fig. 4. The calibration curves and the determination coefficients of the original data (red) and its corrected data, which is corrected by two-point background subtraction method (blue) and wavelet background removal (black), respectively, for sample No. 2.

Furthermore, the results of further study can be found in Table 2. Table 2 shows that the results from predicting the accuracy of the quantitative analysis of the Pb element in the six soil samples by cross validation (what have done in Fig. 4). The poor R2 and prediction accuracy are obtained due to the serious matrix effect in these soil samples. Table 2. The predicted results of the conventional calibration methods.

Sample No.

Pb concentration (ppm)

Original data

Two point data

WT data

1

14

Predict concentration (ppm) 27.2

94%

Predict concentration (ppm) −15.1

207%

Predict concentration (ppm) 25.3

2

110

106.7

3%

84.5

23.2%

97.4

11.5%

3

283

315.1

11.3%

323.6

14.3%

324.9

14.8%

4

552

282.6

48.8%

304.9

44.8%

319.4

42.1%

RE

RE

RE 80.7%

5

675

1277.9

89.3%

1301.5

92.8%

1231.1

82.4%

6

1141

1215.8

6.6%

1298.0

13.8

1201.2

5.3%

#254197 © 2016 OSA

Received 18 Nov 2015; revised 15 Jan 2016; accepted 27 Jan 2016; published 1 Feb 2016 8 Feb 2016 | Vol. 24, No. 3 | DOI:10.1364/OE.24.002607 | OPTICS EXPRESS 2613

Furthermore, Table 2 shows that, the predicted results of Pb in sample No.1 are very poor (even negative data is obtained). The weak signal of Pb I 405.78 nm (14 ppm) was supposed to be the reason of unreliable result in sample No.1, as shown in Fig. 5. To avoid confusion, the data of sample No.1 will not be used in this work. The RMSEP of the three kinds of data are 332.3, 346.0 and 303.0 ppm, respectively. All of them are too large to calculate the concentration accurately, and it is obvious that the background removal method do not affect the prediction error for the conventional calibration method.

Fig. 5. Original spectrum of sample No. 1, which ranges from 404.5 to 406.5 nm.

3.3 Quantitative analysis with the standard addition method Because the matrix of soil is complex, the spectral intensity is affected by the different matrix of the soil samples [15], the accuracy of the conventional calibration methods is poor. For the purpose of improving the accuracy of quantitative analysis, a SAM based on background removal was adopted to analyze the heavy metals in soil. Five groups of samples were prepared to study SAM, they were No. 2, No. 3, No. 4, No. 5 and No. 6, respectively (as previously declared, to avoid confusion, we have not used the data of sample No.1). Each group contains 5 samples, Pb was added through a solution of lead (II) nitrate (Pb(NO3)2), and the added concentrations of the Pb element were 100, 200, 400, 600, and 800 ppm, respectively. The slurries were evenly dispersed using ultrasound, then dried though in vacuum drier. Finally, the soil was grinded in a mortar for 5 min. The whole preparation time is 3-4 hours (it is more convenient and simple when compared with ICP). The experiment parameters were set to be the same as what was used in the conventional calibration method. Figure 6 shows the calibration plot of the group of sample No. 2 (as the figures for sample 2 and the other samples (2-6) are very similar, and the principles of them are all the same, the group of samples No. 2 was set as an example to represent other groups). The R2 of the original data is 0.9956. Meanwhile the R2s of the two-point data and WT data are 0.9954 and 0.9953, respectively. The results of all the 5 groups are shown in Table 3, The R2 values of all samples are larger than 0.9776, Meanwhile, the slopes of the 5 kinds of soils are varies from each other, which indicate that the matrix of the soils are not the same. Furthermore, what should be declared is that, dark current and ambient light have been removed by recording and subtracting a blank spectrum. The background of the original spectrum only contains bremsstrahlung and interference from other element lines.

#254197 © 2016 OSA

Received 18 Nov 2015; revised 15 Jan 2016; accepted 27 Jan 2016; published 1 Feb 2016 8 Feb 2016 | Vol. 24, No. 3 | DOI:10.1364/OE.24.002607 | OPTICS EXPRESS 2614

Fig. 6. Calibration plots of original data (red), the two-point background subtraction data (blue) and wavelet background removal data (black) for sample No. 2, respectively. Table 3. The slope and R2 of the calibration curves.

Sample No.

Original data

Two point data

WT data

Slope ( × 10−4)

R2

Slope ( × 10−4)

R2

Slope ( × 10−4)

R2

1













2

2.89

0.9959

2.87

0.9954

2.92

0.9953

3

2.79

0.9955

2.73

0.9955

2.73

0.9955

4

1.63

0.9963

1.56

0.9945

1.58

0.9956

5

2.99

0.9781

2.99

0.9776

3.03

0.9787

6

1.16

0.9929

1.16

0.9922

1.18

0.9934

Table 4 shows the predicted results using standard addition methods for the Pb element. The RMSEP is 70.3 ppm for the original data. The predicted concentrations of the original data for samples No. 4 (Pb: 552 ppm), No. 5 (Pb: 675 ppm) and No. 6 (Pb: 1141 ppm) are 601.2, 721.4 and 1226.7 ppm, respectively, the relative error are all less than 9%. However, for sample No. 2 (Pb: 110 ppm), the predicted concentration of the original data is 198.6 ppm, the relative error is more than 80%, which is much worse than other samples. The results indicate that the predicted accuracy of SAM had a great relevance with the concentration of the samples. The lower of the concentration is, the larger of the error will be. The phenomenon can be explained by Eq. (4), b  C0 = −  Cx, I = 0 +  ≈ −Cx, I = 0 . (4) k  As previously shown, C0 is the original concentration of the element to be examined in the sample, k is the slope of the curve, and b is the background of the spectrum When C0>> b/k, the parameter b will not affect C0. If C0 is not large enough, the background (parameter b) will seriously influence the results, the smaller the parameter: b/k is, the more accurate the result will be. To solve this problem, two background removal methods were investigated. The RE of two-point data for the five samples are 11.9%, 7.3%, 4.5%, 8.9%, 5.1%, respectively, and the RMSEP is 43.7 ppm. The RE of WT data for the five samples are 0.2%, 7.1%, 6.2%, 3.6%, 2.6%, respectively, and the RMSEP is 25.7 ppm. Compared with the RMSEP of the original

#254197 © 2016 OSA

Received 18 Nov 2015; revised 15 Jan 2016; accepted 27 Jan 2016; published 1 Feb 2016 8 Feb 2016 | Vol. 24, No. 3 | DOI:10.1364/OE.24.002607 | OPTICS EXPRESS 2615

data, there are huge improvements for both two-point data and WT data. For two-point data, it only involves two points on both side of the wavelength, if the spectral backgrounds are curved or interfered by other elements, results given by this technique will be unsatisfactory, and it would make the intensity of removed background higher or lower than the real value [28]. However, these problems do not exist for WT data. Furthermore, a MATLAB program has been written to perform wavelet transform, automatically. According to these results, the WT method is better than the two-point method. All in all, analyzing the soil by SAM can inhibit the serious matrix effect in soil detection, but the drawback of SAM is that, the results of SAM are more sensitive to the background of the spectrum than the results of the conventional calibration method for the low concentration elements. Background removal is an effective process to overcome the drawback of SAM. Therefore, the combination of SAM and Background removal can produce huge improvement in the quantitative analyses of soil. Furthermore, to study the application range of this method, the LODs of Pb for sample No.2, No.3, No.4, No.5, and No.6 are calculated, the LOD can be calculated by Eq. (5), LOD =

3σ , k

(5)

where, k is the slope of the calibration line, σ is the noise of the spectra (relative standard deviation of the continuum background). The calculated LODs are 14.6, 21.5, 34.8, 18 and 37.6 ppm, respectively. The LODs of all the samples are better than the threshold of Pb (50 ppm) in soil in China, which shows a bright prospect of this method. Table 4. The predicted results of the standard addition methods.

Sample No.

Pb concentration (ppm)

1 2 3 4 5 6

14 110 283 552 675 1141

Original data Predict concentration (ppm) — 198.6 336.7 601.2 721.4 1226.7

RE — 80.5% 18.9% 8.9% 6.9% 7.5%

Two-point data Predict concentration RE (ppm) — — 96.9 11.9% 262.2 7.3% 526.9 4.5% 615.1 8.9% 1083.9 5.1%

WT data Predict concentration (ppm) — 110.2 303.2 586.1 650.5 1111.2

RE — 0.2% 7.1% 6.2% 3.6% 2.6%

3.4 Detection of the Cd element Apart from the Pb element in soil samples, the emission spectra of the Cd element are obtained. The line of Cd (Cd II: 214.4 nm) is chosen to predict the concentration of the Cd element in soil, sample No. 3 (Cd concentration: 100 ppm) is quantitatively analyzed. In Fig. 7, the data are dealt with conventional calibration methods (the calibration curve is fitted by the other five samples), the R2s are less than 0.93, and the predicted Cd concentrations of the soil samples are 118.5, 116.6 and 116.3 ppm using these three kinds of data, respectively, indicating the R2 and the accuracy are both poor due to the strong matrix effect from the soil samples.

#254197 © 2016 OSA

Received 18 Nov 2015; revised 15 Jan 2016; accepted 27 Jan 2016; published 1 Feb 2016 8 Feb 2016 | Vol. 24, No. 3 | DOI:10.1364/OE.24.002607 | OPTICS EXPRESS 2616

Fig. 7. Calibration plots of original data (red), the two-point background subtraction data (blue) and wavelet background removal data (black) for Cd element, respectively.

Based on the SAM, Fig. 8 shows that, the predicted concentration of the Cd element in sample No. 3 is 206.3 ppm, and the RE is 106.3%. After background removal by two-point removal method or wavelet transform, the predicted concentrations of the Cd element is 106.4 and 106.1 ppm, respectively. The best RE is 6.1%, Compared with the conventional calibration method with an RE of 16.3%, the accuracy of the Cd element in the soil sample is obviously improved.

Fig. 8. Comparison of the calibration plots using original data (red), two-point data (blue) and WT data (black) for Cd by using SAM.

4. Summary and conclusions In summary, accuracy improvement by background removal based on SAM in laser-induced breakdown spectroscopy was studied and two background removal methods were compared. The Pb and Cd elements were detected in soil samples. For Pb element, the RMSEP of quantitative analysis was more than 300 ppm and R2s were less than 90% using conventional methods due to the soil matrix effect. To overcome the drawbacks of the conventional methods in LIBS, the combination of background removal and SAM was investigated, compared with conventional calibration methods, the matrix effect of the Pb element in soil samples has been reduced effectively, the RMSEP of Pb was lower than 26 ppm and the R2s were more than 97.76%, when using this method, and the detection of Cd can prove the results as well. Furthermore, the mechanism of background removal by wavelet transform algorithm based on SAM was discussed. The results of this study provide a new pathway for

#254197 © 2016 OSA

Received 18 Nov 2015; revised 15 Jan 2016; accepted 27 Jan 2016; published 1 Feb 2016 8 Feb 2016 | Vol. 24, No. 3 | DOI:10.1364/OE.24.002607 | OPTICS EXPRESS 2617

improving the accuracy of quantitative analysis in the detection of powder samples using the LIBS technique. Acknowledgments This research has been financially supported by the National Special Fund for the Development of Major Research Equipments and Instruments (No. 2011YQ160017), by the National Natural Science Foundation of China (No.61575073 and 51429501), by Natural Science Foundation of Hubei province (No. 2015CFB298).

#254197 © 2016 OSA

Received 18 Nov 2015; revised 15 Jan 2016; accepted 27 Jan 2016; published 1 Feb 2016 8 Feb 2016 | Vol. 24, No. 3 | DOI:10.1364/OE.24.002607 | OPTICS EXPRESS 2618

Background removal in soil analysis using laser- induced breakdown spectroscopy combined with standard addition method.

The matrix effect of powder samples, especially for soil samples, is significant in laser-induced breakdown spectroscopy (LIBS), which affects the pre...
564B Sizes 0 Downloads 14 Views