Plasma Science and Technology, Vol.17, No.11, Nov. 2015
Parameters Optimization of Laser-Induced Breakdown Spectroscopy Experimental Setup for the Case with Beam Expander∗ WANG Xin (王鑫), ZHANG Lei (张雷), FAN Juanjuan (樊娟娟), LI Yufang (李郁芳), GONG Yao (弓瑶), DONG Lei (董磊), MA Weiguang (马维光), YIN Wangbao (尹王保), JIA Suotang (贾锁堂) State Key Laboratory of Quantum Optics and Quantum Optics Devices, Institute of Laser Spectroscopy, Shanxi University, Taiyuan 030006, China
Abstract
Improvement of measurement precision and repeatability is one of the issues currently faced by the laser-induced breakdown spectroscopy (LIBS) technique, which is expected to be capable of precise and accurate quantitative analysis. It was found that there was great potential to improve the signal quality and repeatability by reducing the laser beam divergence angle using a suitable beam expander (BE). In the present work, the influences of several experimental parameters for the case with BE are studied in order to optimize the analytical performances: the signal to noise ratio (SNR) and the relative standard deviation (RSD). We demonstrate that by selecting the optimal experimental parameters, the BE-included LIBS setup can give higher SNR and lower RSD values of the line intensity normalized by the whole spectrum area. For validation purposes, support vector machine (SVM) regression combined with principal component analysis (PCA) was used to establish a calibration model to realize the quantitative analysis of the ash content. Good agreement has been found between the laboratory measurement results from the LIBS method and those from the traditional method. The measurement accuracy presented here for ash content analysis is estimated to be 0.31%, while the average relative error is 2.36%.
Keywords: laser-induced breakdown spectroscopy (LIBS), relative standard deviation (RSD), signal to noise ratio (SNR), beam expander (BE), support vector machine (SVM)
PACS: 42.62.Fi, 52.50.Jm, 89.30.Ag, 42.60.−v DOI: 10.1088/1009-0630/17/11/04 (Some figures may appear in colour only in the online journal)
1
Introduction
which is the angular spreading of light waves as it propagates through space. Even a perfect laser beam will experience some beam divergence due to diffraction effects. Diffraction is the effective bending of light rays caused by truncation from an opaque object, such as a knife edge. The spreading arises from secondary wave fronts emitted from the edge of truncations. These secondary waves interfere with the primary wave and also with themselves, sometimes forming quite complicated diffraction patterns [15] . Given that we know that BE is a telescope in which a laser beam could be expanded to a larger diameter, this is useful to the reduction of laser beam divergence angle, resulting in a better focusing effect. Based on the above discussion, a BE was employed in our present work to expand the laser beam to a suitable diameter. Taking into account that the BE is likely to affect the experimental parameters that were previously optimized, we concentrated on the study of the effects of such factors as tablet machine pressure, laser energy, laser repetition rate, and laser focus posi-
Laser-induced breakdown spectroscopy (LIBS) is an emerging analytical spectroscopy technique which can determine the elemental composition from the line emission of laser generated plasma on the basis of elemental and molecular emission intensities. This technology demonstrates numerous appealing features that distinguish it from conventional analytical spectrochemical techniques. However, the relatively low quality spectrum and poor pulse-to-pulse repeatability of measurement with LIBS are always regarded as obstacles for a wide range of applications of this technology [1−4] . Many methods have been developed to improve the signal quality and stability of LIBS, such as, spatial or magnetic confinement [5−9] , microwave-assisted LIBS [10] , laser ablation combined with fast pulse discharge [11] , various data processing methods [12−14] , etc. Laser beam quality is often characterized by a parameter known as beam divergence,
∗ supported by the 973 Program of China (No. 2012CB921603), National Natural Science Foundation of China (Nos. 61475093, 61127017, 61178009, 61108030, 61378047, 61275213, 61475093, and 61205216), the National Key Technology R&D Program of China (No. 2013BAC14B01), the Shanxi Natural Science Foundation (Nos. 2013021004-1 and 2012021022-1), the Shanxi Scholarship Council of China (Nos. 2013-011 and 2013-01), and the Program for the Outstanding Innovative Teams of Higher Learning Institutions of Shanxi, China
914
WANG Xin et al.: Parameters Optimization of LIBS Experimental Setup for the Case with Beam Expander tion. As a result, the BE included experimental LIBS setup for coal quality analysis was modified accordingly, and the improvement in signal quality and stability by using BE was demonstrated.
stepping motor (PK545-NA, Oriental Motor Ltd.) and the rotation speed is set appropriately according to the measurement time.
3
Experimental study
In this study, several experimental parameters were systematically changed and their effects on the quality and repeatability of the characteristic line of C and Si were studied, and as a result the optimal experimental parameters in LIBS measurement could be determined. Here, the signal quality was judged by the signal-tonoise ratio (SNR) which is expressed by the formula SN R = I/(R × d),
Fig.1 A schematic diagram of the experimental setup. XLV: Xenon Lamp power supply voltage, HVP: Q-switched high voltage pulse, HWP: Half-wave plate, PBS: Polarizing beam splitter, PM: Power meter, SM: Stepping motor
2
(1)
where I is the characteristic line integral intensity; R is the root mean square value of noise, which is calculated by linear fitting to the background continuous spectrum within a narrow band around ends of the characteristic lines, and then each end seeking a root mean square value and taking the average; and, d is the characteristic line width. Obviously, the greater the SNR, the better the signal quality. The relative standard deviation (RSD) of spectral line intensity is used to evaluate the repeatability, smaller RSD value indicates more precise LIBS measurement, which can be defined as i hX (xi − M )2 /(n − 1) /M, (2) RSD =
Experimental setup
The experimental LIBS setup used for coal quality analysis is shown in Fig. 1. A Q-switched Nd:YAG pulsed laser (DCR-3, Spectra-physics Ltd.) with 8 ns pulse width operating at 1064 nm is employed for coal ablation and plasma generation. The laser beam passes through a half-wave plate (HWP) and is divided into two beams by a polarizing beam splitter (PBS). The reflected light travels to a power meter. The transmitted light is reflected by a mirror covered with silver and passes through a BE which can reduce the divergence angle, resulting in a better focusing effect. The light is focused at the sample surface using a 15 mm diameter and 200 mm focal length plano-convex quartz lens. A tiny portion of the sample was ablated to form the plasma. As the plasma expands and gradually cools, the atomic and ionic emission lines appear and the emitted radiation is guided to a dual channel spectrometer (AvaSpec-2048-USB2, Avantes Ltd.) which covers the spectral ranges of 190-366 nm and 586-1116 nm, providing a nominal resolution of 0.7 nm and a fixed 1 ms gate-width by means of an optical fiber. The coal samples used in this study were provided by China Coal Yangjian Separations Ltd. and were gathered from different mines in Shuozhou. Following the standard sampling procedure, all of the coal samples were prepared to a grain size of 100 µm. Due to the extremely small ablated mass being vaporized and exited by the laser pulse, the LIBS precision and accuracy are heavily dependent on the homogeneity of the sample. Therefore, the powdery samples were pretreated by pressing them into compact and smooth pellets (40 mm diameter, 7 mm thickness) using a tablet machine. Each pellet weights 5 g, which was measured with an analytical balance whose precision is 0.05 mg at low load. Additionally, in order to make each laser pulse have a fresh action point and achieve uniform sample of points, the samples were fixed on a rotation stage driven by a
where n is the number of a set of measurements; xi is the absolute line intensity of element i in each measurement; and, M is the arithmetic mean line intensity of the repeated measurements.
3.1
Optimization of the tablet machine pressure
As mentioned above, the LIBS precision and accuracy are heavily dependent on the homogeneity of the sample. Therefore, the pretreatment of powdery samples, i.e., pressing them into compact and smooth pellets using a tablet machine, seems to be extremely important. Experimental results demonstrate that different pressures may result in different changes in plasma emission intensities, and thereby affects the measurement precision and accuracy. In order to yield more accurate measurement results, it is necessary to choose an optimum tablet machine pressure. In this work, experiments were performed on the sample with 100 successive laser shots by using pressure of 5 MPa, 10 MPa, 15 MPa, 20 MPa, 25 MPa, and 30 MPa, respectively. Here, the pulse energy is 80 mJ, the laser repetition rate is 10 Hz, the focal plane is 4 mm above sample surface, the integration time is 10 ms, and the delay time is 2 µs. A comparison of the RSD and SNR values is shown in Fig. 2(a) and (b), respectively. With the increase of the tablet machine pressure, the SNR 915
Plasma Science and Technology, Vol.17, No.11, Nov. 2015 values of C(I) 247.9 nm and Si(I) 288.2 nm grow bigger while the RSD values become smaller. When the tablet machine pressure is 25 MPa, the SNR reaches the maximum value while the RSD reaches the minimum value. This is due to the fact that the pulse energy couples to more matter when the tablet machine pressure getting bigger. When the pressure is over 25 MPa, the ablation quality does not increase any further. Thus, it is suggested that the optimum pressure of the tablet machine should be set to 25 MPa to further enhance the stability and reliability of the quantitative analysis results.
The spectral line intensity is saturated when the pulse energy is more than 100 mJ due to the plasma shielding effect. Meanwhile, it can be seen that the bigger the laser repetition rate is, the greater the SNR and lower RSD values of C(I) 247.9 nm and Si(I) 288.2 nm will be. The reason for this might be that with the increase of the repetition rate, the reflected shockwave shakes off more ash off the surface of the sample. The maximum SNR and the minimum RSD values can be obtained simultaneously when the ablation laser energy and the repetition rate are selected to be 100 mJ and 10 Hz, respectively.
Fig.3 RSD and SNR of C(I) 247.9 nm line versus laser energy and repetition rate
3.3
LIBS operates by focusing the laser onto a small area at the surface of the specimen to form a plasma, which atomizes and excites the specimen. This focusing effect affects the morphology and property of the plasma as well as the LIBS signal. In our study, the laser beam is expanded to a lager diameter prior to focusing so as to obtain a smaller laser spot. In this case, the focusing area on the coal sample surface can achieve a relatively high laser fluence, and thus the plasma shielding can become strongly contributed by ionized ambient gas localized in the propagation front of the plume [18] . Therefore, it is essential to optimize the distance between the focusing lens and the sample surface. A comparison between the RSD and SNR values of C(I) 247.9 nm and Si(I) 288.2 nm with various focus positions is shown in Fig. 4, where the focus position of zero indicates the sample surface. It can be seen that with the increase of lens-to-sample distance, the SNR values of C(I) 247.9 nm and Si(I) 288.2 nm increase while the RSD values decrease. In the vicinity of 7.5 mm, the SNR reaches a maximum value but the RSD reaches a minimum value. The reason is likely to be that with the increase of ablation depth, the pulse energy coupled with the sample decreases. In this work, the optimum focus position situates at 7.5 mm above the sample surface.
Fig.2 SNR and the RSD values of the normalized spectral line intensity of (a) C(I) 247.9 nm and (b) Si(I) 288.2 nm
3.2
Optimization of the laser focus position
Optimization of the laser pulse energy and repetition rate
It is reported that both laser pulse energy and repetition rate are important LIBS parameters that can greatly improve the SNR values of characteristic lines [16,17] . In order to adjust the ablation laser energy without changing other laser parameters, such as the flash lamp supply voltage, which could compromise the pulse-to-pulse stability of the laser source, a HWP coupled with a PBS was placed in the laser beam path. In the current work, the effect of the ablation laser energy together with the laser repetition rate with respect to the RSD and SNR values of C(I) 247.9 nm and Si(I) 288.2 nm line has been investigated at fixed laser parameters. The focal plane is 4 mm above sample surface, the integration time is 10 ms, and the delay time is 2 µs. From Fig. 3, we can see that the trends of SNR and RSD are completely opposite. With the increase of pulse energy, the SNR values of C and Si increase accordingly. This happens because the irradiance increases with the increase of the pulse energy. 916
WANG Xin et al.: Parameters Optimization of LIBS Experimental Setup for the Case with Beam Expander experimental BE-included LIBS setup gives a significant improvement in measurement accuracy.
Fig.4 Comparison of the RSD and SNR values of C(I) 247.9 nm and Si(I) 288.2 nm under different laser focus positions
4
Fig.5 Comparison of the SNR values of C(I) 247.9 nm and Si(I) 288.2 nm using seven coal samples for the cases with and without BE
Result and discussion
After the parameter optimization mentioned above, the quality and stability of characteristic spectral lines for the cases with and without BE were compared using seven coal samples. The RSD values of the normalized line intensity with respect to the whole spectrum area and the SNR values of C(I) 247.9 nm and Si(I) 288.2 nm for the cases with and without BE are shown in Figs. 5 and 6, respectively. Our results demonstrate that by using the BE, both the quality and the stability of the plasma spectra can be significantly improved. To validate the improvement in quantitative analysis results, the optimized experimental LIBS setup was employed to analyze the ash contents of coal samples that were certified by the traditional thermo-gravimetric analyzer (TGA). A total of thirty samples were selected as calibration samples and the remaining five samples were selected for predication. Support vector machine (SVM) [19−20] regression combined with principal component analysis (PCA) [21] was used to establish a calibration model to achieve the real ash contents. For comparison, the ash contents measured by LIBS combined with the above mentioned multivariate regression model as a function of the corresponding certified values for the calibration samples are shown in Fig. 7(a). As can be seen, the results are in fairly good agreement between LIBS and traditional method for ash contents, the linear regression coefficient R is equal to 0.9976. To validate the repeatability of the LIBS measurement, each of the five prediction samples was measured four times. A comparison of the certified and the measured ash contents is displayed in Fig. 7(b), where the error bars represent the SD values. The relatively small experimental error bars indicate that we can provide a reasonable repeatability of measurements by using the obtained calibration model with BE. From the quantitative results, the measurement accuracy is estimated to be 0.31% while the average absolute error is 2.36%. Compared to our previous work on on-line pulverized coal quality analysis in power plants [22] , this optimized
Fig.6 Comparison of the RSD values of C(I) 247.9 nm and Si(I) 288.2 nm using seven coal samples for the cases with and without BE
5
Conclusion
The purpose of this work is to investigate the effect of several important parameters related to the signal quality and repeatability, so as to optimize the analytical performances of a LIBS experimental setup for the case with BE. By comparing both the RSD and SNR values of the emission lines, the optimal experimental parameters of the LIBS setup is determined as follows: the tablet machine pressure is 25 MPa, the laser energy is 100 mJ, the laser repetition rate is 10 Hz, and the laser focus position is 7.5 mm above the sample surface. The optimal experimental parameters lead to an increase in signal quality and a decrease in RSD values of the normalized line intensities. To validate the necessity of optimization, SVM regression combined with PCA are used to establish a calibration model to quantitatively analyze the ash contents in coal. Experimental results show that the measured ash contents using 917
Plasma Science and Technology, Vol.17, No.11, Nov. 2015 2
3 4 5 6 7 8 9 10 11 12 13 14 Fig.7 (a) Comparison between LIBS (coupled with a multivariate model) and standard analysis procedures for the ash contents, (b) The calculated ash contents of five validation coal samples compared to the certified standards
15
LIBS are in fairly good agreement with those from the traditional method, the measurement accuracy has been slightly improved compared to our previous work. The measurement accuracy presented here for ash content analysis is estimated to be 0.31% while the average relative error is 2.36%.
18
16 17
19 20 21 22
References 1
Miziolek A W, Palleschi V, Schechter I. 2006, Laser Induced Breakdown Spectroscopy (LIBS): Fundamentals and Applications. Cambridge University Press, New York, p.163 Lee W B, Wu J Y, Lee Y I, et al. 2004, Applied Spectroscopy Reviews, 39: 27 Wang Z, Yuan T B, Hou Z Y, et al. 2014, Frontiers of Physics, 9: 419 Aguilera J A, Aragon C, Bengoechea J, et al. 2003, Applied Optics, 42: 5938 Aguilera J A, Aragon C. 2004, Spectrochimica Acta Part B, 59: 1861 Wang Z, Li L Z, West L, et al. 2012, Spectrochimica Acta Part B, 68: 58 Hou Z Y, Wang Z, Liu J M, et al. 2013, Optics Express, 21: 15974 Shen X K, Lu Y F, Gebre T, et al. 2006, Journal of Applied Physics, 100: 053303 Liu Y, Baudelet M, Richardson M. 2010, Journal of Analytical Atomic Spectrometry, 25: 1316 Zhou W D, Li K X, Shen Q M, et al. 2010, Optics Express, 18: 2573 Li L Z, Wang Z, Yuan T B, et al. 2011, Journal of Analytical Atomic Spectrometry, 26: 2274 Body D, Chadwick B L. 2001, Spectrochimica Acta Part B, 56: 725 Dong M R, Lu J D, Yao S C, et al. 2011, Journal of Analytical Atomic Spectrometry, 26: 2183 Tanaka K, Yoshida K, Taguchi M. 1988, Applied Optics, 27: 1310 Sirven J B, Mauchien P, Salle B. 2008, Spectrochimica Acta Part B, 63: 1077 Robert P, Fabre C, Dubessy J, et al. 2008, Spectrochimica Acta Part B, 63: 1109 Bai X, Ma Q, Perrier M, et al. 2013, Spectrochimica Acta Part B, 87: 27 Goode S R, Morgan S L, Hoskins R, et al. 2000, Journal of Analytical Atomic Spectrometry, 15: 1133 Peng X J. 2010, Neural Networks, 23: 365 Lu Y M, Roy C V. 2008, Knowledge and Information Systems, 14: 233 Yin W B, Zhang L, Dong L, et al. 2009, Applied Spectroscopy, 63: 865
(Manuscript received 26 April 2015) (Manuscript accepted 24 June 2015) E-mail address of corresponding author ZHANG Lei:
[email protected]
Cremers A, Radziemski L J. 2006, Handbook of LaserInduced Breakdown Spectroscopy. John Wiley & Sons Ltd., England, p.437
918
