Correlating laser-induced breakdown spectroscopy with neutron ...

Spectrochimica Acta Part B 138 (2017) 46–53

Contents lists available at ScienceDirect

Spectrochimica Acta Part B journal homepage: www.elsevier.com/locate/sab

Correlating laser-induced breakdown spectroscopy with neutron activation analysis to determine the elemental concentration in the ionome of the Populus trichocarpa leaf☆,☆☆ Madhavi Z. Martin a,⁎, David C. Glasgow b, Timothy J. Tschaplinski a, Gerald A. Tuskan a, Lee E. Gunter a, Nancy L. Engle a, Ann M. Wymore a, David J. Weston a,c a b c

Biological Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA Chemical Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA Climate Change Science Institute, Oak Ridge National Laboratory, Oak Ridge, TN, 37831, USA

a r t i c l e

i n f o

Article history: Received 21 December 2016 Received in revised form 4 October 2017 Accepted 15 October 2017 Available online 17 October 2017

a b s t r a c t The black cottonwood poplar (Populus trichocarpa) leaf ionome (inorganic trace elements and mineral nutrients) is an important aspect for determining the physiological and developmental processes contributing to biomass production. A number of techniques are used to measure the ionome, yet characterizing the leaf spatial heterogeneity remains a challenge, especially in solid samples. Laser-induced breakdown spectroscopy (LIBS) has been used to determine the elemental composition of leaves and is able to raster across solid matrixes at 10 μm resolution. Here, we evaluate the use of LIBS for solid sample leaf elemental characterization in relation to neutron activation. In fact, neutron activation analysis is a laboratory-based technique which is used by the National Institute of Standards and Technology (NIST) to certify trace elements in candidate reference materials including plant leaf matrices. Introduction to the techniques used in this research has been presented in this manuscript. Neutron activation analysis (NAA) data has been correlated to the LIBS spectra to achieve quantification of the elements or ions present within poplar leaves. The regression coefficients of calibration and validation using multivariate analysis (MVA) methodology for six out of seven elements have been determined and vary between 0.810 and 0.998. LIBS and NAA data has been presented for the elements such as, calcium, magnesium, manganese, aluminum, copper, and potassium. Chlorine was also detected but it did not show good correlation between the LIBS and NAA techniques. This research shows that LIBS can be used as a fast, high-spatial resolution technique to quantify elements as part of large-scale field phenotyping projects. © 2017 Elsevier B.V. All rights reserved.

1. Introduction The elemental composition of the plant, or ionome, is influenced by above- and below-ground conditions along with physiological and developmental processes [1]. The ionome is a reflection of plant nutrient

☆ Selected paper from the 9th International Conference on Laser-Induced Breakdown Spectroscopy (LIBS), Chamonix-Mont-Blanc, France, September 12 – September 16 2016. ☆☆ This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC0500OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan). ⁎ Corresponding author at: Biosciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831-6422, USA. E-mail address: [email protected] (M.Z. Martin).

https://doi.org/10.1016/j.sab.2017.10.008 0584-8547/© 2017 Elsevier B.V. All rights reserved.

status which is a critical component for food security programs through climate adaptation of crop systems [2–3], bioenergy feedstock sustainability (e.g., [4]) and in determining the stoichiometry of elements used to predict plant production and community composition change in unmanaged ecosystems (e.g., [5–6]). As a result, there have been numerous studies aimed at understanding environmental and genetic controls on mineral nutrients and trace element accumulation in specific plant organs and tissues ([1,7–11] for reviews). The aforementioned plant ionomic studies have been aided by the relatively highthroughput Inductively Coupled Plasma (ICP) technology to ionize analyte atoms and detection via Optical Emission Spectrometry (ICP-OES) and Mass Spectrometry (ICP-MS). Since ICP-techniques require sample dissolution prior to analysis, it is much less “high throughput” than initially thought and homogenizes across tissues. Despite the avid use ICP related approaches in plant biology for determining elemental composition, there are additional approaches including LIBS. LIBS induces the vaporization of a small volume of sample material with sufficient pulsed laser energy for optical excitation of the elemental species in the

M.Z. Martin et al. / Spectrochimica Acta Part B 138 (2017) 46–53

resultant sample plume. The vaporized species then undergo deexcitation and optical emission on a microsecond time scale, and time-dependent ultraviolet-visible spectroscopy fingerprints the elements associated with the spectral peaks. LIBS is typically a surface analytical technique, with each laser pulse vaporizing submicrogram or nanogram sample masses. However, the rapidity of sampling (typically 10 Hz laser repetition rate) facilitates scanning a sample surface. Repeated same-location laser pulses offer depth profiling by, ablating a hole into a solid sample. Finally, focusing the laser spark below the surface of a liquid sample permits more versatile analyses and provides sufficient statistics for bulk sampling. These methods, taken together illustrate the wide array of analytical power available to LIBS. The greatest advantage of LIBS is its capability for remote chemical analysis of samples since the laser beam can be delivered by optical fibers to the sample under test and the plasma emission from the sample can be collected and delivered to the detector also via fibers which makes it possible to do remote sampling. The instrumentation and operation of a LIBS system is simpler than some of the more sensitive techniques (e.g., ICP-MS), and analysis times on the order of minutes make it more amenable to real-time analysis of chemical processes. Although calibration standards are required for quantitative analysis, the generation of a single calibration curve will suffice for analysis of samples in a similar matrix. Recently we have expended tremendous effort in quantitative analysis of the elemental composition of chemically treated wood and natural wood products [12]. We have used this technique to determine the presence of elements such as, C, N, Ca, Al, Fe, Ti, Si, Mg, Mn, and Na from natural wood, and other biological samples [13,14,15] which contains only a few ppm of the target elements from LIBS spectral data sets. Chemometric methods such as Principal Component Analysis (PCA) are commonly used for this type of analysis. PCA removes the redundancy (inter-correlation) in a data set, transforming it into a few loadings, which contain most of the valuable spectral information while retaining most of the original information content. In this research, partial least squares (PLS) method was used to develop a very robust analytical methodology which is a multivariate analysis technique that provides a model for the relationship between a set of predictor variables M (n objects, m variables) and a set of response variables P (n objects, p response). In this case m variables are the LIBS wavelengths and p responses are properties such as the inorganic content. The p response has to be independently measured for each sample. If the spectral data contain information about the properties of interest, a reliable calibration model can be constructed. This leads to the determination of regression coefficients for the different elements that are correlated using the two techniques of LIBS and NAA. Multivariate analysis of the data is performed using the Unscrambler (vsn. 8.05) software, CAMO, Corvallis, OR. The package has the capability to perform both PCA and PLS (also known as projection latent structures) analyses.

1.1. Advantages of LIBS technique The advantages of LIBS are well documented [12–17]. A few that are pertinent to the research in this article are, the real-time detection with the identification of metals and non-metals in seconds due to the capability of data acquisition over a large spectral range of 200–800 nm, and the ability to obtain the fine spatial resolution of ionomic profiles within samples (e.g., tissue/organs). The ability of LIBS to provide rapid multielemental microanalysis of bulk samples (solid, liquid, gas, and aerosol) in the parts-per-million (ppm) range with little or no sample preparation has been widely demonstrated. Here, for the first time in plants to our knowledge, the technique of laser-induced breakdown spectroscopy correlated to a standard technique of neutron activation analysis has been chosen to determine the abundance of different elements in leaves.

47

2. Materials and methods 2.1. Theory and practice of LIBS The first step in LIBS is occurrence of breakdown followed by plasma formation. Breakdown can occur when a laser beam of sufficient energy is focused onto a small area (~14 × 10−9 cm2), creating a power density inside the targeted area exceeding tens of gigawatts per square centimeter. The experimental configuration is described in detail in reference [12]. A pulsed Nd:YAG Bigsky™ laser that is frequency doubled (532 nm wavelength) is used to generate a spark on the sample that is being tested. During the transitory bombardment of the high-energy laser pulse (3 ns pulse width), a small amount of material (in the range of micrograms to nanograms) can be readily excited and vaporized under the intense energy-matter interaction. This plasma appears as intense light followed by a loud sound and an intense spark as it expands outward at a supersonic speed in all directions. Spectral emissions from ionized, neutral, and molecular species occur sequentially after plasma is formed. For example, the spectral peaks observed in a solid sample ranges anywhere from 0.5 to 2 μs after plasma formation are mostly due to ionized species present in the plasma. Between 2 and 10 μs, the main contributions of emission lines are from the de-excitation of neutral atoms. The emission peaks that occur after 10 μs are due to both the de-excitation of neutral atoms and simple excited molecules resulting from radiative recombination [16,17]. Interference in the spectral measurement of an element can thus be eliminated by carefully selecting the time at which the measurement is taken relative to the time of plasma formation [18]. The optimal time is unique for each element, providing a second characteristic for identification in addition to the spectra. The resolved spectrum is detected by an intensified charge coupled device (ICCD) built by Andor Technology that is delayed and gated by a delay generator that is integrated onto the ICCD. An XYZ stage is used to move the laser beam to different positions on the sample surface so that the laser does not dig a hole in the sample and lose its focus and light collection efficiency when sampling at the same point on the sample surface. The experimental setup is also equipped with a red wavelength auto-focusing laser which makes certain that the height of the sample is always maintained at the laser-lens focal plane. Extensive work in detection and quantification of total elemental concentrations from biological matrices such as plants (roots, stem and leaves) for phytoremediation applications [19] has been performed. We also have detected pollutants in streams by using LIBS on invertebrates that were collected from locations close to the pollution source and comparing these with the samples obtained from pristine areas [20]. In addition, trace elements have been mapped in biological matrices using laser-induced breakdown spectroscopy [21] as shown in Fig. 1. Each dark spot on the leaf surface represents a spectrum which shows a number of palladium elemental peaks. In applications such as these, LIBS provides a potentially attractive alternative to standard methods of analyzing metal accumulation, because sample mass requirements are greatly reduced, measurements can take only a few seconds to complete, and sample preparation is minimal. For elemental quantification it is important to correlate the LIBS measurements to a laboratory standard analytical technique such as NAA. It has also been observed that certain metals will accumulate in the green leafy part of the leaf tissue and others will collect in the midriff and veins of the leaf [21]. 2.2. Theory and practice of NAA NAA is a nuclear analytical method that can be used for nondestructive trace-element analysis of solid materials. NAA has a well-defined uncertainty budget which has led NIST to use NAA techniques extensively to certify trace elements in candidate reference materials [22]. Samples for analysis by NAA are introduced into a strong neutron source

48


Fig. 1. Surface scan of palladium metal from a leaf using LIBS.

(usually a reactor) where they remain for predetermined duration, and then are removed for measurement. Such a neutron irradiation can result in several processes, all initiated by neutron absorption (Fig. 2). While all of these processes can be exploited through various NAA techniques, this work utilized the measurement of delayed gammarays by High Purity Germanium (HPGe) gamma-ray spectrometry. For the set of measurements discussed in this document, NAA was conducted using this well-known activation equation and solving for target atom mass through knowledge of the interrogating neutron flux and cross-section [22]. For this effort, the Comparator method of NAA analysis was employed. In this method, comparators, or known materials, are used to develop reaction rates for the reactions of interest. Reaction rates in samples are then compared to these to calculate the results on a mass basis.

The NAA laboratory at Oak Ridge National Laboratory's High Flux Isotope Reactor (HFIR) contains two pneumatic tubes (PT-1 and 2), which have access to thermal neutron fluxes of about 4·1014 and 4·1013 n/cm2-s, each with a thermal to epithermal ratio of about 45 and 250, respectively (see Fig. 3 for 238-group neutron spectra). In addition, the NAA lab contains a hot cell, fume hoods, delayed neutron counting station, and multiple HPGe detectors for sample preparation and measurement. This combination of high neutron flux and nearby counting equipment allows for rapid irradiation and measurement of trace elements within a wide array of sample matrices. Fig. 4 shows the arrangement of the PT-1 pneumatic tube facility (PT-2 is similar). Sample containers called “rabbits” are sent to the reactor and back on counter-current air columns. These containers are usually made of high-density polyethylene and are fabricated under cleanroom

Fig. 2. Neutron irradiation processes in Neutron activation analysis [23].


49

14

10

PT-1 PT-2

13

Neutron Flux (n*cm-2*s-1)

10

12

10

11

10

10

10

9

10

8

10

7

10 -10 10

-8

10

-6

10

-4

10 Energy (MeV)

-2

10

0

10

Fig. 3. HFIR PT-1 and PT-2 neutron energy spectra.

conditions. The rabbits make the trip to and from the reactor in just a few seconds after remaining in the reactor for a predetermined period. Flux monitor foils (dilute Mn and Au in aluminum) are irradiated each day of use to track the daily flux values. Poplar leaf samples normally represent an excellent matrix for NAA. Tree leaves do not activate very strongly, and do not produce isotopes that emit gamma-rays which could interfere with the analysis. Nuclear heating during irradiation is low, so irradiation times may be extended to maximize sensitivity while not resulting in softening of the polyethylene rabbits. Finally, since both the cross section for neutron absorption and the density of elemental carbon are low, self-shielding problems

during irradiation and during gamma-ray counting may be safely considered negligible. One concern is that the samples should be of uniform moisture content to avoid problems associated with excessive outgassing during irradiation that causes swelling of the irradiation container. In addition, the neutron scattering cross section will vary some in cases where high moisture is present. 2.3. Sample collection Leaf plant material was obtained from a common garden at Clatskanie, Oregon USA that contains a clonally replicated population

Fig. 4. HFIR PT-1 pneumatic transfer tube.


of 1100 black cottonwood (Populus trichocarpa Torr. & Gray) genotypes as previously described [24]. The genotypes of P. trichocarpa are characterized by high levels of molecular and phenotypic variation, with relatively weak interpopulation differentiation making this a suitable population for investigating variation in the leaf ionome. A single leaf on the south side of the tree exposed to full sunlight conditions and 10 ± 1 from the apex was picked within a 4 hour window centering on solar noon, and quickly frozen on dry ice [24]. Genome resequencing reveals multiscale geographic structure and extensive linkage disequilibrium in the forest tree Populus trichocarpa. 2.4. Sample preparation All the poplar leaf samples were finely ground to 40 μm particle size. The ground powder was weighed to 75 mg, left overnight to humidify in a chamber with open beakers of water, and pelletized for 3 min at 2000 psi in a Carver press with a 6 mm evacuable die. Initially 25 poplar samples and 5 NIST standard reference materials were chosen to perform LIBS and NAA. The five NIST standards that were selected were, spinach leaves, orchard leaves, tomato leaves, pine needle, and bovine liver. These two techniques were done on the same samples to check if we could get correlation of the concentrations of the different elements within the two techniques. Two pellets each for the 30 calibration samples were created, one was used for the LIBS measurements and the NAA was performed on the other pellet. The same 30 samples that were used to do the LIBS measurements on were also used to determine the real concentrations of the 7 elements using neutron activation analysis and then used to construct a regression model for the various elements. Multivariate statistical analysis has been completed and shows very promising correlation of LIBS to NAA data. The elements that have been detected and analyzed are, Ca, Mg, Mn, K, Cu, Cl, and Al. Armed with that information we are continuing to conduct LIBS on an additional 450 samples. For NAA analysis, the pellets were weighed on a precision MettlerToledo analytical balance model AE-163 (Mettler Toledo, Columbus OH), and packaged in high-density polyethylene irradiation containers (rabbits). These were inserted into the HFIR PT-2 pneumatic tube for 30 s and counted immediately to reveal the elements that form shortlived radioisotopes. After the shorts were analyzed in all of the samples and comparators, a second irradiation was made for 300 s in PT-1. The second irradiation was followed by additional counting beginning one day later, and a third count was made after one week or so. This effectively divided the trace element makeup into three bins, sorted by half-life. In this way, the individual gamma-ray spectra were simplified. All of the counting was accomplished on an HPGe detector. Counting heights ranged between 30 cm for the first count to 30 mm for the last count. The reaction rates of the samples were compared to the reaction rates of the comparator materials in the method discussed above. The fluence rate values in the irradiation positions were measured on the days of irradiation to provide means to analyze elements that were not satisfactorily covered by one of the comparators. The comparators consisted of NIST reference standard botanical materials. 3. Results and discussions The NAA results for the elements, Mg, Al, Cl, K, Ca, Mn, and Cu are shown in Table 1. These elements were identified and the concentrations for the seven elements were obtained for the 25 leaf samples and 5 NIST standards. As seen from Table 1, no data was obtained for the NIST standard sample for pine needle. This is because pine needles have elevated Mn, which activates strongly and creates potential radiological hazards. Also concentrations for Mg, Cl, and Cu could not be obtained using the NAA analysis for the NIST standard for spinach leaves. The typical LIBS spectra for NIST standards for bovine liver, tomato leaves, and spinach leaves is shown in Fig. 5.

Table 1 The NAA results for elements, Mg, Al, Cl, K, Ca, Mn, and Cu for the 25 samples and 5 NIST standards. NAA results ID

Mg

Al

Cl

K

Ca

Mn

BESC-211 BESC-833 BESC-294 BESC-42 HIRD-12-2 KTMA-12-3 BESC-15 BESC-890 BESC-184 BESC-290 SKWD-24-4 BESC-329 KTMB-12-1 BESC-258 HOMC-21-3 BESC-869 BESC-104 HOMD-21-2 BESC-881 CNYH-28-3 BESC-351 BESC-226 SKWD-24-4 BESC-246 BESC-290 Spinach Orchard leaves Tomato leaves Pine needle Bovine liver

2022.4 1609.9 1932.4 2052.3 2696.0 1649.8 2732.6 2889.1 1185.1 2131.6 1482.2 1640.9 1343.5 2378.6 2005.6 1858.5 1675.6 2311.4 2641.8 2299.6 2588.5 2076.7 1650.4 2566.9 1860.2

28.6 20.2 30.2 12.5 28.2 42.2 20.0 29.3 23.7 55.0 17.1 47.1 99.1 50.5 35.1 48.9 21.3 63.2 14.8 21.0 35.2 17.5 22.5 27.3 23.2 903.8

501.2 300.4 463.6 399.6 559.8 543.0 375.0 421.6 280.7 389.4 216.1 267.4 351.7 348.2 276.4 408.8 334.6 383.1 372.5 280.5 269.2 328.4 463.2 177.7 231.5

7320.4 7456.9 7342.3 4687.2 6097.7 7612.4 6466.0 3680.3 5544.4 8545.9 3553.6 8694.2 6577.0 4821.7 7729.5 5376.4 6737.4 6773.0 4883.6 5969.6 5218.1 5613.3 6884.5 4579.5 5952.1 34,324.4 14,700.0 42,784.9

9135.9 6458.8 11,259.3 7849.6 17,484.2 13,162.7 11,028.6 9821.0 6160.8 8249.7 3047.1 13,849.4 10,755.2 12,912.7 10,984.6 13,852.0 11,306.6 10,149.5 12,962.6 13,713.8 19,308.4 16,214.7 8217.9 14,482.6 9555.8 13,801.7 21,536.4 31,347.0

65.4 62.5 91.5 67.7 156.6 116.1 84.5 73.1 61.8 86.9 34.3 142.9 123.9 145.8 122.5 112.4 46.9 113.1 46.1 83.8 87.9 66.7 32.7 54.7 47.2 168.8 90.0 217.2

6372.2 6746.7

690.0 1075.1

678.6

2387.8

9406.0

10.7

Cu 8.7 8.5 17.6 14.6 9.9

14.5 5.4 10.7 16.3 12.0 10.3 9.7

10.8 17.3 10.1

19.1

194.0

In Fig. 5 the LIBS spectra for three NIST standards is presented which demonstrates that some of the elements, such as Ca, Mg, and K, are present in all biological materials standards. Furthermore, it shows some elements that are present in one but not in the other standards. These differences and similarities are identified by MVA analysis of the LIBS spectra. From Fig. 5 it has been observed that the spectra for the three NIST standards show the differences in the elemental features in those samples and also some similarities. The LIBS intensity that was observed from the bovine liver sample was much weaker than those that were observed for the tomato leaves and spinach leaves. The NIST bovine liver reference material had a sticky and wet texture as compared to the other standards. LIBS is influenced by this difference. If the sample has higher moisture content then a higher amount of laser energy would be needed to for plasma formation, breaking bonds, and ionizing

Bovine Liver

LIBS Intensity (A. U.)

50

15000 Ca Mg

10000

Al Mg

Cu

K Mn

Tomato Leaves

5000 Spinach Leaves

Fe

0 200

300

400

500

600

700

Wavelength (nm) Fig. 5. LIBS spectra for three NIST standard reference materials such as, bovine liver, tomato leaves, and spinach leaves.


51

the neutron activation analysis (NAA) data, twenty five unknown samples mixed with 5 NIST standards were used in the multivariate analysis (MVA).

the elements that are present in the material. Furthermore, LIBS is highly influenced by the matrix that is being tested while NAA is not. In order to obtain statistical correlation for the LIBS measurements with 35x103 NAA vs Ca-LIBS cal NAA vs Ca-LIBS val r2 = 0.99 r2 = 0.94

LIBS Intensity (A. U.) for Ca

30x103 25x103 20x103 15x103 10x103

(a)

5x103 0 5x103

0

10x103 15x103 20x103 25x103 30x103 35x103 Neutron Activation Analysis (Ca)

6.0x103

250 NAA vs Mg-LIBS cal NAA vs Mg-LIBS val r2 = 0.97 r2 = 0.87

4.5x103

3.0x103

(b)

1.5x103

LIBS Intensity (A. U.) for Mn

LIBS Intensity (A. U.) for Mg

7.5x103

0.0 1.5x103

0.0

3.0x103

4.5x103

NAA vs Mn-LIBS cal NAA vs Mn-LIBS val r2 = 0.9 r2 = 0.85

200

150

100

(c)

50

0

6.0x103

0

50

Neutron Activation Analysis (Mg)

100

150

200

250

Neutron Activation Analysis (Mn)

60 NAA vs Al-LIBS-cal NAA vs Al-LIBS val r2 = 0.998 r2 = 0.81

LIBS Intensity (A. U.) for Al

50 40 30 20

(d) 10 0 10

20

30

40

50

Neutron Activation Analysis (Al)

20

50x103

LIBS Intensity (A. U.) for Cu

18 16

LIBS Intensity (A. U.) for K

NAA vs Cu-LIBS cal NAA vs Cu-LIBS val r2 = 0.97 r2 = 0.90

14 12 10

(e)

8

NAA vs K-LIBS cal NAA vs K-LIBS val r2 = 0.95 r2 = 0.87

40x103

30x103

20x103

(f)

10x103

6 4

0 4

6

8

10

12

14

16

Neutron Activation Analysis (Cu)

18

20

0

10x103

20x103

30x103

40x103

50x103

Neutron Activation Analysis (K)

Fig. 6. (i). The regression coefficients for calibration and validation for (a) calcium, (b) magnesium, and (c) manganese. (ii). The regression coefficients for calibration and validation for (d) aluminum, (e) copper, and (f) potassium.

52

M.Z. Martin et al. / Spectrochimica Acta Part B 138 (2017) 46–53 0.025

Table 2 Slopes, offsets, Root Mean Square Error (RMSE), and regression coefficients for calibration and validation for six elements (Ca, Mg, Mn, Al, Cu, and K). Offset

RMSE

Ca (calibration) Ca (validation) Mg (calibration) Mg (validation) Mn (calibration) Mn (validation) Al (calibration) Al (validation) Cu (calibration) Cu (validation) K (calibration) K (validation)

0.990 0.940 0.971 0.870 0.900 0.850 0.998 0.810 0.970 0.900 0.950 0.870

357.25 2630.98 103.2 463.7 12.45 21.07 0.0035 15.26 0.184 3.158 398.12 1978.39

979.81 1892.78 227.00 400.86 15.46 18.39 21.25 146.38 0.18 1.23 1572.72 2714.75

The regression coefficient for Cl is not well correlated to NAA since the concentrations for that element does not increase steadily for the 30 samples. Future experiments will use samples that have larger concentrations for chlorine and aluminum with a larger range to check if the correlation holds for these elements. Therefore the correlation coefficients along with the statistics have not been included here for the element of chlorine. The correlation model for calibration and validation was performed for the elements Ca, Mg, Mn, Al, Cu, K, and Cl. Out of the seven elements the correlation regression coefficients for calibration and validation for the 25 samples and 5 standards for six of the elements are shown in Fig. 6(i) that (shows the correlations for calcium (a), magnesium (b), and manganese (c)). Similarly 6(ii) shows the correlations for aluminum (d), copper (e), and potassium (f). From Fig. 6(i), the regression coefficient (r2) for calibration (NAA) and validation (LIBS) for calcium is 0.99 and 0.94 respectively. The regression coefficient for calibration and validation for magnesium is 0.97 and 0.87 respectively. Similarly, the r2 for calibration and validation for manganese is 0.9 and 0.85 respectively. Similarly from the Fig. 6(ii), the PLS model has calculated the r2 for calibration and validation for aluminum is 0.998 and 0.81 respectively. The r2 for calibration and validation for copper is 0.97 and 0.90 respectively. And finally the r2 for the last element potassium the calibration and validation is 0.95 and 0.87 respectively as shown in Fig. 6(ii). Table 2 shows the slope, offsets, root mean square error (RMSE), and regression coefficients for both calibration and validation sets for calcium, magnesium, manganese, aluminum, copper, and potassium. The loading parameters are shown in Fig. 7 for the six elements were determined by the model. The loading parameters are the peaks that the multivariate statistics has picked out for each element that contribute to the differences in their elemental concentrations. Multivariate approaches are chemometric analytical tools which take into account all the variables in the spectra, remove the redundancy, correlation, and collinearity, and resolve and extract useful information from the large and complex LIBS spectra [25]. Many studies have successfully applied multivariate approaches for analyzing the LIBS spectra of wood [12,25] human and animal bones [20], glass [28], geomaterials [29,30] and plant materials [31]. However, only a few studies applied multivariate approaches for soil spectra analysis, such as classification of distinct soil samples using principal component analysis (PCA) [32] and determination of elemental carbon concentration using partial least square (PLS) [33]. The PLS is one of the most useful and powerful multivariate approaches. It first establishes a linear model by relating the variations of dependent variables to the variations of independent variables, and then uses the established model to predict the dependent variables with the new independent variables. The PLS technique works especially well when independent variables carry common information such as correlations and collinearity [34]. The PLS components extracted from the original independent variables can be used to predict the dependent variables, and the first PLS

LIBS Intensity (A. U. )

r2

Copper

0.015 0.010 0.005 0.000 -0.005 200

300

400

500

600

700

800

Wavelength (nm)

5

4 LIBS Intensity (A. U.)

Element

0.020

Aluminum 3 Potassium 2 Manganese

1 Magnesium Calcium 0 200 300

400

500

600

700

800

Wavelength (nm) Fig. 7. The loading parameters (contributing spectral features) for the six elements.

component contains the most relevant information. In the LIBS spectra analysis, the PLS can simultaneously analyze all the peaks and select the most important and relevant peaks (as shown in Fig. 7) to construct a linear correlation to predict the elemental concentrations from the new measured spectra. With the application of multivariate approaches, there is no need to select a specific peak for the element of interest. Labbé et al. [25] reported that using several peaks in combination could produce dramatically better results than using any of the individual peaks alone. These previous studies showed the potential of multivariate approaches for analyzing the LIBS spectra. Therefore, our main objective is to implement the PLS regression technique to analyze the LIBS spectra of these plant samples for the quantitative analysis of the plant inorganic elements. 4. Conclusions The combination of the two techniques of LIBS and NAA analysis in application to determining the ions present in plant leaf material has been performed for the first time. It was determined that we could easily correlate 6 different elements via LIBS to NAA spectra. The elements, calcium, magnesium, manganese, aluminum, copper, and potassium were observed to be correlated to a large extent across multiple plant sample sets. It was determined that the element chlorine was difficult to determine. This approach will benefit from future studies that addition of more samples with a large range of values for the different elements to get a better correlation of LIBS data to NAA results. The loading parameters that were calculated by the PLS analysis selected the spectral peaks that belong to the six elements shown in this article. The regression coefficients of calibration and validation using multivariate analysis methodology for six out of seven elements have been determined. The range for the coefficients of regression for the six elements was determined to be between 0.81 and 0.99. LIBS has the advantage of sampling at 10 μm resolution allowing one to characterize the vast spatial heterogeneity in plant organ elemental composition.


Here, we demonstrate that LIBS and NAA techniques show very good correlation for the elemental composition of six elements that are present in plant leaf samples. This correlation of the two techniques has been shown for the first time. Acknowledgements The research was funded by the BioEnergy Science Center, a U.S. Department of Energy Bioenergy Research Center supported by the Office of Biological and Environmental Research in the DOE Office of Science. This manuscript has been authored by a contractor of the U.S. Government under contract DE-AC05-00OR22725. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. References [1] D.E. Salt, I. Baxter, B. Lahner, Ionomics and the study of the plant ionome, Annu. Rev. Plant Biol. 59 (2008) 709–733. [2] D. Cordell, J.O. Drangert, S. White, The story of phosphorous: global food security and food for thought, Glob. Environ. Chang. 19 (2009) 292–305. [3] J.S.I. Ingram, P.J. Gregory, A.M. Izac, The role of agronomic research in climate change and food security policy, Agric. Ecosyst. Environ. 126 (2008) 4–12. [4] G.P. Robertson, S.K. Hamilton, S.J. Del Grosso, W.J. Parton, The biogeochemistry of bioenergy landscapes: carbon, nitrogen, and water considerations, Ecol. Appl. 21 (2011) 1055–1067. [5] J.J. Elser, W.F. Fagan, A.J. Kerkhoff, H.G. Swenson, B.J. Enquist, Biological Stoichiometry of plant production: metabolism, scaling and ecological response to global change, New Phytol. 186 (2010) 593–608. [6] A.D. Friend, Terrestrial plant production and climate change, J. Exp. Bot. (2010), https://doi.org/10.1093/jxb/erq019. [7] S. Atwell, Y.S. Huang, B.J. Vilhjálmsson, G. Willems, M. Horton, Y. Li, D. Meng, A. Platt, A.M. Tarone, T.T. Hu, R. Jiang, N.W. Muliyati, X. Zhang, M.A. Amer, I. Baxter, B. Brachi, J. Chory, C. Dean, M. Debieu, J. de Meaux, J.R. Ecker, N. Faure, J.M. Kniskern, J.D. Jones, T. Michael, A. Nemri, F. Roux, D.E. Salt, C. Tang, M. Todesco, M.B. Traw, D. Weigel, P. Marjoram, J.O. Borevitz, J. Bergelson, M. Nordborg, Genome-wide association study of 107 phenotypes in Arabidopsis thaliana inbred lines, Nature 465 (Suppl. 7298) (2010) 627–631. [8] E. Buescher, T. Achberger, I. Amusan, A. Giannini, C. Ochsenfeld, A. Rus, B. Lahner, O. Hoekenga, E. Yakubova, J.F. Harper, M.L. Guerinot, M. Zhang, D.E. Salt, I.R. Baxter, Natural genetic variation in selected populations of Arabidopsis thaliana is associated with ionomic differences, PLoS One 5 (Suppl. 6) (2010), e11081. . [9] A. Rus, I. Baxter, B. Muthukumar, J. Gustin, B. Lahner, E. Yakubova, D.E. Salt, Natural variants of At HKT1 enhance Na + accumulation in two wild populations of Arabidopsis, PLoS Genet. 2 (Suppl. 12) (2013), https://doi.org/10.1371/journal. pgen.0020210. [10] X.Y. Huang, D.E. Salt, Plant ionomics: from elemental profiling to environmental adaptation, Mol. Plant 9 (Suppl. 6) (2016) 787–797. [11] I. Baxter, Ionomics: studying the social network of mineral nutrients, Curr. Opin. Plant Biol. 12 (Suppl. 3) (2009) 381–386. [12] Madhavi Z. Martin, Nicole Labbe, Timothy G. Rials, Stan D. Wullschleger, Analysis of preservative-treated wood by multivariate analysis of LIBS spectra, Spectrochim. Acta B 60 (7–8) (2005) 1179–1185.

53

[13] Rebekah J. Wagner, W. Kaye Margot, Marc D. Abrams, Paul J. Hanson, Madhavi Martin, Tree-ring growth and wood chemistry response to manipulated precipitation variation for two temperate Quercus species, Tree-Ring Res. 68 (1) (Jan 2012) 17–29. [14] Madhavi Z. Martin, Arthur J. Stewart, Kimberley D. Gwinn, John C. Waller, Laserinduced breakdown spectroscopy used to detect endophyte-mediated accumulation of metals by tall fescue, Appl. Opt. 49 (13) (May 1 2010) C161–C167. [15] Steven D. Brown, Madhavi Martin, Sameer Deshpande, Sudipta Seal, Katherine Huang, Eric Alm, Yunfeng Yang, Liyou Wu, Tingfen Yan, Xueduan Liu, Adam Arkin, Karuna Chourey, Jizhong Zhou, Dorothea K. Thompson, Cellular Response of Shewanella oneidensis to Strontium-Stress, Appl. Environ. Microbiol. 72 (1) (2006) 890–900. [16] D.A. Cremers, L.J. Radziemski, Laser plasmas for chemical analysis, in: L.J. Radziemski, R.W. Solarz, J.A. Paisner (Eds.), Laser Spectroscopy and Its Applications, Marcel Dekker, New York 1987, pp. 351–415. [17] L.J. Radziemski, D.A. Cremers, Spectrochemical analysis using laser plasma excitation, in: L.J. Radziemski, D.A. Cremers (Eds.), Laser-Induced Plasmas and Applications, Marcel Dekker, New York, 1989 (Chapter 7). [18] M.Z. Martin, M.D. Cheng, Detection of chromium aerosol using time-resolved laserinduced plasma spectroscopy, Appl. Spectrosc. 54 (2000) 1279–1285. [19] M.Z. Martin, A.J. Stewart, K.D. Gwinn, J.C. Waller, Laser-induced breakdown spectroscopy used to detect endophyte-mediated accumulation of metals by tall fescue, Appl. Opt. 49 (Suppl. 13) (2010) C161–C167. [20] M.Z. Martin, N. Labbé, N. André, R. Harris, M. Ebinger, S.D. Wullschleger, A.A. Vass, High resolution applications of laser-induced breakdown spectroscopy for environmental and forensic applications, Spectrochim. Acta, Part B 62 (Suppl. 12) (2007) 1426–1432. [21] J. Kaiser, K. Novotný, M.Z. Martin, A. Hrdlička, R. Malina, M. Hartl, V. Adam, R. Kizek, Trace elemental analysis by laser-induced breakdown spectroscopy – biological applications, Surf. Sci. Rep. 67 (2012) 233–243. [22] R.R. Greenberg, P. Bode, F. De Nadia, A. Elisabete, Neutron activation analyis: a primary method of measurement, Spectrochim. Acta Part B 66 (2011) 193–241. [23] http://archaeometry.missouri.edu/naa_overview.html, Accessed date: 4 August 2016. [24] G.T. Slavov, S.P. DiFazio, J. Martin, W. Schackwitz, W. Muchero, E. Rodgers-Melnick, M.F. Lipphardt, C.P. Pennacchio, U. Hellsten, L.A. Pennacchio, L.E. Gunter, Genome resequencing reveals multiscale geographic structure and extensive linkage disequilibrium in the forest tree Populus trichocarpa, New Phytol. 196 (Suppl. 3) (2010) 713–725. [25] N. Labbé, I.M. Swamidoss, N. André, M.Z. Martin, T.M. Young, T.G. Rials, Appl. Opt. 47 (Suppl. 31) (2008) G158–G165. [28] M.M. Tripathi, K.E. Eseller, F.-Y. Yueh, J.P. Singh, Spectrochim. Acta Part B 64 (2009) 1212–1218. [29] S.M. Clegg, E. Sklute, M.D. Darby, M. Dyar, J.E. Barefield, R.C. Wiens, Quantitative remote laser-induced breakdown spectroscopy by multivariate analysis, Spectrochim. Acta Part B 64 (Suppl. 1) (2009) 79–88. [30] J.L. Gottfried, F.C. De Lucia Jr, A.W. Miziolek, Laser-induced breakdown spectroscopy (LIBS), for green chemical analysis due to its unique features, Spectrochim. Acta Part B 64 (Suppl. 10) (2009) 1009–1019. [31] J.W.B. Braga, L.C. Trevizan, L.C. Nunes, I.A. Rufini, D. Santos Jr., F.J. Krug, Comparison of univariate and multivariate calibration for the determination of micronutrients in pellets of plant materials by laser induced breakdown spectrometry, Spectrochim. Acta Part B 65 (Suppl. 1) (2010) 66–74. [32] B. Bousquet, G. Travaillé, A. Ismaël, L. Canioni, K. Michel-Le Pierrès, E. Brasseur, S. Roy, I. le Hecho, M. Larregieu, S. Tellier, M. Potin-Gautier, T. Boriachon, P. Wazen, A. Diard, S. Belbèz, Development of a mobile system based on laser-induced breakdown spectroscopy and dedicated to in situ analysis of polluted soils, Spectrochim. Acta Part B 63 (2008) 1085–1090. [33] M.Z. Martin, N. Labbé, N. André, S.D. Wullschleger, R.D. Harris, M.H. Ebinger, Novel multivariate analysis for soil carbon measurements using laser-induced breakdown spectroscopy, Soil Sci. Soc. Am. J. 74 (Suppl. 1) (2010) 87–93. [34] S.V. Romanenko, A.G. Stromberg, Modelling of analytical peaks: peaks modifications, Anal. Chim. Acta 581 (2007) 343–354.