Laser-Induced Breakdown Spectroscopy (LIBS) Measurement of Uranium in Molten Salt

Abstract

In this current study, the molten salt aerosol–laser-induced breakdown spectroscopy (LIBS) system was used to measure the uranium (U) content in a ternary UCl₃–LiCl–KCl salt to investigate and assess a near real-time analytical approach for material safeguards and accountability. Experiments were conducted using five different U concentrations to determine the analytical figures of merit for the system with respect to U. In the analysis, three U lines were used to develop univariate calibration curves at the 367.01 nm, 385.96 nm, and 387.10 nm lines. The 367.01 nm line had the lowest limit of detection (LOD) of 0.065 wt% U. The 385.96 nm line had the best root mean square error of cross-validation (RMSECV) of 0.20 wt% U. In addition to the univariate calibration approach, a multivariate partial least squares (PLS) model was developed to further analyze the data. Using partial least squares (PLS) modeling, an RMSECV of 0.085 wt% U was determined. The RMSECV from the multivariate approach was significantly better than the univariate case and the PLS model is recommended for future LIBS analysis. Overall, the aerosol-LIBS system performed well in monitoring the U concentration and it is expected that the system could be used to quantitatively determine the U compositions within the normal operational concentrations of U in pyroprocessing molten salts.

Keywords

Laser-induced breakdown spectroscopy LIBS nuclear safeguards pyroprocessing uranium process monitoring

Introduction

An important aspect of reprocessing used nuclear fuel (UNF) in a nuclear facility is the material accountancy and safeguards of special nuclear materials (SNM) such as uranium (U) and plutonium (Pu). The ability to track SNM in a timely manner ensures that the process is operating under expected (normal) conditions and that the facility is compliant with policy and restrictions.¹ A reprocessing technology known as pyroprocessing, or electrochemical processing, has been under development in the United States and throughout the world as an alternative to the traditional aqueous reprocessing technology.² This approach offers many advantages including a relatively small facility footprint, radiation tolerance, and proliferation resistance. In pyroprocessing, the UNF is chopped and loaded into an anode basket and immersed into an electrorefiner (ER), where the U is electrochemically extracted and transported through a molten salt electrolyte where it deposits on a cathode. As part of the process, SNM accumulates in the molten salt over time that must be monitored. Due to the high temperature and radiation involved, as well as the hygroscopic nature of the molten salt electrolyte, pyroprocessing is conducted in remote handling inert atmosphere hot cells. These extreme conditions present new challenges for monitoring the SNM concentrations within the ER and in pyroprocessing.

The current approach to monitor the actinide concentrations in the ER and other pyroprocessing operations is to draw a melt sample which is then transported out of the hot cell for a destructive assay analysis, typically inductively coupled plasma–mass spectrometry (ICP-MS).³ This approach is time-consuming and can lead to delays in facility operations. An actinide sensor is currently under development by the Idaho National Laboratory to monitor SNM in situ.⁴ However, to date, the technology has only been demonstrated using surrogates and the sensor will not provide other elemental information useful for process monitoring. An attractive technology for monitoring the ER composition including the SNM is laser-induced breakdown spectroscopy (LIBS). In LIBS, a pulsed laser is used to ablate a sample in the formation of a plasma. As the plasma cools, each ion in the plasma emits characteristic light which is collected to form a spectrum representing the elements present in the sample.⁵ The advantage of LIBS is that it can be done with little to no sample preparation and analytical data can be collected in near real time in many sample phases—including gases,⁶ liquids,^7–9 solids,¹⁰ and aerosols.^11,12

Laser-induced breakdown spectroscopy technology has been applied^13–16 in the past to solid salt samples that have been drawn from a melt. However, this approach still requires an extraction or sample preparation phase, which adds to the required time to acquire a measurement. In addition, dust formation¹⁴ and depth profiling¹⁵ become important concerns. Despite these challenges, LIBS has been used to successfully measure cerium (Ce),^14,15 gadolinium (Gd),¹⁴ and U¹³ in a solid salt matrix. A more direct approach has been used to monitor the molten salt by performing the analysis directly in the molten salt liquid.^17–20 In this approach, the laser is fired downwards onto the top surface of the salt to form the plasma on the liquid surface. The plasma formation creates a shock wave that splashes material that can be damaging to the optical equipment used. In addition, surface inhomogeneity and bubble formation contribute to uncertainty in the results. However, with this approach, chromium (Cr),¹⁷ cobalt (Co),¹⁷ manganese (Mn),^17,18 Ce,¹⁸ europium (Eu),¹⁹ praseodymium (Pr),¹⁹ neptunium (Np),²⁰ U,²⁰ and Pu²⁰ have been successfully measured in the molten salt phase.

Recently, a novel molten salt aerosol approach has been demonstrated using Ce.^21,22 In this approach, a nebulizer with an argon gas carrier was used to generate a molten salt aerosol into which the laser light was focused to create the plasma. The advantages of this system are that splashing is reduced and the optics are isolated from the salt system. The main motivation of this work was to explore the molten salt aerosol–LIBS approach more fully in the measurement of U in LiCl–KCl molten salt. The approach was to optimize system parameters, identify ideal U lines, and to perform univariate and multivariate calibration methods to determine the analytical figures of merit of the aerosol–LIBS system. The success of this work will demonstrate the feasibility of such an aerosol–LIBS system for online monitoring of SNM in molten salt. Consequently, this work has the potential to significantly enhance safeguards and material accountancy in pyroprocessing technology that will further drive the global attractiveness of the technology.

Materials and Methods

In this work, the aerosol–LIBS system used in previous molten salt studies was implemented and a full description of the system can be found in the literature.^21,22 In brief, the system consists of a stainless-steel Collison nebulizer, a sealed sampling chamber (with sapphire windows), and a series of filters to remove the aerosol from the gas stream as shown in Fig. 1. The entire system was set up and operated within an inert atmosphere glove box. At room temperature, the salt was loaded into the nebulizer and sealed. Swagelok fittings were used to assemble and seal the system components before heating. Argon gas was used as the carrier gas. The gas pre-heat and nebulizer were heated using KerrLab furnaces while the sampling chamber and filters were heated using cartridge heaters. The temperatures of the different components were maintained at 550 ℃, 470 ℃, 260 ℃, 500 ℃, and 60 ℃ for the gas pre-heat, nebulizer salt, sampling chamber, 1.0 µm filter, and the 0.5 µm filter, respectively.

The LIBS excitation laser was a Nd:YAG laser (Q-smart 450, Quantel USA) that had been configured to the 532 nm wavelength. The laser was positioned outside the glove box and the light was directed into the glove box via a 10 × 10 cm quartz window. Inside the glove box, the laser light was directed downward through a 3X beam expander before being focused into the aerosol gas stream within the sampling chamber. Light from the sample plasma was collected at 90° from the laser path and focused into a 50 µm fiber optic cable (Ocean Optics). This cable was 4 m long and passed through a sealed port in the glove box wall to a Mechelle 5000 spectrometer (Andor Technologies) equipped with an iStar ICCD detector (Andor Technologies). This optical configuration and equipment set-up was identical to work done by Williams.²² However, the optical components (lenses and mirrors) within the glove box were removed and cleaned before the molten salt U (MSU) experiments. With the re-assembly of the components, the alignment shifted slightly, resulting in a 20% loss in signal. In addition, with the re-assembly of the 3X beam expander, the beam profile shifted, which resulted in hot spots that damaged the sapphire windows over time. Partial re-alignment was done to reduce the hot spots and increase the LIBS signal. However, due to challenges of working within the glove box environment, the system was not re-aligned to the exact condition of operation.

Depleted uranium trichloride (UCl₃) salt (74.4 wt% in LiCl–KCl eutectic) was acquired from the Idaho National Laboratory. Other salts used were LiCl (Alfa Aesar, 99.99%) and KCl (Alfa Aesar, 99.95%). The LiCl and KCl salts were mixed at the eutectic ratio (44 wt% LiCl and 56 wt% KCl) and dried at 250 ℃ for 2–12 h in a stainless-steel crucible. After drying, the LiCl and KCl salts were melted at 500 ℃ for 1 h before being removed to cool. The depleted UCl₃ salt was then added on top of the solid LiCl–KCl salt before being melted at 500 ℃ for 3 h to ensure a homogeneous ternary salt had formed. Figure 2a shows a picture of the prepared salt while still molten. Before the ternary salt had solidified, a cold stainless-steel rod was inserted into the salt to draw a small sample for ICP-MS analysis. Often, a thin oxide layer was observed on the top surface of the salt. This oxide layer formed as a result of small amounts of O₂ and H₂O within the glove box and from corrosion of the stainless-steel crucible. Before collecting the ICP-MS sample, the oxide layer was pushed to the side revealing the clean salt below. Both the stainless-steel rod and the oxide formation can be seen in Fig. 2a. Figure 2b shows the solidified salt. It should be noted that the salt has a dark purple color that is a characteristic of UCl₃. To load the U salt into the nebulizer, the prepared salt (as shown in Fig. 2b) was crushed using wire cutters to pieces approximately 3–5 mm in diameter.

Following each experiment, salt samples were taken from six different locations throughout the aerosol–LIBS system (as shown in Fig. 1). These locations are: (1) the nebulizer jar; (2) the nebulizer spout; (3) the sampling chamber nozzle; (4) the sampling chamber volume cell; (5) the 1.0 µm filter; and (6) the 0.5 µm filter. The amount of salt recovered in these locations varied, in most cases at least 0.01–0.05 g of material was recovered. However, for the sampling chamber volume, typically 0.005 g was recovered. Salt concentrations at the different locations were measured using ICP-MS. For the ICP-MS sample preparation, a range of 0.005–0.03 g of material (0.02 g was standard) was put into 15 mL test tubes and dissolved in 10 mL of 5% nitric acid for at least 24 h. Following dissolution, the samples were diluted between 100 × and 3000 × in order to get the expected concentration of U down below the EPA limit of 30 µg/L. The dilution factor varied based on the amount of salt originally dissolved and the expected concentration of the salt sample. The five wt% samples required a larger dilution factor to get below the EPA limit. In addition, samples from the sampling chamber with small amounts of dissolved salt required less dilution.

To reduce the spread of radioactive material and to decrease U waste, the system components were not removed from the glove box and washed between experiments. The one exception to this was the sapphire windows, after being wiped with chem-wipes in the glove box, the windows were transferred out and cleaned with water and acetone, inspected for damage, and then transferred back into the glove box for re-assembly. The residual salts adhered to the stainless-steel components following an experiment were removed manually using hand scraping tools. In the scraping step, the majority of the material was removed; however, small amounts of salt deposited within the tubing remained. Dust generated from scraping was removed from the components using parafilm. The 1.0 µm filter could not be cleaned by scraping because of salt adhered within the different stainless-steel layers of the filter. To prevent salt build-up and increased pressure drop, this filter was replaced for every experiment. The 0.5 µm filter was not replaced since the fine powder salt collected outside the filter was easily removed.

For each of the UCl₃ concentrations studied, six spectral repetitions were acquired, each comprising 300 laser shots. The laser energy was set at 50 ± 5 mJ and was operated at 2 Hz. The gate width was set constant at 300 µs and the gate delay was varied from 3 to 25 µs. The pressure was controlled to maintain a 12 psi differential pressure between the nebulizer inlet and the sampling chamber. With this pressure differential, the gas flow rate was constant at approximately 0.3 L/min.

Table I provides a summary of the MSU aerosol experiments conducted. The first five experiments (MSU1 through MSU5) were run at 1 wt% UCl₃ in order to optimize the gate delay, to assess the observed U peaks, and to explore the effect of gas temperature and pressure. For the first three experiments, the gas pre-heat furnace was left off and the nebulizer pressure was varied at 12 psi, 20 psi, and 30 psi, respectively. Following these pre-experimental runs, tests MSU4 through MSU9 were operated with the gas pre-heat furnace at 550 ℃ and with a 12 psi pressure differential across the nebulizer. In addition, the salt in these experiments contained 0.3, 1, 2, 3, and 5 wt% UCl₃ in LiCl–KCl eutectic salt.

Table I.

Summary of the UCl₃–LiCl–KCl experiments conducted.

Experiment name	wt% UCl₃	wt% U
MSU1	1.0	0.69
MSU2	1.0	0.69
MSU3	1.0	0.69
MSU4	1.0	0.69
MSU5	1.0	0.69
MSU6	2.0	1.38
MSU7	3.0	2.07
MSU8	5.0	3.46
MSU9	0.3	0.21

Results and Discussion

The UCl₃ salts were prepared and tested in the aerosol–LIBS system. After the MSU1 experiment, the salt in the nebulizer jar was purple-brown from oxide formation; however, salt collected everywhere else in the system was pure white suggesting low concentrations of UCl₃ (see Fig. S1, Supplemental Material). This observation was confirmed using ICP-MS and results are shown in Fig. 3. The UCl₃ concentration dropped drastically as the sampling location moved further from the nebulizer. This change in concentration throughout the system was also observed in the Ce experiments but to a lesser extent.²² This may be because the density of the UCl₃ salt aerosol is higher than that of CeCl₃; thus, a similar sized particle is more likely to fall from the slow aerosol gas stream. To test this postulation, MSU2 and MSU3 were operated at higher differential pressures and flow rates.

For MSU2, the pressure differential was maintained at 20 psi. However, the gas flow rate did not noticeably increase from the 12 psi case. For MSU3, the pressure differential was maintained at 30 psi and the gas flow rate increased to 0.6 L/min. The concentrations of UCl₃ at the different sampling locations are shown in Fig. 3, indicating that the concentration profiles between MSU1 and MSU2 are different but have a similar trend. Though the pressures were different between these experiments, the gas flow rates were essentially the same. MSU3 has a different profile, indicating that the increased flow rate helps reduce the rate at which the UCl₃ falls from the aerosol gas stream. However, at the 0.5 µm filter, all three experiments had nearly the same low concentration. Experiment MSU4 was run with identical conditions to MSU1 with an exception of the gas pre-heat being set at 550 ℃. The MSU4 concentration profile can be seen in Fig. 3. Here, the profile does not drastically drop to low values as is observed with the first three experiments. In addition, the amount of oxide formation within the nebulizer has been significantly reduced in MSU4. The mechanism creating the non-homogeneous concentration profile with cold carrier gas was not explored further because it remained outside the scope of the project; however, it might be the result of melt crystallization effects described for molten salt separations.^23–26 Future studies to explore the aerosol formation and characterization are recommended. The remaining experiments were done using the gas pre-heat furnace set at 550 ℃.

The ICP-MS results for samples MSU5 through MSU9 are summarized in Table II showing the calculated concentrations of the samples based on the added amount of UCl₃ in the sample. There are some significant discrepancies between the calculated and measured concentrations. The differences may result from the salt sampling method and oxide formation. As noted above, a brownish oxide layer often formed on the top surface of the salt. If a significant portion of the oxide layer was sampled, the ICP-MS results would indicate higher resulting values than expected. A good example of this is the MSU5 bulk-melt measurement where the measured concentration is approximately 20% greater than the calculated. To prevent this, when sampling from the melt, the oxide layer was avoided in order to sample the clean salt below the layer. To better understand the actual composition, a portion of the salt was later crushed and sampled to provide a more even distribution of the salt and oxide layer. With this modification, the difference between the calculated and ICP-MS results are similar and, in most cases, not significantly different. The one exception is the MSU9 experiment where only 20% of the calculated amount is measured. However, this measurement value is similar for all of the other sampling locations for MSU9. With this experiment, there is some brown oxide residue left in a ring around the outside of the stainless-steel melt crucible. Being that there was such a small amount of UCl₃ in the sample to start, this residue might account for the difference between the calculated and measured concentrations. Because the crushed ICP-MS measurements are the most consistent with observed and calculated results, they were selected as the accepted true values for this study.

Table II.

ICP-MS results from the MSU experiments, the calculated concentration is the expected value based on the amount of salts added. Results are in terms of the sample wt% of U.

Sampling location	MSU9	MSU5	MSU6	MSU7	MSU8
Calculated	0.20	0.69	1.38	2.074	3.46
Bulk-melt	0.023	0.88	1.23	1.98	3.029
Bulk-crush	0.042	0.66	1.25	1.91	3.24
Nebulizer spout	0.061	0.45	1.26	1.70	2.94
Nozzle	0.022	0.45	1.091	1.92	2.94
Sampling chamber	0.068	0.50	1.33	1.64	2.68
1.0 m filter	0.060	0.063	0.47	1.01	2.80
0.5 m filter	0.087	0.13	0.95	1.66	2.75

The MSU1–MSU4 experiments were analyzed using LIBS, but results were varied as a result of the non-homogeneous aerosol formation and transport. MSU4 was significantly more stable and data sets collected during this experiment were used for peak selection and gate delay optimization. A LIBS spectrum from MSU4 salt was collected with a 16 µs gate delay so a direct comparison could be made with the pure LiCl–KCl salt reported by Williams.²² To make the comparison meaningful, both the pure salt and UCl₃ salt spectra were normalized by dividing the entire spectrum point wise by the average spectral intensity of the entire spectrum. This comparison highlights differences in the spectra as shown in Fig. 4. Using this approach, multiple U lines have been identified and verified from the literature.^27–29 Table III shows some of the observed lines. The strongest U peaks identified are the U II 367.01 nm, U II 385.96 nm, and U I 387.10 nm lines.

Table III.

Observed U peaks in the UCl₃ salt samples.

Wavelength (nm)	Ionization state	Wavelength (nm)	Ionization state
357.70	I	394.28	II
358.09	I	400.52	I
358.45	I	401.77	II
359.37	I	406.25	II
360.63	II	409.01	II
365.92	I	417.16	II
367.01	II	435.57	I
374.64	II	436.21	I
381.19	I	437.20	I
381.41	II	439.39	I
385.46	II	457.68	I
385.96	II	502.74	I
386.52	II	562.08	I
387.10	I	607.73	I
387.85	I

Italicized ionization states were selected for further analysis.

During the MSU4 experiment, the gate delay was varied between 3 µs and 25 µs to determine the optimal value. Figure 5a shows the comparison between the different gate delay values, all with a constant gate width of 300 µs. The overall intensity of the spectra decreases as the gate delay increases. However, it appears that there is significant background present at the shorter gate delay values. When the signal-to-background (S/B) ratios were calculated for these spectra, the trend was relatively flat and the background was not strongly affected by the gate delay, though clearly the overall intensity was affected. Figure 5b shows the intensity decay of several U lines with respect to a region with no U lines. From this comparison, it was apparent that the background was decaying at nearly the same rate as the peaks themselves. With closer observation, the “background” observed in the spectrum decayed at nearly the same rate as the specific spectral lines, which was also noted by Chinni et al.¹⁰ There are a high density of U lines in the spectrum, due to instrument broadening of the lines, the different peaks overlapped to create this apparent background. By using a spectrometer with high spectral resolution as was done by Barefield et al.,²⁹ peak broadening would be reduced, and less “background” and more spectral lines would be observed. As a result of the mid-range spectrometer used in this study, the S/B ratio was not suitable for determining the optimal gate delay. As a result, the 3 µs gate delay was selected in this work as it had the strongest intensity overall. The actual optimal parameter may be at a different gate delay than studied here.

Figure 6 shows the entire spectrum generated for the 1 wt% UCl₃ (MSU4) experiment at a 3 µs gate delay. The spectrum is dominated by the Li, K, and Ar lines as expected due to the large concentrations of the elements in the sample and system. In the close-up region, a cyclic pattern in the apparent background has been observed. This is a result of transmission efficiencies of the echelle grating. The intensity of the U peaks is quite small compared to the apparent background and the Li, K, and Ar peaks. Definitely, increasing the spectral resolution would help with the apparent background. In addition, the fiber optic cable used in this work had low transmission characteristics below 400 nm, which drastically impacted the intensity of the U lines observed.

The precision of the aerosol–LIBS results were analyzed by calculating the percent relative standard deviations (%RSD) between the six spectra measured per experiment. The average %RSD values in the range of 350–to 800 nm were between 2.5% and 8% for the un-normalized spectra. MSU8 (5 wt% UCl₃) and MSU9 (0.3 wt% UCl₃) had the best and worst %RSDs, respectively. Table IV gives a list of the %RSD values for the selected U peaks. Between the three spectral lines explored, the %RSD does not vary significantly; however, it varies from experiment to experiment. The variation was greater in samples containing less U. The limiting factor for the precision in this study was the fluctuation in laser energy, particle interaction with the laser light, and damage to the sapphire windows over time. On average, the %RSD for the different experiments and lines was 4.5%, which is within the 5% fluctuation in the laser energy.

Table IV.

Percent RSD values for the peak intensities for select peaks between the six repetitions.

Experiment	U 367.01 nm	U 385.96 nm	U 387.10 nm
MSU9	7.3	7.2	7.7
MSU5	5.2	4.5	4.7
MSU6	3.2	3.5	2.6
MSU7	2.5	2.7	3.1
MSU8	2.3	2.2	2.7

Next, the repeatability of the system overall was explored between the different MSU experiments. The %RSD between the mean spectra from MSU5 through MSU9. The %RSD averaged approximately 25% in the range of 550–800 nm (where fewer U lines are present). However, by normalizing each spectrum with respect to the average intensity of the Li 610.3 nm peak, the %RSD between the different experiments dropped to around 10–15% (see supplemental material for %RSD plots). The high variation in the spectra range of 350–450 nm was a result of the varying U concentrations between experiments. The variation observed from experiment to experiment was likely due to optical degradation of system components within the glove box as a result of dust and vibrations. Indeed, a small decrease in the signal was observed over time.

To further assess the aerosol–LIBS system to measure UCl₃, univariate calibration curves were generated using the peak areas for the 367.01 nm, 385.96 nm, and 387.10 nm lines. The peak areas were calculated by integrating the area under the peak and then subtracting off the background area. The calibration curves for these lines are shown in Fig. 7a–c. Linear regression was done to determine the slope (m) and y-intercept (b) of the linear fit of the calibration points. These regression coefficients as well as the goodness of fit (R²) are shown in Table V. The standard deviation of the fit (S_y/x) with respect to the y-direction is given as

S_{y / yx x} = \sqrt{\frac{\sum (y_{i} - {\hat{y}}_{i})^{2}}{n - 2}}

(1)

where, y_i is the measured data point,

{\hat{y}}_{i}

is the regression values, and n is the number of data points. In addition, the limits of detection (LOD) for each curve was calculated using

LOD = \frac{3 σ}{m}

(2)

where σ is the standard deviation of the blank sample (or the point with the lowest concentration). For the peak area curves, the U 367.01 nm line had the best LOD of 0.065 wt%. The U 385.96 nm line had a higher LOD at 0.11 wt% despite having a steeper slope. This was because the 385.96 nm line had a larger standard deviation for the MSU9 experiment. In this work, the lowest U concentration studied was 0.042 wt% U, which lies below the calculated LOD. Consequently, very little to no U signal could be expected from the MSU9 sample, this was observed in the spectra and MSU9 essentially was a blank sample for the aerosol–LIBS system.

Table V.

Regression coefficients and LOD for the univariate calibration curves for U in molten salt.

Parameter	367.01 nm	385.96 nm	387.10 nm
m	0.25	0.29	0.11
b	−0.015	0.017	−0.0086
S_y/x	0.032	0.032	0.026
R²	0.992	0.994	0.973
LOD (wt%)	0.065	0.12	0.19
RMSECV (wt%)	0.23	0.20	0.28

To assess the accuracy of the calibration curves, a leave one out cross-validation (LOOCV) analysis was done. The challenge with this analysis was that there were only five data points to start with, performing a LOOCV analysis would leave only four active points that could affect the overall response. The effect of leaving one sample out on the linear regression was explored and indeed the slope and y-intercepts varied as each of the samples was left out. However, the resulting four-point regression curve fits into the confidence intervals of the five-point regression line as shown in Fig. 7d. The largest variation in the study came from leaving out the first point (MSU9) and the last point (MSU8). The results of the cross-validation are best given as the root mean squared error of cross validation (RMSECV) as shown below

RMSECV = \sqrt{\frac{\sum {(x_{i, LIBS} - x_{i, ICP - MS})}^{2}}{n}}

(3)

where

x_{i, LIBS}

is the independently predicted concentration of the left out sample and

x_{i, ICP - MS}

is the actual concentration of the sample as determined using ICP-MS. The RMSECV values for the selected lines are shown in Table V. The U 385.96 nm line had the best RMSECV of 0.20 wt% U, meaning that, overall, there was less variation or error (compared to the other lines) between the predicted and actual concentration values. For samples with very low concentrations of U, the U 367.01 nm line may provide the best indications of the presence of U due to the lower LOD. However, for samples containing approximately 1 wt% U or more, the U 385.96 nm line provides lower uncertainties on the predicted concentrations.

One limitation to the calibration curves reported was the limited number of data points available. With more data points (at least six), better analytical results could be obtained. The error estimation and the confidence intervals depended on the number of data points used in the analysis. With more points, the uncertainty estimation decreases and variations within the experimental setup would more likely to be accounted for. Additional points would also reduce variation in the LOOCV and better cross-validation results would be achievable. In addition, more experiments in the lower concentration range would provide a better estimate on the measurement limit of the aerosol–LIBS system.

The measured spectra were also analyzed using the PLS_Toolbox in Matlab using partial least squares (PLS) regression. Because there were few data points, and only six repetitions per experiment were done (30 spectra total), the data were not divided into calibration and validation sets. One of the MSU9 repetitions had an unusually high Q residual with respect to the Hotelling value (a generalization of Student's t distribution) and was left out of the calibration. As a result, 29 data points were used in the PLS analysis. Figure 8a shows the number of latent variables used with respect to the RMSECV for the generated model. In this model, six latent variables were selected which yields a RMSECV of 0.11 wt% U. Figure 8b shows the PLS calibration curve. The spectral response is linear with some variation between the sample repetitions.

Being that no validation data set was used, the accuracy of the model was verified using an additional cross-validation approach within the PLS_Toolbox. To be consistent with the univariate approach, a LOOCV method was used. Here, the RMSECV was calculated to be 0.085 wt%, which is slightly lower than the venetian blind approach. This was expected because with the LOOCV, 100% of the samples were left out and uncertainty was reduced.

Between the univariate and PLS calibration approaches, the PLS model had a significantly better RMSECV (0.085 wt% versus 0.20 wt%). One reason for this was that the spectral response for U in these experiments was very small. When looking at individual peaks (univariate), the variation due to the %RSD between the spectra combined with the weak signal created high variation and lower correlation between the measured response and the concentration. On the other hand, the PLS model used the entire spectrum, and latent variables that best describe the response were selected for the modeling. As a result, the PLS model performed significantly better and is recommended for future analytical measurements of UCl₃ in molten salts.

Uranium results obtained from the aerosol–LIBS system in molten salts were compared to other U studies found in the literature. The LOD was the most widely reported figure of merit and was used for the comparison. Singh et al.³⁰ found the LOD for U in liquid aerosol to be 1359 parts per million (ppm). In addition, in liquids, Wachter and Cremers⁷ determined the LOD to be approximately 100 ppm. On a substrate, Sarkar et al.³¹ found the LOD of U to be 18 ppm. Finally, Chinni et al.¹⁰ found the LOD of U in soils to be 0.26 wt% at close range and 0.5 wt% at a distance of 30 m. From these studies, it can be seen that results vary for the LOD of U in different materials and methods. The LOD determined using the aerosol–LIBS system was 0.065 wt% (650 ppm), which was located within the mid-range of values reported in the literature. To date, the 0.065 wt% U value is the only reported LOD for U in a molten salt.

Typical concentrations of different components within the ER have been reported by Phongikaroon and Simpson³ and Bean.³² Table VI shows the compositions of the ER salt following electrorefining (the amount of UNF reprocessed up to this point was not reported). The salt compositions reported by Bean were for a small batch process in which 269.45 g of UNF was electrorefined in 790 g of 9 wt% UCl₃–LiCl–KCl salt. Following electrorefining, 0.0817 wt% of U²³⁵, 1.8 wt% U²³⁸, 0.0876 wt% Pu²³⁹, and 0.0199 wt% Pu²⁴⁰ were contained in the salt. In both reported studies, greater than 1 wt% UCl₃ was present in the salt which was well above the calculated LOD and within the range of aerosol–LIBS system's capability to provide good analytical measurements. This indicates that monitoring U in the ER using aerosol–LIBS technique may be possible. Definitely, additional work should be done to explore multi-component salts with complex spectra to further assess the capability of the approach.

Table VI.

Typical distribution of elements and concentrations (wt% shown in parentheses).³

LiCl (21.22)	SrCl₂ (1.25)	NdCl₃ (5.14)	PuCl₃ (5.09)
KCl (26.99)	UCl₃ (2.81)	PmCl₃ (0.07)	AmCl₃ (0.02)
NaCl (23.77)	BaCl₂ (1.67)	SmCl₃ (0.99)	PrCl₃ (1.51)
RbCl (0.44)	LaCl₃ (1.60)	EuCl₃ (0.07)	YCl₃ (0.92)
CsCl (3.20)	CeCl₃ (3.04)	GdCl₃ (0.04)	NpCl₃ (0.14)

Conclusion

The first three U experiments were run to explore the effect of the gas pre-heat temperature with respect to the gas pressure and flow rate. Results indicated that without the gas pre-heat at 550 ℃, the U concentration within the system components decreased as the distance from the nebulizer increased, an effect that was controlled by adjusting the gas flow rate. Concentration profiles with the gas pre-heat were similar to those obtained for the Ce cases.

The %RSD values for the different U experiments were on an average of 4.5%. Considering the optical degradation of the sapphire windows overtime and the large laser energy fluctuation, this %RSD result was reasonable. From the univariate calibration curves, the U 367.01 nm line using the peak areas yielded the lowest LOD of 0.065 wt% U and the U 385.96 nm line giving the best overall RMSECV of 0.20 wt% U. The PLS model performed significantly better than the univariate case with a RMSECV of 0.085 wt%. This is a result of using more latent variables to model the analytical response with respect to the concentration. Results from both the univariate and PLS model showed that samples containing UCl₃ at concentrations above the LOD could be measured with reasonable accuracy.

From the typical ER salt compositions reported, the aerosol-LIBS system should be capable of detecting, monitoring, and quantifying U within the salt at near real-time. The successful competition of this work strongly supported the aerosol–LIBS approach as a viable method for measuring the ER salt used during pyroprocessing of used nuclear fuel.

Supplemental Material

Supplemental material for Laser-Induced Breakdown Spectroscopy (LIBS) Measurement of Uranium in Molten Salt

Supplemental material for Laser-Induced Breakdown Spectroscopy (LIBS) Measurement of Uranium in Molten Salt by Ammon Williams and Supathorn Phongikaroon in Applied Spectroscopy

Footnotes

Acknowledgments

The authors thank Dr. Sarah Rutan for her help and insights on statistical analysis of obtained data sets.

Conflict of Interest

The authors report there are no conflicts of interest.

Funding

This research was performed using funding received from the DOE Office of Nuclear Energy's Nuclear Energy University Programs (NEUP), work package # NU-12-VA-VCU-0203-01. Additional financial support was provided by the Nuclear Regulatory Commission Faculty Development Grant.

Supplemental Material

All supplemental material mentioned in the text, consisting of two figures, S1 and S2, is available in the online version of the journal.

References

M. Hori, S. Li, M. Pellechi. “Proliferation Potential and Safeguards Challenges of Pyroprocessing”. Paper presented at: IAEA Symposium on International Safeguards 2014. Vienna, Austria; October 20–24, 2014.

Nawada

H.P.

Fukuda

“Role of Pyro-Chemical Processes in Advanced Fuel Cycles”. J. Phys. Chem. Solids. 2005. 66(2–4): 647–651.

Phongikaroon

Simpson

“Equilibrium Model for Ion Exchange Between Multi-Valent Cations and Zeolite-A in a Molten Salt”. AIChE J. 2006. 52(5): 1736–1743.

J. Ju, S. Li. Actinide Ion Sensor for Pyroprocessing Monitoring. US Patent 8741119 B1. Filed 2011. Issued 2014.

Pasquini

Cortez

Silva

L.M.C.

Gonzaga

F.B.

“Laser Induced Breakdown Spectroscopy”. J. Braz. Chem. Soc. 2007. 18(3): 463–512.

Tambay

Thareja

“Laser-Induced Breakdown Studies of Laboratory Air at 0.266, 0.355, 0.532, and 1.06 µm”. J. Appl. Phys. 1991. 70(5): 2890–2892.

Wachter

J.R.

Cremers

D.A.

“Determination of Uranium in Solution Using Laser-Induced Breakdown Spectroscopy”. Appl. Spectrosc. 1987. 41(6): 1042–1048.

St-Onge

Kwong

Sabsabi

Vadas

E.B.

“Rapid Analysis of Liquid Formulations Containing Sodium Chloride Using Laser-Induced Breakdown Spectroscopy”. J. Pharm. Biomed. Anal. 2004. 36(2): 277–284.

Barreda

Trichard

Barbier

Gilon

et al.

“Fast Quantitative Determination of Platinum in Liquid Samples by Laser-Induced Breakdown Spectroscopy”. Anal. Bioanal. Chem. 2012. 403(9): 2601–2610.

10.

Chinni

R.C.

Cremers

Radziemski

L.J.

Bostian

et al.

“Detection of Uranium Using Laser-Induced Breakdown Spectroscopy”. Appl. Spectrosc. 2009. 63(11): 1238–1250.

11.

Martin

M.Z.

Cheng

M.-D.

Martin

R.C.

“Aerosol Measurement by Laser-Induced Plasma Technique: A Review”. Aerosol Sci. Technol. 1999. 31(6): 409–421.

12.

Cahoon

E.M.

Almirall

J.R.

“Quantitative Analysis of Liquids from Aerosol and Microdrops Using Laser Induced Breakdown Spectroscopy”. Anal. Chem. 2012. 84(5): 2239–2244.

13.

D. Park, B. Han, S. Park, J. Ku. “Quantitative Analysis of Uranium Spectra in Molten Eutectic LICl–KCl+UCl3 Using LIBS”. Poster presented at: International Pyroprocessing Research Conference. Jeju Island, Korea; September 21–23, 2016.

14.

Williams

A.N.

Bryce

Phongikaroon

“Measurement of Cerium and Gadolinium in Solid LiCl–KCl Salt Using Laser-Induced Breakdown Spectroscopy (LIBS)”. Appl. Spectrosc. 2017. 71(10): 2302–2312.

15.

B. Yoo, S.H. Kim, J. Lee. “Quantitative Analysis of Molten Salt by Laser-Induced Breakdown Spectroscopy”. Paper presented at: GLOBAL: Nuclear Fuel Cycle for a Low-Carbon Future. Paris, France; September 20–24, 2015.

16.

S.H. Kim, J. Lee, S. Phongikaroon. “The Remote Compositional Analysis of Molten Salt (KF–LiF–ZrF4) by Probe Assisted In-Situ LIBS System”. Paper presented at: American Nuclear Society (ANS) Winter Conference. 2015. Marriott Wardman Park, Washington, D.C.; November 8–12, 2015.

17.

A.J. Effenberger. Methods for Measurement of Heterogeneous Materials with Laser-Induced Breakdown Spectroscopy (LIBS). [Ph.D. Dissertation]. San Diego, CA: University of California, 2009.

18.

Hanson

Phongikaroon

Scott

J.R.

“Temperature effect on Laser-Induced Breakdown Spectroscopy Spectra of Molten and Solid Salts”. Spectrochim. Acta B. Atom. Spectros. 2014. 97: 79–85.

19.

Weisberg

Lakis

Simpson

M.F.

Horowitz

et al.

“Measuring Lanthanide Concentrations in Molten Salt Using Laser-Induced Breakdown Spectroscopy (LIBS)”. Appl. Spectrosc. 2014. 68(9): 937–948.

20.

N.A. Smith, J.A. Savina, M.A. Williamson. “Application of Laser-Induced Breakdown Spectroscopy to Electrochemical Process Monitoring of Molten Chloride Salts”. Paper presented at: International Atomic Energy Agency (IAEA): Linking Strategy, Implementation and People Symposium on International Safeguards. Vienna, Austria; October 20–24, 2014.

21.

Williams

A.N.

Phongikaroon

“Laser-Induced Breakdown Spectroscopy (LIBS) in a Novel Molten Salt Aerosol System”. Appl. Spectrosc. 2017. 71(4): 744–749.

22.

A.N. Williams. Measurement of Rare Earth and Uranium Elements Using Laser-Induced Breakdown Spectroscopy (LIBS) in an Aerosol System for Nuclear Safeguards Applications. [Ph.D. Dissertation]. Richmond, VA: Virginia Commonwealth University, 2016.

23.

Cho

Park

Lee

Kim

“Concentration of Cesium and Strontium Elements Involved in a LiCl Waste Salt by Melt Crystallization Process”. Nucl. Technol. 2010. 171(3): 325.

24.

Choi

Cho

Lee

Eun

et al.

“Inclusion Behavior of Cs, Sr, and Ba Impurities in LiCl Crystal Formed by Layer-Melt Crystallization: Combined First-Principals Calculations and Experimental Study”. J. Cryst. Growth. 2013. 371: 84–89.

25.

Williams

A.N.

Phongikaroon

Simpson

“Separation of CsCl from Ternary CsCl–LiCl–KCl salt via a Melt Crystallization Technique for Pyroprocessing Waste Minimization”. Chem. Eng. Sci. 2013. 89: 258–263.

26.

Versey

J.R.

Phongikaroon

Simpson

“Separation of CsCl from LiCl–CsCl molten salt by Cold Finger Crystallization”. Nucl. Eng. Technol. 2014. 46(3): 395–406.

27.

A. Kramida, Yu. Ralchenko, J. Reader, NIST ASD Team. “NIST Atomic Spectra Database (Version 5.3)”. National Institute of Standards and Technology (NIST). http://www.nist.gov/pml/data/asd.cfm [accessed Dec 15 2016].

28.

Judge

E.J.

Barefield

J.E.

II Berg

J.M.

Clegg

S.M.

et al.

“Laser-Induced Breakdown Spectroscopy Measurements of Uranium and Thorium Powders and Uranium Ore”. Spectrochim. Acta B. Atom. Spectros. 2013. 83–84: 28–36.

29.

Barefeld

J.E.

Judge

E.J.

Berg

J.M.

Willson

S.P.

et al.

“Analysis and Spectral Assignments of Mixed Actinide Oxide Samples Using Laser-Induced Breakdown Spectroscopy (LIBS)”. Appl. Spectrosc. 2013. 67(4): 433–440.

30.

Singh

J.P.

Yueh

Zhang

Carney

K.P.

et al.

“A Preliminary Study of the Determination of Uranium, Plutonium, and Neptunium by Laser-Induced Breakdown Spectroscopy”. In: Pandalai

S.G.

(ed). Recent Research Developments in Applied Spectroscopy, Trivandrum (India): Signpost, 1999.

31.

Sarkar

Alamelu

Aggarwal

S.K.

“Determination of Thorium and Uranium in Solution by Laser-Induced Breakdown Spectrometry”. Appl. Opt. 2008. 47(31): G58–G64.

32.

R. Bean. “Project Report on Development of a Safeguards Approach for Pyroprocessing”. Idaho National Laboratory Report, INL/EXT-10-20057, 2010. https://inldigitallibrary.inl.gov/sites/STI/STI/4731805.pdf#search=Project%20Report%20on%20Development%20of%20a%20Safeguards%20Approach%20for%20Pyroprocessing [accessed Jan 30 2018].

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.44 MB