Ion exchange kinetics of fission products between molten salt and zeolite-A

Microporous and Mesoporous Materials 152 (2012) 185–189

Contents lists available at SciVerse ScienceDirect

Microporous and Mesoporous Materials journal homepage: www.elsevier.com/locate/micromeso

Ion exchange kinetics of fission products between molten salt and zeolite-A Michael Shaltry a,⇑, Supathorn Phongikaroon a, Michael F. Simpson b a Center for Advanced Energy Studies, Department of Chemical and Materials Engineering and Nuclear Engineering Program, University of Idaho, 995 University Boulevard, Idaho Falls, ID 83401, United States b Center for Advanced Energy Studies, Pyroprocessing Technology, Idaho National Laboratory, P.O. Box 2528, Idaho Falls, ID 83415, United States

a r t i c l e

i n f o

Article history: Received 21 January 2011 Received in revised form 15 November 2011 Accepted 15 November 2011 Available online 30 November 2011 Keywords: Ion exchange Kinetics Fission products Molten salt Zeolite-A

a b s t r a c t Experimentation, data analysis, and modeling of ion exchange kinetics between fission products (cesium and strontium) and zeolite-A beads in molten LiCl–KCl have been performed to support optimization of an electrochemical process to treat used nuclear fuel. Models based on pseudo-first- and pseudo-secondorder sorption as well as diffusion have been adapted and compared to experimental data to assess their validity in describing the system. Individual experiments were performed with different concentrations of CsCl or SrCl2 in the LiCl–KCl salt. Zeolite beads were removed from the molten salt at prescribed intervals of time and prepared for analysis by inductively coupled plasma-mass spectrometry. Results indicate maximum cesium (Cs) and strontium (Sr) loading occurred at approximately 31 min and 104 min of contact, respectively. The rate of loading and maximum loading were found to increase with increasing initial concentration of Cs or Sr. Data analysis included determination of rate constants and diffusion coefficients of the proposed models for each experimental condition. Results reveal that the diffusion model provides the best fit to the experimental data with average diffusion coefficients of 2.0  1010 m2 s1 for Cs and 6.3  1011 m2 s1 for Sr. This suggests that chemical diffusion is the dominant mechanism of mass transfer. Ó 2011 Elsevier Inc. All rights reserved.

1. Introduction Used nuclear fuel from the Experimental Breeder Reactor-II is currently being treated using an electrochemical process at the Materials and Fuels Complex located at Idaho National Laboratory. The electrorefiner is the most important component of the electrochemical treatment process and is used for the simultaneous dissolution of metal fuel at the anode and deposition of uranium metal at the cathode. This process takes place in molten lithium chloride– potassium chloride (LiCl–KCl) [1–4]. During this process, active metals (including fission products) accumulate in the molten salt. After several batches of fuel are treated, it becomes necessary to remove some of the molten salt from the electrorefiner to avoid the buildup of heat generating fission products, such as cesium (Cs) and strontium (Sr). Sodium accumulation, which increases the melting temperature of the molten salt, and criticality issues associated with increasing concentration of plutonium are also reasons for removing salt [4]. To address this issue, it is currently planned to remove salt from the electrorefiner, combine it with zeolite-A and glass frit, and consolidate the resulting powder into a ceramic waste form. This method leads to the disposal of relatively large volumes of high-level waste. An ion exchange column loaded with zeolite has been

proposed for implementation in the process to reduce the volume of waste generated by enabling the recycle of LiCl–KCl back to the electrorefiner. This could dramatically lower processing costs, especially the cost to dispose of the waste. Numerous studies of fission product removal from molten salt by zeolite-A have focused on equilibrium exchange conditions [5–9]. However, an understanding of the kinetic behavior of this system is fundamental for equipment design and optimization of the process. To explore the kinetic behavior of fission product removal, experiments were tailored to produce zeolite samples at times prior to and at equilibrium conditions. Several batches of salt were prepared with varying concentrations of active metal chlorides to span a wide range of conditions. The zeolite samples were analyzed to determine the change in their elemental composition as a function of time. Data analysis was carried out to assess how well each model fit the data. Ultimately, the objective of this research is to develop a simple model that is capable of predicting the kinetic and equilibrium ion exchange behavior.

2. Experiment 2.1. Materials and equipment

⇑ Corresponding author. Tel.: +1 208 533 8140; fax: +1 208 526 8255. E-mail address: [email protected] (M. Shaltry). 1387-1811/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.micromeso.2011.11.035

All chloride salts used in the experiments were procured from Alfa Aesar (99.9% LiCl, 99.95% KCl, and 99.995% SrCl2) with the

186

M. Shaltry et al. / Microporous and Mesoporous Materials 152 (2012) 185–189

exception of 99.99% CsCl, which was an Acros Organics product. Zeolite-A beads (average diameter of 0.00213 m) were a commercial product obtained from UOP. The beads are composed of 2 lm–3 lm zeolite-4A (sodium-form) crystals held together by a proprietary clay binder. Dense magnesia (MgO) crucibles, with inside diameters of 0.0254 m and heights of 0.0508 m, were obtained from Ozark Technical Ceramics. The experiments were performed within an MBraun sealed-atmosphere glovebox using a KerrLab lab-scale furnace (see Fig. 1(a) and (b)).

3. Results and discussion 3.1. General data interpretation Measurement of the elemental composition of the samples yielded information about the loading of the cations in the zeolite as a function of time. The zeolite loading, yi, was calculated using the equation:

yi ¼ 2.2. Experimental procedure Throughout the experimental duration, oxygen and moisture levels were maintained below 0.1 lL/L. Each salt mixture was prepared by adding 1.0 wt.%, 2.5 wt.%, or 5.0 wt.% of CsCl or SrCl2 to approximately 0.0125 kg of LiCl–KCl salt (72 mol% LiCl). The MgO crucible containing the salt mixture was placed into the furnace and heated to 773 K. The experiment was initiated by lowering a stainless steel mesh basket containing the zeolite-A beads, which was coupled to a rotating motor drive (shown in Fig. 1(b)). The mesh basket was rotated at a rate of 10.47 rad s1 (100 rpm) while in contact with the molten salt. At the prescribed intervals, the mesh basket was raised to a height slightly above the furnace lid and a portion of the zeolite beads was retrieved and stored for analysis. Experiment durations were chosen to capture the kinetic and equilibrium regimes based on information from previous work [5,10]. Samples from the Cs experiments were taken at 5 min intervals during the first 30 min of contact and at 15 min intervals thereafter. Samples from Sr tests were taken at 20 min intervals during the first 120 min of contact and at 40 min intervals thereafter. Stored samples were then removed from the glovebox and prepared for compositional analysis via an inductively coupled plasma-mass spectrometer (ICP-MS). Zeolite beads were prepared for analysis by combining 600 lL of hydrochloric acid, 600 lL of nitric acid, 400 lL of hydrofluoric acid, and 6 mL of a 2.2% boric acid + 0.2% ethylenediaminetetraacetic acid solution [11–13]. Dissolution took place under a fume hood at room temperature. The zeolite beads were not crushed or rinsed prior to analysis. Once dissolved, the zeolite/acid solutions were diluted to a total dissolved solid of less than 0.001 kg L1. Internal and external standards were prepared for calibration of the Agilent 7200c ICP-MS. The diluted sample solutions were placed on the ICP-MS autosampler in random order. Uncertainty associated with the ICPMS measured value is in the range of 7–10%.

12ui wi MAl wAl Mi

ð1Þ

where 12 represents the moles of aluminum per unit cell, ui is the molar equivalent of species i per mole of species i, wi is the mass fraction of species i to the total dissolved solid of the sample, and Mi is the molar mass of species i. Results of the Cs and Sr ion exchange tests are depicted in Fig. 2(a) and (b), respectively. Maximum loading in the zeolite increases with increasing initial concentration of CsCl and SrCl2 in the salt. This observation is consistent with previously reported studies [5,7,8]. Further data analysis was facilitated by determining a critical time (tcritical) in the transition from the kinetic to equilibrium regions of the plots in Fig. 2. The critical time is the value of the time coordinate at the intersection of the average line of regression in the kinetic region and the equilibrium value of a particular data set. This method is illustrated in Fig. 3. In this case, the slope of the line in the kinetic region (mave) is formulated by averaging of the slopes of several lines (dashed lines) fitted to one or multiple data points in the apparent kinetic region of the plot. It was found that the average value for loading in the zeolite at the critical time for the Cs and Sr experiments is approximately 76% and 81% of equilibrium loading, respectively. The critical time value is useful for plotting the data in a dimensionless format, as shown in Fig. 4. From this, it can be seen that the data sets from both the Cs and Sr experiments collapse upon each other indicating universal behavior. Moreover, the time at which equilibrium loading occurs in the zeolite, indicated by the dashed line (teq/tcritical), is easily obtained from Fig. 4. In this particular case, the average values of teq/tcritical for Cs and Sr are 2.38 ± 0.142 and 2.11 ± 0.122, respectively. This result can be extended to systems of any initial concentration of Cs or Sr. For the Cs experiments, equilibrium was achieved at an average of 31 min of contact. For Sr, the average was 104 min. Additionally, the average rate of loading (yeq/teq) prior to equilibrium was calculated for each of the experiments. Table 1 lists the values associated with Figs. 3 and 4 for the Cs and Sr experiments.

Fig. 1. (a) The MBraun glovebox and (b) the KerrLab melting furnace.

187

M. Shaltry et al. / Microporous and Mesoporous Materials 152 (2012) 185–189

a

a

1.6 1.4

0.6

yCs(t)/yeq

yCs (mol eq/unit cell)

0.8

wt% 1 2.5 5

0.4 0.2

1.2 1

wt%

0.8

1

0.6

= 2.38 ± 0.199

2.5

0.4

5

0.2 0 0

0 0

40

80

b

8

8

10

1.4

6 wt% 1 2.5 5

4 2

1

wt%

0.8

1

0.6

2.5

0.4

= 2.11 ± 0.0645

0.2 2

4

120

240

360

t (minutes)

8

Fig. 4. Dimensionless plot of experimental data for (a) cesium and (b) strontium.

Fig. 2. Plot showing the loading of (a) cesium and (b) strontium in the zeolite phase as a function of time. The three curves represent different concentrations of fissionchlorides in the salt initially.

tcritical = yeq/mave

y = mavet

0.8

y = yeq 0.4

0.2

5.0wt% CsCl 0.0 10

20

Table 1 List of calculated values based on experimental data. CsCl

SrCl2

wt.%

1.0

2.5

5.0

1.0

2.5

5.0

tcritical (s) teq (s) yeq (mol equiv./unit cell) yeq/teq  105

873 1500 0.134 8.93

992 2700 0.289 10.7

836 1350 0.533 39.5

3130 5580 0.321 5.75

3350 6000 1.43 23.8

3760 7200 5.04 70.0

removal by zeolite from the molten salt has been performed as an analogy to that of aqueous systems [14]. A form of the pseudofirst- and pseudo-second-order Lagergren sorption models was adapted for this study [15,16]. It should be noted that the utility of the sorption models is general. The models can be applied to a variety of systems allowing for possible comparison to aqueous research and providing additional analyses of the kinetics for the molten salt system. The differential representation of the first-order equation is:

0.6

0

6

t/tcritical

0 0

5

0 0

yCs (mol eq/unit cell)

6

1.2

ySr(t)/yeq

ySr (mol eq/unit cell)

4

t/tcritical

t (minutes)

b

2

120

30

dyi ¼ k1 ðyeq  yi Þ dt

t (minutes) Fig. 3. Scheme for calculating critical time.

ð3Þ

and that of the second-order equation is:

3.2. Proposed models

dyi ¼ k2 ðyeq  yi Þ2 dt

Previous studies utilize an equilibrium model, which assumes the following reactions occur between the salt and the zeolite phases [6–9]:

where k1 (s1) and k2 (unit cell (equiv. s)-1) are the first- and secondorder sorption rate constants. Integration and rearrangement of Eqs. (3) and (4) appear, respectively, in linear form as

 nþ   nþ   þ  þ Ai i þ ni Li z Ai i þ ni Li s

logðyeq  yi Þ ¼ logðyeq Þ 

s

z

ð2Þ

ð4Þ

k1 t 2:303

ð5Þ

and where Ai is the active metal of interest, ni is the ionic charge of species i, and subscripts s and z denote the salt and zeolite phases. This approach is valid only under the conditions of equilibrium, but not the kinetics. Thus, the sorption and diffusion processes have been considered in this study. Since the kinematic viscosities of water and LiCl–KCl at high temperature are similar, modeling sorption rates of fission product

t 1 1 ¼ þ t yi k2 y2eq yeq

ð6Þ

An equation for modeling diffusion into a sphere was also considered [17] and has been used previously in zeolite studies [14,18]. The differential equation for radial diffusion into a sphere, which has been adapted for this study, is:

188

M. Shaltry et al. / Microporous and Mesoporous Materials 152 (2012) 185–189

0.6

ð7Þ yCs (mol eq/unit cell)

@yi @ 2 yi 2 @yi þ ¼D r @r @t @r2

!

where D is the diffusion coefficient (m2 s1) and r is the radius of the sphere (m). The series solution of Eq. (7) in dimensionless form is

FðtÞ ¼

1 yi ðtÞ 6 X 1 2 ¼1 2 expðj BtÞ yeq p j¼1 j2

ð8Þ

where B = p2D/R2, R is the radius of the zeolite bead (m), and values for j are integers.

(a)

1.0 wt% 2.5 wt%

0.4

5.0 wt% 0.2 2nd Order Sorption 0.0 0

40 80 t (minutes)

3.3. Modeling analysis

6.0

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u  m  u1 X ymeasured  ypredicted 2 RMSD ¼ 100%t m l¼1 ymeasured

ySr (mol eq/unit cell)

A root mean square deviation (RMSD) was calculated to gauge the accuracy of each model (predicted values) with reference to experimental data sets (measured values). Having a value from 0% to 100%, the RMSD is defined as,

ð9Þ

where m is the number of data points and l is the summation index of integer values. A model is most reliable when the RMSD is near or equal to zero.The first-order sorption model (Eq. (5)) was compared to the data sets. From a plot of log (yeq  yt) versus t, values for k1 and yeq (predicted) were calculated. The RMSD values were found to be in the range of 73%–98%, indicating a very poor fit to the data. These results have been omitted. Comparison of the data to the second-order sorption model (Eq. (6)) gave better results. From a plot of t/yt versus t, k2 and yeq (predicted) were calculated, again allowing calculation of RMSD values. The second-order sorption model fits the data well compared to the first order model. The second-order model also predicts values for yeq, which are very close to those measured. Plotting the diffusion model was accomplished by taking the first ten terms of the summation in Eq. (8). Fig. 5 shows a plot of the percent relative difference (PRD) versus Bt based on the diffusion equation. PRD is calculated by:

  F jþ1  F 1 PRD ¼ 100% F jþ1

ð10Þ

and indicates the relative accuracy of the model among values of j. The plot indicates that in the region of interest for the Cs and Sr experiments (Bt = 0.381, . . ., 16.0), 10 terms of the summation is fully sufficient to accurately represent the experimental data.

(b)

2.5 wt%

4.0

5.0 wt% 2.0 2nd Order Sorption 0

90

180 270 t (minutes)

360

The diffusion model is comparable in accuracy to that of the second-order sorption model yet displays the lowest RMSD values. Fig. 6 shows the second-order sorption and diffusion models plotted with the experimental data. Table 2 lists the values for second-order rate constant, diffusion coefficient, equilibrium loading, and RMSD based on fitting the previously discussed models to the experimental data. Calculated diffusion coefficients are shown to vary (though they have the same order of magnitude) for the different concentrations of the fission-chlorides tested. This is anomalous considering that in general the diffusion coefficient should be independent of concentration. It is likely this is due to an over-simplification of the observed mass-transfer mechanism(s), and study of this is ongoing. This aside, average values of the diffusion coefficient were calculated to be 2.0  1010 m2 s1 for Cs and 6.3  1011 m2 s1 for Sr. It is important to point out that the zeolite bead radius was used rather than the zeolite crystal radius for calculating the diffusion

j 1 2 3 4 5 6 7 8 9

20 15

Region of Interest PRD

30

10 5

15 0 0.01

0.1

1 Bt

10

Diffusion

Fig. 6. Comparison of experimental data and model generated curves for (a) CsCl experiments and (b) SrCl2 experiments.

60 45

Diffusion

1.0 wt%

0.0

Diffusion Model Accuracy PRD, (Fj+1-Fj)/Fj+1 100%

120

0 0.01

0.1

1 Bt

Fig. 5. Percent relative difference (PRD) versus Bt, displaying the relative accuracy of the diffusion model among values of j.

M. Shaltry et al. / Microporous and Mesoporous Materials 152 (2012) 185–189

support the assumptions that the sorption process in diffusionlimited across the bead.

Table 2 List of model parameters and calculated values. CsCl wt.%

1.0

SrCl2 2.5

5.0

1.0

2.5

5.0

Second-order sorption k2 (unit cell/ 1.03 0.217 0.526 3.25  103 1.05  103 1.47  104 equiv s1) yeq (predicted) 0.150 0.351 0.566 0.344 1.50 5.45 RMSD (%) 12.8 18.4 7.50 11.0 4.09 7.81 Diffusion D  1010 (m2 s1) 1.81 RMSD (%) 9.28

189

1.47 15.5

2.58 6.10

0.836 8.88

0.675 4.15

0.385 6.40

4. Conclusions A series of ion exchange experiments were performed using zeolite-A to remove CsCl or SrCl2 from molten LiCl–KCl. This ion exchange process has application to treating used nuclear fuel via electrorefinement. Molten salt within the electrorefiners becomes contaminated with Cs, Sr, and other fission products and must be purified periodically. While the results of extensive studies on the equilibrium conditions between molten salt and zeolite have been published, little information was previously available on the kinetics of this process. Such information is critical for process parameters and optimal design of equipment. In the tests reported here, CsCl and SrCl2 concentrations were varied from 1.0 wt.% to 5.0 wt.% in the initial salt. Frequent sampling and analysis of zeolite beads indicated that equilibrium loading was reached in approximately 30 min for Cs ion exchange, while for Sr approximately 100 min was required. Three models were derived to explain the observed kinetics. RMSD values were calculated and it was found that the intraparticle diffusion-limited model provided the best fit to the data. Optimal diffusion coefficients varied as a function of initial concentration, but not dramatically. Comparison to literature values [5,19] and XRF images confirm the use of the current modeling approach on calculating the diffusion coefficient. The average diffusion coefficients were found to be 2.0  1010 m2 s1 for Cs and 6.3  1011 m2 s1 for Sr. Acknowledgements

Fig. 7. XRF images of the interior cross-section of a sample bead at 5 min during the 5.0 wt.% Cs experiment (upper image) and at 60 min during the 5.0 wt.% Sr experiment (lower image).

coefficient. This is based on the assumption that the time scale for diffusion across the bead (1 - 2 mm in diameter) is substantially larger than across the crystals (1 - 2 micron in diameter). Previous studies [5,19] made the same assumption while estimating diffusion coefficients based on initial rate of uptake. Their calculated diffusion coefficients are consistent with those reported here. The larger diffusion coefficient for Cs compared to Sr in the molten salt is reasonable considering the mobility of the cations which are complexed with the stoichiometric number of Cl anions (one for Cs+ and two for Sr2+). This relationship between cationic valence and diffusion coefficient has also been observed and reported by Lexa [19]. The values presented in this study are also consistent with typical liquid-state diffusion in a porous solid, as tortuosity can bring the effective diffusion coefficient down by a factor of 10 or more [20]. A plot of Bt versus t was used to determine whether the process was dominated by film or particle diffusion phenomena. Of the experiments, all plots (not shown) are linear and pass through the origin, which indicates a particle diffusion mechanism. In addition, X-ray fluorescence (XRF) analysis was performed on beads that were sectioned in half to validate the concept of using the diffusion model. An XRF image of the interior cross-section of the bead indicates how the molten salt diffused into the bead. Fig. 7 shows XRF images of a sample acquired 5 min into the 5 wt.% Cs experiment and another at 60 min into the 5.0 wt.% Sr experiment. Here, the white pixels of the images indicate the presence of either Cs or Sr. The spatial distribution of Cs and Sr in the beads

This research was supported by the Center for Advanced Energy Studies and the University of Idaho. Funding was provided by the United States Department of Energy’s Office of Nuclear Energy under the Fuel Cycle Research and Development Program and through the Pyroprocessing Technology Department at Idaho National Laboratory. The author would like to give special thanks to Joanna Taylor with the University of Idaho for ICP–MS analyses and laboratory support. References [1] J.J. Laidler, J.E. Battles, W.E. Miller, J.P. Ackerman, E.L. Carls, Prog. Nucl. Energy 31 (1997) 131–140. [2] R.W. Benedict, H.F. McFarlane, Rad. Waste Mag. (1998) 23–30. [3] J.L. Willit, W.E. Miller, J.E. Battles, J. Nucl. Mater. 195 (1992) 229–249. [4] R.W. Benedict;C. Solbrig;B. Westphal;T.A. Johnson;S.X. Li;K. Marsden, K.M. Goff, Proceedings of Global, 9–13 September; Boise; Idaho, USA, 2007. [5] D. Lexa, I. Johnson, Metall. Mater. Trans. B 32B (2001) 429–435. [6] Tae-Sic Yoo, S.M. Frank, M.F. Simspon, P.A. Hahn, T.J. Battisti, S. Phongikaroon, Nucl. Technol. 171 (2010) 306–315. [7] M.F. Simpson, M.L. Dunzik-Gougar, Ind. Eng. Chem. Res. 42 (2003) 4208–4212. [8] M.L. Dunzik-Gougar, M.F. Simpson, B.E. Scheetz, Micropor. Mesopor. Mater. 84 (2005) 366–372. [9] S. Phongikaroon, M.F. Simpson, AIChE J. 52 (2006) 1736–1743. [10] R.K. Ahluwalia, H.K. Geyer, C. Pereira, J.P. Ackerman, Ind. Eng. Chem. Res. 37 (1998) 145–153. [11] J.G. Sen Gupta, J.L. Bouvier, Talanta 42 (1995) 269–281. [12] J.G. Sen Gupta, N.B. Bertrand, Talanta 42 (1995) 1595–1607. [13] J.G. Sen Gupta, N.B. Bertrand, Talanta 42 (1995) 1947–1957. [14] A.M. El-Kamash, J. Hazard. Mater. 151 (2008) 432–445. [15] Y.S. Ho, G. McKay, Process Biochem. 34 (1999) 451–465. [16] S. Lagergren, Zur theorie der sogenannten adsorption geloster stoffe. Kungliga Svenska Vetenskapsakademiens, Handlingar 24 (1898) 1–39. [17] J. Crank, The Mathematics of Diffusion, first ed., Oxford University Press Inc., 1956. [18] R.M. Barrer, S. Barri, J. Klinowski, J. Chem. Soc., Faraday Trans. I (1980) 1038– 1051. [19] D. Lexa, Metall. Mater. Trans. B 34B (2003) 201–208. [20] M.F. Simpson, J. Wei, S. Sundaresan, Ind. Eng. Chem. Res. 35 (1996) 3861–3873.