experimental productivity rate optimization of rare earth element separation through preparative...

C

Esc

HD

a

ARRAA

KRCOMHN

1

mmceliiscl

mfpaia[

h0

ARTICLE IN PRESSG ModelHROMA-355377; No. of Pages 5

Journal of Chromatography A, xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Journal of Chromatography A

j o ur na l ho me page: www.elsev ier .com/ locate /chroma

xperimental productivity rate optimization of rare earth elementeparation through preparative solid phase extractionhromatography

ans-Kristian Knutson, Mark Max-Hansen, Christian Jönsson, Niklas Borg, Bernt Nilsson ∗

epartment of Chemical Engineering, Centre for Chemistry and Chemical Engineering, Lund University, P.O. Box 124, SE-221 00 Lund, Sweden

r t i c l e i n f o

rticle history:eceived 16 December 2013eceived in revised form 17 February 2014ccepted 25 April 2014vailable online xxx

a b s t r a c t

Separating individual rare earth elements from a complex mixture with several elements is difficult andthis is emphasized for the middle elements: Samarium, Europium and Gadolinium. In this study wehave accomplished an overloaded one-step separation of these rare earth elements through prepara-tive ion-exchange high-performance liquid chromatography with an bis (2-ethylhexyl) phosphoric acidimpregnated column and nitric acid as eluent. An inductively coupled plasma mass spectrometry unitwas used for post column element detection. The main focus was to optimize the productivity rate, sub-

eywords:are earth elementshromatographyptimizationonaziteDEHP

ject to a yield requirement of 80% and a purity requirement of 99% for each element, by varying the flowrate and batch load size. The optimal productivity rate in this study was 1.32 kg Samarium/(h m3

column),0.38 kg Europium/(h m3

column) and 0.81 kg Gadolinium/(h m3column).

© 2014 Elsevier B.V. All rights reserved.

itric acid

. Introduction

Rare earth elements (REE) are important components of manyodern technological products [1]. They occur in many types ofinerals. These minerals normally contain all REEs with varying

oncentrations of each [1,2]. The minerals are processed throughxtraction methods [2,3] and must be upgraded to high purityevels before being used for commercial purposes [2]. Separatingndividual REEs from a complex mixture with several elementss difficult [1,3,4], and achieving a high productivity rate for theeparation process is problematic since large feed loads with highoncentration results in difficulties with reaching sufficient purityevels.

The current industry standard is to employ liquid extractionethods due to their ability to handle large and highly concentrated

eeds, and achieve purity levels above 90% [1–3]. Chromatogra-hy, as an alternative method, has the benefit of being able tochieve even higher purity levels. The expenditure of extractantss less demanding than for liquid extraction, and chromatography

Please cite this article in press as: H.-K. Knutson, et al., J. Chromatogr.

lso offers possibilities of recovery and recycling of process media2]. Furthermore, liquid extraction requires several process steps

∗ Corresponding author. Tel.: +46 046 2228088; fax: +46 046 2224526.E-mail address: [email protected] (B. Nilsson).

ttp://dx.doi.org/10.1016/j.chroma.2014.04.085021-9673/© 2014 Elsevier B.V. All rights reserved.

whereas it is possible to reduce the separation to a single stepthrough chromatography [2,5–7].

Since there are many apparent benefits for chromatography itwould be of interest to determine if it is a commercially feasible sep-aration method compared to liquid extraction. However, the detailsfor either method are usually not disclosed in publication [2,3]. Forthis reason we have focused on finding the optimal chromatogra-phy operation point, in terms of productivity, for a REE separationcase including the middle REEs: Samarium (Sm), Europium (Eu)and Gadolinium (Gd), which are particularly difficult to separate[4,8–10].

REE chromatography utilizes the differences in affinity elementshave for a ligand to separate target elements from other elements[11,12]. The affinity will decide the order of elution, and the degreeof separation between the elements can be controlled by adjus-ting the operating conditions. This includes changing mobile phaseproperties, such as media type and concentration, and columnproperties such as length, porosity and bead particle- and pore-size.The retention time of each element will also be decided by temper-ature, batch load size and composition, ligand concentration, flowrate and elution-gradient length [12,13].

Previous work has shown that bis (2-ethylhexyl) phosphoric

A (2014), http://dx.doi.org/10.1016/j.chroma.2014.04.085

acid (HDEHP) is a suitable extractant for liquid extraction of REEs[2,3]. HDEHP also makes it possible to separate all REEs in a singlestep since its affinity for a REE increases with atomic number. It hasalso been shown in analytical REE chromatography studies that

dx.doi.org/10.1016/j.chroma.2014.04.085


http://www.sciencedirect.com/science/journal/00219673

http://www.elsevier.com/locate/chroma

mailto:[email protected]


ARTICLE IN PRESSG ModelCHROMA-355377; No. of Pages 5

2 H.-K. Knutson et al. / J. Chromatogr. A xxx (2014) xxx–xxx

Table 1Decision variables.

Decision variable Lower boundary Upper boundary

Hm[[acptpwr

tRcpwM

2

2

ogcpbkmr

2

g

P

woV

Y

wp

2

Table 2Composition of the mixture used in the experiments.

TSe

Flow rate (ml/min) 0.25 0.75Load (�l) 150 220

DEHP is a suitable ligand which enables separation with differentobile phases such as nitric acid [5,8,9,14,15], hydrochloric acid

8,15], and ˛-hydroxyisobuturic acid (˛-HIBA) [16]. Recent work5,6,17] has shown that it is possible to model both analyticalnd overloaded HPLC chromatography of REEs under demandingonditions with an acid involved. They have also highlighted theotential for chromatography as a large scale separation methodhrough computer simulations. However, experimental studies ofroductivity optimization for overloaded HPLC chromatographyith HDEHP as ligand and nitric acid as mobile phase has not been

eported.In this study we have experimentally shown that it is possible

o accomplish a demanding and overloaded one-step separation ofEEs through preparative ion-exchange high-performance liquidhromatography (IE-HPLC), and provided data regarding expectederformance for chromatography as a REE processing method. Thisas done by focusing on finding the optimal operating point for aonazite middle REE mixture containing Sm, Eu, Gd and impurities.

. Theory

.1. Optimization

When optimizing a process step it is necessary to define thebjective function in order to clarify what the optimization tar-et is. The objective function will depend on several variables thatan be divided into two groups, the decision variables and fixedarameters. The decision variables comprise the conditions that areeing altered during the optimization, and the fixed parameters areept constant. Finally, some constraints are normally introduced toake sure that the optimization results remain within a feasible

egion [18].

.1.1. Objective functionThe objective function is the productivity, P, of component i as

iven in Eq. (1)

i = LiYi

tcVcol(1)

here the load, L, is defined as the product of the feed concentrationf component i and the feed volume, tc is the total cycle time andcol is the total column volume.

The yield, Y, of component i is defined in Eq. (2)

i = cpool,iVpool,i

Li(2)

here cpool,i is the product pool concentration and Vpool,i is the


roduct pool volume of component i.

.1.2. Decision variablesThe decision variables are presented in Table 1.

able 3ummary of the experimental conditions. Experiments 1–3 were used in the flow optimixperiments 2, 5, 7 and 8 were used in the scale-up study.

Experiment 1 2 3

Flow rate (ml/min) 0.25 0.50 0.75

Load (�l) 180 180 180

Column volume (ml) 2.49 2.49 2.49

REE Nd Sm Eu Gd Tb

%(wt) 4.6 58.2 12.0 24.3 0.9

2.1.3. ConstraintsA purity above 99% was required since this is a common com-

mercial purity grade [2], and it was decided that an 80% yield foreach component was necessary to avoid excessive waste.

3. Materials and methods

An Agilent 1200 series HPLC system (Agilent Technologies,Waldbronn, Germany) was used together with two differentcolumns, Kromasil M3 and Kromasil H4 (Eka, Bohus, Sweden). Thecolumns were delivered as is with a stationary phase of spheri-cal silica particles coated with C18, a diameter of 16 �m and a poresize of 100 A. HDEHP was used as ligand due to its versatile ability toseparate REE [2,3,14,16,19] and each column was filled with HDEHP(Sigma–Aldrich, St. Louis, USA) to a concentration of 342 mM. Nitricacid was used as eluent, and the elution concentration gradientwas varied between 6 and 13 vol.% of 7 M acid. The length of theelution gradient was set to 5 column volumes in order to avoiddiluted product pool concentrations while still enabling sufficientseparation. Each elution was followed by a regeneration step of2.5 column volumes 7 M nitric acid and a equilibration step of 2.5column volumes water. An inductively coupled plasma mass spec-trometry (ICP-MS) system (Agilent Technologies, Tokyo, Japan) wasused for in-line post column REE detection due to its documentedcapability for this purpose [1,14,20].

3.1. Experimental study

The REE composition in this study is an approximation of a REEmixture from a Monazite ore [21] that has been pre-treated to iso-late the Sm, Eu and Gd (SEG) part [2]. Neodymium (Nd) and Terbium(Tb) were introduced to make sure that other REE impurities wouldnot interfere with the objective of producing pure SEG pools. Nd andTb are specifically suitable for this purpose since Nd precedes Smand Tb follows Gd in terms of affinity for HDEHP. The experimentfeed composition is given in Table 2. The mixture was dissolvedin nitric acid to reach a lanthanide concentration of 32.56 g/l anda final pH level of 1.51. The two variables that were investigatedwere the flow rate and the batch load size. A smaller scale up trialof close to optimal operating point conditions was also conducted.A summary of the experimental conditions is given in Table 3.

4. Results and discussion

4.1. Optimal flow rate

The load was set to 180 �l and the flow rate was varied. Fig. 1 and


Table 4 show how the production rate, yield and pool concentrationof each component varied with an increased flow rate. It can be seenthat the best yield was achieved for the lowest flow rate and that itworsened with increasing flow rate. We believe that this is due to

zation study, experiments 2 and 4–6 were used in the load optimization study, and

4 5 6 7 8

0.50 0.50 0.50 0.835 0.835150 200 220 300 333

2.49 2.49 2.49 4.15 4.15



H.-K. Knutson et al. / J. Chromatogr. A xxx (2014) xxx–xxx 3

0.2 0.3 0.4 0.5 0.6 0.7 0.80

0.5

1

1.5

2

2.5

Pro

duct

ivit

y(k

g/m

3h)

0.2 0.3 0.4 0.5 0.6 0.7 0.80

25

50

75

100

Flow rate (m l/m in)

Yie

ld(%

)

SmEuGd

F onstanp o the

ai

rrStt

4

Talmww

TRttprT

point, it can be argued that a 180 �l load is better than a 150 �l loadsince the pool concentrations are improved with a marginal loss ofproductivity.

1.5m3h)

ig. 1. The productivity and yield of Sm, Eu and Gd for varying flow rates with a croductivity increases with the flow rate until the decrease in yield is detrimental t

peak broadening effect caused by the decreasing retention timen the column for the higher flow rates.

An increased flow rate will result in an increased productionate, but with a decreasing yield as trade off. Eventually the flowate becomes too high to meet the minimum 80% yield requirement.ince a 0.5 ml/min flow rate gave an 84% yield for Eu, we concludedhat this flow rate was very close to the limit and decided to declarehis as optimal flow rate.

.2. Optimal load

The load was varied for a given flow rate of 0.5 ml/min. Fig. 2 andable 5 show the productivity, yield and pool concentration vari-tion for each component. The production rate increased with the


oad, but the yield decreased. Only the 150 �l and 180 �l load caseset the 80% yield requirement. Out of these two operating pointse achieve the highest production rate for the 150 �l case. Thereould be no point in investigating the effect of a lower load since

able 4esults from the flow rate experiments showing that with an increased flow rate,he yield decreases and the productivity increases until the yield becomes so lowhat it is detrimental to the productivity. This is accentuated for Eu. The optimalroductivity, with respect to the minimum yield constraint, was achieved at a flowate of 0.5 ml/min. The experimental conditions correspond to experiments 1–3 inable 3.

Flow rate (ml/min) Prod (kg/(h m3column

)) Yield (%) cpool (kg/m3)

0.250.66 Sm 99.7 Sm 0.64 Sm0.19 Eu 98.4 Eu 0.35 Eu0.41 Gd 99.6 Gd 0.46 Gd



t batch load. It can be seen that the yield decreases with increased flow rate, andproduction rate.

a base line separation (close to 100% yield for all components) isachieved at this point, and a lower load would automatically resultin a lower production rate. Therefore, 150 �l load should be con-sidered as optimal. The chromatogram for this operation point canbe seen in Fig. 3, and the performance is given in Table 5.

However, it should be noted that the pool concentrationincreases with a higher load and we expect that this will becomean important factor for a complete process design. From this view-


140 150 160 170 180 190 200 210 220 2300

0.5

1

Pro

duct

ivit

y(k

g/

140 150 160 170 180 190 200 210 220 2300

25

50

75

100

Batch loa d (µl)

Yie

ld(%

)

SmEuGd

Fig. 2. The productivity and yield of Sm, Eu and Gd for a constant flow rate withvarying batch loads. It can be seen that the yield decreases with an increased load,and productivity increases with the load until the decrease in yield is unfavourableto the production rate.




4 H.-K. Knutson et al. / J. Chromatogr. A xxx (2014) xxx–xxx

Table 5Results from the load experiments showing that with an increased batch load, the yield decreases and the productivity increases until the yield becomes so low that itnegatively affects the productivity. The optimal productivity was achieved at 150 �l load. The experimental conditions correspond to experiments 4, 2, 5 and 6 in Table 3.

Load (�l) Prod (kg/(h m3column

)) Yield (%) cpool (kg/m3)

1501.32 Sm 99.9 Sm 0.55 Sm0.38 Eu 95.5 Eu 0.32 Eu0.81 Gd 99.0 Gd 0.35 Gd



2201.48 Sm 90.25 Eu 50.99 Gd 9

0 1 2 3 4 5 6 70

0.5

1

1.5

Con

cent

rati

on(m

g/m

l)

Volum e (VColumn)

NdSmEuGdTb

Fig. 3. Chromatogram for the optimal operation point with 150 �l load and0.5 ml/min flow rate. The elution order is Nd, Sm, Eu, Gd and Tb.

160 180 200 220 2400

0.25

0.5

0.75

1

1.25

1.5

Batch

Pro

duct

ivity

(kg/

m3h)

70 75 80 850

0.25

0.5

0.75

1

1.25

1.5

Normalize d batch load (µl load/ml column)

Pro

duct

ivity

(kg/

m3h)

Fig. 4. Comparison of productivity between different column sizes. Residence time andand the load for the larger column. It can be seen that productivity rates were intact whexperiments 2, 5, 7 and 8 in Table 3.

6.7 Sm 1.09 Sm1.9 Eu 0.61 Eu1.4 Gd 0.41 Gd

For all the operating points, a yield above 90% was achieved forSm and Gd. This indicates that the actual challenge is to achieve agood yield for Eu.

4.3. Scale-up experiment

A test was carried out to see if it was possible to linearly scaleup operation points from the 150 mm long Kromasil M3 column tothe larger 250 mm long Kromasil H4 column. The inner diameter,stationary phase properties and ligand density were the same forboth columns. The flow rate was increased from 0.5 ml/min for the150 mm column to 0.835 ml/min for the 250 mm column to main-tain a constant residence time. The 180 �l and 200 �l loads for the150 mm column were compared with 300 �l and 333 �l loads forthe 250 mm column to keep the load versus column volume ratioconstant.

Fig. 4 shows each components productivity rate for each columnand load case. The smaller plot also shows the productivity for the


same experiments but with the load normalized against the volumeof each column to emphasize that identical productivity rates perm3 column were achieved when the system was linearly increased.This indicates that the production rate can be linearly increased by

260 280 300 320 340loa d (µl)

Sm on 150 mm columnEu on 150 mm columnGd on 15 0 mm columnSm on 250 mm columnEu on 250 mm columnGd on 25 0 mm column

batch load versus column volume are kept constant by increasing the flow rateen the system was linearly increased. The experimental conditions correspond to


ING ModelC

romato

ip

5

Scapbtl

ddppippwtwtr

bopspcotp

[[

[

[

[[[

[

[

ARTICLEHROMA-355377; No. of Pages 5

H.-K. Knutson et al. / J. Ch

ncreasing the column volume, and that process scale up should beossible.

. Conclusions

It was possible to achieve a difficult one-step purification of aEG mix with other REE impurities through HPLC ion-exchangehromatography with a high load. We have also demonstrated that

process scale-up could be possible, and given that chromatogra-hy is a proven large scale production method [11–13] this shoulde considered as achievable. However, it is difficult to benchmarkhe process performance against liquid extraction methods due toack of process data specifics.

The process objective in this study was to maximize the REE pro-uctivity. However, our results indicate that this is not a sufficientlyefined objective when considering a comprehensive purificationrocess chain. The productivity rate and yield are related, and it isossible to find an optimal relation between the two by produc-

ng a Pareto front [5,6]. Nonetheless, this approach will neglect theroduct pool concentration which will have an impact on the totalroduction cost due to down stream process steps. For this reason,e propose that pool concentration should be a part of the objec-

ive function for process optimizations. Although we believe that itill be difficult to define such an objective function without turning

o financial specifics of an extensive purification process. Anotheroute could be to decide a minimum pool concentration.

There is potential to improve the process performance furthery investigating other parameters that were not within the scopef this study, such as particle diameter and pore size. It is alsoossible to improve the performance by utilizing other processchemes such as multi column counter current solvent gradienturification (MCSGP) that has shown an improved performance


ompared to batch production [17]. We believe that further studiesf alternative ligand types and choosing optimal ligand concen-ration will have the highest impact on improving the processerformance.

[

[[

PRESSgr. A xxx (2014) xxx–xxx 5

Acknowledgements

This study has been performed within Process Industry Centre atLund University organized by the Swedish Foundation of StrategicResearch. K.A. Rasmussen AS, Hamar, Norway is acknowledged forfinancial support.

References

[1] S.B. Castor, J.B. Hedrick, Ind. Miner. Soc. Min., Metall., Explor., 7th ed., Littleton,Colorado, 2006, p. 769.

[2] C.K. Gupta, N. Krishnamurthy, Extractive Metallurgy of Rare Earths, CRC Press,Boca Raton, 2004.

[3] N. Thakur, Miner. Process. Extr. Metall. Rev. 21 (2000) 277.[4] S. Siekierski, I. Fidelis, in: T. Braun, G. Ghersini (Eds.), Extraction Chromatogra-

phy, vol. 2 of J. Chromatogr. Libr., Elsevier, Budapestn, 1975, p. 226.[5] F. Ojala, M. Max-Hansen, D. Kifle, N. Borg, B. Nilsson, J. Chromatogr. A 1220

(2012) 21.[6] M. Max-Hansen, F. Ojala, D. Kifle, N. Borg, B. Nilsson, J. Chromatogr. A 1218

(2011) 9155.[7] D. Kifle, G. Wibetoe, M. Frøseth, J. Bigelius, Solvent Extr. Ion Exch. 31 (2013)

668.[8] R.J. Sochacka, S. Siekierski, J. Chromatogr. 16 (1964) 376.[9] T.J. Ireland, F.L.H. Tissot, R. Yokochi, N. Dauphas, Chem. Geol. 357 (2013) 203.10] T.B. Pierce, P.F. Peck, Analyst 88 (1963) 217.11] G. Carta, A. Jungbauer, Protein Chromatography: Process Development and

Scale-up, John Wiley & Sons, Weinheim, 2010.12] D.G. Shirazi, A. Felinger, A.M. Katti, Fundamentals of Preparative and Nonlinear

Chromatography, Academic Press, Amsterdam, 2006.13] H. Schmidt-Traub, M. Schulte, A. Seidel-Morgenstern, Preparative Chromatog-

raphy, John Wiley & Sons, Weinheim, 2012.14] C. Hernández González, A.J.Q. Cabezas, M.F. Díaz, Talanta 68 (2005) 47.15] S. Siekierski, R.J. Sochacka, J. Chromatogr. 16 (1964) 385.16] N. Sivaraman, R. Kumar, S. Subramaniam, P.R. Vasudeva Rao, J. Radioanal. Nucl.

Chem. 252 (2002) 491.17] M. Krattli, T. Muller-Spath, N. Ulmer, G. Strohlein, M. Morbidelli, Ind. Eng. Chem.

Res. 52 (26) (2013) 8880–8886.18] J.A. Snyman, Practical Mathematical Optimization: An Introduction to Basic

Optimization Theory and Classical and New Gradient-Based Algorithms, vol.


97, Springer, New York, 2005.19] M. Rehkämper, M. Gärtner, S.J.G. Galer, S.L. Goldstein, Chem. Geol. 129 (1996)

201.20] D.S. Braverman, J. Anal. At. Spectrom. 7 (1992) 43.21] X.K. Zhu, R.K. O’Nions, Contrib. Mineral. Petrol. 137 (1999) 351.


http://refhub.elsevier.com/S0021-9673(14)00689-X/sbref0005












































































































































































































































































































experimental productivity rate optimization of rare earth element separation through preparative...

Documents