You are on page 1of 103

Lithium Extraction from Brines Using Ion Concentration Polarization

By
Alex Barksdale

B.S. in Electrical Engineering (2017)


Stanford University

Submitted to the Department of Electrical Engineering and Computer Science in Partial


Fulfillment of the Requirements for the Degree of

Master of Science in Electrical Engineering and Computer Science


At the
Massachusetts Institute of Technology
May 2020

© 2020 Massachusetts Institute of Technology. All rights reserved.

The author hereby grants MIT permission to reproduce and to distribute publicly paper and
electronic copies of this thesis document in whole or in part in any medium now known or
hereafter created.

Signature of Author…………………………………………………………………………………
Alex Barksdale
Author, Department of Electrical Engineering and Computer Science
March 13, 2020

Certified by…………………………………………………………………………………………
Jongyoon Han
Professor of Electrical Engineering and Biological Engineering
Thesis Supervisor

Accepted by………………………………………………………………………………………...
Leslie A. Kolodziejski
Professor of Electrical Engineering and Computer Science
Chair, Department Committee on Graduate Students
2
Lithium Extraction from Brines Using Ion Concentration Polarization
By
Alex Barksdale
B.S. in Electrical Engineering (2017)
Stanford University
Submitted to the Department of Electrical Engineering and Computer Science on
March 13, 2020 in Partial Fulfillment of the
Requirements for the Degree of Master of Science in Electrical Engineering and Computer
Science at the Massachusetts Institute of Technology
ABSTRACT
The surge of electric vehicle deployment in response to the global climate crisis has marked
tremendous increase in the demand of lithium ion batteries. Lithium brine has gained much
attention as a critical primary lithium resource over mineral sources due to cheaper processing and
greater abundance. Although lithium brine resources are estimated to account for over 60% of
global lithium resources, lithium extraction from brine remains a challenge, as high magnesium
concentration in lithium brine severely reduces the purity of precipitated lithium carbonate from
brine. Thus, the Mg2+:Li+ mass ratio is an important metric when considering the economic
viability of brine. Because lithium ion batteries require >99.5% purity of lithium carbonate, a
technique for selectively reducing the magnesium content in brine while retaining lithium is
desired.
Here we present a continuous, scalable ion concentration polarization based technique for
reducing the Mg2+:Li+ ratio of brine. The device utilizes ion concentration polarization, a
fundamental phenomenon occurring at the interface of ion exchange membranes, to induce a
locally amplified electric field when a potential is applied across the membrane system. The
amplified electric field separates ions into streams according to electrophoretic mobility, which
can be separately collected and analyzed.
In this study, we demonstrate a proof of concept continuous platform to achieve reduction
of the Mg2+:Li+ of brine, over a variety of flow and initial Mg2+:Li+ compositions of a lab brine.
We demonstrate reduction of a 25:1 and 60:1 Mg2+:Li+ brine to below 10:1 Mg2+:Li+, which is
identified by industrial standards to allow adequate precipitation of high purity lithium carbonate
for lithium ion battery production. We also demonstrate reduction of a 100:1 Mg2+:Li+ brine to
approximately 20:1, despite the short active membrane length used in the experiments. We then
provide a techno-economic analysis of the platform, developing a method for comparing the
continuous ICP device, with prevalent batch mode membrane technologies in the literature. By
accounting for longer effective membrane length, to which energy consumption is well known
through experiment and theory to be inversely proportional to, a fair energy consumption
comparison can be made. Ultimately, this work demonstrates a potentially powerful tool for
continuous lithium extraction from brines, and possibly a generalizable ion separation tool utilizing
differences in electrophoretic mobility.

Thesis Supervisor: Jongyoon Han


Title: Professor of Electrical Engineering and Biological Engineering

3
4
Acknowledgements

I would like to first thank Prof. Jongyoon Han for giving me the opportunity to study in his
group, and advising me through this project. From joining, he has been extremely supportive of
exploring completely new interests outside of my background. He has always encouraged me to
ask questions, and I have always felt that my own learning through this project was his primary
goal in his mentorship. I cannot thank him enough for deeply impacting my way of thinking and
approach to engineering problems, which I will greatly rely upon in both my near- and long-term
endeavors.

I would like to thank Dr. Junghyo Yoon and Dr. Hyukjin Kwon for their mentorship
throughout all phases of this project. I am also extremely grateful for group members who have
provided so much guidance and experience to me through my various research pursuits, including
Wei Ouyang, Kyungyong Choi, Matthew Flavin, Debbie Yu, Dr. Taehong Kwon, Dr. Hyunkook
Jeon, Prof. Chenhui Peng, Dr. David Collins, Dr. Hyunryul Ryu, Prof. Aniruddh Sarkar, Ching
Ann Tee, and Prof. Xiwei Huang. These group members shared so much of their own experience
to help me find answers and solutions to problems I faced every day in the lab, along with creating
a truly friendly and pleasant lab environment to work in.

I have been quite fortunate to develop many new, deep friendships since studying at MIT,
while maintaining old friendships of the past that have given my experience here so much color. I
would like to express deep gratitude to all of these friends who I hold also as great mentors, near
and far.

Lastly, I am so thankful to have the greatest family I could possibly ask for. Their
unconditional support and love have always given me so much confidence in my decisions and
pursuits. The lessons and values they have inscribed in me have made me who I am today, and I
could not imagine experiencing life without them. I would especially like to thank Cameron, who
I am proud to call my brother, as the passion he shows for his endeavors drives me each day to
seek passion in my own.

This work is made possible by the financial support of the Kuwait-MIT Signature Project
on Brine Desalination.

5
6
Table of Contents

1. Introduction ...........................................................................................................................12
1.1 Growing Demand for Lithium Ion Batteries ..........................................................12
1.2 Primary Resources for Lithium Extraction ............................................................15
1.2.1 Lithium Brine ............................................................................................15
1.2.2 Lithium Containing Minerals ....................................................................17
1.2.3 Lithium Extraction from Sea Water ..........................................................19
1.3 Challenges Associated with Lithium Extraction from Brines ...............................20
1.3.1 Current Industrial Extraction .....................................................................20
1.3.2 Mg2+:Li+ Ratio: A Critical Composition Metric ........................................21
1.3.3 Mg2+:Li+ Separation Techniques in the Literature .....................................22
1.4 Ion Concentration Polarization Based Devices for Lithium Extraction from
Brines ..........................................................................................................................24
1.5 Thesis Scope and Outline ..........................................................................................25
1.6 References ..................................................................................................................26
2 Theoretical Background .......................................................................................................28
2.1 Ion Transport in Electrolytes ...................................................................................28
2.1.1 Diffusive Flux of Particles .........................................................................28
2.1.2 Electrical Migration of Charged Particles in a Potential Field ...................29
2.1.3 Transport by Fluid Convection and Extended Nernst Planck Equation .....29
2.1.4 Complete Model for Ion Transport in a Fluid System ................................30
2.2 Ion Exchange Membranes ........................................................................................32
2.3 Ion Concentration Polarization (ICP) .....................................................................34
2.3.1 ICP as a Fundamental Phenomenon in IEMs .............................................34
2.3.2 Current-Voltage Characteristic Curve .......................................................36
2.3.3 Local Amplification of the Electric Field in the Depletion Zone ..............40
2.4 Ternary Electrolyte Concentration Profiles in a 1D AEM System .......................42
2.5 Microfluidic Simulations of Li Extraction from Brines Using ICP .......................48
2.6 Concept Behind the Continuous, ICP Based Li Extraction System ......................49
2.6.1 ICP Desalination System ...........................................................................49
2.6.2 ICP Based Li Extraction System ................................................................50
2.7 References ..................................................................................................................53
3 Experimental Setup .............................................................................................................. 55
3.1 Methods and Materials .............................................................................................55
3.1.1 ICP Schematic and Device Setup ..............................................................55
3.1.2 Preparation of Feed Brines and Electrode Rinse Solution ........................56

7
3.2 Device Operation .......................................................................................................57
3.3 Analytical Methods ....................................................................................................63
3.4 Data Analysis Metrics ...............................................................................................64
3.4.1 Separation Factor .......................................................................................64
3.4.2 Li Recovery Ratio ......................................................................................64
3.4.3 Specific Energy Consumption ...................................................................65
4 Results and Discussion ..........................................................................................................66
4.1 Effect of Applied Current Density ...........................................................................66
4.2 Effect of Convective Flow .........................................................................................68
4.3 Effect of Brine Composition .....................................................................................70
4.4 Summary of Experimental Results ..........................................................................71
5 Techno-Economic Analysis ...................................................................................................73
5.1 Membrane-based Techniques found in Literature ................................................74
5.1.1 Overview of Electrodialysis and Selective Electrodialysis .......................74
5.1.2 Reported Performance of Selective Electrodialysis in the Literature ........76
5.2 Specific Energy Consumption as a Function of Output Mg2+:Li+ Ratio ..............79
5.3 Energy Consumption in ED Systems as a Model for the ICP Based System .......81
5.4 Scaling the Continuous, ICP Based System for Energy Consumption Comparison
......................................................................................................................................83
5.4.1 Comparison of Batch Mode and Continuous Technologies ......................83
5.4.2 Main Channel Power Consumption ...........................................................85
5.4.3 Scaling the ICP Device, and Comparison with S-ED ................................89
5.5 References ..................................................................................................................90
6 Concluding Remarks and Future Directions ......................................................................91
6.1 Concluding Remarks .................................................................................................91
6.2 Future Directions .......................................................................................................93
6.2.1 Addressing Device Limitations .................................................................93
6.2.2 Generalized Ion Separation Technique .....................................................94
6.3 References ..................................................................................................................96
7 Appendix ................................................................................................................................97
Appendix A: Lab Brine Compositions ...........................................................................97
Appendix B: Source Conditions .....................................................................................98
Appendix C: ICP-OES Analysis and Verification ........................................................99
Appendix D: Scaled Power Consumptions of ICP Device by Effective Length .........103

8
List of Figures

Figure 1.1 USGS Lithium Global End Use Markets ................................................................13


Figure 1.2 Ragone Plot of Battery Formulations .....................................................................13
Figure 1.3 Projected Demand of Lithium Carbonate ...............................................................14
Figure 1.4 World Map of Major Li Brine Deposits ..................................................................16
Figure 1.5 Example Process Flow of Li Brine Processing ........................................................18
Figure 1.6 Image of Li Brine Evaporation Ponds .....................................................................21
Figure 1.7 Selective Electrodialysis Schematic .......................................................................23
Figure 2.1 Ion Exchange Membrane Illustration .....................................................................33
Figure 2.2 Ion Concentration Polarization Illustration .............................................................35
Figure 2.3 Current Voltage Characteristic Curve of IEM Systems ..........................................40
Figure 2.4 H-channel Microfluidic ICP System ......................................................................41
Figure 2.5 1D AEM System in Contact with Ternary Electrolyte Solution ............................42
Figure 2.6 Potential and Concentration profiles for Ternary Electrolyte AEM System .........47
Figure 2.7 ICP Desalination System .......................................................................................50
Figure 2.8 Schematic of Idealized ICP System for Li Extraction from Brine ........................52
Figure 3.1 Schematic of Implemented ICP System ................................................................58
Figure 3.2 Image of Built ICP System ....................................................................................59
Figure 3.3 Image of ICP System Pumping Mechanism ..........................................................60
Figure 3.4 Image of Electromembrane Stack ..........................................................................61
Figure 3.5 Image of Flow Measurement and Output Solution Collection ................................62
Figure 4.1 Results of Applied Current Investigation and Flow Rate Variation .......................67
Figure 4.2 Results of Feed Mg2+:Li+ Variation .........................................................................71
Figure 5.1 Schematic of Electrodialysis System ......................................................................75
Figure 5.2 Specific Energy Consumption vs. Output Mg2+:Li+ Ratio ......................................80
Figure 5.3 Schematic of Main Channel Power Consumption Measurement ...........................86
Figure 5.4 Image of Main Channel Power Consumption Measurement Electrode Placement 82
Figure 5.5 25:1 Brine Main Channel Power Consumption with Flow Variation ....................83
Figure C1 ICP-OES Example Calibration Curves ...................................................................95
Figure C2 Normalized Concentration of Collect and Waste Streams .....................................97

9
List of Tables

Table 1.1 Estimated Lithium Content of Significant Brine Deposits .......................................16


Table 1.2 Chemical Composition of Major Li Mineral Sources ..............................................18
Table 5.1 Summary of Reviewed Selective Electrodialysis Membranes Parameters ..............78
Table 5.2 Summary of Reviewed S-ED Systems in Literature and Results of This Study .....78
Table A1 Ideal and Measured Brine Compositions of Feed Solutions by Salt Mass ..............97
Table B1 Source Conditions for Experiments ..........................................................................98
Table C1 Compositions of ICP-OES Calibration Standards ...................................................99
Table C2 Measured Brine Compositions of Feed Solutions by ICP-OES ..............................101
Table D1 Extended Experimental Conditions Energy Consumption Scaling Comparison ....103

10
11
1. Introduction

1.1 Growing Demand for Lithium Ion Batteries

Lithium is an extraordinarily versatile material with many of its compounds providing a wide

array of uses and applications. Figure 1.1 shows a breakdown of lithium uses from the 2019 US

Geological Survey [1]. Despite the many uses of lithium compounds, lithium for batteries

dominates this breakdown, namely lithium ion batteries (LIBs).

LIBs currently offer the highest performance in terms of gravimetric power and energy

densities among all battery formulations on the market [2]. This particularly suits LIBs for mobile

electronics applications, such as electric vehicles (EVs), where the weight of battery cells plays a

critical role in performance of the vehicles. Figure 1.2 shows a Ragone plot comparing gravimetric

power and energy densities across different battery formulations for automotive applications [2].

Among efforts to mitigate the global climate crisis, the development and deployment of

electric vehicles has experienced a dramatic increase in recent years, and is predicted to continue

to grow as part of a global solution to this pressing issue. In fact, the International Energy Agency

predicts a roughly 40-fold increase in the number of EV ownership by 2030 [3]. Because of their

critical role as the energy storage device of choice for Evs, this trend is similarly reflected by the

growth in demand projected in coming years for LIBs. Figure 1.3 shows projected demand of

LIBs in Li2CO3 equivalent provided by the consultancy Stormcrow [4]. It is clear that we must

match the increasing production pace of raw battery materials like Li2CO3 to ensure battery

production for the future of EV applications.

12
LITHIUM GLOBAL END USE MARKETS
2019
Air Treatment
Continuous 2%
Casting Mold Other
Flux Powders 6%
3%
Polymer
Production Batteries
4% Ceramics and 56%
Lubricating Glass
Greases 23%
6%

Figure 1.1 Breakdown of use of lithium by USGS [1].

Figure 1.2 Ragone plot of various battery technologies. Reprinted with permission from [2].

13
Figure 1.3 Projected demand of Li2CO3 by Stormcrow from 2017 through 2025 [4].

14
1.2 Primary Resources for Lithium Extraction

Three major resources exist for lithium extraction: brine lakes, minerals (from mining),

and sea water. This section briefly introduces each of the resources, and discusses industrial

practices for obtaining lithium from each resource. Ultimately, this thesis will discuss the

challenges associated with lithium extraction with brine, however this section presents general

information for the remaining primary resources as well.

1.7.3 Lithium Brine

Lithium brine is found in several regions across the globe, with the most significant brine

resources containing economical sources of lithium located in South America and China [6]. A

map highlighting these lithium brines is shown in Figure 1.4 [7]. The source of lithium in these

brines has not been clearly identified, however many experts suggest local volcanic activity

provided rock formations rich in lithium, which, upon weathering, formed the sedimentary

lacrustine formations that lie beneath these lakes saturating each lake with its particular salt content

[6]. Average lithium content in these brines can range from 200-700ppm, with the highest

estimated exceeding 4000ppm [8]. Brines are considered the most important lithium resource, as

they account for approximately 60% of available global lithium resources [9,10]. Table 1.1 shows

estimates of overall lithium content by brine where estimates are available [6]. Among these

brines, the South American region consisting of Argentina, Bolivia, and Chile has been named the

“Lithium Triangle”, due to containing roughly ¾ of lithium resources of all brines [10]. The

extraction techniques of brines are discussed in Section 1.3.

15
Figure 1.4 Map of brine lakes with significant lithium deposits. Reprinted with permission from [7].

Table 1.1 Table of most significant brines with estimated lithium content in metric tons. Reprinted with permission
from [6].

16
1.2.2 Lithium Containing Minerals

Approximately 130 different lithium containing minerals have been identified so far.

Roughly 10 of those minerals have large enough quantity of deposits found in earth’s crust, along

with existing technological techniques for obtaining the mineral deposit and extracting lithium

from the raw resource. Table 1.2 presents the composition of selected such minerals [5,8,10]. Of

all minerals, spodumene leads in overall lithium production contribution primarily from deposits

in Australia.

Production of lithium products from mineral sources is quite a challenging task. Due to the

wide variety in composition of each mineral, each particular manufacturing and processing

procedure must be developed and tailored for a specific composition [5]. Figure 1.5 shows an

example process flow of producing lithium carbonate, a critical material for production of LIBs,

from spodumene.

Mineral processing is also somewhat expensive due to energy intensive steps required for

liberating lithium from mineral crystal structures. In the case of spodumene, which is the primary

mineral source of lithium, the naturally occurring crystal structure is a-spodumene. This crystal

structure resists sulfuric acid leaching which is used to extract the lithium from the mineral due to

extremely tight packing of the atoms in the crystal structure. To overcome this, converting the a-

spodumene to b-spodumene decreases the specific gravity of the mineral, increasing the lithium’s

susceptibility to sulfuric acid leaching [31]. However, conversion of the crystal structure to favor

acid leaching requires highly elevated temperatures. In practice, this is performed between 900°C

-1300°C as seen in the process flow in Figure 1.5, consuming large amounts of energy to achieve

properly elevated temperatures. Because of these energy intensive steps, much attention has

recently been directed towards brine processing for production of lithium compounds [8].

17
Table 1.2 Table of most significant mineral sources of lithium broken down by chemical composition. Reprinted with
permission from [8].

Figure 1.5 Example process flow of production of lithium carbonate from spodumene. Reprinted with permission
from [8].

18
1.2.3 Lithium Extraction from Sea Water

Sea water has been a consideration as a resource for many materials such as gold, in which

nearly 7000000Mt is estimated in the oceans. In the case of lithium, some authors estimate that it

is a nearly “unlimited” resource containing 2000000Mt of lithium due to the sheer abundance of

sea water [8]. However, the extremely low concentrations, around 100ppb, demand extremely

challenging and costly separation processing. Some authors have examined technologies for sea

water extraction, but ultimately their energy consumption estimates place these technologies

around 10-30 times costlier than lithium extraction from brines [8].

19
1.3 Challenges Associated with Lithium Extraction from Brines

1.3.1 Current Industrial Extraction

Though lithium extraction from brines is currently the cheapest option from those

discussed in Section 1.2, it is still a challenging and expensive process. Brine compositions vary

widely between brines, and each extraction process must be separately developed for each brine

[9]. Nevertheless, a very rough overview of the necessary processes is presented below, generally

following the commercial processed outlined by Tam Tran and Van T. Luong in [11].

The ultimate goal of lithium extraction is to produce an end product of extremely high

purity, and in the case of LIBs, a purity of 99.5-99.99% purity Li2CO3 is required for battery

production. The first stage in brine processing is a series of solar evaporation ponds, which

enhances lithium concentration, but also facilitates precipitation of other cations, like Na and K,

which could act as contaminants in downstream processes. An image of solar evaporation ponds

for brines is shown in Figure 1.8. The brines are pumped between each lake as the undergo stages

of solar evaporations, where the final stages produce the yellowish green slurries as seen in the

image.

After solar evaporation, the brines are then processed to remove various contaminating

ions through a series of chemical precipitation steps, along with other techniques to prevent

reduction of end product purity. The most critical of these contaminants, magnesium, will be

discussed in the following subsection. After contaminants are removed, finally the desired product

can be precipitated or crystallized for use.

20
Figure 1.6 Solar evaporation ponds at the Salar de Atacama in Chile.
<https://www.forbes.com/sites/jamesellsmoor/2019/06/10/electric-vehicles-are-driving-demand-for-lithium-with-
environmental-consequences/#337e258862e2 >

1.3.2 Mg2+:Li+ Ratio: A Critical Composition Metric

One of the most challenging contaminants found in some brines is magnesium. Magnesium

tends to form compounds that contaminate LiCl formation, which often is an intermediate in the

production of lithium carbonate [5]. For example, lithium carnallite (LiCl•MgCl2•6H2O) tends to

precipitate rather than as pure LiCl crystals, as magnesium behaves quite similarly to lithium with

respect to precipitation chemistries [11]. Thus, the Mg2+:Li+ ratio of brines is a very important

metric characterizing the economic feasibility of brines. A higher Mg2+:Li+ ratio indicates a brine

that is costlier due to additional magnesium removal processing, while a lower Mg2+:Li+ ratio

indicates more economical brine [5]. Generally, a Mg2+:Li+ ratio <10 is required for a brine to be

economically exploitable through the commercial precipitation processes [8,12].

Due to a relatively high Mg2+:Li+ ratio of ~20, the salt flats in Bolivia known as the Salar

de Uyuni remain unexploited compared to their Chilean and Argentinian counterparts, with

Mg2+:Li+ ratio brines of 6:1 and 1:1 respectively [12]. The Salar de Uyuni is a critical lithium

resource, as it alone contains over one quarter of accessible global lithium resources [12]. Thus a

21
technology for cost effective reduction of the Mg2+:Li+ ratio of brines, capable of high performance

across a variety of brine compositions must be developed in order to utilize these integral brine

resources such as the Salar de Uyuni.

1.3.3 Mg2+:Li+ Separation Techniques in the Literature

Currently, two major branches of Mg2+:Li+ separation technologies have emerged in the

literature – chemical and membrane based methods. Chemical methods have been reviewed by

several authors [9,13,14,15]. While chemical methods are widely employed, there are many

drawbacks associated with them such as reagent cost, long processing times, waste disposal, and

environmental unfriendliness, along with often brine specific process development [16]. Thus,

membrane-based methods are recently emerging as an alternative to circumvent these challenges.

These include electrolysis [17], electrodialysis using monovalent ion exchange membranes

(selective electrodialysis or S-ED) [18,19,20,21,22], electro-electrodialysis with bipolar

membranes [23], nanofiltration [24,25,26], reverse osmosis [24], and membrane capacitive

deionization [27] for treatment of brines and subsequent production of lithium compounds. Of

these techniques found in the literature, electrodialysis appears to be a very promising technique

for reduction of the Mg2+:Li+ ratio of brines due to scalability and performance over a range of

brine compositions. However, the high cost of very specialized monovalent IEMs, along with

potential steric blocking of monovalent cations by divalent cations poses limitations to this

technique that must be overcome for commercial scale lithium extraction from brines. Table 5.1

and Table 5.2 gives a brief summary of monovalent ED investigations at optimized performance

found in the literature, along with a more detailed description in Chapter 5. A schematic of S-ED

is shown in Figure 1.7.

22
Figure 1.7 Schematic of S-ED operation. In S-ED, monovalent ion exchange membranes are stacked as shown,
alternating between monovalent cation (CIMS) and anion (ACS) exchange membranes, which selectively allow
monovalent cation and anion species respectively. As a voltage is applied across the membrane stack, lithium can
travel from the Dil (Diluate) stream into the Con (Concentrate) stream, where magnesium is retained in the Dil stream.
Through this, the Li transferred into the Con stream is collected as a low Mg2+:Li+ lithium solution. Reproduced with
permission from [20].

23
1.4 Ion Concentration Polarization Based Devices for Lithium Extraction

from Brines

A novel membrane-based technique can be realized through ion concentration polarization

(ICP). Using ion exchange membranes, the highly amplified electric fields induced by ICP

achieved mobility-based separation of ionic species with differing electrophoretic mobilities. In

fact, Gong et al. have simulated continuous extraction of lithium from high Mg2+:Li+ brines using

ICP based techniques in microchannels [16,28]. The simulations achieved a Li+:Mg2+ flux of

2.81⨉103 (AU), with 25.6% Li+ recovery in their initial work, and improved lithium recovery in

their second work by modifying the channel structure and collection method to achieve a 38%

recovery of lithium while also demonstrating 13 fold enrichment of the brine at a 10 fold reduction

of Mg2+:Li+ ratio. While microfluidic simulations provide valuable insights into the experimental

parameters’ effects on the output of this system, herein a scalable system aimed at continuous ion

separation for lithium extraction from high Mg2+:Li+ brines is desired.

Yoon et al. have developed scalable ICP systems for desalination applications [29]. These

desalination devices operate utilizing unipolar ICP using only CEMs to perform more energy

efficient desalination over traditional ICP employing both anion and cation exchange membranes

by utilizing the higher diffusivity of chloride over the sodium ion to increase current utilization

above unity [29]. In this work, we explore combining the concepts of these two technologies to

perform a preliminary investigation into a scalable ICP based system for continuous ion separation.

24
1.5 Thesis Scope and Outline

The objective of this master’s thesis is to demonstrate a proof of concept membrane-based

technology for continuous ion separation using ion concentration polarization. Lithium extraction

from brine lakes is the focus of this study, in which the separation and removal of magnesium from

lithium containing brine solutions is a critical challenge. To achieve this, we first introduce critical

ion selective materials used to generate the ion concentration polarization phenomena along with

the relevant physical phenomena in Section 2 to build a basis for the function of the platform. In

Section 3, the experimental methods and techniques are laid out to perform verification of the

proposed method. Following in Section 4, the results of the investigations are shown.

Continuous mode operating platforms are generally more highly sought after than batch

mode operation. In batch mode operation, a defined volume of input solution or material is

processed for a set duration, and thus requires replacement of input solution or materials along

with resetting of the device for the next batch of input. This makes integration into upstream and

downstream processes challenging, as the replacement and resetting step introduces a required

interruption of the process flow. Continuous mode operation devices allow for seamless integration

with upstream and downstream processes, as the output from the previous step can be integrated

into the input of the device, and the processed output is immediately ready for the next processing

stage. However, it is challenging to compare certain performance specifications between batch and

continuous mode technologies due to the inherently differing operating mechanisms. In Section 5,

a simple method for comparing continuous and batch mode technologies is presented for an

economic evaluation of the platform in this thesis to existing methods in the literature. Finally, in

Section 6, concluding remarks are made, along with potential future directions for the work in this

thesis.

25
1.6 References

[1] I. Ore, I. O. Pigments, P. Rock, Q. Crystal, R. Earths, and S. Ash, Mineral Commodity Summaries. 2019.
[2] H. Budde-Meiwes et al., “A review of current automotive battery technology and future prospects,” Proc. Inst.
Mech. Eng. Part D J. Automob. Eng., vol. 227, no. 5, pp. 761–776, 2013.
[3] J. Hykawy and T. Chudnovsky, “LITHIUM DEMAND, SUPPLY AND PRICE FORECASTS-Who do we throw
out of the boat?,” pp. 1–20, 2017.
[4] DiChristopher T., “Electric vehicles will grow from 3 million to 125 million by 2030: IEA,” [Online] CNB. 2018.
[5] H. Vikström, S. Davidsson, and M. Höök, “Lithium availability and future production outlooks,” Appl. Energy,
vol. 110, pp. 252–266, 2013.
[6] S. E. Kesler, P. W. Gruber, P. A. Medina, G. A. Keoleian, M. P. Everson, and T. J. Wallington, “Global lithium
resources: Relative importance of pegmatite, brine and other deposits,” Ore Geol. Rev., vol. 48, pp. 55–69, 2012.v
[7] L. A. Munk, S. A. Hynek, D. Bradley, D. F. Boutt, K. Labay, and H. Jochens, “Lithium brines: A global
perspective,” Rev. Econ. Geol., vol. 18, no. January, pp. 339–365, 2016.
[8] P. K. Choubey, M. Kim, R. R. Srivastava, J. Lee, and J. Lee, “Advance review on the exploitation of the prominent
energy-storage element: Lithium. Part I: From mineral and brine resources.” Minerals Engineering, vol. 89, pp. 119-
137, 2016
[9] B. Swain, “Recovery and recycling of lithium: A review,” Sep. Purif. Technol., vol. 172, no. November, pp. 388–
403, 2017.
[10] P. Christmann, E. Gloaguen, J.-F. Labbé, J. Melleton, and P. Piantone, Global Lithium Resources
and Sustainability Issues. 2015.
[11] T. Tran and V. T. Luong, Lithium Production Processes. Elsevier Inc., 2015.
[12] R. Mills, “Bolivia: Where revolutionaries and lithium miners go to die”, MINING.com, 2018
[13] P. Meshram, B. D. Pandey, and T. R. Mankhand, “Extraction of lithium from primary and secondary sources by
pre-treatment, leaching and separation: A comprehensive review,” Hydrometallurgy, vol. 150, pp. 192–208, 2014.
[14] P. K. Choubey, K. S. Chung, M. seuk Kim, J. chun Lee, and R. R. Srivastava, “Advance review on the exploitation
of the prominent energy-storage element Lithium. Part II: From sea water and spent lithium ion batteries (LIBs),”
Miner. Eng., vol. 110, pp. 104–121, 2017.
[15] J. F. Song, L. D. Nghiem, X. M. Li, and T. He, “Lithium extraction from Chinese salt-lake brines: Opportunities,
challenges, and future outlook,” Environ. Sci. Water Res. Technol., vol. 3, no. 4, pp. 593–597, 2017.
[16] L. Gong, W. Ouyang, Z. Li, and J. Han, “Direct numerical simulation of continuous lithium extraction from high
Mg2+/Li+ ratio brines using microfluidic channels with ion concentration polarization,” J. Memb. Sci., vol. 556, no.
May, pp. 34–41, 2018.
[17] X. Liu, X. Chen, L. He, and Z. Zhao, “Study on extraction of lithium from salt lake brine by membrane
electrolysis,” Desalination, vol. 376, pp. 35–40, 2015.
[18] X. Y. Nie, S. Y. Sun, Z. Sun, X. Song, and J. G. Yu, “Ion-fractionation of lithium ions from magnesium ions by
electrodialysis using monovalent selective ion-exchange membranes,” Desalination, vol. 403, pp. 128–135, 2017.
[19] X. Y. Nie, S. Y. Sun, X. Song, and J. G. Yu, “Further investigation into lithium recovery from salt lake brines
with different feed characteristics by electrodialysis,” J. Memb. Sci., vol. 530, no. February, pp. 185–191, 2017.
[20] Z. yong Ji, Q. bai Chen, J. sheng Yuan, J. Liu, Y. ying Zhao, and W. xian Feng, “Preliminary study on recovering
lithium from high Mg 2+ /Li + ratio brines by electrodialysis,” Sep. Purif. Technol., vol. 172, pp. 168–177, 2017.
[21] Q. B. Chen, Z. Y. Ji, J. Liu, Y. Y. Zhao, S. Z. Wang, and J. S. Yuan, “Development of recovering lithium from
brines by selective-electrodialysis: Effect of coexisting cations on the migration of lithium,” J. Memb. Sci., vol. 548,
no. 8, pp. 408–420, 2018.
[22] Z. Y. Guo et al., “Prefractionation of LiCl from concentrated seawater/salt lake brines by electrodialysis with
monovalent selective ion exchange membranes,” J. Clean. Prod., vol. 193, pp. 338–350, 2018.
[23] C. Jiang, Y. Wang, Q. Wang, H. Feng, and T. Xu, “Production of lithium hydroxide from lake brines through
electro-electrodialysis with bipolar membranes (EEDBM),” Ind. Eng. Chem. Res., vol. 53, no. 14, pp. 6103–6112,
2014.
[24] A. Somrani, A. H. Hamzaoui, and M. Pontie, “Study on lithium separation from salt lake brines by nanofiltration
(NF) and low pressure reverse osmosis (LPRO),” Desalination, vol. 317, pp. 184–192, 2013.
[25] X. Wen, P. Ma, C. Zhu, Q. He, and X. Deng, “Preliminary study on recovering lithium chloride from lithium-
containing waters by nanofiltration,” Sep. Purif. Technol., vol. 49, no. 3, pp. 230–236, 2006.
[26] G. Yang et al., “Investigation of Mg2+/Li+ Separation by Nanofiltration,” Sep. Sci. and Eng., vol 19., pp 586-
591, 2011.

26
[27] W. Shi et al., “Efficient lithium extraction by membrane capacitive deionization incorporated with monovalent
selective cation exchange membrane,” Sep. Purif. Technol., vol 210, pp. 885-890, 2019.
[28] L. Gong, Z. Li, and J. Han, “Numerical simulation of continuous extraction of highly concentrated Li + from
high Mg 2+ /Li + ratio brines in an ion concentration polarization-based microfluidic system,” Sep. Purif. Technol.,
no. January, pp. 174–182, 2019.
[29] J. Yoon, V. Q. Do, V. S. Pham, and J. Han, “Return flow ion concentration polarization desalination: A new way
to enhance electromembrane desalination,” Water Res., vol. 159, pp. 501–510, 2019.
[30] S. V. Pham, H. Kwon, B. Kim, J. K. White, G. Lim, and J. Han, “Helical vortex formation in three-dimensional
electrochemical systems with ion-selective membranes,” Phys. Rev. E, vol. 93, no. 3, pp. 1–5, 2016.
[31] L. Extraction and F. Spodumene, “HARD ROCK LITHIUM PROCESSING LITHIUM EXTRACTION,”
2010.

27
2 Theoretical Background

2.1 Ion Transport in Electrolytes

The laws governing the transport of charged species in electrolytes form the physical

foundation of this work. In this section, a simplified transport discussion is presented guided by

the derivations in the textbook by Alan J. Grozdinsky [30] and the Ph.D. thesis by Sang Van Pham,

in [31].

2.1.1 Diffusive Flux of Particles

For a solute species in a dilute, isotropic medium, Fick’s first law of diffusion states that

spatial concentration gradients result in a net flux of the species to reduce the spatial differences

in concentration. This is expressed by:

N# = −D# ∇c# (2.1)

where N# is the molar flux of the solute (mol m-2 s-1), D# is the diffusivity of the ith species (m2 s-

1
) , and c# is the concentration of the ith species (mol m3).

Considering an overall fluid system containing the solute, conservation of mass states that

the overall change in concentration of the solute within the system due solely to diffusion effects

can arise only as a result of flux in or out of the system. This can be represented by the divergence

of the molar flux of solute into or out of the boundary of a system, and the above statement imposes

that this equals the overall change in concentration within the system with respect to time, which

is known as the continuity law of solutes in a stationary fluid.

∂c# (2.2)
= −∇ ∙ N#
∂t

28
We can combine Fick’s first law of diffusion and the continuity law to yield Fick’s second

law of diffusion, or the diffusion equation by substituting the expression for the molar flux of each

species into the continuity equation:

∂c# (2.3)
= −D# ∇- c#
∂t

2.1.2 Electrical Migration of Charged Particles in a Potential Field

A charged particle in solution will attain a maximum drift velocity when placed in a

potential due to a complex interaction of many coupled phenomena, comprising a complicated

theory of electrokinetics. The drift velocity is given by:

v/ = µ(−∇φ) (2.4)

where v/ is the drift velocity (m s-1), µ is the mobility for a given solute in a particular solvent (m2

V-1 s-1), and φ is the potential throughout the system (V). While a seemingly simple expression,

the determination of the mobility of an ionic species analytically is incredibly intricate as

mentioned above. Considering a volume of ions with concentration c# , the overall flux of ions

under applied potential is given by the product of the drift velocity under a given potential and the

concentration:

N# = c# v/4 = − µ# c# ∇φ (2.5)

2.1.3 Transport by Fluid Convection and Extended Nernst Planck Equation

Combining the flux contributions as a result of diffusive and electrical migration effects

results in the Nernst Planck equation in a stationary medium, ultimately describing the motion of

29
particles relative to fluid background. However, the addition of convection accounting for fluid

motion results in the extended Nernst Planck equation, given by:

N# = −D# ∇c# − µ# c# ∇φ + c# u (2.6)

where u is the fluid velocity (m s-1).

2.1.4 Complete Model for Ion Transport in a Fluid System

The ion distribution as a result of the potential distribution with the fluid system, along

with a description of the fluid itself complete the overall description of ion transport within a fluid.

The potential and ion distribution are related by the Poisson equation:

−ε∇- φ = 8 z# Fc# (2.7)


#

where ε is the permittivity of the fluid (F m-1), Fis Faraday’s constant (C mol-1), and z# is the

valence of the ith ionic species. The Navier-Stokes equation for an incompressible fluid describes

the fluid motion:

∂ (2.8)
ρ< + u ∙ ∇= u = −∇p + γ∇- u + F@
∂t

where ρ is the density of the fluid (kg m-3), p is the pressure throughout the fluid system (kg m-1 s-
2
), γ is the dynamic viscosity of the fluid (kg s-1 m-1), and F@ is an electric force density term due

to the ions distribution throughout the fluid. Together, the extended Nernst-Planck equation,

Poisson equation, and Navier-Stokes equation describe the nature of a fluidic system containing

charged species. In many fluidic systems with charge carrying species, the intense coupling

between equations 2.6 – 2.8 results in a set of governing equations that are quite challenging to

30
solve. However, recent demonstrated advances in numerical engineering have allowed for an

analytical, theoretical model to capture the behavior of these highly coupled, nonlinear system, to

the great benefit of engineers utilizing these platforms [33].

31
2.2 Ion Exchange Membranes

Ion exchange membranes (IEMs) are essential materials for many electromembrane based

processes, including electrodialysis and ion concentration polarization (ICP)-based devices to

name a few. IEMs are typically constructed from a dense polymer material which has fixed

charges. It forms a nanoporous matrix, which, due to the fixed charge density on the surface of the

nanopores, has the ability to selectively allow certain ionic species to conduct through the matrix

[1]. This property is referred to as permselectivity, or the ability to pass oppositely charged ions

(counter-ions), while blocking identically charged ions (co-ions) with respect to the polarity of the

fixed charges on the polymer material [1].

Two main classes of IEMs exist: cation exchange membranes (CEMs) contain negative

fixed charged groups, while anion exchange membranes (AEMs) contain positive fixed charge

groups[2]. For example, -SO3- groups are commonly used in CEM materials, while -N+(CH3)3 are

commonly used in AEM materials [3]. An illustration of AEM and CEMs is shown in Figure 2.1

[4]. Further characterization of CEMs and AEMs is based on how strongly the fixed charge groups

dissociate when placed in solution, determining whether they are strong acid or based membrane

in the case of a high degree of dissociation, or weak acid or base membranes in the case of a lower

degree of dissociation [2]. IEMs can also be characterized by how they are made. Homogeneous

IEMs are made by chemically functionalizing a polymer material with groups that result in

dissociating species resulting in fixed charged groups when placed in solution [2]. Heterogeneous

IEMs are created by mixing an ion exchange resin with very fine size into a polymeric binder [2].

32
Figure 2.1 Illustration of the function of ion exchange membranes (IEM). Fixed charge groups in a nanoporous
material effectively screen the passage of ions with the same charge of the fixed groups in the material. Source:
<http://piano.chem.yamaguchi-u.ac.jp/1_5_mosaic(English).html>

33
2.3 Ion Concentration Polarization (ICP)

2.3.1 ICP as a Fundamental Phenomenon in IEMs

Ion Concentration Polarization refers to a phenomenon that occurs in electromembrane

processes (in addition to other selective membrane/interface systems) that arises due to a mismatch

in transport properties between the bulk solution side of the membrane system, and the membrane

itself. Upon an externally applied voltage across a bulk solution-membrane interface, a current will

form that is carried by the ions in the solution. In the bulk solution, all ions contribute current

according to their conductivities, whereas in the IEMs, only cations (anions) contribute to the

current, since anions (cations) cannot readily enter the CEM (AEM) because of the membrane

permselectivity [2].

In the bulk solution, both anions and cations can be transported by electric migration (drift),

and diffusion, while in the IEM, the ions can mainly be transported by drift [5]. Thus, the single

ionic species in the IEM must contribute all of the current through the membrane resulting in an

overall higher transport rate when compared to the bulk solution for this particular species. A

simplified illustration of the ICP phenomenon is shown in Figure 2.2 for a CEM in an electrolyte

system with external applied potential [5]. Because the flux of cations through the membrane is

higher than the flux of cations in the bulk solution for a given current, the concentration of cations

at the surface of the membrane drops on the anodic side of the membrane as shown in Figure 2.2,

forming the so called “ion depletion region” [6]. Similarly, cations accumulate on the cathodic

side of the membrane as they are transported more quickly through the membrane to the membrane

surface than removed towards the cathode, forming the so called “ion enrichment region” [6]. Due

to the electroneutrality condition in the bulk solution, the concentration of anions follows the

34
concentration of cations (for a symmetric, binary electrolyte), resulting in the concentration profile

in black as shown below.

Figure 2.2 Illustration of the ICP phenomenon for a cation exchange membrane immersed in electrolyte. Note the
formation of the depletion and enrichment zones under applied voltage. Overall anion diffusion and drift contributions
to the current cancel out, as the cations in this system are responsible for the total current carried between the electrodes
with zero net contribution to current from anions. Reproduced with permission from [5].

35
2.3.2 Current-Voltage Characteristic Curve

2.3.2.1 Linear Regime

The concentration profile above strongly dictates the nature of the IV characteristic seen in

electrochemical systems involving IEMs in contact with electrolyte solutions, as shown in Figure

2.2. At low applied voltages, the system generally behaves linearly as a simple resistor-like IV

characteristic determined by the conductivity of the solution as shown in portion I of the IV curve

in Figure 2.4 [7].

2.3.2.2 Limiting Regime

Upon further increase of voltage, experiments characterizing these systems noted a sharp

increase in the resistance of the electrolyte solution, which is known as the limiting current regime

as seen by segment II in Figure 2.4 [6]. We recall that the cations in the CEM system in Figure

2.2 make different contributions to the current in the bulk and the membrane phases of the system.

In the bulk phase, both the anions and cations share contribution to the current while in the

membrane phase only the cations carry the current. Thus, the flux of cations leaving bulk-

membrane interface through the membrane exceeds the flux of cations approaching the interface

from the bulk, resulting in eventual depletion of the ion concentration at the interface [6]. In order

to supply higher current, diffusion is the main mechanism (except convection) by which additional

ions can be transported to the interface and ultimately rate-limits the system, accounting for the

sharp increase in resistance [6]. In fact, classical theory of ICP predicted the dashed line showing

an ideal saturation of the current in these systems [8]. Here, we present a simple derivation of the

limiting current density of a 1-D system involving a CEM immersed in a symmetric, binary

electrolyte, largely following the derivation outlined by Ronald F. Probstein in Physiochemical

Hydrodynamics and Rhokyun Kwak’s thesis presentation [25,32].

36
The molar flux of ionic species i as presented in Section 2.1.2 can be related to the current

density, given by I, carried by the flux of ions by the following relation [30]:

I# = z# FN# (2.9)

Furthermore, the relationship between the diffusivity of species i, and mobility of species

i is given by [30]:

z# FD# (2.10)
µ# =
RT

where R is the gas constant and T is temperature. Substituting these expressions into the Nernst-

Planck equation (ignoring convection) can yield the current density contribution by cations and

anions in solution is given by:

dc± F - z±- D± c± dφ (2.11)


I± = −D± Fz± −
dx RT dx

where I± represents current density, D± is the diffusivity, F is Faraday’s constant, z± is the valence

of the ions in the electrolyte, c± is the concentration as a function of distance from the reservoir to

the membrane, R is the ideal gas constant, T is temperature, and φ is the potential from the

reservoir to the membrane. The + subscript denotes cations, and the – subscript denotes anions.

The first term in (1) represents the diffusive flux of species contributing to current, while the

second term represents the migration term of ions resulting from the external applied potential. We

ignore current flux from convection. We assume that because of the ideal permselectivity of the

CEM, the anions do not contribute to the overall current, and thus:

37
dcG F - zG- DG cG dφ (2.12)
IG = 0 = −DG FzG −
dx RT dx

which yields the equation of the electric potential from the reservoir to the membrane:

dφ RT dcG (2.13)
= −
dx FzG cG dx

Substituting (3) into (1), and using assumptions about the symmetric, binary electrolyte (and the

condition of electroneutrality):

zI = −zG , cI = cG = c (2.14)

yields an expression for the cation current which comprises all of the current through the system:

dc (2.15)
II = 2FDI zI
dx

In the case of limiting current, we assume that the concentration of ions at the electrolyte,

membrane interface goes to zero, and thus the rate of diffusion should dictate the current. Imposing

boundary conditions of an infinite reservoir and vanishing ion concentration at the electrolyte

membrane interface:

c(x = 0) = c (2.16)

c ( x = L) = 0 (2.17)

yields a linear concentration profile from the reservoir to the interface when solving the steady

state 1-D diffusion equation, resulting in the limiting current density:

38
2FDI zI c (2.18)
IM#N =
L

2.3.2.3 Overlimiting Regime

Many electrochemical experimentalists observed a large departure from the classically

predicted limiting current, as measured currents at very high voltages appeared like segment III of

the IV curve in Figure 2.3 [8]. While several models were proposed to explain this phenomenon

like water splitting to support increased current, or loss of membrane permselectivity, Rubinstein

et al. proposed a mechanism for over-limiting current which eventually became widely accepted.

They suggested strong electroconvection at high voltages can act as a mixing mechanism in which

the depletion layer collapses, as strong vortex like motion is induced near the surface of the

membranes [9]. Advancement of microfluidic technologies and techniques eventually gave way

for experimental observations of Rubinstein’s predictions through the use of fluorescent

microscopy tracing fluorescent particles in microfluidic channels integrated with ion exchange

materials [10,11].

39
Figure 2.3 Illustration of IV characteristic typical of IEM systems. Three distinct current regions can be observed: I
– ohmic region, II – limiting current, and III – overlimiting current. Reproduced with permission from [8].

2.3.3 Local Amplification of the Electric Field in the Depletion Zone

Microfluidic technologies gave engineers much success in utilizing the phenomena

associated with ICP, as the microfluidic environment offered many advantages, since low

Reynolds number flow regime yields powerful control over various electrokinetic phenomena.

Some applications include particle preconcentration for sensing and enhanced detection

applications [12-17], mobility-based assays [18,19], and desalination [20,21].

The ICP phenomenon allows for these techniques due to the development of an amplified

electric field in the depletion zone, resulting from the extremely low conductance of the low ion

concentration region comprising the depletion zone exerting a strong electric force on charged

particles [22]. As simulated by Gong et al., particles can experience a force balance by creating

40
opposing flow force fields through electroosmotic or pressure driven flow, and the electrical force

in the depletion zone, as shown in Figure 2.4 [22].

Figure 2.4 (a). Schematic of H-channel microfluidic device to illustrate ICP phenomenon used for preconcentration
of blue charged particles. (b). Plot of force magnitudes including fluid drag force (green curve) on ions and electric
force (pink curve) on ions versus distance along the microchannel. Local amplification of the electric field near the
nanochannels may create a force balance region in which particles can accumulate, as given by the green and pink
curve intersection in (b). Reproduced with permission from [22].

41
2.4 Ternary Electrolyte Concentration Profiles in a 1D AEM System

Because we are interested in cation separation, it is important to extend the theoretical

development above to systems containing multiple ionic species. In the case of more complex

solutions, interesting relationships between concentration profiles can arise due to properties of

the species involved. For example, studying the impact of electrophoretic mobility and diffusivity

on the concentration distribution of ions will ultimately form the foundation for the separation

process utilized in the experimental sections of this work. To do so, we adopt the approach of Alan

C. West’s textbook [34] for a ternary solution and electrode reaction, to develop a simple analytical

model for the ion selective membrane system in one dimension, which is capable of demonstrating

the cation separation phenomena.

Figure 2.5. 1D system considered in this derivation: a solution of three ionic species in contact with an ideally
permselective AEM. A constant current I is applied through the system, and the resulting potential distribution and
ion concentration profiles are studied.

In this 1D system, an ideally permselective AEM is placed at the zero-position location. A

current is applied such that a diffusion boundary at location L develops, with depletion zone

between locations 0 and L. Here, the transport of anions, which are the only species contributing

to current is limited by diffusion from the diffusion boundary to the surface of the AEM. We

assume that there is no convection in this system, and that the system is in steady state. We also

assume that the current is applied such that the concentration of anions goes to zero at the surface

42
of the AEM, corresponding to a limiting current scenario. We assume that the corresponding

potential is defined to be zero at the diffusion boundary, with interest to determine the

concentration profiles of each ionic species, along with the potential distribution from the surface

of the AEM to the diffusion boundary. We will adopt the notation of the preceding sections, with

the species subscripts identifying the corresponding species (ex: µO corresponding to the

electrophoretic mobility of Li+, µ- to Mg2+, µP to Cl- ...) as identified in Figure 2.5. In the bulk

solution, the concentration of all species is c#,Q . We assume electroneutrality also holds through

the solution system, ignoring space charge effects that arise due to fixed charge of the membrane.

To begin, we consider the no flux boundary condition imposed by the permselectivity of

the AEM on the cations throughout the system. Their distribution is governed by the potential

developed in the system. Considering equation (2.6) in the absence of convection and in 1D

dc# dφ (2.19)
N# = −D# − µ# c# =0
dx dx

for i = 1,2 representing only the cationic species. The solution to this equation yields the cation

distribution across the diffusion boundary as a function of potential:


µ# (2.20)
c# (φ) = c#,Q exp (− φ)
D#

Here, c#,Q represents the bulk concentration of species outside of the diffusion boundary layer.

The statement of electroneutrality is as follows:

0 = 8 z# c# (2.21)
#

43
And substituting the expressions for the concentration of cations in equation (2.20), we can yield

an expression for the concentration of the anion as a function of potential:

zO µO z- µ- (2.22)
cP (φ) = − cO,Q exp <− φ= − c-,Q exp <− φ=
zP DO zP D-

Assuming we apply a constant current I, which is carried only by the anion species, such that the

concentration of anion drops to zero at the surface of the AEM, we can write the flux condition of

the anion using equation (2.9):

I dcP dφ (2.23)
NP = = −DP − µP cP
zP F dx dx

Because we have determined the anion concentration as a function of potential, we can express

equation (2.23) as:

I dcP dφ dφ (2.24)
= −DP − µP cP
zP F dφ dx dx

Integrating this expression from the diffusion boundary to the AEM surface yields the following

expression:
W T
I dcP (2.25)
S dx = S (−DP − µP cP )dφ
WUX zP F TUV dφ

Evaluating the integrals and substituting the expression for the anion concentration in equation

(2.22), and dividing through by -D3:

44
I z µ z µ
(L − x) = <− O cO,Q Yexp <− O φ= − 1[ − - c-,Q Yexp <− - φ= − 1[=
zP DP F zP DO zP D-

µP zO DO µO z- D- µ-
+ < cO,Q Yexp <− φ= − 1[ + c-,Q Yexp <− φ= − 1[=
DP zP µO DO zP µ- D-

(2.26)

Utilizing equation (2.10), we can further simplify this expression to yield an implicit equation of

the potential as a function of distance from the AEM surface:

I z µ z µ
(L − x) = (1− O ) cO,Q exp <− O φ= + (1 − - )c-,Q exp <− - φ=
zP DP F zP DO zP D-

− cO,Q − c-,Q − cP,Q

(2.27)

From here, we can calculate the characteristic diffusion distance of the overall system, given by

the variable L by substituting the value φ\ at 𝑥 = 0 which approaches infinity to drive equation

(2.20) to zero into (2.27):

zP DP F (2.28)
L = (−cO,Q − c-,Q − cP,Q )
I

This follows our assumption of zero anion concentration at the membrane solution interface, along

with our assumption of electroneutrality. In reality, due to the fixed charge density of the AEM

material, a layer of counter-ions develops in the solution near the interface to maintain

electroneutrality [35]. This region, typically existing on order of nanometers from within the

membrane surface is called the electric double layer (EDL). While many EDL models exist such

45
as those covered by Schoch et al. in [35], in this model we assume zero concentration of all species

resulting in infinite potential at the interface.

Driving the ideal system with a current of 10mA, with a bulk concentration of 0.01M LiCl

and 0.06M MgCl2, with ionic parameters DMg2+ = 0.706 ⨉ 10-5 cm2/s, µMg2+ = 5.457 ⨉ 10-8 m2/Vs,

DLi+ = 1.03 ⨉ 10-5 cm2/s, µLi+ = 3.98 ⨉ 10-8 m2/Vs, DCl- =2.030 ⨉ 10-5 cm2/s, the resulting potential

distribution with respect to distance from the membrane normalized to the characteristic diffusion

length is shown in Figure 2.6a). Similarly, the concentration profiles of the cationic species, along

with the cation concentration ratio as a function of distance from the membrane are overlaid in

Figure 2.6b). We see that within a short distance of the membrane magnesium is displaced farther

away from the membrane than lithium. This can be explained by examining equation (2.20):

because of the higher electrophoretic mobility and lower diffusivity, we see that the rapidly

increasingly potential moving close to the surface of the membrane results in a more rapid decay

of the magnesium concentration than lithium. This effect can facilitate separation of cationic

species. Engineering a system capable of extracting the solution close to the membrane would

allow for collection of a solution with much lower magnesium to lithium ratio than the solution in

the bulk. This principle will be explored in simulated systems in the following section, and will

form the basis of the experimental work in Section 3 of this thesis.

46
Figure 2.6. (a). Solution for potential for simple ternary cation solution for system outlined in Figure 2.5. Note the
rapidly increasing potential close to 0 distance from the membrane, driving the concentration of cations to zero as a
result of depleting the concentration of anions at the surface of the membrane. This corresponds with a region of highly
amplified electric field over the rapidly changing potential in space (b). Concentration profiles of Li+ and Mg2+ as a
function of distance from the membrane surface, with overlaid Mg2+:Li+ molar concentration ratio.

47
2.5 Microfluidic Simulations of Li Extraction from Brines using ICP

Zirui Li and Linyang Gong et al. developed simulations utilizing the amplification of the

electric field in order to perform continuous ion separation by differences in mobility in

microfluidic channels applied to lithium extraction in brine [23,24]. Here we briefly present the

setup and results of their work in order to motivate experimental development of an ICP based

system for this purpose.

In their first simulation, the authors develop an AEM based microfluidic simulation model

and demonstrate its ability to deplete majority of magnesium while allowing some lithium to be

collected. The solution containing a mixture of ions in a ~70:1 Mg2+:Li+ mass ratio is passed from

left to right. Voltage is applied such that a depletion zone forms in the middle of the channel.

Because the electrophoretic mobility of magnesium is higher than lithium (µbc


^_`a =

7.616 × 10Gh m- /Vs, µbc Gh -


X#a = 4.007 × 10 m /Vs), magnesium experiences a higher

electrophoretic force by the amplified electric field in the depletion layer than lithium. In this

scenario, upon increasing the convective flow, the authors demonstrate a flux ratio of Li+:Mg2+ of

~2800:1 with an Li recovery of ~25.6% (resulting in a low Mg2+:Li+ output solution). However,

further modification of their channel structure and collection method ultimately resulted in the

ability to concentrate the lithium while simultaneously reducing the magnesium content, yielding

a 50-fold reduction in magnesium and 13-fold increase in concentration of lithium.

48
2.6 Concept Behind the Continuous, ICP Based Li Extraction System

2.6.1 ICP Desalination System

In the Micro/Nano BioMEMS group (Han group) at MIT, scalable desalination

technologies based on ICP have been developed [25-29]. In these systems, a brine solution is

passed through a membrane stack as shown in Figure 2.7 [27]. While the shear flow of solution

convectively transports ions tangentially across the membrane surface, a voltage is applied from

the anode to the cathode to generate an electric field perpendicular to the membrane surface. Due

to the permselectivity of the membranes, the ICP phenomenon generates a depletion zone on the

anodic side of the membranes. As the fluid travels across the membrane, the size of the depletion

zone increases from the inlet to the outlet. A bifurcation at the end of the channel splits the channel

into a desalted stream and concentrated stream, which can be separately collected and analyzed

for salinity. By simply stacking unit cells consisting of CEM pairs, throughput can be scaled almost

indefinitely, in a similar manner as conventional electrodialysis (ED) systems.

49
Figure 2.7 Schematic of ICP device utilizing CEMs for desalination. Between the CEMs, a depletion zone (enrichment
zone) forms on the anodic (cathodic) side of the CEMs, resulting in a region of depleted (concentrated) solution.
Bifurcations at the end of the channel allow for collection of both the concentrate and diluate stream to take advantage
of the separation. Reproduced with permission from [27].

2.6.2 ICP Based Li Extraction System

Combining the mobility-based separation of cations from Zirui and Gong et al.’s

microchannel simulations, and the high throughput desalination system developed by Han and

coworkers in the Micro/Nano BioMEMs group would allow for a novel technique to reduce the

Mg2+:Li+ ratio of solutions for lithium extraction from brines. To do this, an AEM based device is

constructed as shown in Figure 2.8. While shear convection to the membrane transports feed

solution through the device, electrical migration and diffusion are the primary transport

mechanisms perpendicular to the membrane surface (ignoring electroconvection or reactions at the

membrane-solution interface). Considering the zoomed main channel potion of the figure, ICP can

generate depletion zones on the cathodic side of the AEM membranes. As the resistance grows

within the depletion zone due to decreasing conductivity, a larger fraction of voltage will drop

across this region, inducing a locally amplified electric field. As this electric field strength grows,

50
it increasingly drives cations toward the cathodic AEM. And based upon analysis of the ternary

solution in Section 2.4, the relative change in concentrations will differ due to the differences in

the electrophoretic mobility and diffusivity of the ionic species in response to the amplified electric

field.

This overall strategy of ion separation is studied in the microchannel simulations of Section

2.5; however convection and electrical migration are the primary transport mechanisms

determining the average zero velocity positions for each ionic species in the simulations. Band-

like structure concentration distribution emerg, with ions having the lowest mobility occupying

the region closer to the front of the ion depletion zone, and ions with the highest mobility farther

from the ion depletion zone as the electric field more strongly impacts them [23,24]. Applying this

principle to the shear flow structure in Figure 2.8, we believe continuous ion separation can be

achieved as depicted in the illustration. Furthermore, scaling of this system can be achieved by

repeatedly stacking unit cells consisting of AEMs. In the following sections, an experimental

scheme demonstrating this ability, along with results is presented.

51
Figure 2.8 Schematic of electromembrane stack of the ICP based system. The overall system on the left shows the
stack consisting of electrodes (black), electrode rinse, buffer channels for reducing concentration gradients between
the main channel and electrode rinse channels, and the main channel with nanoporous spacer separating the collet and
waste bifurcated streams. The right hand zoomed idealized illustration of the main channel shows the mechanism of
ICP based separation through depleting the magnesium to a greater extent than lithium in the collect channel. An
amplified electric field is formed on the cathodic side of the bottom AEM due to the depletion region, and based on
the difference in electrophoretic mobility, the individual ion depletion boundaries allow for collection of a stream
purified of magnesium.

52
2.7 References

[1] T. Luo, S. Abdu, and M. Wessling, “Selectivity of ion exchange membranes : A review,” Journal of Mem. Sci.
vol. 555, no. December 2017, pp. 429–454, 2018.
[2] H. Strathmann, “Electrodialysis , a mature technology with a multitude of new applications,” DES, vol. 264, no.
3, pp. 268–288, 2010.
[3] N. P. Berezina, N. A. Kononenko, O. A. Dyomina, and N. P. Gnusin, “Characterization of ion-exchange membrane
materials : Properties vs structure,” Adv. In Colloid and Interface Sci., vol. 139, pp. 3–28, 2008.
[4] Higa Lab, “Performance evaluation of charged membrane developed by using ion-track technology”, [Online]
Available: http://piano.chem.yamaguchi-u.ac.jp/1_5_mosaic(English).html [Accessed: 10-1-2019]
[5] S. J. Kim, Y-A. Song, J. Han, “Nanofluidic concentration devices for biomolecules utilizing ion concentration
polarization: theory, fabrication, and applications,” Chem. Soc. Rev., vol 39, pp 912-922, 2010
[6] R. Kwak, “Nonlinear Ion Concentration Polarization: Fundamentals and Applications,” [Thesis], pp. 1–162, 2013.
[7] S. S. Dukhin, “Electrokinetic Phenomena of the Second Kind and Their Applications,” Adv. In Colloid and
Interface Sci., vol. 35, pp. 173-196, 1991
[8] I. Rubinstein, “Mechanism for an Electrodiffusional Instability in Concentration Polarization,” J. Chem. Soc., vol
77, pp 1595-1609, 1981.
[9] I. Rubinstein and B. Zaltzman, “Electro-osmotically induced convection at a permselective membrane,” Phys.
Rev. E., vol. 62, no. 2, pp. 2238–2251, 2000.
[10] S. J. Kim, Y. Wang, J. H. Lee, H. Jang, and J. Han, “Concentration Polarization and Nonlinear Electrokinetic
Flow near a Nanofluidic Channel,” Phys. Rev. Let., vol. 99, no. July, pp. (044501-3)1–4, 2007.
[11] G. Yossifon and H. Chang, “Selection of Nonequilibrium Overlimiting Currents : Universal Depletion Layer
Formation Dynamics and Vortex Instability,” Phys. Rev. Let., vol. 101, pp. (254501)1–4, 2008.
[12] S. J. Kim, L. D. Li, and J. Han, “Amplified Electrokinetic Response by Concentration Polarization near
Nanofluidic Channel,” Langmuir, vol. 25, no. 22, pp. 7759–7765, 2009.
[13] L. F. Cheow, S. H. Ko, S. J. Kim, K. H. Kang, and J. Han, “Increasing the Sensitivity of Enzyme-Linked
Immunosorbent Assay Using Multiplexed Electrokinetic Concentrator,” Anal. Chem., vol. 82, no. 8, pp. 3383–3388,
2010.
[14] P. Dextras et al., “Fabrication and Characterization of an Integrated Microsystem for Protein Preconcentration
and Sensing,” J. Microelectromechanical Syst., vol. 20, no. 1, pp. 221–230, 2011.
[15] S. H. Ko, et al., “Massively parallel concentration device for multiplexed immunoassays,” Lab on a Chip, vol.
11, pp. 1351-1358, 2011
[16] A. Sarkar, J. Han, “Non-linear and linear enhancement of enzymatic reaction kinetics using a biomolecule
concentrator,” Lab on a Chip, vol. 11, pp. 2569-2576, 2011
[17] S. H. Ko et al., “Nanofluidic preconcentration device in a straight microchannel using ion concentration
polarization ,” Lab on a Chip, pp. 4472–4482, 2012.
[18] L. F. Cheow, A. Sarkar, S. Kolitz, and J. Han, “Detecting Kinase Activities from Single Cell Lysate Using
Concentration-Enhanced Mobility Shift Assay,” Anal. Chem., vol. 86, pp. 7455-7462, 2014.
[19] W. Ouyang, S. H. Ko, D. Wu, A. Y. Wang, P. W. Barone, and W. S. Hancock, “Micro fluidic Platform for
Assessment of Therapeutic Proteins Using Molecular Charge Modulation Enhanced Electrokinetic Concentration
Assays,” Anal. Chem., vol. 88, pp. 9669-9677, 2016.
[20] R. Kwak, G. Guan, W. Kung, and J. Han, “Microscale electrodialysis : Concentration pro fi ling and vortex
visualization,” DES, vol. 308, pp. 138–146, 2013.
[21] S. J. Kim, S. H. Ko, K. H. Kang, and J. Han, “Direct seawater desalination by ion concentration polarization,”
Nat. Nanotechnol., vol. 5, no. 4, pp. 297–301, 2010.
[22] L. Gong et al., “Force fields of charged particles in micro-nanofluidic preconcentration systems Force fields of
charged particles in micro-nanofluidic preconcentration systems,” AIP Advances, vol. 7, no. 125020, pp. 0–9, 2018.
[23] L. Gong, W. Ouyang, Z. Li, and J. Han, “Direct numerical simulation of continuous lithium extraction from high
Mg2+/Li+ ratio brines using microfluidic channels with ion concentration polarization,” J. Memb. Sci., vol. 556, no.
May, pp. 34–41, 2018.
[24] L. Gong, Z. Li, and J. Han, “Numerical simulation of continuous extraction of highly concentrated Li + from
high Mg 2+ /Li + ratio brines in an ion concentration polarization-based microfluidic system,” Sep. Purif. Technol.,
no. January, pp. 174–182, 2019.
[25] R. Kwak, V. S. Pham, B. Kim, L. Chen, and J. Han, “Enhanced Salt Removal by Unipolar Ion Conduction in Ion
Concentration Polarization Desalination,” Nat. Publ. Gr., pp. 1–11.

53
[26] B. Kim, R. Kwak, H. J. Kwon, V. S. Pham, and M. Kim, “Purification of High Salinity Brine by Multi-Stage Ion
Concentration Polarization Desalination,” Nat. Publ. Gr., pp. 1–12.
[27] B. Kim, S. Choi, V. S. Pham, R. Kwak, and J. Han, “Energy efficiency enhancement of electromembrane
desalination systems by local flow redistribution optimized for the asymmetry of cation / anion diffusivity,” J. Memb.
Sci., vol. 524, pp. 280–287, 2017.
[28] B. Kim, H. Kwon, S. Hee, G. Lim, and J. Han, “Partial desalination of hypersaline brine by lab-scale ion
concentration polarization device,” Desalination, vol. 412, pp. 20–31, 2017.
[29] J. Yoon, V. Q. Do, V. S. Pham, and J. Han, “Return flow ion concentration polarization desalination: A new way
to enhance electromembrane desalination,” Water Res., vol. 159, pp. 501–510, 2019.
[30] A. J. Grodzinsky, Fields Forces and Flows in Biological Systems (Garland Science, New York, 2011)
[31] V. S. Pham, “Nonlinear Electrokinetic Flow Near Permselective Membrane,” [Thesis], pp. 14-16, 2012.
[32] R. F. Probstein, Physicochemical Hydrodynamics: An Introduction (Wiley-Interscience, New York, 2003), 2 edn.
[33] W. Ouyang, X. Ye, Z. Li, and J. Han, “Deciphering ion concentration polarization-based electrokinetic molecular
concentration at the micro-nanofluidic interface: theoretical limits and scaling laws,” Nanoscale, vol. 10, no. 32, pp.
15187–15194, 2018.
[34] A. C. West, Electrochemistry and Electrochemical Engineering: An Introduction. Alan C. West, 2012.
[35] R. B. Schoch, J. Han, and P. Renaud, “Transport phenomena in nanofluidics,” Rev. Mod. Phys., vol. 80, no. 3,
pp. 839–883, 2008.

54
3 Experimental Setup

3.1 Methods and Materials

3.1.1 ICP Schematic and Device Setup

The schematic of the ICP electromembrane stack is shown in Figure 2.6. The device is

comprised of two electrode rinse channels, and a single main channel bifurcated into collect and

waste streams. To reduce the large concentration gradients across the electrode rinse and main

channel, additional buffer channels with the same composition of the feed solution are inserted

between them. The main function of the buffer channels is to minimize the diffusive flux of ions

across the IEMs, which may arise because of the dramatically differing compositions of the

electrode rinse and feed solutions. Both the collect and waste streams within the main channel are

800 µm in height, and are separated by a nanoporous (200 nm pore size polycarbonate membrane,

Sterlitech, USA) for an overall height of 1.6mm in the main channel. The nanoporous membrane

reduces electroconvective mixing between the waste and collect channels, which would eliminate

any separation effect during processing through the device [29]. Electronvective instabilities arise

as a fundamental phenomenon in ICP, and are currently accepted as the mechanism for overliming

currents exceeding the theoretical limiting current [30]. Buffer channels are also 1.6 mm in height

but possess no nanoporous membrane. AEM membranes (FTAM-E, FuMa-Tech GmbH,

Germany) separate the main channel and buffer channels, and are critical to the function of this

ICP device as discussed in Section 2.5.2. CEM (FTCM-E, FuMa-Tech GmbH, Germany)

membranes separate the buffer channels and electrode rinse channels, preventing sulfate anions

from the electrode rinse solutions from mixing with the main and buffer channels, as well as

preventing chloride from entering the electrode rinse solutions and oxidizing at the cathode. For

both the buffer and main channels, the channel width and length are 5 mm and 75 mm respectively.

55
Thus, the effective membrane area is 375 mm2. Electrodes (Ru-Ir coated Titantium Plates, Baoji

Qixin Titanium Co., LTD., China) were secured in 3D printed housing (Clear Resin v4, FormLabs,

MA, USA). A schematic of the overall connected system and images of the experimental setup are

provided in Figure 3.1 – Figure 3.5.

3.1.2 Preparation of Feed Brines and Electrode Rinse Solution

Feed brines were prepared by mixing LiCl, NaCl, and MgCl2 in DI water. For each desired

brine composition, ideal and measured mass values are presented in Appendix A. These measured

values were used to verify ICP-OES measurements as well. 0.6M Na2SO4 was also prepared by

dissolving the 340.896 g of salt into 4 L of DI water for the electrode rinse solution. All reagents

were purchased from Millipore Sigma unless otherwise stated.

56
3.2 Device Operation

For each experiment, an appropriate Mg2+:Li+ brine was selected as the feed solution

according to the desired composition. Hydrodynamic pressure was suppled using peristaltic pumps

(Masterflex® L/S pump, Cole-Parmer Instrument Company, LLC., Vernon Hills, IL, USA) with

output connected to pulsation dampeners to reduce flow fluctuations created by the peristaltic

pumps. Flow rates in the main channel were measured using flow meters, and adjusted by valves

at each output to reach the desired flow rate condition for each experiment. The buffer channel

flow rate was selected to be 5 times the main channel flow rate to minimize increased power

consumption due to concentration polarization effects. A recirculated reservoir of volume 400mL

was used to supply solution to the buffer channels. Electrode rinse solution was run at 10mL/min

across all experiments. Experiments were stopped if precipitation began to appear in the

recirculating buffer channel tank.

A source meter (Keithley 2400, OH, USA) was used to supply current sourcing, along with

voltage measurements for experiments. LabVIEW 2018 (National Instruments, TX, USA)

software was used to acquire data. Source and measurement conditions for each experiment is

presented in Appendix B. A flow-through conductivity meter (Microelectrode Inc., Bedford, NH,

USA) was connected to the output of the collect channel to verify proper functioning of the ICP

device by detecting overall reduction in conductivity as a result of the depletion zone growing.

After stabilizing each current step according to the source conditions in Appendix B,

approximately 200µL of solution was collected at the end of the source step. 200µL of feed

solution was also collected before each experiment for comparison with the output brines. After

each sample collection at the end of the current source step, the flow rate was readjusted to the

proper experimental conditions.

57
Figure 3.1 High level schematic of overall fluidic system. Appropriate solutions from reservoirs are pumped through
corresponding channels using peristaltic pumps. Pulsation dampeners are coupled to the peristaltic pumps to reduce
pressure fluctuations at the output of the pump. Pulsations in the flow can drastically disrupt the electrochemical
process within the cell, however this disruption is minimized due to the pulsation dampeners. While the buffer and
electrode rinse channels are recirculated through an experiment, the processed solution in the main channel is collected
in collect and waste stream reservoirs for composition analysis with ICP-OES. Flow and conductivity meters allow
for real time measurement of flow conditions and overall conductivity of processed solution. A source measurement
unit at the terminals denoted V+, V- allows for real time application of current and measurement of resulting voltage.
Images of the implemented system are provided in Figures 3.2-3.5.

58
Figure 3.2 Image of overall experimental setup. Figures 3.3-3.5 give closer details of each major component, which
includes the pumping, feed, and electrode rinse solutions (red dashed box), the electromembrane stack (green dashed
box), and the flow and conductivity region at the collection region in the image (yellow dashed box).

59
Figure 3.3 Delivery component for the buffer channel solution, the main channel solution, electrode rinse solution,
and pumps. (a) Recirculating buffer solution reservoir for reducing the concentration gradient between the electrode
channel and the main channel. (b) Main channel reservoir. (c), (d) Pulsation dampeners for recirculating buffer solution
and main channel solution for reducing flow fluctuations from the peristaltic pumps. (e) Recirculating electrode rinse
reservoir. (f), (g) Peristaltic pumps for the buffer and main channels, and electrode rinse channel respectively. (h)
Pulsation dampener for electrode rinse solution for reducing flow fluctuations from the electrode rinse peristaltic
pump.

60
Figure 3.4 Electromembrane stack and device. This device is constructed according to the schematic in Figure 2.6.
The outer electrodes allow for sourcing current and measuring resultant voltage across the entire membrane stack. The
inner electrodes allow for voltage measurement within the channels, primarily used for power consumption
determination of the main channel.

61
Figure 3.5 Measurement component of experiment. (a) Conductivity meter for real time monitoring of depletion zone
(through monitoring overall desalting of the collection channel) connected to the collect output as shown in Figure
3.2. (b), (c) Collect and waste channel flow meters. (d) Output tubing for collect and waste channels allowing for
solution collection and downstream ion concentration analysis.

62
3.3 Analytical Methods

Ion concentration for cations (Li+, Mg2+, Na+) measurements were performed using ICP-

OES (Agilent ICP-OES 5100, MIT MRL, MA, USA). 2.1% w/w HNO3 was prepared by mixing

15mL of 70% w/w HNO3 into 500mL of DI water and used as the sample matrix for ICP-OES

measurement. 100µL of sample were diluted into 5mL of 2.1% w/w HNO3, and filtered using

200nm polyethersulfone membrane filters before processing with the ICP-OES. Standard solutions

of cations were also prepared for tool calibration by serial dilution of purchased ICP-OES

standards for Li+, Mg2+, and Na+. Standards were measured at least 3 times over the course of each

measurement batch on the tool to ensure consistent behavior throughout the run. These standard

measurements were averaged, and a linear calibration curve was generated to map tool intensity

measurements to concentrations. Concentrations were calculated considering dilution factor

during sample preparation. Appendix C. gives some calibration specifics and discusses

verification of ICP-OES calculated concentrations.

63
3.4 Data Analysis Metrics

This section introduces the metrics used to characterize device performance for separation

capability well as energy consumption. In Section 4, they are used to analyze the experimental

data over a range of applied current densities, which is defined as the total applied current divided

by the effective membrane area.

3.4.1 Separation Factor

The separation factor (SMg/Li) is used to characterize how selectively magnesium is

removed over lithium. It is calculated as the ratio of the input and output Mg2+:Li+ mass ratios

respectively. The output brine is collected from the collect channel for a given applied current

density, and the input brine is described as the brine without processing, or zero applied current

density. The Mg2+:Li+ mass is calculated as the ratio of the calculated [Mg2+] and [Li+] from ICP-

OES measurements, where [Az+] represents the concentration of cation Az+. Subscripts o and i

denote output and input respectively.

[Mg-I ]
< [ I] #=
Li #
S^_/X# = (3.1)
[Mg-I ]\
< [ I] =
Li \

3.4.2 Li Recovery Ratio

The lithium recovery ratio (RLi) is calculated as the ratio of the output and input lithium

mass flow rates. The mass flow rate is determined as the product of the fluid volumetric flow rate

and the lithium concentration. The output flow rate in all experiments is half of the input flow rate

due to the bifuracation, as both the collect and waste streams after the bifurcation are held at a 1:1

flow rate. The lithium concentrations are determined from ICP-OES measurements at the output

and input. [Li+] represents the measured lithium concentration. The subscript o denotes the collect

channel, and the subscript i denotes the input or feed.

64
Q \ [LiI ]\
R X# = (3.2)
Q # [LiI ]#

3.4.3 Specific Energy Consumption

Specific energy consumption (ELi) is used as an economic analysis metric. Is calculated as

the energy consumed to process enough volume of feed solution for 1g of Li+ to be extracted into

the collected solution.

1 w
T ∫V I(t)V(t)dt
EX# = (3.3)
Q \ [LiI ]\

In Equation (3.3), the numerator term represents average power over an applied current

step, calculated as the time averaged power over the entire interval of an applied current step. The

interval duration is given by T. I(t) and V(t) represent the applied current and measured voltage as

a function of time, where I(t) is constant over a given current step. The denominator in Equation

(3.3) represents the mass flow rate of Li+, calculated as the product of the volumetric flow rate,

Q \ , and the measured output concentration of Li+, [Li+]o, at a particular applied current step.

65
4 Results and Discussion

4.1 Effect of Applied Current Density

The first investigation carried out in this work was the effect of applied current density on

the separation factor and lithium recovery. For this scenario, a 25:1 Mg2+:Li+ brine was selected,

and the linear flow rate was kept at a constant 1 mm/sec corresponding to a volumetric flow rate

of 250 µL/min across three trials. The effect of increasing the applied current given a constant

linear flow velocity results in a decreasing Mg2+:Li+ ratio at the collected output (Figure 4.1a).

Thus, to achieve a high separation factor (SMg/Li), higher current densities must be applied (Figure

4.1b). However, we also see that a high separation factor comes at the expense of lithium recovery

(RLi), which we see decreases with increasing current density (Figure 4.1c). To understand these

results, we note that in ICP, the amplified electric field resulting from decreasing conductivity in

the depletion layer grows in strength with increasing current density. As discussed in Section 2.5.2,

increasing the applied current density results in larger field strengths generated in the increasingly

depleted region at the cathodic side of the AEM. Thus, as we continually increase the applied

current density, we expect increasing electric field magnitude as seen by the ions passing through

the channel near the depletion region. This increasing field strength supports the experimental

evidence of reduced Mg2+:Li+ ratio and thus higher SMg/Li with increasing current. Furthermore,

the growth of the depletion layer also supports the reduction of RLi as current increases, as the

overall salinity of the collect stream decreases as current density increases.

With respect to specific energy consumption (ELi), we see that as the applied current

density increases, ELi dramatically increases (Figure 4.1d). As the applied current density

increases, the strength of the depletion zone increases both with respect to decreasing conductivity

of the depletion zone, along with size of the depletion zone. These strongly coupled effects result

66
in highly increasing resistance of the depletion zone. In order to maintain the overall applied

current, the resulting region of highly amplified electric field requires increasingly greater voltage

across the region to maintain the current, resulting in greater power consumption due to this region

in the channel. To compound this effect, because RLi simultaneously decreases with increasing

current, the amount of time required to pass a specified amount of lithium ions in solution also

grows. This results in increasing amounts of time applying higher levels of power resulting in the

dramatically increasing ELi trend we see in Figure 4.1d.

Figure 4.1 Comparison of performance with respect to applied current density under varying flow velocity conditions.
(a) The effect of flow velocity on Mg2+:Li+ ratio. (b) The effect of the flow velocity on the separation factor. (c) The
effect of flow velocity on the lithium recovery ratio. (d) The effect of flow velocity on the specific energy consumption.

67
4.2 Effect of Convective Flow

The next investigation carried out in this work was the effect of convective flow on the

separation factor (SMg/Li), lithium recovery (RLi), and specific energy consumption (ELi) (Figure

4.1). Three convective flow conditions were studied for the case of a 25:1 Mg2+:Li+ brine, including

0.5, 1, and 2 mm/s. In Figure 4.1, there is a clear trend that with increasing convective flow, higher

applied currents are required to generate the same level of reduction of the Mg2+:Li+ ratio for the

given brine (Figure 4.1a). As a result, higher currents must be applied to achieve the same SMg/Li

for higher convective flow (Figure 4.1b), and similarly reduction of RLi (Figure 4.1c). To gain

insight into how linear flow velocity results in this effect, we can consider the transport

mechanisms of Cl- ions close to the bulk-AEM interface. Ignoring convection, Cl- ions migrate

through the AEM from cathodic to anodic side. Diffusion is the primary transport mechanism of

replenishing Cl- ions at the interface, however this process is much slower than the migration

removing them from the interface through the AEM. Because of the mismatch in transport rates,

the Cl- is depleted at the interface resulting in the depletion zone and ICP effect as mentioned

above. Next by considering convection, we add another transport mechanism to replace Cl- close

to the bulk-AEM interface. Now, migration through the AEM must compete with both the

diffusion and convection of Cl- to produce the depletion zone for amplification of the electric field,

requiring higher overall currents to increase the rate of migration to account for the additional Cl-

transported by convection. This explains the trends regarding SMg/Li and RLi seen in this

investigation.

We now consider the effect of convective flow on ELi. From Figure 4.1d, while each

individual ELi plot follows the same trend as in section 4.1, we see that the effect of increasing

convective flow shifts the dramatic rise in ELi with respect to applied current density. As

68
convective flow increases, the magnitude of applied current density necessary to generate the

increasingly resistive region following the argument in Section 4.1. Similarly, the decrease in RLi

is also shifted with respect to applied current density over increasing flow rates. Together, these

effects can account for the overall trend of ELi in this investigation.

69
4.3 Effect of Brine Composition

The final investigation carried out in this work was the effect of brine composition on

separation factor (SMg/Li) lithium recovery (RLi), and specific energy consumption (ELi ) (Error!

Reference source not found..2). To perform this investigation, the Mg2+:Li+ ratio of the feed brine

was adjusted by changing the amount of MgCl2 added to the brine while keeping the amount of

sodium and lithium constant. The flow rate in this set of experiments was also maintained constant

at 1 mm/s. From the plots in Error! Reference source not found..2, we notice similar behavior of

all brines in response to increasing current density. However, the onset of the depletion zone is

delayed with respect to increasing current due to the higher conductivity of the solution in the 60:1

and 100:1 brines, compared to the 25:1 brine. This results in higher current densities required to

achieve the same reduction of Mg2+:Li+ (Figure 4.2a) as the feed ratio increases. This yields the

trend in SMg/Li suggested by Error! Reference source not found..2b. While the 25:1 brine was

reduced below the required ~10:1 for commercial processing, the 60:1 brine was reduced to just

above 10:1 and the 100:1 brine was reduced to ~20:1 for the duration of these experiments. Though

it is unclear whether these brines can be reduced to below 10:1 from this data, it can be possible

to reduce the Mg2+:Li+ ratio of these brines to satisfactory level in a two-stage continuous process,

or by decreasing the convective flow rate. RLi follows the similar trend, where for a given RLi, the

data suggests that increasing current density is required to reach the given recovery ratio with

increasing Mg2+:Li+ ratio (Figure 4.2c).

In Figure 4.2d, we see that increasing applied current density produces a dramatic rise in

ELi as in previous sections. The effect of the feed composition results in a similar trend as in

convection; within increasing Mg2+:Li+ feed composition, the rise in ELi is delayed with respect to

applied current density. We can again follow the arguments made for the Mg2+:Li+ ratio, SMg/Li,

70
and RLi to understand this trend. With increasing feed Mg2+:Li+, we require higher applied current

levels to achieve the same amount of depletion, and thus electric field strength within the depletion

zone, ultimately delaying the sharp increase in resistivity of the depletion zone with respect to

applied current. Similarly, the RLi sees the same delay with respect to applied current density,

causing the trend we see in ELi versus the feed ratio.

Figure 4.2 Comparison of performance with respect to applied current density under composition of feed solution. (a)
The effect of feed composition on Mg2+:Li+ ratio. (b) The effect of the feed composition on the separation factor. (c)
The effect of feed composition on the lithium recovery ratio. (d) The effect of feed composition on the specific energy
consumption.

71
4.4 Summary of Experimental Results

In this chapter, the results of several experimental conditions were presented for selectively

removing magnesium from a simplified lithium brine model. Applied current density was varied

over the set of experimental parameters, which included feed flow velocity and feed composition.

For a 25:1 Mg2+:Li+ brine, this data suggests that one can successfully reduce the Mg2+:Li+ ratio of

a brine to below 10:1 as required for industrial processing for lithium with appropriate selection

of experimental parameters. While the data also suggests that for higher grade of magnesium brines

up to 60:1 Mg2+:Li+ reduction to below 10:1 can be possible, for the 100:1 Mg2+:Li+ brine,

processing of a single pass through the device is not feasible with the limitations of the lab scale

device used in this study. However, a multiple stage, continuous processing or slower convective

flow rates could offer solutions for handling higher grade brines. While the technical feasibility is

well demonstrated by the data presented here, an assessment of economic feasibility is critical, and

will be built in the following chapter.

72
5. Techno-Economic Analysis

Comparing performance of the ICP based system for lithium extraction with other

membrane-based techniques found in literature is critical for economic evaluation of the

technology. In this section, we review selected membrane based techniques for reduction of the

Mg2+:Li+ from brine solutions, briefly touching upon operation mechanisms and conditions, to

compare the results from this preliminary investigation with the performance of those found in

literature. Ultimately, scaling the size of the lab-scale device will alter the performance in terms

of power and energy consumption, as the effective membrane area and separation processes can

be implemented more gradually. Power scaling laws based on device geometry from

electrodialysis systems will be applied to scaling the lab-scale ICP system as a method of economic

evaluation for future practical applications. In addition, as the ICP based system operates in

continuous mode, a comparison between batch mode membrane technologies will be presented.

73
5.1 Membrane-based Techniques found in Literature

5.1.1 Overview of Electrodialysis and Selective Electrodialysis [1-5]

The critical components of a lab-scale, electrodialysis (ED) system for desalination are the

concentrate and diluate reservoirs, the electrode rinse solution, the pumping mechanism, the power

supply, and the electromembrane stack where the desalting (and concentrating) occurs [7]. In a

batch mode system, the concentrate and diluate reservoirs are constantly recirculated, passing the

feed solution through the electromembrane stack multiple times. After the experiment duration,

the diluate reservoir (in the case of ED for desalination) can be measured or analyzed for salt

concentration to determine desalination performance. In contrast, continuous mode ED systems

do not recirculate feed solution and process the solution with a single pass through the

electromembrane stack, before collection and analysis of the diluate and concentrate streams.

Critical to the function of the ED system is the electromembrane stack, which consists of

alternating CEMs and AEMs. Channels are formed between the alternating membrane types by

inserting a spacer layer of specified height to allow the feed solution to flow through the membrane

stack [8]. Figure 5.1 illustrates a schematic of a typical ED membrane stack. In this system, there

are two types of channels: a channel with a CEM on the anodic side (and AEM on cathodic side)

and vice versa. As a DC voltage is applied across the stack, anions migrate towards the anode, and

cations migrate towards the cathode. In the cathodic CEM channels, cations migrate through the

CEM towards the cathode, and anions migrate through the AEM towards the anode. This stream

becomes largely desalted and thus forms the diluate stream. The ions from the diluate stream enter

the adjacent anodic CEM channels. However, in these channels, ions are trapped from leaving as

the permselectivity of the membranes impedes the migration in response to the applied potential.

Anions migrating towards the anode are inhibited by the CEM, and similarly cations migrating

74
towards the cathode are inhibited by the AEM, forming the concentrate channel. Distinct reservoirs

for the two channel types in a batch mode operation allows the diluate reservoir to desalt over time

by forcing salt into the concentrate reservoir at the expense of electrical energy consumption.

Figure 5.1 Schematic of an ED electromembrane stack. In this figure, the inlet water is split into two streams, the
product stream (or diluate stream) and the concentrate stream. Reproduced with permission from [6].

Researchers have utilized the ED platform for ion separation by replacing the CEM and

AEM materials with monovalent selective CEM and AEM materials, which allow only

monovalent species to conduct through the membrane (in addition to satisfying the

permselectivity). These systems are referred to as selective electrodialysis, or S-ED. In this case,

the diluate and concentrate channels take on a new role based on the application. When dealing

with separation of the Mg2+ and Li+ ions, because the lithium ions are the only cation species with

75
the ability to transfer channels due to the monovalent CEMs, now the diluate channel becomes

diluted of lithium ions, whereas the concentrate channel becomes concentrated with lithium ions.

Similarly, the magnesium ions are retained within the diluate channel. Typically some form of

recovery solution is used as the initial solution, such as DI water or a NaCl solution, which is free

of both lithium and magnesium for recovery of lithium from the diluate channel.

5.1.2 Reported Performance of Selective Electrodialysis in the Literature [1-5]

Xiao-Yao Nie et al. performed a series of investigations using an electrodialysis system

utilizing monovalent ion exchange membranes for the purpose of reducing the Mg2+:Li+ ratio of

brines. To do so, they replaced the stacks of alternating CEMs and AEMs in traditional

electrodialysis membrane stacks with mono-CEMs and mono-AEMs (Selemion CSO and ASA),

which preferentially conduct monovalent ions over divalent and multivalent ions. Their membrane

stack consisted of 20 pairs of membranes, with effective membrane area of 507cm2, and separation

between membranes of 750um. Table 5.1 provides a summary of some membrane parameters

used in their ED pilot scale system.

In their preliminary investigation, they varied the Mg2+:Li+ ratio of a lab-made brine from

66.7 to 400, while keeping a constant concentration of lithium at 150ppm. The concentrate and

diluate reservoirs were 25L, and they operated the device at a linear flow rate of feed solution at

7.1cm/s, applying a constant current density of 5.9A/m2 in a batch mode configuration over a 3

hour period. They also varied the temperature, linear flow velocity, volume of concentrate and

diluate reservoirs. The applied current density was also varied, for the brine with a constant

Mg2+:Li+ ratio of 150 with a 150ppm Li concentration. Finally, they introduced coexisting cations

(Na+, K+, and Ca2+), at constant current density, temperature, and feed velocity to determine the

effect of coexisting cations on the performance of the device.

76
In their preliminary investigation, they were able to achieve satisfactory Mg2+:Li+ reduction

to below 10:1 for feed solutions of Mg2+:Li+ 150:1 or less, but still demonstrated up to 21.8 fold

reduction of the Mg2+:Li+ ratio for a feed ratio of 400:1. In their optimized scenario, they achieved

a reduction of a 150:1 brine to 8.0:1, with a Li+ recovery ratio of 95.3%, reporting a specific energy

consumption of 0.0019kWh/g Li+. In a follow up publication, they investigated the effect of

operating in constant applied voltage scheme for a brine containing 4.42g/L Li+, 1.6g/L Na+,

1.64g/L K+, 87.0g/L Mg2+, and 30.1g/4 SO42- achieving a reduction of the 20.7:1 Mg2+:Li+ feed

ratio to 2.07, with a Li recovery ratio of 90.5% and specific energy consumption of 0.0045kWh/g

Li using a brine from the East-Taijiner lake in China. Table 5.2 summarizes some optimized

conditions and performance specs of their S-ED system for these investigations.

In another series of publications, Zhi-Yong Ji and colleagues performed several

investigations utilizing a selective electrodialysis platform for the purpose of Mg2+:Li+ reduction

of brines. In their system, they employed 10 membrane pairs (ASTOM ACS and CIMS), with an

effective membrane area of 140.7cm2 with a spacing between membranes of 900um. In their

operation, they used a concentration compartment volume of 2.5L, a desalting compartment of

1.5L, varying the applied voltage, linear flow velocity, feed Mg2+:Li+ composition, and pH to

determine the effect on performance. Optimizing these parameters at an applied voltage of 5V at

6.2cm/s linear flow velocity and a 60:1 feed ratio, they reduced the Mg2+:Li+ ratio by a factor of

12.48 to approximately 4.81 with a Li recovery ratio of 72.46%. In subsequent investigations, they

determined the effect of separation with coexisting cations, and in a final investigation examined

the performance of the device on real brines, achieving a Mg2+:Li+ reduction from 35.18 to 3.91

(feed Li concentration of 1.03g/L), with a Li recovery of 76.45% and a reported specific energy

consumption of 0.66kWh/mol Li or equivalently 0.095kWh/g Li in optimized operating

77
conditions. Table 5.1 provides some membrane parameters for their system, and Table 5.2

provides some optimized conditions and performance for these investigations. Additionally, Table

5.2 contains selected data regarding input and output parameters along with specific energy

consumption from this study for comparison with the literature results, which is discussed in the

following section.

Table 5.1 Summary of membrane geometry used in different S-ED literature. Authors present membrane dimensions
and effective membrane area. Membrane width and length fields are calculated by assuming the membrane dimension
ratios are identical to the effective membrane dimensions*, as presented in the table.

Reference Membrane Membrane Channel Stack Membrane Used


Width* Length* Height Size
[1], [2] 12.9cm 39.4cm 750um 20(CEM) Asahi Glass co. CSO (CEM),
20(AEM) Asahi Glass co. ASA (AEM)

[3], [4], [5] 6.84cm 19.5cm 900um 10(CEM) ASTOM CIMS (CEM)
10(AEM) ASTOM ACS (AEM)

Table 5.2 Summary of feed compositions of optimized experiments and corresponding conditions over selected
references. All references operate in batch mode condition at over the course of the corresponding process time.
Lithium is recovered in the conc. stream of the devices, whereas the dil. stream is the waste stream.
Reference Li Mg:Li Mg:Li RLi Vconc/Vdil Flow Mem. Process Leff ELi Scaled ELi
Grade Feed Output (%) (L/L) Rate Pairs Time (cm) (kWh/g Li+) (kWh/g Li+)
(g/L) (m3/hr) (hr)
[1] 0.150 150:1 8:1 95.3 25/25 1.1 20 3 5200 0.0019 -
(Batch)
[2] 4.42 20.7:1 2.07:1 90.5 25/25 1.0 20 3 5200 0.0045 -
(Batch)
[3] 0.140 60:1 7:1 75.44 2.5/2.5 0.130 10 2 3042 - -
(Batch)
[5] 1.03 35:1 3.91 76.45 1.5/2.5 0.130 10 3 3042 0.0951 -
(Batch)
9.95 25.1:1 8.78:1 9.9 - 250 1 - 7.5 8.804 0.0127 ([1],[2])
This Study (mg/L) (uL/min) 0.0218 ([3].[5])
(Continuous) 9.95 25.1:1 6.46:1 6.55 - 250 1 - 7.5 17.06 0.0246 ([1],[2])
(mg/L) (uL/min) 0.0421 ([3].[5])

78
5.2 Specific Energy Consumption as a Function of Output Mg2+:Li+ Ratio

The output Mg2+:Li+ ratio can be used to select a performance with respect to specific

energy consumption. Figure 5.2 shows the energy consumption performance of the device

parameterized by the output Mg2+:Li+ ratio. Interestingly in Figure 5.2 a, we see that for the 25:1

brine case, that specific energy consumption as a function of desired output Mg2+:Li+ ratio does

not depend on flow rate. This may be explained by the linear relationship between energy

consumption and flow rate, and the linear relationship between power consumption and energy

consumption. As the flow rate increases, the power consumption and energy consumption both

increase linearly with flow rate, as the current required to generate ICP increases linearly with flow

rate. Similarly, as the flow rate increases, the amount of time required to process a given mass of

lithium decreases linearly with flow rate. Thus, the time a power must be applied through the

device decreases, resulting in linearly decreasing energy consumption with flow rate. Thus, the

overall energy consumption with respect to processing 1g of Li through the device does not depend

on the flow rate of the feed solution. We see that only for the 25:1 case in this investigation, we

can reliably reduce to below the 10:1 desired output threshold. For the purposes of economic

evaluation, we will consider the 25:1 brine moving forward, which could closely resemble the

brine found in the Salar de Uyuni in Bolivia.

In stark contrast to the S-ED batch mode systems introduced in Section 5.1.2, the

continuous mode ICP system has 1000-10000 fold higher energy consumption (for sufficient

output Mg2+:Li+ reduction) (see Table 5.2, column ELi). In the following sections, we explain some

of the most significant identified discrepancies, largely arising due to differences between

continuous mode and batch mode operating mechanisms. These sections develop an energy

consumption scaling argument to fairly asses the power consumption between the two operating

79
modes to produce the Scaled ELi column of Table 5.2. We use this metric to argue that the ICP

technology is within an order of magnitude power consumption of S-ED platforms, along with a

discussion some future directions of improvement.

Figure 5.2 (a) Specific Energy consumption versus output Mg2+:Li+ ratio of a 25:1 brine over varying flow rates. (b)
Specific energy consumption versus output Mg2+:Li+ ratio at 1mm/sec flow conditions over varying feed Mg2+:Li+
ratio. In both plots the reported ELi is measured for the entire device including electrode rinse and buffer channels.

80
5.3 Energy Consumption in ED Systems as a Model for the ICP Based System

With respect to average linear fluid velocity, cell height between membranes, and the

effective membrane length, H.-J. Lee et al. provide a practical relationship of these parameters

into the determination of the specific energy consumption of desalination in an ED system, which

is given by the following scaling relationship (in the author’s notation) [6]:

P/@x u ∆
Exy@z = ~ (5.1)
Q y|}z Ly|}z

where Exy@z is the energy required to desalinate a volume of water, P/@x is the power requirement

for desalination, Q y|}z is the practical flow rate selected by the electrodialysis system designer. u

is the linear flow velocity of the solution through the device, ∆ is the cell height between the

membranes, and Ly|}z is the practical path length for the membrane. Path length refers to the length

of membrane the solution passes over when traveling thorough a device (some membrane may be

compressed between channel forming spacers and thus does not actively contribute to processing

of the solution).

The major limitation of modeling an ICP based system by this ED scaling law is the

differing current regimes that these systems operate in. In ED, designers apply current below the

limiting current value to mitigate excess energy consumption resulting from concentration

polarization. However, in ICP, the performance of the device relies upon the formation of the

depletion layer to induce amplification of the electric field in the depletion zone. Though the above

scaling law assumes operation in the ohmic current regime, we use this relationship as a scaling

argument for energy consumption in the ICP system in order to make a comparison between the

S-ED systems in literature for Mg2+:Li+ reduction in brines. Despite channel architecture

81
difference, we can utilize this relation to account for differences in length scales to more closely

compare the different technologies.

82
5.4 Scaling the Continuous, ICP Based System for Energy Consumption

Comparison

5.4.1 Comparison of Batch Mode and Continuous Technologies

The wide array of operating conditions and system designs provides difficulties when

comparing membrane-based technologies. A major challenge is comparing batch mode

technologies with continuous mode technologies. In batch mode technologies, the solution is

recirculated many times, thus a single membrane pair processes the solution many times, whereas

feed solution in a continuous mode technology passes over the membranes once. We can

equivalently relate a batch mode platform to a continuous mode platform with extended membrane

length. A simple way of doing so is to create an effective membrane length by multiplying the

average number of passes a solution passes through a batch mode device, by the length of

membrane utilized in a single pass. By doing so, it is possible to better compare these technologies

utilizing the energy consumption scaling relations presented in Section 5.2. This is because it is

well established (experimentally) that longer membrane lengths are beneficial in terms of reducing

the current density, which leads to higher energy efficiency. Most of the non-batch mode ED

systems (pilot scale and beyond) are utilizing much longer channel lengths, even up to 1m or

longer. This inverse relationship between energy consumption and membrane length is seen in

Equation 5.1.

For the S-ED system utilized by Xiao-Yao Nie et al. in [1,2], the flow rates examined were

between 0.6m3/hr to 1.2m3/h (4.3cm/s-8.5cm/s linear flow velocity). Both the concentrate and

diluate reservoirs had volumes of 25L in the optimal process. For the S-ED system utilized by Zhi-

Yong Liu et al. in [3-5] were between 40L/hr to 160L/hr, or equivalently 0.04m3/hr to 0.16m3/hr

(1.9cm/s-7.6cm/s linear flow velocity). The concentrate and diluate reservoirs were 1.5L and 2.5L

83
respectively, however in subsequent investigations the higher the concentration reservoir volume

showed better performance (2L concentration reservoir). Because the electrochemical separation

effects occur only when the solution passes across the membrane, a simple method to calculate an

equivalent effective membrane length for an equivalent single pass system (continuous operation)

is to calculate the total volume processed over the experiment period divided by the volume of the

solution in the reservoirs.

For the purpose of calculating the effective membrane length, we will conservatively

estimate that the volume of both the concentrate and diluate reservoirs in the second group were

2.5L. A greater reservoir volume means fewer passes for a given flow rate and given device

dimensions. This allows for a simplification that both the concentrate and diluate reservoirs are

cycled through the device the same number of times. The effective membrane length is given as

follows:

Q •\M t
L@€€ = ∗L (5.2)
V|@x

where L@€€ is the effective membrane length, Q •\M is the volumetric flow rate, t is the duration of

experiment, and V|@x is the volume of each reservoir, and L is the length of membrane in a single

pass of solution through the device. Table 5.2 presents the effective length calculations of the S-

ED references in the previous section, including the length of the device used in this study.

Using the above effective length calculation, an energy scaling factor is developed for a

length-matched continuous system to a corresponding batch mode system. In equation 5.1, the

length of an electromembrane system critically affects the energy consumption, as energy

consumption is inversely proportional to membrane length. The energy scaling factor for

84
membrane length comparing the ICP continuous mode technology in this study to batch mode S-

ED technologies in literature is given as:

Lb€€ †Gb‡ (5.3)


SFM@ƒ_„… =
Lb€€ ˆ‰c

5.4.2 Main Channel Power Consumption

A major benefit of stacked membrane devices such as the ICP system or ED is the ability

to scale throughput but stacking additional channels at the expense of additional membrane cost.

When scaling the ICP system, we can estimate the power consumption of a device with a larger

number of stacked channels by multiplying the power contribution of a single channel with the

number of channels. As the number of channels increases, we expect that the power consumption

by the electrodes and buffer channels would diminish, and the overall power consumption would

tend towards the product of the main channels power consumption and the number of channels.

To determine the power contribution of the main channels, we can directly measure the

voltage across the main channel by inserting electrodes into the membrane stack. As opposed to

the experimental setup in Chapter 3, rather than measuring only the power consumption across the

entire device including the electrode rinse channels and the buffer channels, this allows us to more

precisely scale the device with respect to power consumption when additional channels are added.

Figure 5.3 shows a schematic of the device detailing the position of the main channel voltage

measurement electrodes with respect to the membrane and spacer locations, with Figure 5.4

showing pictures of experimental implementation of this schematic. In this case, the power

consumption can then be calculated with the measured voltage and total applied current. Figure

5.5 shows results of this experiment with resepct to the voltage ratio defined in Figure 5.3 of a

25:1 brine under three flow rates considered for this economic evaluation.

85
𝐕𝐦𝐚𝐢𝐧 𝐜𝐡𝐚𝐧𝐧𝐞𝐥
𝐕𝐫𝐚𝐭𝐢𝐨 =
𝐕𝐞𝐧𝐭𝐢𝐫𝐞 𝐝𝐞𝐯𝐢𝐜𝐞

Vmain channel

Ventire device

Figure 5.3 Device schematic with location of electrode for voltage measurement. Measurement of the main channel
voltage (Vmain channel) allows for more precise determination of power consumption of the main channel. Because a one
channel device is used in this system, a significant portion of power is consumed by the buffer channels and electrode
rinse channels. Precise measurement of power consumption of the main channel allows for more accurate energy
consumption scaling of the device in the case of a many channel system within a single pair of electrode channels.

86
Figure 5.4 Images of device with main channel electrodes. Sintered Ag/AgCl electrodes are inserted as shown into
the rubber spacers forming the channels (left). Because the depletion zone develops as the fluid flows, the voltage
drop across the channel increases with the development of the depletion zone due to increasing resistance of the
growing depletion zones. Placing the electrodes in the center of the channel allows to roughly capture the average
voltage drop which is used for power consumption calculations. As the rubber channels are secured within the 3D
printed electrode structure, the main channel electrodes can be easily accessed for potentiometric measurement (right).

87
Figure 5.5 Measurements of main channel voltage fraction versus applied current over varying flow rates. Voltage
fraction is calculated as the ratio of Vmain channel to Ventire device as in the schematic of Figure 5.3.

In many electro-membrane systems, increasing the number of membrane stack units

between electrodes allows for space efficient enhancement of throughput. By the above plot, we

see that for the 1mm/s and 2mm/s case, the voltage ratio of the main channel to the entire device

approaches 0.4. As a conservative estimate, we will use 0.6 as the tending behavior of the

0.5mm/sec condition. As the number of channels increases, the average power consumption

fraction of the main channels including the electrode and buffer channels approaches 0.4, resulting

in 0.4 ⨉ ELi as the resulting specific energy consumption. For example, a device with 20 main

channels and 2 buffer and 2 electrode rinse channels, the total power consumption normalized to

the power consumption of a single channel device is 0.6+20⨉0.4, where the 0.6 is the contribution

of the buffer and electrode rinse channels, and the 20⨉0.4 term is the contribution of the 20 main

channels. Averaging this power consumption over the 20 main channels and buffer and electrodes

88
(considering these as a single channel as they are constant across all devices) results in an average

power consumption of 0.4095 kWh/g Li+, indicating that the effect of the electrode rinse and buffer

channels grows negligible to the overall power consumption. While the energy consumption of the

device grows linearly with the number of channels, the mass flow rate of lithium through the device

also increases linearly with the number of channels, and thus the specific energy consumption is

not affected in this way, but tends to the single channel energy consumption with growing number

of channels.

5.4.3 Scaling the ICP Device, and Comparison with S-ED

Table 5.2 reports the scaled results of several performances over output Mg2+:Li+ brines

that satisfy the requirement of below 10:1 post processing the ICP device. The brines are scaled

both according to the effective length as outlined through Section 5.4, and number of membrane

stacks that the respective S-ED references provide. This serves to provide an order of magnitude

estimate to determine the competitiveness of the ICP platform for reduction of the Mg2+:Li ratio

of brines compared to S-ED.

By examining the scaled values of ELi in Table 5.2, we see that for the given performances

in terms of Mg2+:Li+ reduction are within 1-2 orders of magnitude of the performances of some S-

ED investigations (see Appendix D for additional experimental parameters producing output

brines below 10:1 Mg2+:Li+ and scaled energy consumptions). However, one major limitation of

this investigation is the extremely low grade of Li used in the feed brine (10mg/L). Due to the

short membrane length used in the ICP device, a relatively low concentration of brine can only be

processed. In the concluding section, this and other limitations are discussed, along with possible

routes for addressing these limitations for improvement in future work.

89
5.5 References

[1] X. Y. Nie, S. Y. Sun, Z. Sun, X. Song, and J. G. Yu, “Ion-fractionation of lithium ions from magnesium ions by
electrodialysis using monovalent selective ion-exchange membranes,” Desalination, vol. 403, pp. 128–135, 2017.
[2] X. Y. Nie, S. Y. Sun, X. Song, and J. G. Yu, “Further investigation into lithium recovery from salt lake brines
with different feed characteristics by electrodialysis,” J. Memb. Sci., vol. 530, no. February, pp. 185–191, 2017.
[3] Z. yong Ji, Q. bai Chen, J. sheng Yuan, J. Liu, Y. ying Zhao, and W. xian Feng, “Preliminary study on
recovering lithium from high Mg 2+ /Li + ratio brines by electrodialysis,” Sep. Purif. Technol., vol. 172, pp. 168–
177, 2017.
[4] Q. B. Chen, Z. Y. Ji, J. Liu, Y. Y. Zhao, S. Z. Wang, and J. S. Yuan, “Development of recovering lithium from
brines by selective-electrodialysis: Effect of coexisting cations on the migration of lithium,” J. Memb. Sci., vol. 548,
no. 8, pp. 408–420, 2018.
[5] Z. Y. Guo et al., “Prefractionation of LiCl from concentrated seawater/salt lake brines by electrodialysis with
monovalent selective ion exchange membranes,” J. Clean. Prod., vol. 193, pp. 338–350, 2018.
[6] H-J. Lee et al., “Designing of an electrodialysis desalination plant,” Desalination, vol. 142, pp. 267-286, 2002

90
6 Concluding Remarks and Future Directions

6.1 Concluding Remarks

In this work, we present a novel membrane-based technique for reducing the Mg2+:Li+ ratio

of brines. Utilizing locally amplified electric fields induced by ion concentration polarization, a

scalable, continuous technique for separation of magnesium ions from lithium brine solutions can

be realized based on the differences in electrophoretic mobility of the ionic species. This

preliminary study investigated the effects of applied current, convection rate, and feed Mg2+:Li+

ratio on the performance of the reduction of the Mg2+:Li+ ratio of the output processed brine.

Generally increasing applied currents resulted in enhanced reduction of the Mg2+:Li+ ratio at the

expense of Li recovery, along with increasing specific energy consumption. While the device used

was capable of reducing the 25:1 brines in this study to less than 5:1 over a variety of flow rates

satisfying the below 10:1 reduction criteria for subsequent lithium precipitation as discussed in the

introduction, the device was able to reduce a 60:1 Mg2+:Li+ brine to ~10:1 and a 100:1 Mg2+:Li+

brine to ~20:1.

Furthermore, a comparison between the novel ICP platform, and the prominent membrane

based technique in the literature, selective electrodialysis utilizing monovalent ion exchange

membranes for Mg2+:Li+ reduction, was made to economically evaluate the energy consumption

of the ICP device. A rudimentary technique for comparisons between batch mode technologies

and continuous mode technologies was presented to account for the effective membrane length

experienced by the feed solution, along with energy consumption scaling arguments as a function

of membrane length for order of magnitude energy consumption estimates. Considering the 25:1

brine, this continuous technology was within 1-2 orders of magnitude when compared with some

91
selective electrodialysis reported specific energy consumptions, despite being a proof of concept

demonstration.

92
6.2 Future Directions

6.2.1 Addressing Device Limitations

While the proof of concept investigation in this study demonstrated the possibility for a

continuous ion separation platform, there are limitations that must be addressed during future

investigations and scaling of the platform. The brine composition of the brines followed a 1:R:2R

Li+:Mg2+:Na+ ratio, where R is the Mg2+:Li+ ratio of the feed brine. Often the sodium is 1-2 orders

of magnitude higher in concentration compared to other ions like magnesium. While scaling the

overall brine concentration (10:10R:20R) may result in minimal changes to the specific energy

consumption (the energy consumption is approximately linear with respect to concentration, along

with the mass flow rate increasing linearly with concentration of lithium, thus these effects cancel

[1]), increasing the concentration of ions other than lithium (1:R:20R) would dramatically increase

the specific energy consumption as the absolute energy consumption would rise, without the

additional mass flow rate of lithium increase that cancels in the overall scaling of concentration of

the brine. Nevertheless, demonstration of continuous ion separation based on electrophoretic

mobility can be evidenced as by the normalized concentrations shown in Appendix C, Figure C2.

The normalized collect channel concentrations can give information about the relative rate of

depletion of the ions. Here, we see a persistent trend of magnesium being depleted the greatest for

a given current, followed by sodium, and finally lithium with respect to normalized concentration.

Another limitation that must be addressed is the lithium recovery of the device. Compared

to the selective electrodialysis, the lithium recovery is quite low (>90% for some reported selective

electrodialysis investigations verses <10% for the ICP investigation). Improving the lithium

recovery could account for a near order of magnitude improvement in the specific energy

consumption. In fact, incorporation of a concentration mechanism of the lithium could well exceed

93
an order of magnitude improvement in energy consumption. However, the device architecture

limits the lithium recovery in this case, and must be altered to improve this limitation. In the current

device setup, the lithium recovery is restricted to 50%, as the feed brine in the waste channel has

no opportunity to be collected. The Mg2+:Li+ ratio occurs at the expense of Li recovery due to the

nature of the migration of ions towards the waste channel in the ICP phenomenon. One possible

technique could be to adopt a similar waste feed free of Li (like an NaCl solution) which would

act as the waste collector versus using the feed brine as the waste collection solution. A similar

technique is used in the S-ED references discussed in Section 5, in which an NaCl solution was

used to collect the Li+ ions removed from the Li brine solution, while the Mg2+ ions were retained

in the feed.

Moving forward with this device, increasing the length scale will be critical to improving

performance. Both the specific energy consumption will improve as given by the scaling relation

in Section 5 (equation 5.1), but additionally it may be possible to handle a larger variety of feed

compositions with respect to the Mg2+:Li+ ratio. As mentioned above, improvement of the lithium

recovery ratio is also crucial to improvement of the device, perhaps through the ideas mentioned

in the preceding paragraph, or by considering alternative membrane stack geometries or

configurations for improving the lithium recovery.

6.2.2 Generalized Ion Separation Technique

One major benefit of the ICP based technology over other membrane-based technologies

lies in utilization of mobility differences in ions for separation, as opposed to properties of the

membrane itself. This allows for the technique to be potentially applied to a wide array of problems

not restricted to monovalent ions vs multivalent as made capable by monovalent IEMs. While the

94
technology in the investigation is young, maturation of this technique would yield powerful

separation and purification tools to tackle many challenges in a continuous, scalable fashion.

One pressing example of the need for a generalizable ion separation technique is in that of

recycling spent lithium ion batteries (LIBs). Many precious metals are used in addition to lithium

in LIBs, such as nickel and cobalt. In fact, LiCoO2 cathodes are one of the most common

chemistries employed in LIBs [2]. During the recycling process, a liquor of various metals is

leached from the mechanically ground and separated battery components with the goal of

recovering the precious metal content [3]. This can involve a variety of chemical additions and

pH-controlled steps for selective precipitations from the leached solution containing a mixture of

these precious metals. Often this requires many reagents that can be environmentally unfriendly

or difficult to dispose of. In the case of separation of cobalt from lithium in leach liquor solutions,

Cyanex 272 is used, which is a chemical capable of selectively extracting cobalt with minimal co-

extraction of lithium, but results in high operating costs of the process [3]. Furthermore, in the case

of nickel extraction through electrolysis, aluminum and copper, which are used in various battery

components, must be first separated to prevent contamination of the nickel product during the

electrolysis. A fully matured ICP technology has potential to perform all of these purifications and

separation steps in a single, unified technology, which could dramatically simplify this complex

procedure, as well as simply plant scale operations involving battery recycling.

95
6.3 References

[1] H-J. Lee et al., “Designing of an electrodialysis desalination plant,” Desalination, vol. 142, pp. 267-286, 2002
[2] M. Vanitha, N. Balasubramanian, “Waste minimization and recovery of valuable metals from spent lithium-ion
batteries – a review,” Environmental Technology Reviews, vol. 2, pp. 101-114, 2013
[3] C. Ekberg, M. Petranikova, Lithium Production Processes. Elsevier Inc., 2015.

96
7. Appendix

Appendix A: Lab Brine Compositions

Table A1. Ideal brine compositions and concentrations of ions. Measured values of salts mixed
into DI water are presented additionally.

Brine Experiments LiCl MgCl2 NaCl Calculated Calculated Calculated


Used (mg/4L) (g/4L) (g/4L) [Li+] (mg/L) [Mg2+] (mg/L) [Na+] (mg/L)
Ideal (25:1 - 244.3 3.917 5.084 250 500
Mg2+:Li+ )
Measured .5mm/sec 3 243.7 3.927 5.087 9.975 250.6 500.3
25:1 trials
Measured 1mm/sec 3 250.2 3.953 5.105 10.24 252.3 502.1
25:1 trials
Measured 2mm/sec 3 250.5 3.932 5.091 10.25 251.0 500.7
25:1 trial
Ideal (60:1 - 244.3 9.401 5.084 10 600 500
Mg2+:Li+ )
Measured 1mm/sec 1 245.2 9.404 5.082 10.04 600.2 499.8
60:1 trials
Measured 1mm/sec 2 245.6 9.405 5.091 10.05 600.3 500.7
60:1 trials
Ideal (100:1 - 244.3 15.67 5.084 10 1000 500
Mg2+:Li+ )
Measured 1mm/sec 3 244.5 15.71 5.087 10.01 1002 500.3
100:1 trials

97
Appendix B: Source Conditions:

Table B1. Keithley 2400 source conditions used for applying constant current steps over the
duration of the experiments, along with data acquisition parameters.

Brine Flow Rate Sampling Current Step Current Step Replicates


Period Size Duration
25:1 0.5mm/sec 5s 6mA 10min 3
25:1 1mm/sec 5s 8mA 8min 3
25:1 2mm/sec 5s 10mA 8min 3
60:1 1mm/sec 5s 12mA 8min 3
100:1 1mm/sec 5s 12mA 8min 3

98
Appendix C: ICP-OES Analysis and Verification

In order to produce a calibration curve for analyzing collected sample concentrations, first a
standard curve must be produced mapping concentration to intensity. Known concentrations of
Na+, Li+, and Mg2+ were prepared through serial dilution of purchased ICP standards from Sigma
Aldrich according to the following concentrations. Multielement (ME) standards were used
composed of the three ions characterized below. Standards were measured periodically (at least
three times) throughout the measurement runs to verify consistency in measurement of the ICP-
OES tool. MEi refers to the ith dilution step of the standard.

Table C1. Table of prepared multielement calibration standards through standard dilution for ICP-
OES calibration.

Ion ME0 ME1 ME2 ME3 ME4 ME5 ME6 ME7 ME8 ME9
Mg 50 25 10 5 2.5 1 .5 .25 .1 .05
(ppm)
Li 5 2.5 1 .5 .25 .1 .05 .025 .01 .005
(ppm)
Na 50 25 10 5 2.5 1 .5 .25 .1 .05
(ppm)

Calibration curves were produced by fitting a linear regression to the resulting intensity measured
by the standards to the corresponding concentrations. Examples of these are shown below.

99
Figure C1. Example calibration curves for ICP-OES analysis. Standard solutions of 1000mg/L
were purchased from Milipore Sigma, and several standard concentrations were made by serial
dilution. The intensities measured from these solutions of known concentrations were used to
analyze the concentrations of collected solutions with unknown concentrations using linear fits to
the known intensity-concentration curves.

100
Verification 1: Reservoir Measurement
The first verification of ICP-OES measurement is by comparing the calculated concentration of
samples passed through the device with zero applied field. These should closely agree with the
calculated concentrations from the salt measurements in Appendix A. Table C2 below presents the
calculated concentrations for comparison with Table A1. The values in each field represent the
average over all collect and waste measurements for the set of trials listed in the left-most column.

Table C2. Measured concentrations of feed solutions by ICP-OES for verification of ICP-OES
measured values. In comparison with Table A1, these values indicate good matches with the
expected values based on measured salt masses. Note that the sodium concentrations appear to
have the largest variation and deviation from the expected values.

Experiment Description Li (mg/L) Mg (mg/L) Na (mg/L)


25:1 Ideal 10 250 500
25:1 0.5mm/sec Trial 1,2,3 9.958 239.727 451.508
25:1 1mm/sec, Trial 1,2,3 9.952 254.728 471.692
25:1 2mm/sec Trial 1,2,3 9.636 234.110 443.775
60:1 Ideal 10 600 500
60:1 Trial 1 10.668 611.966 501.961
60:1 Trial 2,3 10.082 589.627 463.080
100:1 Ideal 10 1000 500
100:1 Trial 1,2,3 10.473 1019.653 481.797

Verification 2: Waste Channel Measurement


Another verification of ICP-OES that can be used is by studying the measurements of the waste
channel. As the applied current increases, the size of the depletion layer continues to expand. At
the highest values of applied current in this experiment, we then expect that the collet channel is
near to complete depletion, and thus the concentrations values of the waste channel approach twice
the value of the feed concentration. By plotting the normalized concentrations of each species, we
can verify this trend, which gives additional confidence in the ICP-OES measurements. As seen
in the normalized concentration plots of Figure C1, we can see this trend.

101
Figure C2. Plots of normalized concentration versus applied current across all experimental
conditions. The collect stream lines are solid, while the waste stream lines are dotted. Lithium
collect and waste streams are depicted in red, while magnesium is in blue, and sodium is in green.

102
Appendix D: Scaled Power Consumptions of ICP Device by Effective Length

Table D1 Selected results of scaled specific energy consumption for an order of magnitude
comparison between the ICP technology and S-ED. Results are from 25:1 brines that yielded <10
Mg:Li post processing in the ICP device. To produce the scaled ELi field, the measured ELi from
the experimental data in Figure 5.2a. was divided by the ratio of the Leff between the S-ED and
ICP devices according to the compared reference, and multiplied by the power consumption ratio
as measured in the experiment in Section 5.3.2.
This Study Scaling Comparison with Scaling Comparison
Reference [1,2] with Reference [3,5]
Input Flow Rate Ouput ELi Leff, Leff Leff, Leff
RLi Scaled ELi Scaled ELi
Mg2+:Li+ (uL/min) Mg2+:Li+ (kWh/g Scale Factor Scale
(%) (kWh/g Li+) (kWh/g Li+)
Li+) Factor
8.02 7.15 20.31 0.0293 0.0501
6.6 5.95 31.19 0.045 0.0769
125 6 5.40 45.06 0.065 0.1111
5.29 4.87 69.19 0.0998 0.1705
4.97 4.41 96.29 0.1389 0.2374
5200cm, 3042cm
24.8±0.4 8.78 9.40 8.804 0.0127 0.0218
693.3 405.6
250 6.46 6.55 17.05 0.0246 0.0421
5.14 4.2 34.31 0.0495 0.0846
9.3 9.55 13.58 0.0196 0.0336
500 6.84 6.90 21.97 0.0317 0.0542
4.08 4.54 40.35 0.0582 0.0995

103

You might also like