g model article in press - universiteit twente · for a variety of colloids [15–20], including...

C

Ce

YMP

h

•

•

•

•

•

a

ARRAA

KSCZFSR

h0

ARTICLE IN PRESSG ModelOLSUA-20132; No. of Pages 8

Colloids and Surfaces A: Physicochem. Eng. Aspects xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Colloids and Surfaces A: Physicochemical andEngineering Aspects

journa l homepage: www.e lsev ier .com/ locate /co lsur fa

harge inversion and colloidal stability of carbon black in batterylectrolyte solutions

an Zhang, Aditya Narayanan, Frieder Mugele, Martien A. Cohen Stuart,ichel H.G. Duits ∗

hysics of Complex Fluids Group, University of Twente and MESA+ Institute, PO Box 217, 7500 AE Enschede, The Netherlands

i g h l i g h t s

Specific adsorption of cations on car-bon black in carbonate solvent.Charge inversion at 2 mM for Li+,20 mM for Na+.First flocculation regime at0.1–0.5 mM.Re-entrant colloidal stability for Li+

around 10 mM.All systems flocculated at high saltconcentration.

g r a p h i c a l a b s t r a c t

r t i c l e i n f o

rticle history:eceived 15 July 2015eceived in revised form 24 August 2015ccepted 26 August 2015vailable online xxx

eywords:emi solid flow batteriesarbon blacketa potentiallocculation

a b s t r a c t

We studied the influence of salt on a commercially available carbon black (Ketjenblack 600, KB) in car-bonate solvents commonly applied in rechargeable batteries. Adopting the typically used salts: lithiumhexa-fluorophosphate (LiPF6), lithium bis(trifluoromethane sulfonyl) imide (LiTFSI), as well as sodiumhexafluorophosphate (NaPF6) dissolved in mixtures of ethylene carbonate and propylene carbonate, weinvestigated both the zeta potential and the flocculation kinetics of the KB particles as a function of saltconcentration between 0.01 mM and 1.0 M. Clear evidence was found for the preferential adsorption ofcations. In the absence of salt, KB was found to carry a negative surface charge, but this gets neutralizedby Li+ at very low concentrations (∼1 mM), and by Na+ at intermediate concentrations (∼30 mM). In thecase of lithium ions, the increased adsorption at higher concentration led to a recovery of the colloidalstability around 3–30 mM, depending on the anion. At high concentrations exceeding 30–100 mM, all

tability ratioe-entrant stability

salts cause flocculation of the KB particles, due to a reduction of the electric double layer thickness. Sincethe charge neutralization of the KB by Na+ takes place in the same concentration regime, no re-entrantstability is found for Na+. These findings could have implications in formulation protocols for semi-solidflow batteries, or other systems where an intermediate stable regime could assist mixing and/or structureformation at small length scales.

© 2015 Elsevier B.V. All rights reserved.

Please cite this article in press as: Y. Zhang, et al., Charge inversion andColloids Surf. A: Physicochem. Eng. Aspects (2015), http://dx.doi.org/1

∗ Corresponding author. Fax: +31 53 4891096.

ttp://dx.doi.org/10.1016/j.colsurfa.2015.08.041927-7757/© 2015 Elsevier B.V. All rights reserved.

colloidal stability of carbon black in battery electrolyte solutions,0.1016/j.colsurfa.2015.08.041

1. Introduction

The development of sustainable energy storage systems hasbecome an urgent issue due to the limited amount of fossil

dx.doi.org/10.1016/j.colsurfa.2015.08.041


http://www.sciencedirect.com/science/journal/09277757

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


ING ModelC

2 Physic

fapemflsaftotaitc

cmebvdlcfabtsiC

ttspwcVe

edtisenaiaBsl(ilbcst

itt

ARTICLEOLSUA-20132; No. of Pages 8

Y. Zhang et al. / Colloids and Surfaces A:

uels/natural gas and their huge rate of consumption. Recharge-ble lithium ion batteries are one of the most developed in theast few decades and commonly applied in portable electronics,lectric vehicles, aerospace devices, etc. [1]. Flow batteries offerore flexibility than solid form batteries, but the energy density of

ow batteries is relatively low [2]. A new type, the so-called semi-olid flow battery (SSFB), was invented by MIT and has both thedvantage of flow battery’s flexibility and the high energy densityound in solid batteries [3]. SSFBs differ from other batteries in thatheir electrodes are dispersions or suspensions, often composedf conductive nanoparticles (CNPs, often carbon black) and elec-rochemically active particles (EAPs, normally metal oxides whichccommodate or release lithium ions during charge and discharge)n electrolyte solutions. Electrons reacting at the redox active par-icles are conducted by the CNPs to and from current collectorsonnected to an external circuit.

Since the emergence of SSFBs in 2011, optimization of theiromposition (EAPs, CNPs, solvent, salt) has been explored [4,5] byeasuring the rheological, electrochemical and conductive prop-

rties of the fluid electrode in order to understand and improveattery performance. Here colloidal interactions play a crucial role,ia their determining influence on both the structure and theynamics of the fluid. The CNPs, in spite of being present at much

ower volume fractions than the EAPs, often dominate both theonductive and rheological behavior of the entire fluid via theormation of a flocculated network. Besides that, the colloidal inter-ctions between active and conductive particles can influence theattery performance [6,7]. However, the situation at present ishat only little is known (and much is left to optimize) about thetructure and strength of the (EAP surrounded) CNP network, anmportant reason being that the colloidal interactions between theNPs are yet to be fully understood.

Both the specific solvent and the used salts (the electrolyte solu-ions) in SSFBs are the reason for this lack of knowledge. Similaro classical lithium batteries, the electrolytes in SSFBs are typicallyolutions of Li or Na salts dissolved in carbonate solvents. For exam-le, Li(Na)PF6, Li(Na)TFSI, Li(Na)BF4, Li(Na)ClO4 are used as salts,hile ethylene carbonate (EC), propylene carbonate (PC), dimethyl

arbonate (DMC) and diethyl carbonate (DEC) are used as solvents.arious combinations of these salts and solvents result in differentlectrochemical properties [8–10].

In more common solvents such as water, ethanol, hydrocarbons,tc., the behavior of Carbon Black has been better studied. Since vaner Waals attractions between CBs are ubiquitous, surface oxida-ion followed by surface modification generally has to be appliedn order to disperse carbon black [11–13]. In aqueous systems, theurface charge likely originates from carboxylate groups, while thextent of screening can be controlled via the salt concentrations. Inon-polar solvents there is little screening of the electrostatic inter-ctions, while the surface potential can vary per system (the origins not always clear). In the battery community, the mixture of ECnd PC is recognized as one of the best performing solvents [8,14].oth EC and PC cannot be ranked amongst conventional solvents,ince they are aprotic while having a high dielectric constant. Theatter allows dissociation of certain salts up to high concentrationsas needed for achieving high conductivity). In turn, the dissociatedons can change both the surface potential and the electric doubleayer thickness. This could mean that the electrostatic interactionsear similarity to those in aqueous systems. However, even in thatase there might be practical differences in the behavior at lowalt concentrations, because the sensitivity of carbonate solventso impurities is quite high.


A particular process that could modify the electrostatic particlenteractions, is specific ion adsorption. Depending on the signs ofhe native surface charge and the adsorbing ion, even reversal ofhe surface charge is possible. This phenomenon has been studied

PRESSochem. Eng. Aspects xxx (2015) xxx–xxx

for a variety of colloids [15–20], including many oxides in aqueousenvironments [15–17]. Sign reversal of the zeta potential inducedby alkali metal cations such as Li+ and Na+ has been found in a fewcases (e.g. [21]), while for protons it is a well-known phenomenon,also in non-aqueous solvents [15].

In this paper we focus on the principal aspects of the colloidalstability of Carbon Black suspensions in electrolyte solutions con-taining the same ingredients as in a typical SSFB. We chose threeelectrolyte solutions, namely LiTFSI, LiPF6 and NaPF6 dissolved in a1:1 binary mixture of EC and PC. Measurements of the zeta potentialas a function of salt concentration are combined with determi-nations of the time-dependent average hydrodynamic (aggregate)diameter with Dynamic Light Scattering (DLS). By starting the DLSexperiments with the suspensions in a non-flocculated state, wemeasure the flocculation rate, and calculate the stability ratio. Com-bining the observations from the two types of experiments allowsus to draw conclusions about the ion adsorption and stabilizationmechanism.

2. Theory

In this section we summarize the approach and key equations asused to measure the stability ratio of the KB particles, as a functionof salt concentration. The initial stage of the flocculation of particleswith uniform size and shape can be described via the second-orderreaction:

dNdt

= −kN2 (1)

where N is the concentration of the non-flocculated particles andk is the flocculation rate constant. Early stages of the flocculationprocess can be analyzed by measuring the average hydrodynamicradius (Rh) with DLS as a function of time [22]. The linear depen-dence of Rh on time in this regime allows to determine dRh/dt,which is proportional to the flocculation rate. The average hydro-dynamic radius is calculated using the Stokes–Einstein equation:

D = kBT

6��Rh(2)

where D is the measured diffusion coefficient, � the viscosity ofthe medium and kB the Boltzmann constant and T the absolutetemperature. Unlike many aqueous salt solutions, the viscosity ofsolutions in carbonate solvent depends significantly on the saltconcentration: it nearly triples when the concentration of LiTFSI[23], LiPF6 [24] and NaPF6 [25] reaches 1 M (see also Appendix 1).This is of importance when calculating Rh, but also for the propernormalization of the flocculation rate: in more viscous media, thediffusion-limited flocculation of particles takes longer. This aspect,expressed by the factor KSmol = 8kBT/3� in the Smoluchowskiequation is taken into account by multiplying �r and dRh/dt, where(�r = �sa/�so with �sa the viscosity of the salt solution and �so like-wise for the solvent). The stability ratio [26,27] is then definedas:

W = [d(Rh × �r)/dt]max

d(Rh × �r)/dt(3)

This quantity equals 1 (regardless of the viscosity) in all casesof unhindered flocculation. Repulsive interactions give rise to anenergy barrier, which slows down the flocculation and henceincreases W. In the limit where the particles do not flocculate atall, W approaches infinity.


We remark that the use of dRh/dt to calculate the stability ratiorelies on the assumption that Rh increases linearly with time, whichcan only be expected in the early stages of the flocculation [28]. Thisimplies that the analysis should be performed well within the so-


IN PRESSG ModelC

Physicochem. Eng. Aspects xxx (2015) xxx–xxx 3

cf

t

3

3

A2<pNP(aNiia

pHpiEtcmszg

3

apscmf≈Swrf

W

wduRiufl

siatm

Table 1Particle radius (from SEM and DLS) and zeta potential of KB in EC PC 1:1 in absenceof salt.

RSEM (nm) Rh (nm) Zeta potential (mV)

KB ∼15 350 ± 88 −47



alled half-time, in which the number of entities is reduced by aactor of two:

1/2 = 3�/4kBTN0 (4)

. Experiments

.1. Materials and sample preparation

Carbon black (Ketjenblack 600JDP) (KB) was obtained fromkzoNobel (the Netherlands) and dried in a vacuum oven for4 h before use. Ethylene carbonate (EC) (anhydrous 99%+ purity,0.005% water) and propylene carbonate (PC) (anhydrous 99.7%+urity, <0.002% water) were obtained from Sigma–Aldrich (theetherlands) and used as received. The binary mixture of EC andC was 1:1 by volume (henceforth referred to as EC PC 1:1). LiTFSI99% purity) was obtained as a gift from Solvionic (France). LiPF6nd NaPF6 (99%+ purity) were both purchased from Alfa Aesar (theetherlands). All salts were used as received. Samples for SEM

maging were prepared by spreading a paste-like mixture of KBn ethanol on cleaned silicon wafers followed by drying in ambientir for half an hour.

KB stock suspensions (0.01%wt) were prepared by mixing theowder with EC PC 1:1 in an MBraun Argon-filled glove box (O2,2O < 0.5 ppm), followed by ultrasonic treatment for 15 min. Sam-les for DLS and zeta potential measurements were prepared by

nitially mixing the required amount of the stock dispersion withC PC 1:1 in polystyrene cuvettes; they were then removed fromhe glove box and sonicated for 15 min. Finally, salt solution (theoncentration varied from 1.5 × 10−4 M to 1.5 M) was added. Theixtures were then allowed to equilibrate for at least 24 h and

onicated for another 15 min before the experiments. DLS andeta potential measurements could only be performed outside thelovebox.

.2. Measurement of hydrodynamic radius and zeta potential

Hydrodynamic radii of the KB aggregates were measured as function of time in a Malvern Nano ZS zetasizer using sealedolystyrene containers. Immediately prior to each experiment, theample was sonicated for 5 min to obtain a well-defined, non floc-ulated starting state. Considering that the start-up time of the DLSeasurement amounts ≈2 min, and that about 30 min are needed

or a reliable measurement of dRh/dt, we have aimed t1/2 to be1 h. According to Eq. (4), N0 should then be ≈3 × 1014 particles/m3.ince our unit particles are sintered (i.e. permanent) aggregatesith complicated fractal-like shapes (see below), the quantitative

elation between the theoretical N0 and the experimental weightraction w is not known. In the expression:

= N0vp�p

�s(5)

ith vp the solid volume of a unit particle, and �p, �s the massensities of KB and solvent (respectively ≈1.8 and 1.25 g/mL), vp isnknown. Measurement of the hydrodynamic radius with DLS gaveh ≈ 350 nm. Assuming that the contour of the fractal-like particle

s a sphere with this radius, and estimating the internal filling vol-me fraction to be ≈0.06, we obtain w = 1.3 × 10−6. Using this mass

raction in the preparation of the suspensions, we obtained a goodinearity in the data of Rh vs time (see Fig. 4).

Zeta potentials were measured in the same zetasizer using aolvent-resistant dip cell. Samples were transferred into the cell


nside the glovebox, sealed during transportation to the Zetasizer,nd measured under Nitrogen protection. A thermal equlibrationime of 120 s at 25 ◦C was allowed before starting the measure-

ents. The measurement voltage was always in the range of

Fig. 1. SEM image of Ketjenblack.

4.5–20 V. The instrument calculates the electrophoretic mobilityUE from the average particle velocity and the electric field strength,where the latter is calculated from the measured ion current and theconductivity of the continuous phase. The zeta potential � is thencalculated from the electrophoretic mobility, using an appropriatemodel. We take the simplest equation from Smoluchowski:

UE = ε�

�(6)

with ε the dielectric constant of the continuous phase [29],for which we have taken that of the solvent. Considering theaforementioned complicated particle shape, we refrain from tak-ing electrophoretic retardation and relaxation corrections intoaccount; values reported here are therefore to be considered asestimates rather than precise determinations.

4. Results and discussion

4.1. KB in pure EC PC solvent

Carbon black contains 90–99% carbon; this makes it stronglyhydrophobic [30] and hardly dispersible in water. Generally, CB insuspension exists as units of chemically connected nanoparticles.Particle sizes have been reported to be 794 nm when suspended inwater, and 181 nm in ethanol, in both cases much larger than thesize of subunits that can be distinguished in electron micrographs[31]. Surface modifications such as oxidation introduce carboxylicgroups and significantly improve the colloidal stability in bothaqueous and non-aqueous solvents [32]. The adsorption of surfac-tants such as poly (oxometalate) is also effective to disperse carbonblack in water, as aggregates smaller than 100 nm [33].

Some key properties of the Ketjenblack particles used in ourexperiments are summarized in Table 1, and are further discussedbelow. SEM images of (Fig. 1) show structural subunits with diam-eters of around 20–30 nm. In contrast, the hydrodynamic radius Rhas measured by DLS was 350 ± 88 nm. This clearly indicates that


in EC PC 1:1 (without added salt), KB exists as structures that aremuch larger than the 20–30 nm units. To examine the nature ofconnections between the small building blocks, we subjected thesuspensions to dilution (down to the detection limit of the DLS) and


IN PRESSG ModelC

4 Physicochem. Eng. Aspects xxx (2015) xxx–xxx

ucwte

itsitctts

c(sbdpiswccdhacptast

sgofc

4

4

rbarhc(rl

wspctti

Fig. 2. Apparent molar conductivity of LiTFSI, LiPF6 and NaPF6 versus concentrationof salt. The background conductivity of the solvent was subtracted; For c < 0.1 mM,large error bars result after the subtraction and normalization. For c = 0.01 mM some



ltrasonication (for at least 1 h). In either experiment, no significanthanges in Rh were observed. Additional tests in which the particlesere left to settle due to gravity (a behavior that was manifested in

he flocculated systems) did not show any signs of sedimentation,ven after several days.

Each of these observations indicates that KB is colloidally stablen EC PC 1:1 (without added salt). The zeta potential of KB, measuredo be −47 mV, shows that the particles carry a significant negativeurface charge, which is likely responsible for the colloidal stabil-ty. The fact that the overall hydrodynamic radius is much largerhan that of the 20–30 nm structures indicates that the latter arehemically bound (rather than flocculated). No further effort wasaken to disperse KB into smaller sizes; the fractal-like 350-nm par-icles were treated as permanent and inseparable in this particularolvent.

The origin of surface charge for particles suspended in a liquidan be protonation or deprotonation of functional surface groupstypically hydroxyl) [34], adsorption of ions from solution [35],olvation and release of the ions from the particle [16], or a com-ination of these [36,37]. Which mechanism(s) is(are) applicableepends on the particle surface chemistry, whether the solvent isrotic or aprotic, the dielectric permittivity of the solvent (affect-

ng the solvation of ions), etc. In case of particles in non-aqueousolvents with a low dielectric constant, it is generally less clearhere the surface charge comes from. Impurity ions and moisture

ould play a role [38]. For carbon blacks, the origin of the surfaceharge is even more complex, due to its source or batch depen-ent surface chemistry. For example, carbon atoms may bond withydrogen and oxygen containing groups, such as quinone, ether,ldehyde and phenol, which act as defects. The zeta potential ofarbon nanotubes depends on the dielectric constant of the dis-ersing medium, and varies in both value and sign because electronransfer occurs at different defect sites [39]; this may apply to CBs well. Furthermore, Kosmulski [21] and Xu et al. [31] ascribed theurface charge and potential in non-aqueous solvents to an electronransfer between the particles and the solvent.

Specific for our system is that EC and PC have high dielectric con-tant (EC: 90; PC: 64) which could allow protons of the carboxylicroups on the CB surface to dissociate. [12]. It is also possible for thexygen lone pairs in EC and PC to transfer to the carbon black sur-ace. We speculate that the negative surface charge of KB in EC:PComes from a combination of these two mechanisms.

.2. KB in electrolyte solutions

.2.1. Ion dissociationAnticipating that the colloidal stability of KB will be intimately

elated to the presence of ions, we first examine the relationetween the amount of dissolved (1:1) salt per unit volume (c),nd the actual free ion concentrations. There are two extremeegimes where these two quantities may differ significantly. At theigh concentrations corresponding to SSFBs, the salts may not beompletely dissociated anymore, while at very low concentrations,trace) impurities from the materials themselves or from the envi-onment may significantly contribute to ionic composition of theiquid.

Measurements of the electric conductivity provide a suitableay to examine both regimes. They were performed with the Zeta-

izer as part of the determination of the zeta potential of the KBarticles. To assess whether in these measurements also the KB


ould contribute to the conductivity, we compared the conductivi-ies the salt-free solvent with and without KB. Both were measuredo be 1.7 ± 0.1 �S/cm, indicating that the contribution of KB isndeed negligible.

data points are shifted horizontally to allow distinction between different experi-ments.

4.2.2. Equivalent conductivitiesFig. 2 shows for each of the three salts, how the apparent molar

conductivities (conductivity due to added salt, divided by the con-centration) depend on the salt concentration. For concentrationsroughly between 0.1 and 100 mM, the (normalized) conductivitiesare constant per salt, indicating that the salts are completely disso-ciated, and contributions of other components than the salt alone,are negligible.

In the low salt concentration regime below 0.2 mM, the appar-ent molar conductivities fall slightly above the values expected forthe amount of dissolved salt. This could mean that (for the PF6 salts)some additional ions were via impurities. Assuming a specific con-ductivity comparable to HCl, Kortschot et al. [40,41], found impurityconcentrations of around 0.02 mM.

Water, as present in the starting materials, or as taken upfrom the ambient atmosphere might be a source of impurities.According to manufacturer, the water content of a EC–PC mixtureshould be <≈35 ppm, while other investigators have measured itto be <20 ppm [42]. Moisture in the system could change the sur-face potential of particles dispersed in nonaqueous systems [21].Furthermore, LiPF6 and NaPF6 are known to react with water toproduce HF and F− [43,44] while LiTFSI is rather stable. The rel-atively high equivalent conductivity of Li(Na)PF6 at <0.1 mM saltcompared with that of LiTFSI might thus be attributed to water.However, experiments aimed at corroborating this hypothesis byadding water, turned out inconclusive.

We now turn to the regime of high salt concentrations. BothEC and PC have a high dielectric constant, which favors ionization.However for c > 10 mM, it is indicated from the conductivities thation pair formation becomes significant. In literature, the degree ofdissociation (�) has been estimated to be 0.67 for 0.6 M LiTFSI [45],0.67 for 1.0 M LiPF6 [46] and 0.34 for 0.7 M NaPF6 [25] in EC PC 1:1.Inserting these numbers into the expression for the dissociationconstant (K):

K = ˛2c

1 − ˛(7)

and calculating the actual ion concentrations as a function of theconcentration of added salt, we find that the dissociation is almost


complete below 10 mM for each of the salts, while for higher con-centrations the degree of dissociation (at the same c) follows theorder: NaPF6 < LiTFSI < LiPF6. We remark that the equivalent con-


ARTICLE IN PRESSG ModelCOLSUA-20132; No. of Pages 8

Y. Zhang et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects xxx (2015) xxx–xxx 5

Ffs

dt

4

ccspPt1

paAsdl

saptrazSD

TF

ig. 3. Debye length against the inverse of the square root of the salt concentrationor LiTFSI, LiPF6 and NaPF6. Solid lines are drawn to guide the eye. Inset shows theame data in a log–log plot.

uctivities in Fig. 2 do not follow this trend; suggesting that in EC:PChe mobility of Na+ is higher than that of Li+.

.2.3. Debye lengthsFrom the measurements in Fig. 2 and and the assumed specific

onductivities, the Debye length �−1 can be calculated. In the cal-ulations we included the both the added ions and impurities in theolvent. The concentration of the latter was estimated from a com-arison with the conductivity of pure solvent, similar to [40,41]).ossible ion concentration changes due to the reaction with mois-ure were not taken into consideration. In Fig. 3 we plot it against/√c for the three electrolyte solutions.

In case of complete dissociation and absence of impurities, thislot should show a straight line with a slope of ∼10 mM0.5nm. Thisppears to be the case for concentrations in the range 0.25–5 mM.t lower concentrations (higher c−1/2), the effects of the impuritieshow up, while at higher concentrations the effect of incompleteissociation becomes visible as an upward curvature (see the

og–log plot in the inset).Clearly, the thickness of the electric double layer undergoes

trong variations as the salt concentration is varied between 10 �Mnd 1 M, reaching the nm range (where van der Waals interactionslay a role) at concentrations around 0.1 M (similar to aqueous sys-ems). A calculation of � ̨ (assuming the aggregate radius to beelevant) reveals that � ̨ � 1 (i.e. relatively thin double layers) for


ll salt concentrations investigated. This justifies the calculation ofeta potentials using the Smoluchowski equation (as was done inection 2). Another observation from Fig. 3 is that the variations inebye length between the different salts at the same concentration,

able 2locculation rate of Ketjenblack in LiTFSI solutions with different concentrations.

Concentration of LiTFSI (mM) Viscosity (mPa s)

0.16 2.32

1 2.32

2 2.33

4 2.33

8 2.35

16 2.38

64 2.53

128 2.75

256 3.21

512 4.21

1000 7.33

Fig. 4. Hydrodynamic radii multiplied by the relative viscosity for KB as a functionof time at different concentrations of LiTFSI. The lines are the linear fits.

are generally minor; suggesting that any differences in colloidal sta-bility of KB in presence of the different salts are not be related todifferences in screening.

4.2.4. Zeta potential and colloidal stability of KBFig. 4 shows some typical curves of the time-dependent hydro-

dynamic radius of KB particles in solutions of EC PC 1:1 containingdifferent amounts of salt (in this case LiTFSI). In this graph Rh ismultiplied with the relative viscosity of the salt solution (com-pared to salt-free solvent), to allow direct comparison of theslopes representing the flocculation rates (see Section 2). Theconcentration-dependent viscosities of the different salt solutionsare given in the Appendix 1.

The Rh values themselves are all close to 350 nm at the beginningof the experiment; this confirms that all experiments were startedwith non-flocculated particles. Table 2 lists the numerical data forall flocculation experiments with LiTFSI, together with estimatederrors.

It can be concluded from Fig. 4 and Table 2 that different regionsof colloidal stability of KB are found on varying the concentrationof the LiTFSI salt. At very low salt concentration (≤0.1 mM) thesize of the KB particles remained constant, indicating that the sus-pensions were colloidally stable. Between 0.1 and 10 mM LiTFSIflocculation occurred, corresponding to a first critical flocculationconcentration (CFC). However on the addition of more salt, colloidalstability was recovered, as evidenced by a constant Rh at concen-trations between 10 and 100 mM. Visual observation of samples in


this regime did not show any significant sedimentation, even afterweeks, thus confirming colloidal stability.

To allow a closer inspection and enable a mechanistic explana-tion, we now consider the stability ratio in conjunction with zeta

Normalized flocculation rate (nm/min) Stability

0.58 ± 0.20 Stable6.29 ± 0.21 Unstable7.85 ± 0.30 Unstable6.95 ± 0.14 Unstable6.92 ± 0.20 Unstable1.03 ± 0.28 Unstable0.22 ± 0.28 Stable0.81 ± 0.19 Stable1.36 ± 0.51 Unstable6.01 ± 0.57 Unstable6.89 ± 0.84 Unstable


ARTICLE ING ModelCOLSUA-20132; No. of Pages 8

6 Y. Zhang et al. / Colloids and Surfaces A: Physic

Fig. 5. Stability ratio (open blue circles) and zeta potential (filled black squares) of KBas a function of (a, top) LiTFSI, (b, middle) LiPF6 and (c, bottom) NaPF6 concentration.Si

pAis11Tcocd

occurs is around 1 mM for both LiPF6 and LiTFSI although there are

olid lines are drawn to guide the eye. (For interpretation of the references to colorn this figure legend, the reader is referred to the web version of this article.)

otential. Fig. 5a shows these data for KB in solutions of LiTFSI. very large change in zeta potential is observed, from –47 mV

n absence of salt (not shown) to +28 mV at 10 mM LiTFSI. Inver-ion of the sign of the zeta potential takes place between 1 and0 mM, while the loss of colloidal stability occurs between 0.1 and

mM, where the zeta potential is already close to zero (−15 mV).hese observations appear to be consistent with each other. Thehange in zeta potential is clearly caused by (preferred) adsorption


f cations, in this case (predominantly) Li+. Increasing the salt con-entration beyond 10 mM causes the zeta potential to graduallyecrease again, leading to a second loss of colloidal stability and


hence a second CFC at around 100 mM. The zeta potential is foundto be (now +) 15 mV in this regime. The reason for the second floc-culation is the strong electrostatic screening; the Debye length isonly ≈1 nm (see inset of Fig. 3). A CFC of around 100 mM is a typ-ical value for solvents of high dielectric constant; many aqueoussystems have similar CFCs for monovalent salts.

It is interesting to compare the findings for LiTFSI with those forLiPF6. If the adsorption of cations is the main reason for chargeinversion and re-entrant colloidal stability, similar findings areexpected for the two salts. Fig. 5b suggests that this is indeed thecase. However the salt concentrations where the transitions occur,appear to be somewhat smaller for LiPF6. Additionally, at the samesalt concentration, the zeta potentials of KB were higher in the caseof LiPF6 as compared to LiTFSI. Both differences might be due to thecreation of protons, through the reaction of PF6 anions with tracewater. Due to their small size, protons can be expected to adsorbrelatively easily as compared to larger ions. Moisture has also beenreported to cause charge reversal of silica particles in non-aqueoussuspensions [47].

Interestingly, a similar sign reversal of zeta potential from neg-ative to positive as that reported in Fig. 5 was induced by alkalimetal cations in other nonaqueous solvents like alcohols and diox-ane for silica, titania, and other materials [17]. Remarkably, the signreversal was observed at relatively low concentrations of cesiumand potassium, while lithium and sodium salts were less efficient.Recent papers [48,49] have addressed intercalation of several alkalimetals into carbon via DFT calculations.

The measurements on NaPF6 solutions allow another compar-ison with the case of LiPF6. The data shown Fig. 5c indicate thatremarkable differences exist. Although charge (and zeta potential)inversion also occurs with NaPF6, a significantly larger amount ofsalt is needed, suggesting that the tendency for Na+ to adsorb ontoKB, is weaker. Additionally the finding that the zeta potentials aregenerally lower for the NaPF6 case is consistent. The reason for thisweaker adsorption may be attributed to the much larger ion radiusof Na+ (0.19 nm) as compared to Li+ (0.07 nm). Remarkably, there-entrant colloidal stability found for LiPF6 is not observed withNaPF6. As the charge inversion for NaPF6 occurs at much highersalt concentration, screening due to the additional salt preventsre-stabilization in this regime.

As a final comment we briefly report a series of zeta potentialmeasurements with NaPF6, in which less strict precautions weretaken, i.e. using the dip cell in open air (rather than an atmosphereof nitrogen). In this case, where the samples become more easilycontaminated, the zeta potential was found to be more or less stableat –10 mV for NaPF6 concentrations below 1 mM. This less negativepotential (as compared to −20 to −30 mV in Fig. 5c) could suggestthat proton adsorption play a role. The fact that now only ∼1 mMNaPF6 was needed to achieve charge inversion, would be consistentwith that.

5. Conclusions

Carbon black particles carry negative a surface charge and formcolloidally stable suspensions in the binary solvent of ethylene car-bonate and propylene carbonate. The addition of salt neutralizesthe negative surface charge and causes flocculation of particles, fol-lowed by charge inversion at higher salt concentrations, due to thefurther adsorption of cations (lithium or sodium ions). For somesystems this gives rise to a re-entrant stability at salt concentra-tions around 10 mM. The salt concentration where charge inversion


minor differences due to specific cation-anion interactions. Sodiumions show a much lower adsorption affinity to our carbon black,causing the NaPF6 concentration needed for charge inversion to


ARTICLE ING ModelCOLSUA-20132; No. of Pages 8

Y. Zhang et al. / Colloids and Surfaces A: Physic

Table A1Concentration dependence of the viscosity of the 3 salt solutions. Literature valuesfor the viscosities of LiTFSI [23] and LiPF6 [24], and measured values for NaPF6 werefitted with second order polynomials, with concentrations expressed in mM andviscosities in mPa s. The mean relative error indicates the quality of the fit.

Salt Offset (mPa s) Linear term Quadratic term Mean relativeviscosity error

LiTFSI 2.39 0.0022 2.56 × 10−6 2%

bsSc

A

hrEu

A

A

t0

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

LiPF6 2.3 0.0015 3.53 × 10−6 8%NaPF6 2.33 0.0017 1.48 × 10−6 4%

e much higher (10–100 mM). Related to the stronger electrostaticcreening, a re-entrant stability is not found for the sodium salt.imilar patterns of re-entrant stability occur in aqueous systemsontaining specifically adsorbed (cat) ions.

cknowledgements

The authors thank Prof. J. Lyklema (Wageningen University) forelpful discussions, and ing. Daniel Wijnperlé for SEM imaging. Theesearch leading to these results has received funding from theuropean Union Seventh Framework Programme (FP7/2007-2013)nder grant agreement n◦ 608621.

ppendix 1.

Table A1.

ppendix A. Supplementary data

Supplementary data associated with this article can be found, inhe online version, at http://dx.doi.org/10.1016/j.colsurfa.2015.08.41.

eferences

[1] B. Scrosati, J. Garche, Lithium batteries: status, prospects and future, J. PowerSources 195 (9) (2010) 2419–2430.

[2] S. Hamelet, et al., Non-aqueous Li-based redox flow batteries, J. Electrochem.Soc. 159 (8) (2012) A1360–A1367.

[3] M. Duduta, et al., Semi-solid lithium rechargeable flow battery, Adv. EnergyMater. 1 (4) (2011) 511–516.

[4] M. Youssry, et al., Formulation of flowable anolyte for redox flow batteries:Rheo-electrical study, J. Power Sources 274 (2015) 424–431.

[5] L. Madec, et al., Electronic vs ionic limitations to electrochemical performancein Li4Ti5O12-based organic suspensions for lithium-redox flow batteries, J.Electrochem. Soc. 161 (5) (2014) A693–A699.

[6] M. Zhu, et al., Numerical study of interaction and aggregation of non-sphericalparticles in forming Li-Ion battery cathodes, J. Electrochem. Soc. 161 (9)(2014) A1247–A1252.

[7] M. Cerbelaud, et al., Numerical and experimental study of suspensionscontaining carbon blacks used as conductive additives in compositeelectrodes for lithium batteries, Langmuir 30 (10) (2014)2660–2669.

[8] K. Xu, Electrolytes and Interphases in Li-ion batteries and beyond, Chem. Rev.114 (23) (2014) 11503–11618.

[9] A. Ponrouch, et al., Non-aqueous electrolytes for sodium-ion batteries, J.Mater. Chem. A 3 (1) (2015) 22–42.

10] L.G. Chagas, et al., Unexpected performance of layered sodium-ion cathodematerial in ionic liquid-based electrolyte, J. Power Sources 247 (2014)377–383.

11] H. Liu, et al., Charging behavior of carbon black in a low-permittivity mediumbased on acid-base charging theory, J. Mater. Chem. C 3 (16) (2015)3980–3988.

12] M.M. Gacek, J.C. Berg, The role of acid–base effects on particle charging inapolar media, Adv. Colloid Interface Sci. 220 (2015) 108–123.

13] C. Eisermann, et al., Stabilization of carbon black particles withCetyltrimethylammonium bromide in aqueous media, Powder Technol. 253(2014) 338–346.


14] A. Ponrouch, et al., In search of an optimized electrolyte for Na-ion batteries,Energy Environ. Sci. 5 (9) (2012)8572–8583.

15] N. Boon, R. van Roij, Charge reversal of moisturous porous silica colloids bytake-up of protons, J. Colloid Interface Sci. 385 (2012) 66–72.

[

PRESSochem. Eng. Aspects xxx (2015) xxx–xxx 7

16] M.L. Jimenez, A.V. Delgado, J. Lyklema, Hydrolysis versus ion correlationmodels in electrokinetic charge inversion: establishing application ranges,Langmuir 28 (17) (2012) 6786–6793.

17] M. Kosmulski, E. Matijevic, Microelectrophoresis of silica in mixed-solvents oflow dielectric-constant, Langmuir 7 (10) (1991)2066–2071.

18] I. Semenov, et al., Electrophoretic mobility and charge inversion of a colloidalparticle studied by single-colloid electrophoresis and molecular dynamicssimulations, Phys. Rev. E 87 (2) (2013)022302-1–022302-7.

19] A. Martin-Molina, et al., The hydrophobic effect as a driving force for chargeinversion in colloids, Soft Matter 5 (7) (2009)1350–1353.

20] C. Calero, J. Faraudo, D. Bastos-Gonzalez, Interaction of monovalent ions withhydrophobic and hydrophilic colloids: charge inversion and ionic specificity,J. Am. Chem. Soc. 133 (38) (2011) 15025–15035.

21] M. Kosmulski, Zeta potentials in nonaqueous media: how to measure andcontrol them, Colloids Surf. A: Physicochem. Eng. Aspects 159 (2–3) (1999)277–281.

22] A.Y. Kim, J.C. Berg, Fractal aggregation: scaling of fractal dimension withstability ratio, Langmuir 16 (5) (2000) 2101–2104.

23] W. Xu, et al., Optimization of nonaqueous electrolytes for primary lithium/airbatteries operated in ambient environment, J. Electrochem. Soc. 156 (10)(2009) A773–A779.

24] K. Kondo, et al., Conductivity and solvation of Li+ ions of LiPF6 in propylenecarbonate solutions, J. Phys. Chem. B 104 (20) (2000) 5040–5044.

25] H.F. Bauman, J.E. Chilton, R. Mauri, Lithium-anode limited cycle batteryinvestigation, Lockheed Missiles and space Company, Technical reportAFAPL-TR-66-3S, 1966.

26] V.A. Tolpekin, et al., Stability ratio in binary hard sphere suspensions:measured via time-resolved microscopy, Langmuir 19 (10) (2003)4127–4137.

27] W.B. Russel, D.A. Saville, W.R. Schowalter, Colloidal Dispersions, CambridgeUniversity Press, 1989.

28] J.W. Virden, J.C. Berg, the use of photon-correlation spectroscopy forestimating the rate-constant for doublet formation in an aggregating colloidaldispersion, J. Colloid Interface Sci. 149 (2) (1992) 528–535.

29] A.V. Delgado, et al., Measurement and interpretation of electrokineticphenomena (IUPAC technical report), in Pure and Applied Chemistry, J.Colloid Interface Sci. 309 (2007) 194.

30] Y.N. Lin, P. Alexandridis, Temperature-dependent adsorption of pluronic F127block copolymers onto carbon black particles dispersed in aqueous media, J.Phys. Chem. B 106 (42) (2002) 10834–10844.

31] R. Xu, C. Wu, H. Xu, Particle size and zeta potential of carbon black in liquidmedia, Carbon 45 (14) (2007) 2806–2809.

32] H.S. Jang, D.-W. Park, S.E. Shim, Measurement of dispersion stability ofsurface-modified nanosized carbon black in various liquids, J. Nanosci.Nanotechnol. 7 (11) (2007) 3827–3829.

33] P. Garrigue, et al., Top-down approach for the preparation of colloidal carbonnanoparticles, Chem. Mater. 16 (16) (2004)2984–2986.

34] B.M. Lowe, C.-K. Skylaris, N.G. Green, Acid-base dissociation mechanisms andenergetics at the silica-water interface: an activationless process, J. ColloidInterface Sci. 451 (2015) 231–244.

35] T. Hiemstra, Variable charge and electrical double layer of mineral-waterinterfaces: silver halides versus metal (Hydr) oxides, Langmuir 28 (44) (2012)15614–15623.

36] A. Selmani, et al., Surface and zeta-potentials of silver halide single crystals:pH-dependence in comparison to particle systems, J. Phys. – CondensedMatter 26 (24) (2014) 244104-1–244104-10.

37] R.J. Hunter, Zeta Potential in Colloid Science, Principles and Applications,Academic Press, New York, 1981.

38] J. Lyklema, Principles of interactions in non-aqueous electrolyte solutions,Curr. Opin. Colloid Interface Sci. 18 (2) (2013) 116–128.

39] P.-C. Ma, et al., Correlation between electrokinetic potential: dispersibility,surface chemistry and energy of carbon nanotubes, Compos. Sci. Technol. 71(14) (2011) 1644–1651.

40] R.J. Kortschot, et al., Diverging electrophoretic and dynamic mobility of modelsilica colloids at low ionic strength in ethanol, J. Colloid. Interface Sci. 422(2014) 65–70.

41] V.S. Bhat, A.K. Srivastava, Ionic conductivity in binary solvent mixtures. 5.Behavior of selected 1:1 electrolytes in ethylene carbonate plus water at25 ◦C, J. Chem. Eng. Data 46 (5) (2001) 1215–1221.

42] M. Youssry, et al., Non-aqueous carbon black suspensions for lithium-basedredox flow batteries: rheology and simultaneous rheo-electrical behavior,Phys. Chem. Chem. Phys. 15 (34) (2013)14476–14486.

43] Z. Chen, et al., New class of nonaqueous electrolytes for long-life and safelithium-ion batteries, Nat. Commun. 4 (2013) 1513.

44] C.G. Barlow, Reaction of water with hexafluorophosphates and with Libis(perfluoroethylsulfonyl) imide salt, Electrochem. Solid State Lett. 2 (8)


(1999) 362–364.45] K. Hayamizu, et al., Pulse-gradient spin-echo H-1, Li-7, and F-19 NMR

diffusion and ionic conductivity measurements of 14 organic electrolytescontaining LiN(SO2CF3)(2), J. Phys. Chem. B 103 (3) (1999)519–524.


http://dx.doi.org/10.1016/j.colsurfa.2015.08.041











http://refhub.elsevier.com/S0927-7757(15)30184-9/sbref0005








































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































ING ModelC

8 Physic

[

[

[48] P.C. Tsai, et al., Ab initio study of sodium intercalation into disordered carbon,J. Mater. Chem. A 3 (18) (2015) 9763–9768.

[49] Z.H. Wang, et al., Diffusion of alkali metals in the first stage graphite



46] K. Hayamizu, Temperature dependence of self-diffusion coefficients of ionsand solvents in ethylene carbonate: propylene carbonate, and diethylcarbonate single solutions and ethylene carbonate plus diethyl carbonate


binary solutions of LiPF6 studied by NMR, J. Chem. Eng. Data 57 (7) (2012)2012–2017.

47] K.E. Tettey, D. Lee, Effect of thermal treatment and moisture content on thecharge of silica particles in non-polar media, Soft Matter 9 (30) (2013)7242–7250.



intercalation compounds by vdW-DFT calculations, RSC Adv. 5 (21) (2015)15985–15992.



























































































































g model article in press - universiteit twente · for a variety of colloids [15–20], including...

Documents