This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 17837--17845 17837

Cite this: Phys. Chem.Chem.Phys.,2013,15, 17837

Reference electrode assembly and its use in the studyof fluorohydrogenate ionic liquidsilicon electrochemistry

Boris Shvartsev,a Gil Cohn,a Hila Shasha,a Rudiger-A. Eichelb and Yair Ein-Eli*a

Silicon electrochemistry in fluorohydrogenate ionic liquids is partly hampered owing to the incapability

of producing an accurate and reproducible potential measurement due to a lack of appropriate

reference electrodes. This research work describes a simple assembly of a stable external reference

electrode enabling accurate studies of silicon electrochemistry in fluorohydrogenate ionic liquids. The

electrode configuration is based on the ferrocene/ferrocenium (Fc|Fc+) couple dissolved in the

EMIm(HF)2.3F (1-ethyl-3-methyl-imidazolium fluorohydrogenate)/Carbopol 941 gel. A stable potential of

2.5 wt% Carbopol-based electrode was measured versus a calomel reference electrode at 250 � 3 mV.

By utilizing the constructed electrode, an intensive electrochemical investigation on n-type silicon in

EMIm(HF)2.3F was conducted. Flat-band and open circuit potentials were measured, along with Si–air

half- and full-cell electrochemical measurements. A suggested mechanism for the n-type Si dissolution

process in the EMIm(HF)2.3F electrolyte, without illumination, is discussed as well.

Introduction

Room temperature ionic liquids (RTILs) offer unique propertieswhile serving both as solvents and as electrolytes. Their lowvapour pressure, even at considerably elevated temperatures,provides advantages such as easy containment and recycling.The high intrinsic ionic conductivity, non-flammability, thermal andchemical stability with considerably large electrochemical windows(in some cases as much as 5 V) make them suitable candidates forhigh end electrolytes in electrochemical applications.1,2

In general, one of the major setbacks in the research of RTILelectrochemistry is related to the incapability of generatingaccurate and reproducible potential measurements, due tothe lack of appropriate reference electrodes. In most cases,the electrochemical measurements are conducted with the useof ‘‘quasi reference’’ electrodes.3 The use of ‘‘quasi reference’’or ‘‘pseudo reference’’ electrodes, mainly a Pt or an Ag wire,4 issomewhat problematic as these electrodes are not strictlyelectrochemically defined. The potentials of the electrodes arehighly influenced by impurities in the RTIL itself or by oxideformation at the electrode surface, and thus considered lesssuitable to serve as reference electrodes.5,6 Another associatedproblem with the use of such electrodes is potential drift with

time, even during a single measurement. Another electrodetype being used in RTILs is the one based on the metals Hg andAg, with the appropriate insoluble salt of the metal and thecorresponding electrolyte containing the same anion.7,8 Utiliza-tion of such electrodes in RTILs is possible and the usage of Ag/Ag+ or Ag/AgCl electrodes in such media was reported by severalresearchers.9–13 Nevertheless, one of the problems with suchreference electrode constructions is associated with relativelycomplicated fabrication and production procedures. In somecases, macromolecules need to be used as cryptate agents.9 Inother cases, the need calls for an additional solution (whetherorganic or aqueous) to be used as a salt bridge or as an internalphase for proper Ag based salt dissolution.10 Moreover, possibleleakage of Cl� or other anions from the reference electrode intothe RTIL based solution may alter the electrochemical propertiesof the tested electrolyte, leading to fallacious measurements. Stablereference potentials may also be obtained with the use of differentred/ox couples, mainly ferrocene (Fc|Fc+) and cobaltocene (Cc|Cc+)as internal standards,14–16 rather than separate reference electrodesintroduced externally into each electrochemical system. Althoughthis can be easily achieved and applied in almost any electro-chemical experiment, it possesses major disadvantage whenresults obtained from more than a single working electrodeare compared.17

In previous studies, Raz et al. investigated the electrochemicalbehaviour of silicon in RTIL 1-ethyl-3-methyl-imidazoliumoligofluorohydrogenate (EMIm(HF)2.3F, Scheme 1).18,19 An anodicpolarization applied to the n-type silicon electrode ((100),

a Department of Materials Science and Engineering, Technion - Israeli Institute of

Technology, Haifa 32000, Israel. E-mail: eineli@technion.ac.ilb Forschungszentrum Julich GmbH, Institut fur Energie- und Klimaforschung (IEK-9),

D-52425 Julich, Germany

Received 26th June 2013,Accepted 20th August 2013

DOI: 10.1039/c3cp52661e

www.rsc.org/pccp

PCCP

PAPER

Publ

ishe

d on

21

Aug

ust 2

013.

Dow

nloa

ded

by T

echn

ion

- Is

rael

Ins

titut

e of

Tec

hnol

ogy

on 0

9/11

/201

3 18

:47:

23.

View Article OnlineView Journal | View Issue

17838 Phys. Chem. Chem. Phys., 2013, 15, 17837--17845 This journal is c the Owner Societies 2013

r = 8–12 O cm) immersed in EMIm(HF)2.3F resulted in porousstructure formation at the surface of the Si electrode, with noillumination being applied. Width, length and distribution ofthe pores were controlled by modification in both the appliedpotential and time. The suggested mechanism for the formationof the porous structure relates to Si–F bond formation at thesilicon surface. When time progressed anodic potential is beingapplied at the silicon surface, Si–F bonds are generated at localspots, substituting the pristine Si–H bonds, yielding anincreased potential drop (i.e. electric field) at the silicon surface.With sufficient Si–F bonds being formed, the potential drop atthe surface is increased significantly, yielding initiation of pitsformation. Pore growth is possible due to successive cycles ofSi–F bond formation at the silicon surface and a competitivesilicon etching. This phenomenon agrees with the current burstmodel, suggested by Carstensen et al., which explains theformation of pore structure on n-Si in the presence of a strongoxidizing environment without the need for illumination.20,21

Following the work by Raz et al., a unique metal–air batterywas developed based on the coupling between a silicon waferanode, an air cathode, and an EMIm(HF)2.3F electrolyte.22,23

Among the optional silicon wafer types, the heavily dopedn-type was found to be the most potent for battery operation,as it benefits from the low self-discharge rate along withrelatively high cell potentials (1.1 to 0.8 V), with currentdensities of 0.01 to 0.3 mA cm�2, accordingly. The anodicelectrochemical dissolution of n-type silicon in EMIm(HF)2.3FRTIL during battery discharge is performed following thesuggested reaction scheme:

Si + 12(HF)2F� - SiF4 + 8(HF)3F� + 4e� (1)

However, all anodic half-cell investigations of Si in EMIm(HF)2.3FRTIL were performed with the use of a Pt wire as a quasi-referenceelectrode. These measurements were accompanied by inconsistency,and were difficult to compare, due to instability of the Pt wire voltagein the RTIL electrolyte.

In this paper, we describe a rather simple fabrication andconstruction of a stable external reference electrode, which fullycomplies with the requirements needed for an electrochemical

reference electrode. The reference electrode is based on ferro-cene (Fc|Fc+) dissolved in an EMIm(HF)2.3F gel. The stabilityand electrochemical advantages of the suggested method arediscussed. High quality results obtained with this referenceelectrode are presented as well.

ExperimentalGel based electrode fabrication

Ferrocene (98%, Aldrich) was used in all the described experi-ments. EMIm(HF)2.3F was received from the laboratory of Prof.Rika Hagiwara at the Department of Energy, Kyoto Universityand from Boulder Ionics (USA). The synthesis procedure of theRTIL is described elsewhere.24 Carbopol 941 (polyacrylic acid,Mw B 1800, Aldrich) was utilized as a gelling additive in thiswork. Polyethylene Pasteur pipettes (Bar-Naor, Israel) wereapplied as compartment cells for the EMIm(HF)2.3F gel-basedelectrodes. Platinum wires and foils were cleaned by immersingthem in concentrated nitric acid for 20 seconds, rinsing indeionized (DI) water and drying under a nitrogen gas stream.The glassy carbon (GC) electrode (ALS, Japan) was cleaned inacetone, DI water and then dried under a nitrogen gas stream.

Ionic liquid gels were prepared by adding different amountsof Carbopol 941 to 10 mM of Fc|Fc+ in EMIm(HF)2.3F. The10 mM concentration was chosen due to the fact that such anamount of Fc|Fc+ was already proved to be possible to dissolvein this specific RTIL.24 A higher concentration of Fc|Fc+ mayalso be applied, but without any additional information on thesolubility limit of Fc|Fc+ in EMIm(HF)2.3F the presented paperis focused only on 10 mM Fc|Fc+/EMIm(HF)2.3F solutions. Themixture was allowed to rest for 1 hour, afterwards 0.6 ml of thegelled solution was dispensed into a polyethylene Pasteurpipette, serving as the electrode housing (illustration togetherwith geometrical dimensions of the cell are presented inScheme 2). Subsequently, a platinum wire was immersed intothe electrode housing. The fabricated electrode compartmentwas separated from the cell containing the electrolyte usingfilter paper (Whatmans No. 2) as shown in Scheme 2. In orderto establish a baseline for cyclic voltammetry studies, two other

Scheme 1 Schematic drawing of EMIm(HF)2.3F RTIL chemical structure.

Paper PCCP

Publ

ishe

d on

21

Aug

ust 2

013.

Dow

nloa

ded

by T

echn

ion

- Is

rael

Ins

titut

e of

Tec

hnol

ogy

on 0

9/11

/201

3 18

:47:

23.

View Article Online

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 17837--17845 17839

Fc|Fc+-free electrodes were similarly prepared by adding theappropriate Carbopol content to EMIm(HF)2.3F.

Apparatus and procedures

Electrochemical measurements were performed using an EG&Gpotentiostat/galvanostat 2273 (Princeton Applied Research).Cyclic voltammetry (CV) was carried out in a three electrodeconfiguration, with Pt wire serving as the working electrode(Scheme 3). During initial evaluation studies of the referenceelectrode configuration, Pt wire also served as a quasi-referenceelectrode. For advanced studies, this fabricated gel-based electrodewas applied as the reference electrode. All CV measurementscans were initiated from open circuit potential (OCP) to cathodicpotentials. OCP measurements were obtained in a two electrode cellconfiguration, as presented in Scheme 4. The experiments wereconducted in specially constructed two half-cell compartments (a, b)with a Teflons capillary connector (c) filled with RTIL. In the firsthalf cell (a) a 0.1 M HCl solution with a calomel reference electrode(VL-PC) was placed. EMIm(HF)2.3F RTIL with Pt or gel basedreference electrodes were placed in the second half cell (b).

For Si electrochemical studies, potentiodynamic tests wereconducted with a three-electrode cell configuration. A single-crystal silicon wafer or an air electrode (Electric Fuel Ltd.,

Israel) served as a working electrode, a platinum wire as acounter electrode, and a gel-based electrode as a referenceelectrode. Si wafers were (100) As-doped (University Wafer,USA) with different doping levels: n++ – 0.001–0.005 O cm,n+ – 0.1–0.6 O cm, and (100) B-doped: p++ – 0.001–0.005 O cm,p+ – 1–3 O cm. All silicon samples were cleaved to approximately1 cm � 1 cm pieces, with the exposed surface controlled by anO-ring. Prior to any experiment, all silicon samples were cleanedin 1 : 1 H2SO4 : H2O2 solution to remove organic contaminantsand washed with DI water. Afterwards, the samples were dippedin 20 vol% HF solution in order to remove native oxide, thenwashed again with DI water and dried with N2 stream. Informa-tion on silicon–air cell construction can be found in ref. 23.Experiments with n-type samples were conducted in the dark inorder to avoid photocurrent contribution. Ohmic contacts wereobtained by applying Ag paste on the backsides of the samples.

The electrochemical data were analyzed using EC-Lab (Biologic, V10.18). The peak currents and potentials were manually determinedwith a background current considered as a linear extrapolation ofthe end peak curvature (graphical method).25 Si–air discharge

Scheme 2 Schematic drawing of the reference electrode configuration.

Scheme 3 Schematic drawing of the three electrode cell configuration.

Scheme 4 Schematic drawing of two electrodes cell configuration: (a) 0.1 MHCl solution with the calomel electrode, (b) EMIm(HF)2.3F RTIL with the experi-mental electrode, and (c) capillary filled with EMIm(HF)2.3F RTIL.

PCCP Paper

Publ

ishe

d on

21

Aug

ust 2

013.

Dow

nloa

ded

by T

echn

ion

- Is

rael

Ins

titut

e of

Tec

hnol

ogy

on 0

9/11

/201

3 18

:47:

23.

View Article Online

17840 Phys. Chem. Chem. Phys., 2013, 15, 17837--17845 This journal is c the Owner Societies 2013

experiments were performed using an Arbin BT2000 battery testsystem.

Results and discussion

Pt quasi-reference electrodes suffer from a drift of the potentialduring electrochemical experiments. This behavior is welldemonstrated in Fig. 1, which illustrates a comparison betweenthe potential of a Pt electrode and a gel-based electrodeconfiguration measured vs. a standard calomel reference electrode(SCE), in a cell assembly, as described in Scheme 4. Fig. 1 demon-strates the difficulty associated with the application of Pt as a quasi-reference electrode, as opposed to the stable gel based electrodeconfiguration. The constructed gel based electrode exhibits a stablepotential measured at 250 � 3 mV vs. SCE. On the other hand, thePt quasi reference electrode does not hold a stable potential, as itshows a drift of B0.1 V over 1 hour of measurement. The Fig. 1inset presents the first 600 seconds of the potential measurementsfor both electrodes. It can be clearly seen that the gel based electrodepotential reading is stable and steady even in this short period,while the Pt electrode potential drifts by approximately 50 mV inthis time frame.

The dominance of the gel-based electrode over the Pt wire asa reference is reflected in CV measurements, as well. CVexperiments can provide information on electrode potentialstability, subjected to cycles of oxidation and reduction, whichis determined by the electrochemical equilibrium of the red/oxspecies in the solution. Moreover, the electrode reversibility, interms of the oxidation-to-reduction area peak ratio as reflectedin CV curves, is crucial in determination of electrode stabilityand reliability. In other reference electrode configurations,such as Ag/AgCl, this issue is not acute, since no reversibilityis expected.25 For this reason, fundamental CV investigationswere performed in two configurations: one with the Pt wire asquasi-reference, and in the other cell configuration a gel-basedelectrode served as the reference electrode. The results for these

two cases are summarized in Table 1 (for a Pt quasi reference),and 2 (for a gel-based reference electrode). The CV scans wereconducted consecutively from the lowest scan rate (5 mV s�1) tothe highest one (50 mV s�1). As can be seen in Table 1, for thered and the ox reaction waves, both peak height and peak-to-peak potential separation, DEp, increase with scan rates. Thehigh peak-to-peak potential change observed with the Pt quasireference implies the presence of an uncompensated solutionresistance, most probably due to the relatively high viscosity ofthe solution. The relatively high spread of measured potentialspresented in Table 1, in particular the reversible potential E0

f

(the average of the Fc|Fc+ reaction peak maxima, which may also beconsidered as mid-potential of the reaction), is attributed toinsufficient stability of Pt wire potential, which served as a referenceelectrode in these measurements. Additionally, reversibility ofthe Fc|Fc+ Red/Ox reaction is also being demonstrated by thenear unity ratio of the reductive and oxidative peak currentdensities, jox

p /jredp respectively.

This reversibility is also maintained when the amount ofgelling agent is decreased from 5 to 2.5 wt%. The results of thisstudy are shown in Table 2, where the working electrode (2.5 wt%Carbopol) was measured against the 5 wt% Carbopol configu-ration, which was discussed in the previous section. The 2.5wt% gel electrode shows similar features in both peak heightand peak-to-peak separation, as the data in Fig. 2 and Table 2demonstrate. A different E0

f value of the Fc|Fc+ reaction wasmeasured, as expected, due to the different percentage ofgelling agent. However, the spreading (or error) in the measuredE0

f is much lower this time, owing to the absence of a non-accurate Pt wire and to the improved stability of the gel-basedelectrode.

For further evaluation of the gelation influence on thereaction and in order to ensure that only Fc|Fc+ contributes

Fig. 1 Open circuit potential measurements of the suggested 2.5 wt% gelbased electrode (circle, red) and the Pt quasi-reference electrode (square, black),inset figure – first 600 seconds of the experiment; Calomel was used as thereference electrode.

Table 1 Cyclic voltammograms data for different scan rates of the constructed 5 wt%gel based electrode, vs. Pt quasi reference (electrode surface area 1.88 cm2)

Scan rate

mA cm�2

joxp /jred

p

mV

joxp jred

p Eoxp Ered

p DEp E0f

5 0.038 �0.038 1.01 16 �388 404 �18610 0.04 �0.039 1.05 58 �399 457 �17025 0.045 �0.048 0.94 151 �456 607 �15350 0.073 �0.076 0.96 227 �536 763 �155

Avg. E0f �166 mV � 15 mV

Table 2 Cyclic voltammograms data for different scan rates of the constructed2.5 wt% gel based electrode, vs. 5 wt% reference (electrode surface area 1.88 cm2)

Scan rate

mA cm�2

joxp /jred

p

mV

joxp jred

p Eoxp Ered

p DEp E0f

5 0.067 �0.063 1.07 178 �122 300 2810 0.077 �0.076 1.01 194 �150 344 2225 0.09 �0.081 1.10 266 �218 484 2450 0.108 �0.101 1.07 355 �219 626 22

Avg. E0f 24 mV � 3 mV

Paper PCCP

Publ

ishe

d on

21

Aug

ust 2

013.

Dow

nloa

ded

by T

echn

ion

- Is

rael

Ins

titut

e of

Tec

hnol

ogy

on 0

9/11

/201

3 18

:47:

23.

View Article Online

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 17837--17845 17841

to the CV’s cathodic and anodic peaks, 10 mM Fc|Fc+ inEMIm(HF)2.3F solution was introduced into the pipette con-struction, without the addition of a gelling agent. In addition,the pipette containing the EMIm(HF)2.3F gel (2.5 wt% gellingagent) without any Fc|Fc+ was studied, as well. Fig. 3a presentsa comparison of the two electrodes, one containing a gellingagent and the other is gelling agent-free, while Fig. 3b presents acomparison of an additional two gel based electrodes containing10 mM Fc|Fc+ and Fc|Fc+-free gel. Fig. 3a shows that althoughthe same anodic and cathodic peaks appear in both solutions,the peaks recorded from the gel based electrode are broader andless defined. This may be attributed to the higher viscosity of theproduced gel and lower diffusivity of Fc|Fc+ species in a gelelectrolyte.26 On the other hand, in the Fc|Fc+-deficient gelledelectrode, no peaks are observed, as can be seen in Fig. 3b. Thisserves as further evidence that the sole electrochemical active

species is the Fc|Fc+ Red/Ox couple, which is responsible for thereduction and oxidation current peaks.

The possibility of Fc|Fc+ leakage from the constructedgel based electrode configuration was evaluated, as well. CVmeasurements were conducted upon immersion of the electrodeassembly into the RTIL (free of gelling agent) and subsequent toone hour immersion in the RTIL, as shown in Fig. 4. The sweeps(at a scan rate of 5 mV s�1) were conducted with a glassy carbon(GC) working electrode and Pt wire, as a counter electrode. Fig. 4clearly excludes Fc|Fc+ leakage from the assembled electrodeduring the electrochemical experiment. Thus, the dense gelstructure prevents leakage of the red/ox species from the con-structed reference electrode and therefore impedes electrolytecontamination and measurement error due to the presence ofimpurities.

The reliability and reproducibility of the proposed gel-basedreference electrode was demonstrated in various experiments.Fig. 5 shows half-cell potentiodynamic measurements obtainedwith different silicon types (anodic curves) and an air cathode(cathodic curve) in the EMIm(HF)2.3F electrolyte, polarized vs. agelled reference electrode. The potential difference under OCPconditions between the two half cells is 1.4 V for n-type Si and1 V for p-type. These results are in perfect agreement with theOCP value that a full Si–air cell presents,22,23 which corre-sponds to the difference between the OCP values of eachelectrode on its own.

The stability of potential measurements obtained with agelled reference electrode, in contrast to the instability of the Ptquasi-reference electrode is well illustrated in Fig. 6. Fig. 6ashows a partial discharge curve for an Si–air cell (includinginitial 4 hours of rest time), together with the potential of eachelectrode (Si anode or air cathode), measured separately versusthe Pt wire quasi-reference electrode. First, the rest potential isvery close to 1.4 V, the same as the potential difference shownin Fig. 5. Second, it can be seen that there are fluctuations in

Fig. 2 Cyclic voltammograms of 10 mM Fc|Fc+ in EMIm(HF)2.3F RTIL at differentscan rates with the addition of 2.5 wt% of the gelling agent.

Fig. 3 (a) Cyclic voltammograms of 10 mM Fc|Fc+ in EMIm(HF)2.3F RTIL with and without addition of 2.5 wt% of the gelling agent, (b) cyclic voltammograms of the2.5 wt% gelling agent with and without the addition of 10 mM Fc|Fc+.

PCCP Paper

Publ

ishe

d on

21

Aug

ust 2

013.

Dow

nloa

ded

by T

echn

ion

- Is

rael

Ins

titut

e of

Tec

hnol

ogy

on 0

9/11

/201

3 18

:47:

23.

View Article Online

17842 Phys. Chem. Chem. Phys., 2013, 15, 17837--17845 This journal is c the Owner Societies 2013

the potential of the electrodes when measured separately, whilefor the complete cell the potential is steady and smooth.Moreover, the fluctuations in the distinct electrode potentialsare simultaneous in time, and the magnitude of the anodicpolarization of the silicon anode is equal to the cathodicpolarization at the air cathode. Therefore, those fluctuationsare a result of unstable potential of the Pt reference electrode.On the other hand, Fig. 6b shows a similar study, but in thiscase the potentials are measured against the reported gel-basedreference electrode. For these measurements, all potentials aresmooth and without any acute fluctuations. This furtherdemonstrates the stability of the gel-based reference electrodeduring the extended discharge process, which lasted for morethan 20 days (not shown).

Additional utilization of the gelled reference electrode canbe found in studying the fundamental electrochemical propertiesof silicon in RTILs. The anodic dissolution of silicon inEMIm(HF)2.3F was previously discussed by Raz et al.,18,19 and areaction process was established [reaction 1]. However, since

silicon is a semiconductor, charge transfer in the silicon–electrolyte interface takes place via conduction and valence bands,and is governed by the generated electric field at the semi-conductor’s space charge layer. On the electrolyte side, alayer of highly concentrated charges is formed, known as theHelmholtz layer. One of the most important characteristics of asemiconductor–electrolyte system is the flat-band potential. Itis a key factor in the determination of the charge transfermechanism in the electrochemical reaction. The most commonmethod used to determine the flat-band potential is the Mott–Schottky method, which is based on the relation betweenapplied potential and the resulted interfacial charge capacity,as written in eqn (2):

1

csc2¼ 2

qee0Nd

� �V � Vfb �

kT

q

� �(2)

where Csc is the space-charge capacitance, V is the appliedpotential, Vfb is the flat-band potential, q is the electron charge,e is the dielectric constant of the semiconductor, e0 is thepermittivity of free space and Nd is the donor density (for ann-type semiconductor, assuming full ionization). Extracting Vfb

from eqn (2) is valid only when the capacitance of the spacecharge layer is assumed to be much smaller than the Helmholtzlayer capacitance CH, i.e. the potential drop over the Helmholtzlayer can be neglected, as stated in eqn (3) and (4):

|Csc| { |Csc|, Ctotal D Csc (3)

|Fsc| c |FH| (4)

Fsc and FH are the potential drops in the space charge andHelmholtz layers, respectively. However, when high dopinglevel silicon is being considered, (n++ or p++), the potential dropin the Helmholtz layer is quite significant.27 When taking intoaccount both FH and Fsc, the linearity of the Mott–Schottkyrelation still holds, however, there is a shift in the voltage fromthe flat band potential at C�2 = 0, as eqn (5) shows:28

1

c2¼ 1

cH21þ 2cH

2

qee0Nd

� �V � Vfb �

kT

q

� �(5)

The Mott–Schottky method was used to determine Vfb for n+

and n++ silicon samples in contact with EMIm(HF)2.3F, utilizinga gel-based reference electrode. The capacitance values, C, wereobtained from fitting an equivalent circuit to electrochemicalimpedance spectroscopy (EIS) measurements of the anodichalf-cell, in the high frequency range, from 1 MHz to 100 Hz,under different polarizations. The equivalent circuit that wasused to extract the capacitance values was a Randles circuit(shown in Fig. 7a), containing a resistance element Rb, repre-senting electrolyte and contact resistances, in series with twoparallel elements Rsc and Csc, corresponding to the spacecharge layer resistance and capacitance. The Mott–Schottkyplots, based on the acquired data for n+ and n++ siliconsamples, are shown in Fig. 7b and c. As can be seen, in thecase of the n+ sample (Fig. 7b), the Mott–Schottky approxi-mation is valid, with a clear linear relation between Csc

�2 and V.Extrapolation of Csc

�2 vs. V to intersection C�2 = 0 provides a

Fig. 4 Cyclic voltammograms of EMIm(HF)2.3F RTIL upon immersion of theFc|Fc+ containing gel based electrode (square, black) and after 1 hour (circle, red).

Fig. 5 Potentiodynamic voltammograms obtained with various silicon electrodesand the air cathode in EMIm(HF)2.3F RTIL. Scan rate 5 mV s�1.

Paper PCCP

Publ

ishe

d on

21

Aug

ust 2

013.

Dow

nloa

ded

by T

echn

ion

- Is

rael

Ins

titut

e of

Tec

hnol

ogy

on 0

9/11

/201

3 18

:47:

23.

View Article Online

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 17837--17845 17843

flat-band potential value of �713 � 29 mV. As the gelledreference electrode was used in these measurements the poten-tials stated and written are versus this reference electrode. Fromthe slope of the Mott–Schottky plot shown in Fig. 7b, the donorconcentration Nd can be determined. For the n+ silicon thisleads to a value of 3.80 � 0.12 � 1016 cm�3, which is in verygood agreement with the conductivity of 0.1–0.6 O cm of thesilicon wafers, corresponding to a doping level of 8.45 � 1015 to7.77 � 1016 cm�3. This implies that the contribution of theHelmholtz layer capacitance to the total measured interfacial

capacitance is indeed negligible when n+ silicon is beingconsidered.

Eqn (5) shows that even for n++ silicon, the slope of thelinear part in the Mott–Schottky plot yields the doping concen-tration. Employing eqn (5) for the results in Fig. 7c provides adoping level of 2.77 � 0.26 � 1018 cm�3, a value that is lowerthan the expected doping of 1.24 � 7.36 � 1019 cm�3, based onthe calculations according to the wafer conductivity. This mightbe due to a relatively high number of inactive doping sites inthe case of high doping concentration. For the calculation of

Fig. 6 Discharge of the Si–air cell (0.3 mA cm�2) together with the corresponding n++ silicon anode and air cathode membrane polarizations, recorded with 3 –electrode configuration using (a) the Pt wire quasi-reference electrode and (b) the gel-based reference electrode.

Fig. 7 (a) Randles equivalent circuit consists of an electrolyte resistance with the parallel combination of the space charge capacitance and resistance of faradaicreaction. (b) and (c) Mott–Schottky plots for n+ and n++ silicon anodes in EMIm(HF)2.3F RTIL, evaluating the flat-band potentials.

PCCP Paper

Publ

ishe

d on

21

Aug

ust 2

013.

Dow

nloa

ded

by T

echn

ion

- Is

rael

Ins

titut

e of

Tec

hnol

ogy

on 0

9/11

/201

3 18

:47:

23.

View Article Online

17844 Phys. Chem. Chem. Phys., 2013, 15, 17837--17845 This journal is c the Owner Societies 2013

Vfb of n++ Si according to eqn (5), we assumed a Helmholtz valuecapacitance layer on the order of 10 mF cm�2. This value is basedon measurements of double-layer electrochemical capacitors,employing different ionic liquids electrolytes, among themEMIm(HF)2.3F.29 Although the linearity between C�2 vs. V doesnot hold at low bias as seen in Fig. 7c, the extrapolation of thelinear portion at high bias still provides the flat-band potential,as was suggested by Uosaki et al.27 Using the obtained dopingconcentration and the assumed Helmholtz capacity, this calcu-lation yields a flat-band potential of �772 � 90 mV.

The above presented results may provide an explanationfor previously reported phenomena related to n-type silicondissolution under an anodic polarization without any illumina-tion and to the potentiodymanic behavior of n-Si inEMIm(HF)2.3F, presented in Fig. 5.18,19,22,23 From the samemeasured OCP potential both for n+ and n++ at �750 �10 mV vs. gel based electrode, it can be concluded that similarinitial conditions occur at the silicon surface, regardless of thedoping level. This suggests that the silicon surface is coveredwith a similar layer, the nature of which was already reported tobe F-terminated, i.e. Si–Fx.19 This layer is constructed duringsilicon wafer immersion in the solution, followed by siliconpotential stabilization in EMIm(HF)2.3F RTIL. Additional con-formation for such Si–Fx passivation layer at the Si surface iswell illustrated by the relatively low corrosion currents (at therange of 4–8 mA cm�2 22,30) for both n+ and n++, suggestinginhibition of the silicon electrochemical dissolution at OCP.

However, as Fig. 7 indicates, there is a difference in Vfb

potential for both types of silicon, which shows the differentanodic behavior of these electrodes in EMIm(HF)2.3F RTIL.When considering dissolution at potentials above 0.1 V vs.OCP, one should take into account the F-passivated siliconsurface. The dissolution process is through SiF2 at the surface,and can be described as follows:

SiF2(surface) + 6(HF)2F� - SiF4 + 4(HF)3F� + 2e� (6)

The SiF2 compound at the Si surface serves as a passivationcomplex, which pins the edge of the conduction band at theinterface, and moreover allows additional surface states forelectron injection. As shown in Fig. 5, active dissolution insmall anodic polarizations (o0.5 V) is possible only with n++ Si.On the other hand, as reported by Raz et al., high magnitudeactive dissolution of n+ silicon without illumination is possibleonly at high polarizations, above 1 V.18,19 Such a difference inthe anodic behavior may be attributed to different band bend-ing at the silicon surface. For n++, Vfb was measured at �772 �90 mV vs. gel based electrode (Fig. 7c), and it is only slightlymore cathodic with respect to OCP. Therefore, at OCP, the bandbending at the silicon surface provides only a small depletionregion, hence low corrosion currents in this state. However, thisbend banding structure is invalid while considering silicondissolution under anodic polarization conditions, where thepotential of the n++ Si is even more anodic than Vfb and n++

silicon is under deeper depletion. In this case, the anodicdissolution in the dark is possible due to the surface states

induced by the SiF2 layer. Polarizations of the potentials above0.5 V vs. OCP lead to a shift in the anodic current to highervalues, most probably due to modification in the surface layercomposition.

Nevertheless, different models are considered for n+. Vfb ismore anodic to OCP, and therefore the accumulation region isconstructed at the surface at OCP. Such an accumulation layerelevates the possibility for a dissolution process at OCP, andthus, the corrosion currents measured for n+ are two timeshigher than the values measured for n++. Moreover, such anaccumulation condition exists at very low anodic polarizationas well, up to Vfb, leading to high anodic currents, comparableto n++, as can be seen in a region close to OCP (VOCP o V o�0.7 V) in Fig. 5. The potential difference in this region is in thesame order of Vfb � VOCP. At anodic polarizations above Vfb, adepletion layer is formed on n+, in a similar manner as in n++ Si,however, the obtained anodic currents are much smaller. Thisphenomenon can be explained by either a thicker space chargelayer in n+ compared to n++, or by a smaller overlap between theSiF2 surface states and the electronic states in the n+ Si.Modification in the surface layer composition, as was statedfor n++, can be observed in Fig. 5 as well as for n+ Si at the samepotential, albeit this modification still does not permit ananodic current enhancement. Active dissolution for n+ is onlypossible at high anodic polarizations, approximately >0.9 V,due to an inversion region formed on the silicon surface. Atthat stage, dissolution will occur through holes consumptionmechanism, as was already proposed for p-type silicon.31

Conclusions

This work shows that the Fc|Fc+/EMIm(HF)2.3F gel basedelectrode can be successfully used as an external referenceelectrode for electrochemical measurements, enabling a deeperunderstanding of Si electrochemistry in fluorohydrogenatedionic liquids. The vast stability of this reference potential andthe durability of Fc|Fc+ (no leakage from the constructedelectrode) were presented. The reproducibility of the electro-chemical measurements, as demonstrated in silicon potentio-dynamic measurements, as well as in Si–air battery application(during full cell discharge process), justifies the use of theconstructed electrode as the most appropriate and suitablereference electrode. Flat-band potential and OCP measure-ments were accomplished, while utilizing the gel-based refer-ence electrode. The results show that n++ silicon is underdepletion at OCP, while n+ silicon is in the low accumulationstate. Based on these results, illustration of the mechanism forn+ and n++ Si electrochemical dissolution in EMIm(HF)2.3Fionic liquid was provided. The mechanism calls for thepresence of an F-terminated layer at the Si surface in the activedissolution of n++ Si without illumination and under a rather smallanodic polarization, below 0.5 V. The same F-termination surfacelayer is the responsible one for passivation of n+ Si under the sameconditions. In general, the rather simple preparation method of thegel-based reference electrode and the high-quality results designatethis configuration (Red/Ox couple in ionic liquid with the addition

Paper PCCP

Publ

ishe

d on

21

Aug

ust 2

013.

Dow

nloa

ded

by T

echn

ion

- Is

rael

Ins

titut

e of

Tec

hnol

ogy

on 0

9/11

/201

3 18

:47:

23.

View Article Online

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 17837--17845 17845

of a gelling agent) as the most suitable and reliable referenceelectrode in future electrochemical studies involving ionic liquids.

Acknowledgements

Support for this work was provided by the German IsraelFoundation (GIF), the Grand Technion Energy Program (GTEP)and the Leona & Harry B. Helmsley Charitable Trust.

Notes and references

1 F. Endres and S. Zein El Abedin, Phys. Chem. Chem. Phys.,2006, 8(18), 2101–2116.

2 M. Armand, F. Endres, D. R. MacFarlane, H. Ohno andB. Scrosati, Nat. Mater., 2009, 8(8), 621–629.

3 J. Zhang and A. M. Bond, Analyst, 2005, 130(8), 1132–1147.4 L. Barrosse-Antle, A. Bond, R. Compton, A. O’Mahony, E. Rogers

and D. Silvester, Chem.–Asian J., 2010, 5(2), 202–230.5 K. R. Seddon, A. Stark and M. Torres, Pure Appl. Chem., 2000,

72(12), 2275–2287.6 F. C. Anson and J. J. Lingane, J. Am. Chem. Soc., 1957, 79(5),

1015–1020.7 M. Waleed Shinwari, D. Zhitomirsky, I. A. Deen, P. R.

Selvaganapathy, M. Jamal Deen and D. Landheer, Sensors,2010, 10(3), 1679–1715.

8 C. G. Zoski, Handbook of electrochemistry, Elsevier B.V., 2007.9 A. Lewandowski, M. Osinska, A. Swiderska-Mocek and

M. Galinski, Electroanalysis, 2008, 20(17), 1903–1908.10 A. Saheb, J. Janata and M. Josowicz, Electroanalysis, 2006,

18(4), 405–409.11 G. A. Snook, A. S. Best, A. G. Pandolfo and A. F. Hollenkamp,

Electrochem. Commun., 2006, 8(9), 1405–1411.12 M. Shibata, M. Yamanuki, Y. Iwamoto, S. Nomura and

T. Kakiuchi, Anal. Sci., 2010, 26(11), 1203–1209.13 B. Huber and B. Roling, Electrochim. Acta, 2011, 56(19),

6569–6572.14 A. Lewandowski, L. Waligora and M. Galinski, Electroanalysis,

2009, 21(20), 2221–2227.

15 E. I. Rogers, D. S. Silvester, D. L. Poole, L. Aldous,C. Hardacre and R. G. Compton, J. Phys. Chem. C, 2008,112(7), 2729–2735.

16 S. K. Sukardi, J. Zhang and I. Burgar, et al., Electrochem.Commun., 2008, 10(2), 250–254.

17 L. Waligora, A. Lewandowski and G. Gritzner, Electrochim.Acta, 2009, 54(5), 1414–1419.

18 O. Raz, D. Starosvetsky, T. Tsuda, T. Nohira, R. Hagiwaraand Y. Ein-Eli, Electrochem. Solid-State Lett., 2007, 10(3),D25–D28.

19 O. Raz, Z. Shmueli, R. Hagiwara and Y. Ein-Eli, J. Electrochem.Soc., 2010, 157, H281–H286.

20 J. Carstensen, M. Christophersen, G. Hasse and H. Foll,Phys. Status Solidi A, 2000, 182(1), 63–69.

21 J. Carstensen, M. Christophersen and H. Foll, Mater. Sci.Eng., B, 2000, 69–70, 23–28.

22 G. Cohn, D. Starosvetsky, R. Hagiwara, D. D. Macdonald andY. Ein-Eli, Electrochem. Commun., 2009, 11(10),1916–1918.

23 G. Cohn and Y. Ein-Eli, J. Power Sources, 2010, 195(15),4963–4970.

24 R. Hagiwara, T. Hirashige, T. Tsuda and Y. Ito, Electrochem.Soc., 2002, 149(1), D1–D6.

25 E. Gileadi, Physical electrochemistry. Fundamentals,techniques and applications, Angew. Chem., Int. Ed., 2011,17–25, 221–227.

26 J. S. Lee, T. Nohira and R. Hagiwara, J. Power Sources, 2007,171(2), 535–539.

27 K. Uosaki and H. Kita, J. Electrochem. Soc., 1983, 130(4),895–897.

28 R. De Gryse, W. P. Gomes, F. Cardon and J. Vennik,J. Electrochem. Soc., 1975, 122(5), 711–712.

29 M. Ue, M. Takeda, A. Toriumi, A. Kominato, R. Hagiwaraand Y. Ito, J. Electrochem. Soc., 2003, 150(4), A499–A502.

30 G. Cohn, R. A. Eichel and Y. Ein-Eli, Phys. Chem. Chem.Phys., 2013, 15(9), 3256–3263.

31 T. Tsuda, T. Nohira and K. Amezawa, et al., Electrochim.Acta, 2008, 53(10), 3650–3655.

PCCP Paper

Publ

ishe

d on

21

Aug

ust 2

013.

Dow

nloa

ded

by T

echn

ion

- Is

rael

Ins

titut

e of

Tec

hnol

ogy

on 0

9/11

/201

3 18

:47:

23.

View Article Online