Table Of ContentUnderscreening in concentrated electrolytes
Alpha A. Lee,1,∗ Carla Perez-Martinez,2 Alexander M. Smith,2,3 and Susan Perkin2,†
1John A. Paulson School of Engineering and Applied Sciences,
Harvard University, Cambridge, MA 02138
2Department of Chemistry, Physical and Theoretical Chemistry Laboratory,
7
1
0 University of Oxford, Oxford OX1 3QZ, U.K.
2
3Department of Inorganic and Analytical Chemistry,
n
a
J University of Geneva, 1205 Geneva, Switzerland
7
2 Screeningofasurfacechargebyelectrolyteandtheresultinginteractionenergybe-
] tween charged objects is of fundamental importance in scenarios from bio-molecular
h
p
- interactions to energy storage. The conventional wisdom is that the interaction en-
m
ergy decays exponentially with object separation and the decay length is a decreas-
e
h
c ing function of ion concentration; the interaction is thus negligible in a concentrated
.
s
c electrolyte. Contrary to this conventional wisdom, we have shown by surface force
i
s
y measurements that the decay length is an increasing function of ion concentration
h
p and Bjerrum length for concentrated electrolytes. In this paper we report surface
[
force measurements to test directly the scaling of the screening length with Bjerrum
1
v
1 length. Furthermore, we identify a relationship between the concentration depen-
5
1 dence of this screening length and empirical measurements of activity coefficient and
8
0 differential capacitance. The dependence of the screening length on the ion con-
.
1
0 centration and the Bjerrum length can be explained by a simple scaling conjecture
7
1 based on the physical intuition that solvent molecules, rather than ions, are charge
:
v
carriers in a concentrated electrolyte.
i
X
r
a
The structure of electrolytes near a charged surface underpins a plethora of applications,
from supercapacitors [1] to colloidal self-assembly [2] and electroactive materials such as ion-
monomeric polymer-metal composites [3]. The structure of dilute electrolytes is relatively
well-understood [4–6]. However, dilute electrolytes have a low conductivity because the con-
ductivity is proportional to the concentration of charge carriers. As such, dilute electrolytes
∗ [email protected]
† [email protected]
2
are generally unsuitable for many electroactive materials and concentrated electrolytes are
preferred up to the point when the viscosity increases significantly. Understanding elec-
trolytes at high concentrations remains a conceptual challenge because the ion-ion Coulomb
interaction is strong and long-ranged. The most extreme case of concentrate electrolytes are
ionic liquids — liquids at room temperature which comprise pure ions without any solvent
[7–10].
Tosegueintoexploringthephysicsofconcentratedelectrolytes, wefirstrevisitthephysics
of dilute electrolytes. The seminal Debye-Hu¨ckel theory [4] predicts that the interaction be-
tweentwochargedsurfacesinanelectrolytedecaysexponentiallywiththesurfaceseparation
[11]. The characteristic decay length, known as the Debye length, is given by
(cid:115)
(cid:15)k T 1
λ = B ≡ √ , (1)
D 4πq2c 4πl c
ion B ion
where (cid:15) is the dielectric constant of the medium, k the Boltzmann constant, T the tem-
B
perature, q the ion charge, c the ion concentration (which is twice the salt concentration
ion
for a 1:1 electrolyte), and
q2
l = (2)
B
(cid:15)k T
B
is the Bjerrum length. The Bjerrum length is the distance at which the interaction energy
between two ions equals the thermal energy unit k T. The Debye-Hu¨ckel theory is a mean-
B
field theory valid when l3c (cid:28) 1, i.e. when the ion-ion separation is far greater than the
B ion
Bjerrum length and thus the Coulomb interactions can be treated as a perturbation to ideal
gas behaviour. Therefore, the Debye-Hu¨ckel theory is only applicable for dilute electrolytes.
For concentrated electrolytes, only a handful of analytical results are known. A well-
known result pertains to the pair correlation function, g (r), which is the probability of
ij
finding a particle of component j at a distance r from another particle of component i
[12]. Mathematical analysis of the Ornstein-Zernicke equation reveals that, for particles
interacting via a short-ranged [13, 14] or Coulomb [15] potential, the asymptotic decay of
the correlation function takes the form
r(g (r)−1) ∼ A e−α0rcos(α r+θ ), as r → ∞. (3)
ij ij 1 ij
Crucially, Equation (3) implies that all correlation functions in the system decay with the
same rate α and oscillate with same wavelength 2π/α in the asymptotic limit; only the
0 1
3
amplitude A and phase θ are species-dependent. In an electrolyte solution, 1/α is the
ij ij 0
electrostatic screening length and 2π/α the characteristic correlation wavelength. Equation
1
(3) is a general asymptotic result for the decay of correlations that is independent of the
electrolyte model. Analytical expressions for α and α could be obtained for the restricted
0 1
primitive model [15, 16]. However, the restricted primitive model does not explicitly account
for space-filling solvent molecules and thus may not capture certain important features of
screening in electrolytes (c.f. Section III). Without considering specific models to compute
α and α , we will use the Equation (3) to organise our discussion about different theories.
0 1
The decay of correlations in the bulk electrolyte is directly related to the decay of in-
teractions between charged surfaces, measurable via techniques such as the Surface Force
Balance (SFB). To illustrate why this is the case, recall from Equation (3) that the asymp-
totic wavelength is the same for all correlation functions. Therefore, if we consider two large
charged spheres of radius R immersed in the electrolyte, their asymptotic pair correlation
function is given by r(g (r) − 1) ∼ A e−α0rcos(α r + θ ). Thus the potential of mean
ss ss 1 ss
force v(r) ∼ −k T logg(r) ∼ A e−α0rcos(α r + θ )/r. As the concentration of the large
B ss 1 ss
spheres is negligible compared to the ions and solvent, α and α are independent of the
0 1
properties of the spheres. The interactions between the charged plates in the SFB decays
in the same way as the same as the interactions between two charged spheres of radius
R → ∞. Therefore, within this picture, the electrostatic screening length and characteristic
correlation wavelength measured by SFB is the same as that for the bulk electrolyte.
We first consider the characteristic correlation wavelength 2π/α . The Debye-Hu¨ckel
1
theory for dilute electrolytes corresponds to the limiting case α = 0 and α = 1/λ .
1 0 D
√ √
However, a finite oscillatory period emerges (α > 0) when a/λ = 4πl c a2 (cid:38) 2,
1 D B 0
where a is the ion diameter [5, 15, 16]. This threshold value of a/λ is widely known as
D
the Kirkwood line [17], first reported by John Kirkwood in 1936. In other words, pass the
Kirkwoodline, thedecayofion-ioncorrelationsswitchesfromamonotonicexponentialdecay
to a damped oscillatory decay. In the context of ionic liquids, the presence of an oscillatory
decayofionchargedensityawayfromchargedinterfaceshasbeencalled“overscreening”[18].
Integral equation theories also predict that the density-density correlation function becomes
oscillatory and has a decay length that is longer than the charge-charge correlation function
atanevenhigherelectrolyteconcentration[5,15,16]; thisistermed“core-dominated”decay.
The subject of this paper is the electrostatic screening length 1/α . The Debye-Hu¨ckel
0
4
theory (1) predicts that the electrostatic screening length decreases as the electrolyte con-
centration increases. Direct experimental measurements of this screening length for con-
centrated electrolytes is relatively scarce, perhaps a surprise as the theory of electrolyte
solutions has received significant attention in the past century [19]. The first sign that the
Debye-Hu¨ckel screening length is qualitatively awry for concentrated electrolytes is a series
of SFB studies showing that the interaction force between charged surfaces in an ionic liquid
decays exponentially, but with a decay length that is orders of magnitude larger than the
Debye length or the ion diameter [20, 21]. It was then shown, via SFB measurements of
the screening length in ionic liquid-solvent mixtures and alkali halide salt solutions, that
the long electrostatic screening length is not unique to pure ionic liquids: the electrostatic
screening length in concentrated electrolytes increases with ion concentration, contrary to
the predictions of the Debye-Hu¨ckel theory [22]. Moreover, we provide empirical evidence
that the screening length scales as
λ ∼ l c a3. (4)
S B ion
In the remainder of this paper, we will term Equation (4) “underscreening”. The electrolyte
solution “underscreens” charged surfaces in the sense that the interaction between charged
surfaces is significantly longer-ranged than the Debye-Hu¨ckel regime of ions behaving as a
weakly interacting gas.
To allay potential confusion, we emphasize that “underscreening” and its cognate “over-
screening” [18] are two distinct parameters in the decay of ion-ion correlation, Equation
(3). Underscreening pertains to the anomalously long electrostatic screening length and
overscreening pertains to a finite oscillatory period. Therefore, mathematically speaking,
overscreening and underscreening could occur together if an electrolyte has an oscillatory
decay of ion-ion correlation with a decay length that follows the scaling (4). However, ex-
perimentally oscillations are measured only in the near-surface region and no oscillatory
component is detected in the long-ranged component of the surface force [20–22].
In this paper, we first discuss the experimental evidence for underscreening and report a
newsetofexperimentsverifyingthescalingrelationship(4). Wethenshowhowunderscreen-
ingisreflectedintwoclassicpropertiesofelectrolytes: theactivitycoefficientanddifferential
capacitance. Finally, we propose a scaling conjecture to understand the phenomenology of
underscreening.
5
I. EXPERIMENTAL MEASUREMENTS OF THE SCREENING LENGTH
The screening lengths, λ , of electrolyte solutions were determined from direct measure-
S
ments of the change in the interaction force with distance between two charged mica plates
across the electrolyte. The apparatus used for such measurements, called the surface force
balance (SFB; see Figure 1), employs white light interferometry to determine the separa-
tion between the plates to ∼ 0.1 nm. The mica plates are supported on cylindrical lenses
(each of radius ∼ 1 cm) and mounted in crossed-cylinder configuration; the arrangement is
geometrically equivalent to a sphere of radius 1 cm approaching a flat plate. The symmetry
and well-defined geometry make the setup particularly useful for quantitative comparison
to theory; the technique has been used over the past few decades to study forces across
dilute electrolytes [23], simple molecular liquids [24], and soft matter [11]. In the case of
dilute electrolytes the surface force is dominated by a repulsive osmotic pressure, increasing
exponentially as D decreases, with decay length equal to λ which decreases with increasing
D
concentration in accordance with Equation (1).
In contrast to the measurements in dilute electrolytes, surface force measurements across
pure ionic liquids have revealed short-range oscillations reminiscent of structural forces in
molecular liquids [25] and, beyond the oscillatory region, monotonic screening extending to
distances far greater than predicted by simple application of Debye-Hu¨ckel theory [20]. In a
study aimed at connecting up the dilute electrolyte and ionic liquid ends of the electrolyte
spectrum, some of us recently reported a non-monotonic trend in the asymptotic screening
length with concentration [22]. In this section we describe those experiments in detail and
investigate the scaling behaviour of λ with c and, separately, the scaling of λ with l
S ion S B
achieved by varying solvent (cid:15) at constant c .
ion
A. Experimental details
IntheSFBexperiments,whitelightinterferometryisusedtodeterminetheforcesbetween
two molecularly smooth mica surfaces separated by a thin film of electrolyte (see Figure 1)
[26]. The mica sheets of equal thickness are backsilvered to create a partially reflecting
and partially transmitting mirror before gluing onto the lenses and injection of liquid in the
gap between the mica surfaces. The resulting silver-mica-liquid-mica-silver stack acts as an
6
FIG. 1. Schematic diagram (a) showing the essential features of the SFB experiment. White
light is passed through an interferometric cavity comprising two hemi-cylindrical lenses in crossed-
cylinder orientation. Mica sheets (not shown) are back-silvered to allow for white light reflection
before mounting on the lenses. The mica sheets are immersed in the liquid electrolyte of interest.
Analysis of the fringes of equal chromatic order (FECO) are analysed during approach of the
cylindrical lenses in order to calculate the mica-mica separation distance (D), force (F ), and
N
effective radii of curvature of the lenses (R). An example of a single measurement is shown in (b)
forthepureionicliquid[C C Pyrr][NTf ]; thedataareshownonlog-linearplottodemonstratethe
4 1 2
exponential decay at longer range and, in the inset, on a linear-linear plot to reveal the oscillatory
region with negative force minima at small D. Replotted from Ref. [22].
interferometric cavity. Bright columnated white light incident on interferometer, dispersed
with a spectrometer, emerges as a set of bright fringes of equal chromatic order (FECO).
The bottom lens is mounted on a horizontal leaf spring, while the top lens is mounted on
a piezo-electric tube (PZT). By expanding the PZT, the top surface is brought at constant
velocity towards the bottom surface from separations D of 200-400 nm to D of one or a few
7
molecular diameters. The rate of approach is sufficiently slow that there is no measurable
hydrodynamic contribution to the force, as evidenced by the insensitivity of the measured
forces to small changes of rate of approach. The FECO pattern is captured by a camera
at rates of approximately 10 frames per second. At large separations, there is no normal
force on the spring, but as the lenses are brought in to contact, the normal forces arising
from interactions between the surfaces will cause bending of the spring. The deflection of the
spring, andthusthenormalforce, canbeinferredfromtheinterferometricpattern. Theforce
between the surfaces can then be related to the interaction energy E using the Derjaguin
approximation[11]. Therefore, E = F/2πR, whereRisthelocalradiusofcurvaturebetween
the lenses, of the order of 1 cm for these experiments.
In order to vary solvent (cid:15) at constant c we used solutions of ionic liquid at fixed 2M con-
ion
centrationinsolventsofvaryingpolarity. Theionicliquidwas1-butyl-1-methypyrrolydinium
bis[(trifluoromethyl)sulfonyl]imide (abbreviated [C C Pyrr][NTf ], Iolitec 99.5 %), and the
4 1 2
molecular solvents were propylene carbonate (Sigma-Aldrich, anhydrous 99.7 %), dimethyl
sulfoxide (Sigma-Aldrich, anhydrous 99.9%), acetonitrile (Sigma Aldrich, anhydrous 99.8%),
benzonitrile (Sigma-Aldrich, anhydrous 99%) and butyronitrile (Fluka, purity ≥99%).
The FECO fringes were analysed using the method outlined by Israelachvilii [27]; our
analysis uses the refractive index values of the bulk mixture to compute the separation be-
tween the mica surfaces. The refractive index of the mixtures of 2M [C C Pyrr][NTf ] in
4 1 2
dimethyl sulfoxide and in benzonitrile were measured to be 1.441 and 1.461, respectively, us-
ing an Abbe 60 refractometer; for mixtures of 2M [C C Pyrr][NTf ] in propylene carbonate,
4 1 2
butyronitrile, and acetonitrile, the FECO analysis used estimated refractive index values of
1.422, 1.408, and 1.380, respectively. These estimated values are weighted average values
between the refractive index of the pure ionic liquid (1.425; measured by supplier), and the
refractive index of the solvents (1.4189 for propylene carbonate and 1.3842 for butyronitrile,
bothfromCRChandbook[28], and1.344foracetonitrile(providedbysupplier). Calculation
of such a weighted average for dimethyl sulfoxide and benzonitrile solutions led to values in
very good agreement with our direct measurements.
Several precautions are taken to ensure the purity and stability of the liquid mixtures
during the measurements. In all experiments, the ionic liquid [C C Pyrr][NTf ] was dried
4 1 2
in vacuo (10−2 mbar, 70◦C) for several hours to remove residual water. In the case of
acetonitrile, butyronitrile and propylene carbonate experiments, the liquid was obtained
8
from freshly opened bottles, while for benzonitrile measurements and some of the dimethyl
sulfoxide experiments, the bottles had been opened within two weeks of the measurement.
The dried ionic liquid was then mixed with the solvents and introduced in between the
lenses within a few minutes, in order to minimise exposure of the mixture to atmospheric
moisture. In all cases, the liquid film between the mica surfaces was in contact with a
large bulk reservoir. For solutions of ionic liquid with dimethyl sulfoxide, benzonitrile, and
propylene carbonate, a droplet of solution of approximately 20 µL was injected between the
lenses. In the case of acetonitrile and butyronitrile solutions, which are volatile, the bottom
lens was immersed in a bath of the solution, and for acetonitrile tests, additional solvent
was introduced in the SFB chamber to create a saturated solvent vapour and thus minimise
evaporation. The drying agent P O was also introduced in the chamber to capture any
2 5
residual water vapour. Different glues were used to attach the mica sheets to the lenses,
depending on the compatibility of the solvents: glucose (Sigma-Aldrich, 99.5%) was used
as glue for propylene carbonate experiments, EPON 1004 (Shell Chemicals) was used for
benzonitrile, acetonitrile and butyronitrile soltuions, and paraffin (Aldrich, melting point
53-57◦) was used for dimethyl sulfoxide experiments.
B. Experimental measurements varying c
ion
An example of the measured interaction force between two mica plates as a function of
separation, D, across a pure ionic liquid, [C C Pyrr][NTf ], is shown in Figure 1(b). As the
4 1 2
surfaces approach from large D they experience a repulsive force, exponentially increasing
with decreasing D, eventually giving way to an oscillatory region at D (cid:46) 5 − 8 nm. The
key signature of oscillatory forces is the presence of minima in the profile as detected on
retraction of the surfaces; these are shown using a linear scale in the inset to Figure 1(b).
Interpretation of the structural features in ionic liquids leading to such oscillatory forces has
been discussed in the past [10, 29], the details of this near-surface region depend on ionic
liquidmolecularfeaturessuchascation-anionsizeasymmetryandionamphiphilicity, surface
chemistry and surface charge. Here we focus instead on the monotonic tail of the interaction
force—in this pure ionic liquid, the tail is measurable above our resolution limit from about
20 nm (several tens of ion diameters)—which we will show to be relatively insensitive to the
molecular features of the ionic liquid or electrolyte. The exponential decay length in the
9
asymptotic limit is taken as the screening length λ .
S
We studied the variation of λ with c in a mixture of [C C Pyrr][NTf ] with propylene
S ion 4 1 2
carbonate (molecular solvent), chosen for their miscibility and liquidity over the full range
of mole fraction, from pure solvent to pure salt, at room temperature. Figure 2(a) shows
three force profiles chosen at concentration points to demonstrate the clear decrease in λ
S
between 0.01 M (λ = 2.7±0.3 nm) and 1.0 M (λ = 1.05±0.4 nm), and the subsequent
S S
increase in λ between 1.0 M and 2.0 M (λ = 5.4 ± 0.7 nm). Figure 2(b) shows how λ
S S S
varies with c1/2, and also shows similar measurements made for NaCl in water. It is clear
ion
that in both cases there exists a minimum in λ at intermediate concentration.
S
The realisation that NaCl in water at sufficiently high concentration shows the same di-
vergence of screening length as observed in ionic liquids led us to hypothesise that the origin
of the anomalous λ lies in electrostatic interactions between ions, rather than in a mecha-
S
nism dependent on chemical features such as hydrogen bonding or nanoscale aggregation of
non-polar domains. This indeed appears to be the case, as demonstrated by the collapse of
alldatapointswhenthescreeninglengthisscaledbytheDebyelengthandtheconcentration
is scaled by the dielectric constant and ion diameter, as shown in Figure 3. We note that
the dielectric constant varies substantially as a function of ion concentration; the dielectric
constants of ionic liquid solutions are calculated using effective medium theory [30], and the
dielectric constant of alkali halide solutions are taken from the literature [31, 32]. Included
in Figure 3 are also a wide range of pure ionic liquids, and some further 1:1 inorganic salts
in water; the common scaling appears to be general across these electrolytes. The abscissa
in Figure 3 is the nondimensional quantity a/λ , where a is the mean ion diameter in the
D
electrolyte; the ion diameter of ionic liquid is estimated from X-ray scattering experiments
[33], and we take the ion diameter of alkali halide salts to be the unhydrated ion diameter
[22]. a/λ scales as (c /(cid:15))1/2. As we will show in the following section, a/λ is also an im-
D ion D
portant parameter describing the scaling of the chemical potential and activity in electrolyte
solutions.
There are two distinct scaling regimes in Figure 3, which we call “low” and “high”
concentration. At low concentration, where a/λ < 1, the measured screening length is
D
the Debye length i.e. λ /λ = 1. This persists until the point at which the Debye length
S D
shrinks to the ion diameter, a = λ . At high concentration, when a/λ > 1, the scaling
D D
switches to a power law:
10
Concentration
a
12
10
[CCPyrr]+
Debye-Hückel 4 1
8 screening length!
m) Propylene
(n 6 [NTf2] _ carbonate
λS
4
Na+ Cl – HO
2
2
0
0.0 0.4 0.8 1.2 1.6 2.0
b
1/2 1/2
c (M )
FIG. 2. (a) Example measurements of the normalised force between mica sheets as a function of
separation, D, across mixtures of [C C Pyrr][NTf ] and propylene carbonate at c = 0.01 M, 1.0 M
4 1 2
and 2.0 M. The concentrations are chosen to demonstrate the non-monotonic variation in the long-
range decay with concentration. (b) Screening length of the long-range (asymptotic) component of
the surface force plotted as a function of c1/2, and for two different electrolytes: [C C Pyrr][NTf ]
4 1 2
in propylene carbonate, and NaCl in water. Data in (b) are replotted from Ref. [22]. The circled
data points arise from the three force profiles in (a).
λ (cid:18) a (cid:19)3
S
∼ (5)
λ λ
D D
which is equivalent to λ ∼ c a3l . That is to say, our measurements suggest that in the
S ion B
high-concentration regime the screening length scales linearly with Bjerrum length.
