Ionic structure in the aqueous electrolyte glass LiCl · 4D2O

Article abstract of DOI:10.1088/0953-8984/9/42/002

The difference methods of neutron diffraction and isotopic substitution were applied to Li⁺ and Cl^- in the aqueous lithium chloride glass LiCl · 4D₂O at 120 K (the glass transition T_g = 150 K). Results for the ionic hydration show that structural changes are more evident for the Cl^- coordination than for the Li⁺ coordination. At the level of the anion-anion structure there is an appreciable increase in the nearest-neighbour coordination number over that in the liquid which suggests a greater degree of correlation between the Cl^- anion and the water molecules. A model of the ionic structure is proposed to explain the relative ease of glassification of concentrated aqueous lithium chloride solutions.

Full text of DOI:10.1088/0953-8984/9/42/002

Home Search Collections Journals About Contact us My IOPscience
Ionic structure in the aqueous electrolyte glass
This content has been downloaded from IOPscience. Please scroll down to see the full text.
1997 J. Phys.: Condens. Matter 9 8835
(http://iopscience.iop.org/0953-8984/9/42/002)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 132.203.227.62
This content was downloaded on 03/10/2015 at 19:34
Please note that terms and conditions apply.

J. Phys.: Condens. Matter 9 (1997) 8835–8847. Printed in the UK
PII: S0953-8984(97)82643-4
Ionic structure in the aqueous electrolyte glass LiCl · 4D2O
S Ansell†, J Dupuy-Philon‡, J-F Jal‡ and G W Neilson§k
† Material Science Division, Argonne National Laboratory, 9700 South Cass Avenue, Argonne,
IL 60439, USA
‡ Departement de Physique de Materiaux, Universite Claude Bernard, 69622 Villeurbanne Ce´dex,
France
§ H H Wills Physics Laboratory, University of Bristol, Tyndall Avenue, Bristol BS8 1TL, UK
Received 17 March 1997, in final form 14 July 1997
Abstract. The difference methods of neutron diffraction and isotopic substitution were applied
to Li+ and Cl− in the aqueous lithium chloride glass LiCl · 4D2O at 120 K (the glass transition
Tg = 150 K). Results for the ionic hydration show that structural changes are more evident for the
Cl− coordination than for the Li+ coordination. At the level of the anion–anion structure there
is an appreciable increase in the nearest-neighbour coordination number over that in the liquid
which suggests a greater degree of correlation between the Cl− anion and the water molecules.
A model of the ionic structure is proposed to explain the relative ease of glassification of
concentrated aqueous lithium chloride solutions.
1. Introduction
Lithium chloride–water mixtures are well suited to the study of fragile glass formation [1];
a range of concentrations exists over which glass formation is relatively easy [2]. For two
mixtures in particular, LiCl · 4H2O and LiCl · 6H2O, a variety of methods have been used
to characterize the glassification process. The atomic structure of both systems has been
recently studied by the method of neutron diffraction and isotopic substitution (NDIS) [3, 4].
The use of H/D substitution shows that the water structure, as defined through the partial pair
radial distribution function gHH (r), and linear combinations of the other radial distributions
functions, is ‘even more severely disrupted’ in the tetrahydrate than in the hexahydrate [4].
As a continuation of this work, attention is focused on the structure of the ions in the
glassy state, and is aimed at identifying a structural signature associated with glassification
such as might be anticipated from the work of Angell and Sare [1]. Recall that in their
classic paper on fragile glasses they demonstrated that aqueous electrolyte glasses could
be classified according to the anions present in the system. In particular they showed that
metal chloride hydrates and metal nitrate hydrates formed two groups of glassy compounds.
Because NDIS is particularly well adapted to the study of systems containing chlorine [5],
a study was made of Cl− hydration and the solute structure defined by the pair distribution
function gClCl(r) in LiCl · 4D2O. A complementary study of Li+ coordination was also
made to investigate the extent of changes in Li+ hydration on glassification. The results
were compared with those obtained in previous NDIS experiments on liquid lithium chloride
solutions [6–13], and used to provide new insights into the nature of glassification of aqueous
electrolytes.
k Please send correspondence and reprint requests to G W Neilson, University of Bristol.
c
0953-8984/97/428835+13$19.50 ꢀ 1997 IOP Publishing Ltd
8835

8836
S Ansell et al
2. Experimental details
2.1. Sample preparation
Five sample solutions of ∗Li∗Cl (table 1) were prepared directly from a mixture of
isotopically labelled solutions of lithium hydroxide, LiOH(aq), and hydrochloric acid,
∗
HCl(aq). The isotopically labelled lithium hydroxide was prepared by the dissolution of
∗Li metal in water: the Li was washed in ethyl alcohol to remove the oil in which the
∗
metal was stored, dried in air and added in small pieces to chilled distilled water in the
manner described by Vogel [14]. The isotopically labelled hydrochloric acid was made by
elution of sodium chloride (Na∗Cl) solution through an Amberlite 120H ion exchange resin,
described in∗detail by BDH [15]. Excess acid was added to the hydroxide solution and the
∗
resultant Li Cl solution was filtered through C-18 filters to remove traces of oil and other
hydrophobic contaminants. The purified solution was evaporated to dryness under vacuum
∗
∗
∗
∗
at a temperature ∼300 ◦C to produce solid Li Cl(s). Each Li Cl(s) was added to excess
D2O and gently boiled down to near dryness, with heavy water being added about four or
five times until the H2O content (checked by IR spectroscopy) was sufficiently low (< 0.4%)
(table 1). Each solution was then boiled down to the correct molality. Samples were then
checked for isotopic content by mass spectrometry, light-water impurity (IR spectrometry)
and ionic concentration (conductivity and colorimetry) (tables 1 and 2). They were then
sealed in glass containers and transported to Saclay.
Table 1. The atom−ic3 fractions of the samples ∗Li∗Cl · 4D2O. The sample number density, ρ,
was 0.092 atoms A at 20 ◦C and was assumed to be unchanged in the glass.
˚
Solute
c7
c6
c35
c37
Cl
cD
cH
cO
Li
Li
Cl
7Li35Cl
0.0688
0.0714
0.0713
0.0330
0.007
0.0026
0.000
0.0001
0.0384
0.0707
0.0680
0.0541
0.0645
0.0541
0.0541
0.0035
0.0173
0.00689
0.0173
0.0173
0.567
0.567
0.567
0.567
0.567
0.004
0.004
0.004
0.004
0.004
0.285
0.285
0.285
0.285
0.285
7LiNAT Cl
7Li37Cl
φLiNAT Cl
6LiNAT Cl
Table 2. The mean neutron coherent scattering length b (fm) of the constituent atoms in each
of the five samples. The deuterium was contaminated with 0.4% H.
Lithium
Chlorine
Oxygen
Deuterium
375F(Q)
NA7T F(Q)
−2.06
−2.22
−2.22
1.96
11.49
9.579
3.59
5.805
5.805
5.805
5.805
5.805
6.60
6.60
6.60
6.60
6.60
377F(Q)
NA6T F(Q)
NAφT F(Q)
9.579
9.579
0.01
2.2. Instrumental set-up
The experiment was carried out on the 7C2 diffractometer at Saclay with an incident neutron
˚
beam of 0.701 A which was collimated by two cadmium insets of 10 mm width and 50 mm

Ionic structure in LiCl · 4D2O
8837
height. Another cadmium inset was used to shield the detectors from scattering from the
aluminium walls of the front of the bell jar. This was the only collimation used during the
experiment. (Because there was no secondary collimation on the detectors, the background
was large and found to have an appreciable sample dependent component.) The wavelength
of the neutrons and orientation of the banana detector were calculated from nickel powder
diffraction data.
The samples were maintained in their cold glassy state at 120 K in an ‘orange’ cryostat.
The design of this cryostat allows the sample to be positioned in the beam between two
1 mm thick vanadium foil heat shields at the neutron beam level. The temperature was
controlled by dripping liquid helium onto a small heater which was thermally anchored to
the inner heat shield and the top of the sample. This arrangement allowed the temperature
and the temperature gradient across the sample to be controlled to ±1 K and ±0.2 K cm−1
respectively.
The liquid samples were glassified as follows. The vanadium container with its liquid
sample was attached to the candlestick mounting of the cryostat. The container and end
part of the candlestick mounting were then plunged into liquid nitrogen; the rapid cooling
ensured no crystallization took place (see figure 1). The candlestick with its sample on the
end were quickly transferred to the cryostat and brought to thermal equilibrium at 120 K in
about half an hour. Data were collected for all samples at this temperature which is well
below the glass transition temperature, Tg = 150 K [2].
Figure 1. The total structure functions, ∗∗F(Q), for the five isotopically labelled ∗Li∗Cl · 4D2O
φ
6
7
7
glasses at 120 K: (a) NAT F(Q) + 0.3; (b) NAT F(Q) + 0.2; (c) 35F(Q) + 0.1; (d) NAT F(Q);
7
(e) 37F(Q) − 0.1.
2.3. Data acquisition and treatment
Neutron diffraction data were collected for five solutions in the glassy state (7LiNAT Cl,
7
7
6LiNAT Cl, φLiNAT Cl, Li37Cl, Li35Cl, each with 4D2O per molecule of salt), the empty
cryostat, the empty cell in the cryostat, a cadmium rod in the cryostat and a 0.6 cm
diameter vanadium rod which was used for data normalization. The data were treated
by standard methods for multiple scattering, normalization and absorption correction [16],
φ
7
6
7
7
and the respective structure factors (NAT F(Q), NAT F(Q), NAT F(Q), 37F(Q) and 35F(Q))
were calculated (figure 1). Two main problems were encountered in the data reduction—
one associated with the background subtraction and the other arising from kinematic or
‘Placzek’ corrections.

8838
S Ansell et al
The background subtraction, even when large, is of minor importance in a difference
experiment if the total cross-sections are similar in all the samples. This is because the real
background is similar for all samples, thus inaccuracies in its determination are removed
when taking differences. However, lithium and chlorine isotopes have significantly different
absorption cross-sections, resulting in larger beam attenuation in samples containing 6Li and
35Cl, which corresponds to a smaller real background scattering. In order to obtain some
indication of the extent of this effect, data were collected for a cadmium rod of exactly the
same diameter as the sample. The background was found to be approximately one and a half
times bigger for a non-absorbing sample than a completely absorbing sample. A modified
correction procedure, which gave self-consistent results, was introduced to remove most of
the background from the very highly absorbing samples, the remainder having no significant
effect on the structural information extracted from the data. The procedure involved making
the cadmium spectrum closer to that for a perfectly absorbing spectrum by the removal of the
scattering at high angles. For 2θ < 90◦ the cadmium spectrum was unchanged; however,
above this angle it was possible for neutrons to scatter directly from the surface of the
cadmium rod into the detectors without being absorbed. The area of unattenuated cadmium
scattering seen by the cadmium rod was
ꢀ
ꢀ
ꢁꢁ
r
area = hr 2θ − sin−1
≈ hr 2θ
(1)
D
where h is the beam height, r is the radius of the cadmium rod and D is the distance from
the sample to the detector. The equivalent scattering from a plate of thickness 0.5 mm and
area hr 2θ with a cross-section equivalent to the cadmium high-Q limit was subtracted from
the measured cadmium spectrum. This procedure enabled the calculation of the ‘corrected’
cadmium background, which was assumed to be the background due to a total absorbed. A
linear combination of the ‘corrected’ cadmium background and experimentally determined
empty cryostat background weighted to the calculated sample transmission was then used
to produce the sample backgrounds. These new backgrounds replaced the experimentally
measured background in the correction procedures.
The Placzek correction is a major problem for systems which contain water because the
Placzek series for the deuterium atom does not converge very rapidly and that for hydrogen
does not converge at all. This problem becomes compounded when 6Li and 7Li are present,
because of the relatively large mass difference between the two isotopes, and the high
6
absorption of the Li. Because of this, it was necessary to apply Placzek corrections. It
was understood that these would not correct for the deuterium atoms adequately but this
was not significant as all the samples had the same deuterium content. The overall effect
6
7
7
of this correct−io1n was to give NAT F(Q), NAT F(Q) and 35F(Q) the same average slopes
˚
for Q > 10 A (figure 1).
∗
The final structure factors F(Q) are related to the partial structure factors Sαβ of the
∗
system as follows:
X X
∗∗F(Q) =
cαcβbα bβ(Sαβ (Q) − 1)
(2)
α
β
where cα is the atomic fraction of species α whose mean coherent neutron scattering length
is bα, and the double sum extends over the four atomic species D, O, Li and Cl.
¯
3. Results
The experimental data enabled the direct determination of GLi(r), GCl(r), and gClCl(r) in
the aqueous lithium chloride glass (LiCl · 4D2O) at 120 K.

Ionic structure in LiCl · 4D2O
3.1. Lithium hydration
8839
6
7
The total scattering cross-sections, NAT F(Q) and NAT F(Q), were used to determine the
first-order difference function 1Li(Q) given by
6
7
1Li(Q) = NAT F(Q) − NAT F(Q).
(3)
The 1Li(Q) was corrected for residual self-scattering and Placzek effects by subtraction of
a second-order Chebychev polynomial, and can be written explicitly as
1Li(Q) = A1[SLiO(Q) − 1] + B1[SLiD(Q) − 1] + C1[SLiCl(Q) − 1]
+D1[SLiLi(Q) − 1]
(4)
(5)
where
¯
¯
¯
¯
¯
¯
2
¯
6
7
6
7
A1 = 2cLicObO(b Li − b Li
)
B1 = 2cLicDbD(b Li − b Li
)
2
Li
2
¯
¯
¯
¯
n
7
7
6
6
C1 = 2cLicClb Cl(b − b Li
)
D1 = c ((b ) − (b Li) ).
Li
Li
A cosine window function was applied between Q = 12.0 A and Q = 16.0 A−1 to remove
termination ripple from the transform. The Fourier transformation of 1Li(Q) (see figure 2)
gave the Li+ radial distribution function (GLi(r)) (figure 3(a)), which can be expressed as
−1
˚
˚
GLi(r) = A1[gLiO(r) − 1] + B1[gLiD(r) − 1] + C1[gLiCl(r) − 1] + D1[gLiLi(r) − 1]
= 9.90gLiO(r) + 22.8gLiD(r) + 4.1gLiCl(r) − 0.6gLiLi(r) − 26.2.
(6)
˚
The features in GLi(r) below r = 1.7 A are considered unreal and were set equal to
GLi(0) = [−(A + B + C + D)]. The back transformation was calculated and is shown with
the original 1Li(Q) (figure 2). The difference between the direct Fourier transform and the
back Fourier transform (figure 3) is a small slowly varying function which allows a high
O
Li
degree of confidence in the calculated GLi(r). The coordination (or hydration) number n¯
associated with the O atoms of the water correlations can be calculated by integration over
the peaks due to Li . . . O
Z
r2
4πρ
nOLi
=
cO
r2(GLi(r) + E) dr . . . .
(7)
A1
r1
D
Li
Similar calculations can be made for n¯
.
Figure 2. The first-order difference function 1Li(Q) for LiCl · 4D2O glass at 120 K. The open
circles were calculated from the F(Q) values directly (3), and the full curve is the result of
‘back’ Fourier transformation of the full curve for GLi(r) (figure 3(a)). The slowly varying
dashed curve at the bottom is the difference between the two functions shifted by −0.08 barns.

8840
S Ansell et al
(a)
(b)
Figure 3. (a) The total Li+ radial distribution function GLi(r) for LiCl · 4D2O glass at 120 K.
The open circles are obtained by direct Fourier transformation of the open-circle curve in figure 2,
˚
and the full curve is GLi(r) edited with the data for r < 1.8 A set equal to GLi(0). (b) The
total Li+ radial distribution function GLi(r) for LiCl · 4D2O liquid at 295 K [7, 9].
3.2. Chloride hydration
7
7
The total scattering cross-sections 35F(Q) and 37F(Q) were used to determine the first-order
difference function 1Cl(Q) given by
7
7
1Cl(Q) = 35F(Q) − 37F(Q).
(8)
The self-scattering term and residual of the Placzek effects were corrected by subtraction of
a second-order Chebychev polynomial. The resulting 1Cl(Q) (figure 4) can be expressed
as
1Cl(Q) = A2[SClO(Q) − 1] + B2[SClD(Q) − 1] + C2[SLiCl(Q) − 1] + D2[SClCl(Q) − 1]
+E2[SLiO(Q) − 1] + F2[SLiD(Q) − 1] + G2[SLiLi(Q) − 1]
(9)

Ionic structure in LiCl · 4D2O
8841
Figure 4. The first-order difference function, 1Cl(Q), for LiCl·4D2O glass at 120 K. The open
circles were calculated directly from the F(Q) values (8), and the full curve is the ‘back’ Fourier
transformation of the full curve for GCl(r) shown in figure 5(a). The difference between these
two functions is the slowly varying dashed line shifted by −0.2 barns.
where
¯
¯
¯
¯
¯
¯
35
37
35
37
Cl
A2 = 2cClcObO(b
− b
)
B2 = 2cClcDbD(b
− b
)
Cl
Cl
Cl
2
2
2
¯
¯
¯
¯
¯
7
35
37
35
37
C2 = 2cLicClb Li(b
− b
)
Cl
D2 = 2cCl((b Cl) − (b Cl) )
F2 = 2cLicDbD(b Li − b7Li
Cl
(10)
0
0
¯
¯
¯
¯
¯
¯
7
7
E2 = 2cLicObO(b Li − b7Li
)
)
2
2
2
¯
¯
7
7
G2 = 2cLi((b Li) − (b Li) ).
−1
˚
1Cl(Q) had a cosine window function applied from 12 A to the end of the data at
16 A and was Fourier transformed to give the chloride radial distribution function GCl(r)
(figure 5(a)):
−1
˚
GCl(r) = A2[gClO(r) − 1] + B2[gClD(r) − 1] + C2[gLiCl(r) − 1] + D2[gClCl(r) − 1]
+E2[gLiO(r) − 1] + F2[gLiD(r) − 1] + G2[gLiLi(r) − 1]
= 17.3gClO(r) + 39.7gClD(r) − 1.5gLiCl(r) + 5.7gClCl(r) − 0.36gLiO(r)
+0.82gLiD(r) − 0.03gLiLi(r) − 61.63.
(11)
The last three terms involving the lithium ion are non-zero because of a small (4%)
35
37
7
7
difference between the lithium isotopic compositions of the Li Cl and Li Cl samples.
It is seen from (11) that A2, B2 > C2, D2 ꢁ E2, F2, G2 and consequently the last three
terms will not contribute significantly to GCl(r).
˚
The features in the radial distribution function below r = 1.95 A were deemed to
˚
be unphysical and GCl(r) was set to GCl(0) for r < 1.95 A and its back transform was
determined (figure 4). It shows good agreement with the 1Cl(r) calculated directly from the
F(Q) values with only low-frequency differences. This suggests that no serious systematic
error had arisen in the correction procedures.
D(1)
Cl
The hydration number n
of Cl− was calculated from
Z
ρ
D(1)
Cl
n
= 4π cD (GCl(r) + H)r2 dr . . .
(12)
B2
where H = (A2 + B2 + C2 + D2 + E2 + F2 + G2) = 61.63 mb. Evaluation of the integral
D(1)
˚
over the range of the first peak at 2.23 A gives a value of 4.8 for nCl
.

8842
S Ansell et al
(a)
(b)
Figure 5. (a) The total Cl− radial distribution function GCl(r) for LiCl · 4D2O glass at 120 K.
The circles are the Fourier transformation of the open-circle curve for 1Cl(Q) in figure 4, and
˚
the solid line is GCl(r) edited with the data for r < 1.95 A set equal to GCl(0). (b) The total
Cl+ radial distribution function GCl(r) for LiCl · 4D2O liquid at 295 K [13].
3.3. Anion–anion coordination
The pair correlation function SClCl(Q) was calculated from a combination of all five factors
F(Q):
7
7
S
ClCl(Q) = α F(Q) + βNAT F(Q) + γ 377F(Q) + δ φ F(Q) + εNAT6 F(Q) + corrections
35
NAT
φ
7
7
7
≡ 162735F(Q) − 1936NAT F(Q) + 476.537F(Q) − 227N F(Q)
+60NN F(Q) + corrections.
(13)
7
7
7
Usually only three, 35F(Q), NAT F(Q) and 37F(Q), are required [6]. However, because
6
7
7
of different amounts of Li contamination in the nominally pure Li samples of Li35Cl
φ
6
7
and Li37Cl, we had to include the two additional functions, NAT F(Q) and NAT F(Q), to
eliminate large contributions from Li+ . . . water correlations.
The ‘corrections’ term in (13) is primarily due to the Cl− self-scattering, and was
determined by an empirical fit of a Chebychev polynomial of order three. In addition,
SClCl(Q) was influenced by two other deficiencies in the data. The first was caused by the

Ionic structure in LiCl · 4D2O
8843
limited accuracy of the values of the scattering lengths bCl and bLi, which introduced a large
uncertainty (∼6%) in the constants in (13). This problem was corrected by allowing the
scattering lengths to fluctuate within error, and the use of a fitting routine to find the ‘best’
values of α, β, δ, γ and ε, that minimized SClCl for 6 < Q (A−1) < 16. (The assumption
˚
−1
˚
that there is little or no significant structural information in SClCl(Q) for Q < 6.0 A
has already been demonstrated [6].) SClCl(Q) determined by this procedure is shown in
figure 6.
Figure 6. The partial structure factor (SClCl(Q) − 1); the circles were obtained directly from
the factors F(Q) (13), and the full curve from self-consistently checks and a maximum-entropy
procedure [17] (see the text for details).
The second problem wi−th1 the data concerned the presence of two sharp peaks, which
−1
˚
˚
were clearly visible at 3.3 A and 4.0 A in the original SClCl(Q). It is justifiably claimed
that these peaks arose from ice forming on the aluminium wall of the cryostat’s vacuum
chamber during the course of the experiment. Consequently their presence in the data could
not be properly accounted for in the background subtractions. (It is worth noting that the
effect of the ice on the data is sufficiently small that it is not observable in the first-order
differences). These Bragg peaks of the ice are of comparable size to the real peaks in
SClCl(Q) and it was assumed that all information in the Q space region of the peaks had
been destroyed. As a result, SClCl(Q), in the region from 3.0 to 4.5 A−1, was fitted with a
˚
cosine window function.
The final SClCl(Q), together with its maximum-entropy fit, is shown in figure 6 [18].
This fit was necessary due to the noise that resulted from multiplying the factors F(Q) by
the five coefficients in (13). The Fourier transform of the maximum-entropy fit is shown in
figure 7. The maximum-entropy fit to the data is sufficiently close that a reasonable degree
of confidence in the data can be assumed.

8844
S Ansell et al
Figure 7. The radial pair distribution function gClCl(r) for LiCl · 4D2O glass at 120 K.
4. Discussion
4.1. Li+ hydration structure
To understand the hydration structure of Li+ in the glass, it is helpful to compare it with
its structure in the liquid (figure 3). The main features of GLi(r) in the glass are the two
˚
peaks centred at 1.94 ± 0.03 and 2.53 ± 0.03 A, which are identified with oxygen and
O
Li
deuterium atoms respectively. Integration under the first peak gives n¯ = 3.0 ± 0.2 atoms
(table 3). From this result it is expected that the second peak will contain six deuterium
˚
atoms. However integration over the range 2.25 6 r (A) 6 2.90 shows that the peak can
accommodate significantly more than this. In order to explain the additional amount, one
concludes that other atoms must contribute to this peak. It is unlikely to be other hydrogen
or oxygen atoms because of the high concentration of the solution (table 1). This is because
1.5 extra hydrogen or oxygen atoms would be required to be bound to the lithium ion and
this would result in a hydration shell of 4.5 water molecules with very little ability of the
water to share with other lithium ions. A more likely and preferred situation is one in which
direct contact occurs between the positively charged Li+ central ion and negatively charged
near-neighbour Cl− ions.
Similar results are obtained for the liquid [7], albeit with more water molecules in the
first hydration shell and correspondingly less difference between the first and second peaks
(table 3), a result which implies less encroachment by Cl− in the liquid phase. A much
larger difference was seen by Yamagami et al [18] whose results for LiCl · 5D2O showed a
decrease of coordination from four to three water molecules in the supercooled liquid phase
˚
between 213 and 173 K. They also report rLiO and rLiD distances of 2.02(5) and 2.61(5) A
respectively, which are significantly greater than the experimental results obtained here
(table 3).
˚
A second coordination shell is clearly evident, centred ≈4.3 A. It is much more
prominent than in the liquid (figure 3(b)) and signifies a sharpening of structure in the
glass. This shell can accommodate seven chloride ions evenly distributed in the volume
together with eight water molecules, whereas in the liquid this shell can only accommodate
four chloride atoms and eight water molecules. This second coordination shell has also been

Ionic structure in LiCl · 4D2O
8845
Table 3. The Li+ hydration parameters derived from the present study and previous experimental
studies of aqueous solutions of LiCl at ambient temperature unless stated otherwise.
Molality in
D2O
R (moles
D2O/salt) rLiO (A) rLiD (A) n¯Li
D O
2
˚
˚
Reference
24.8 (130 ◦C)
16.66
14.0
12.5
12.5 (120 K)
10
6.25
3.60
3.57
1.0
2.02
3
3.57
4
4
5
8
13.9
14
1.95
2.5
2.31
2.71
2.52
2.68
2.53
2.50
2.43
2.52
2.55
2.52
2.3
[11]
[10]
[7]
[10]
this work
[8]
[10]
[7]
[8]
4 ± 1
1.96
2.22
1.93
1.95
1.95
1.95
1.95
1.96
3.2 ± 0.21
4 ± 1
3.0 ± 0.2
3.3 ± 0.5
4 ± 1
6.0 ± 0.4
5.5 ± 0.3
6.5 ± 1.0
50.0
[7]
identified by Yamagami et al (18) in the low temperature liquid, and was interpreted as an
increase in the number of water molecules on cooling. However, we feel that the structural
enhancement must be due to an increase in the number of Cl− ions as this interpretation is
more consistent with our results for gClCl(r) (section 4.3).
4.2. Cl− hydration structure
The total chloride ion radial distribution function, GCl(r), for the aqueous lithium chloride
glass (figure 5(a)) is similar in shape to that obtained in the liquid at 295 K (figure 5(b)), and
to that in other electrolyte chloride solutions [5, 19]. However, for the glass, the first peak,
˚
at 2.23 A, which is identified with nearest-neighbour hydrogen atoms, is much sharper than
˚
in the liquid and drops to a value close to GCl(0) at r = 2.8 A, suggesting a high degree of
stability for this correlation. Integration over this peak gives a value for the Cl− hydration
of 4.8(3). As there are only four water molecules per lithium chloride unit, the Cl− ions
must share water molecules in the form of non-contact pairs bridged by water molecules.
Evidence for bridged Cl− pairs has already been demonstrated in the less concentrated
aqueous solution of 8.6 molal lithium chloride [13], and in a more concentrated solution
of 14.9 molal lithium chloride [6]. By simple interpolation of the data of the 14.9 and
H
Cl
∼
=
9.98 molal solutions, one calculates that in the liquid n¯
4.7 atoms, a result in close
agreement with the experimental data here (table 4).
˚
The second feature of the GCl(r) is centred at 3.3 A and contains predominantly O
and D atoms. In the glass, this compound peak has become partially resolved, with a first
˚
maximum at 3.29 ± 0.03 A, which is identified with oxygen atoms from water molecules
˚
associated with the deuterium atom (D1) at 2.23 A. This implies stretching of O–D due
−
˚
to the strong polarizing effect of Cl . Recall in pure water O–D ≈ 0.98 A with a half
˚
˚
width of ≈0.5 A. The low-r part of the peak at 3.29 A was reflected about its maxima and
O
integrated to give n¯ = 5.5 ± 0.5, which is in reasonable agreement with that calculated
Cl
˚
for the first peak at 2.23 A.
glass
Cl
˚
Generally one observes that G
(r) has correlations to 12 A, whereas the liquid GCl(r)
˚
of both 14 molal [16] and 9.98 molal solutions [15] terminates at ∼4.0 A. The additional
structure in the glass must be related to indirect correlations between chloride ions. In total
˚
three hydration shells out to 12 A can be identified; these are smeared out in the liquid,
presumably because the water molecules are in fast exchange around the Cl−.

8846
S Ansell et al
Table 4. The Cl− hydration parameters derived from the present study and previous experimental
studies of aqueous solutions of LiCl at ambient temperature unless stated otherwise.
Molality in
D2O
R (moles
D2O/salt) rClD (A) rClO (A) rClD (A)
D O
2
˚
˚
˚
n¯Cl
Reference
1
2
14.9
14.0
12.5 (120 K)
9.98
8.3
3.35
3.57
4
5
6
2.24
2.24
2.23
2.22
2.225
2.225
2.25
3.25
3.27
3.29
3.29
3.2
3.56
3.54
3.50
4.4 ± 0.3
4.4 ± 0.3
[6]
[9]
4.8 ± 0.3 this work
3.50–3.68 5.3 ± 0.2 [19]
3.625
3.625
—
5.4
5.4
[12]
[12]
8.3 (125 K)
3.57
6
14
3.175
3.34
5.9 ± 0.2 [19]
Table 5. The properties of the gClCl(r) measured in different lithium chloride solutions: r1, rM
1
and rM are respectively the cut-off distance and the positions of the first and second maxima.
2
Cl
Cl
˚
˚
˚
r1 (A) rM (A) n¯Cl
rM (A) n¯Cl
Reference
1
2
12.5 molal 120 K 3.1
14.9 molal 293 K 3.0
8.6 molal 295 K 3.9
3.90
3.75
4.2
3.6 ± 0.5 6.4
2.3 ± 0.3 6.38
1.2 ± 0.5 6.4
8.5 ± 1.5
10.0 ± 1.5
this work
[9]
7.0 ± 0.5 [13]
4.3. The ion–ion structure
6
Although a determination was made of gLiLi(r) from the three functions NAT F(Q),
φ
7
NAT F(Q) and NAT F(Q), it was found that the resulting function was unphysically large
and no useful information could be extracted.
The pair correlation function gClCl(r) (figure 7) derived from Fourier transformation of
˚
SClCl(Q) (13) shows two prominent peaks at 3.9 ± 0.2 and 6.4 ± 0.2 A with coordination
numbers, n¯Cl, of 3.6 ± 0.5 and 8.5 ± 1.5 respectively. The discussion of gClCl(r) is best
Cl
conducted in comparison with gClCl(r) for the two liquids: 14.9 molal LiCl(aq) [6], and
8.6 molal LiCl(aq) [15]. The main characteristics of gClCl(r) in these three systems are
˚
listed in table 5. The first peak at 3.90 A in the glass is almost exactly at the same position
as that calculated from a linear interpolation from the peak positions of 14.9 molal and
8.6 molal solutions. However, the coordination number is twice that for the liquid at a
similar concentration. This can be explained by considering the water molecules, anions
and cations together. In the glass, the effective free volume of the Cl− coordination shell will
be larger due to a reduction in thermal vibration of the water molecules, which will facilitate
direct Cl− . . . Cl− contacts without a reduction in hydration number. In contrast, the smaller
Li+ with its relatively stronger electrostatic interaction will not benefit in free volume to
the same extent. However, the reduction in n¯ from ∼4 to ∼3 suggests that Li+ . . . Cl−
O
Li
contacts can occur. The picture overall therefore is one in which the direct Cl− . . . Cl−
contacts are stabilized by the Li+ hydration shell and a water bridge. The implication
for this model is that in cases where the cation is not strongly hydrated, e.g. K+, Sr2+, no
additional direct contacts will take place when a system is glassified. Conversely in the case
of even stronger cation hydration, e.g. Ni2+, Fe3+, . . ., we would predict that in the glass
there will be an increased probability of direct Cl− . . . Cl− contacts over that in the liquid.
The second peak in the gClCl(r) is at a similar distance to Cl−–Cl− distances seen at the
14.9 molal and 8.6 molal LiCl solution and is interpreted as indirect chloride pairs which

Ionic structure in LiCl · 4D2O
8847
are bridged by a water cation complex. In the glass, the coordination number of 8.5 is,
as might be expected, intermediate between the values of ten and seven in the two liquids
(table 5).
5. Conclusions
The most important observation from the above study is that, compared to the liquid, the
glass has a higher number of ion pairs and ion clusters, which force the water molecules
into an ‘edge-on’ position in the cluster. The readiness of glass formation is conditioned
by the ability of the water molecules abetted by the strongly hydrated Li+ ion to produce
relatively stable clusters of ions and water molecules. Good glass forming concentrations
of aqueous lithium chloride are, therefore, those in which there is a sufficient number of
water molecules to fill the first hydration spheres of the ions but not enough to allow an
extended hydrogen bond network to be formed between the ion clusters. This hypothesis
could explain the complicated glass phase diagram of LiCl·nH2O and could be investigated
experimentally by a study of the hydrogen bond network, in the crystalline, liquid and glass
states, at different concentrations.
Acknowledgments
The authors wish to thank Dr Marie-Claire Bellissent-Funel and M Jean-Pierre Ambrose
(CEN Saclay, France) for their help with the neutron diffraction experiments and Dr Ian
Howell for access to unpublished results. We also thank EPSRC for financial support.
References
[1] Angell C A and Sare E J 1970 J. Chem. Phys. 52 1058
[2] Dupuy J, Elarby-Aouizerat A, Claudy P, Jal J-F, Letoffe J M and Chieux P 1987 Aqueous Ionic Solutions
NATO ASI C205, ed M-C Bellissent-Funel and G W Nielson (Dordrecht: Reisel) p 447
[3] Preval B, Jal J-F, Dupuy-Philon J and Soper A K 1995 J. Chem. Phys. 103 1897
[4] Preval B, Jal J-F, Dupuy-Philon J and Soper A K 1995 J. Chem. Phys. 103 1886
[5] Neilson G W and Enderby J E 1996 J. Phys.: Condens. Matter 100 1317
[6] Copestake A P, Nielson G W and Enderby J E 1985 J. Phys. C: Solid State Phys. 18 4211
[7] Howell I and Neilson G W 1996 J. Phys.: Condens. Matter 8 4455
[8] Newsome J R, Neilson G W and Enderby J E 1980 J. Phys. C: Solid State Phys. 13 L923
[9] Howell I 1993 PhD Thesis University of Bristol
[10] Narten A H, Vaslow F and Levy H A 1973 J. Chem. Phys. 58 5017
[11] Ichikawa K, Kameda Y, Matsumoto T and Misawa M 1984 J. Phys. C: Solid State Phys. 17 L725
[12] Jal J-F, Soper A K, Carmona P and Dupuy J 1991 J. Phys.: Condens. Matter 3 551
[13] Ansell S 1995 PhD Thesis University of Bristol
Ansell S and Neilson G W in preparation
[14] Vogel I 1953 Macro and Semimicro Qualitative Inorganic Analysis (London: Longmans)
[15] BDH 1980 Ion Exchange Methods (Poole, UK: BDH)
[16] Powell D H, Neilson G W and Enderby J E 1989 J. Phys.: Condens. Matter 1 8721
[17] Soper A K 1989 Inst. Phys. Proc. 97 (Bristol: Institute of Physics) p 711
[18] Yamagami M, Yamaguchi T, Wakita H and Misawa M 1994 J. Chem. Phys. 100 3122
[19] Cummings S, Enderby J E, Neilson G W, Newsome J R, Howe R A, Howells W S and Soper A K 1980
Nature 287 5784