R.

VOSZKA
K. RAKS~NYI,
Research Laboratory for Crystal Physics, Hungaria>n
Academy of Sciences, Budapest

Further Data on the Preparation

of Extremely Pure Sodium Chloride Crystals

The following methods for purification of sodium chloride were compared in relation
to their effectiveness : ion exchange, extraction, crystallization from solution,
precipitation and distillation. For the purification of sodium chloride the ion
exchange followed by a crystallization from solution in a teflon vessel appeared t o
be the most effective method giving a sodium chloride with an alkaline earth content
about 1 x mole/mole.
Es werden verschiedene Reinigungsmethoden wie Extraktion, Filllung, Ionenaus-
tausch, Kristallisation aus Losung und Vakuumverdampfen hinsichtlich ihrer Wirk-
samkeit verglichen. Die besten Resultate lieferte die Ionenaustauschmethode mit
einer darauffolgenden Kristallisation im TeflongefaB aus der Losung, wobei lo-'
Mol/Mol Reinheit erreicht werden konnte.

1. Introduction
I n a previous paper (VOSZKA, T A R J ~ NBERKES,
, KRAJSOVSZKY) we presented
some data on employing the zone melting technique and distillation for growing
ultrapure alkali halide monocrystals. For sodium chloride where the zone melt-
ing technique seems t o be ineffective, the second method of distillation, by
means of which a purity of 9 x 10-7 was attainable, proved to be inadequate
for the production of higher amounts of pure substance. For the latter purpose
we have investigated a lot of purification procedures to find the best one. I n
this paper some details on our experiments are presented.

2. Analysis
For the determination of alkaline earth ions we have chosen the method of
PARRY and DOLLMANN, i.e. a complexometric titrating procedure with an UV-
spectrophotometric end-point detection. The sensitivity of the method was
enhanced by a pre-concentrating process in an ion exchange column filled with
Dowex A-1. By this manner it became possible to measure the concentration
of the above mentioned ions up t o mole/mole with an error of 3 x
mole/mole in conc. sodium chloride solution.

3. The effect of the wall of the vessels and of the purity of water
I n the course of the various purification procedures the elimination of the im-
purities dissolved from the wall of vessels is of utmost importance. For this
purpose we have carried out investigations on usual laboratory equipment

15 *
228 R.I ,VOSZKA
K. R A K S ~ N Y

materials, as follows : sodium chloride with an impurity content about

mole/mole dissolved in redistilled water in various vessels, was dried during
a fixed time, and the alkaline earth content measured in three parallel experi-
ments. The data obtained are shown in Table 1.
Table 1
Metal (11)content Increasc in
used vessels 10-7 mole/mole metal (11)
before after content
experiments mole/mole
3.6 47.6 44.0
3.8 102.8 99.0
3.7 78.7 75.0
pyrover glass 3.4 20.5 17.1
(Hungary) 3.4 30.8 25.4
3.2 37.2 34.0
quartz glass 3.9 9.9 6.6
(G.D.R.) 3.7 10.7 7.0
3.5 10.7 7.2
durari glass 3.6 11.1 7.5
(G.F.R) 3.6 5.6 2.1
3.6 14.6 11.0
rasothcrm glass 3.0 8.0 5.0
(G.F.R) 3.6 9.2 5.6
3.3 5.2 2.0
teflon 3.5 4.0 0.5
(France) 3.4 3.2 -0 . 2
3.7 3.4 - 0.3
polyethylene 3.5 4.0 0.5
3.2 3.8 0.6
3.2 3.3 0.1

Before the experiments, the glass and quartz vessels were treated with an
1 : l O mixture of conc. nitric and sulphuric acid. The teflon and polyethylene
vessels were purified with conc. hydrochloric acid.
As it can be seen from the results, in the following experiments we used only
polyethylene and teflon vessels with one exception. It is interesting that the
unfavourable effect of the wall appears during the melting of sodium chloride
in a quartz crucible too. I n Table 2 some data are given on the alkaline earth
content getting into the melt during a two hours melting process in various
atmospheres. The three data quoted refer t o three independent experiments.
It follows directly from the data that the various atmospheres have no signifi-
cant effect, although we have to reckon with an increase of impurities up to
6.5 x mole/mole on are average. The great fluctuation of the data, far
exceeding the limits of error in the analysis shows, that here we encountered
a concentration limit depending on the very complex working conditions
playing a decisive part during the growth of crystals in a quartz crucible.
We have determined the alkaline earth content in water purified by various
procedures as follows: 1000 ml of water was evaporated, the used teflon vessel
Further D a t a on the Preparation of Extremly Pure Sodium Chloride Crystals 229

Table 2

metal (11)contcnt increase in

Atmosphere mole/mole metal (11)
before after content
melting mole/mole
~- - ~~ ~~ ~- ~ ~

3.2 ~ 4.0 0.8

air 2.8 5.3 2.5
2.9 14.0 11.1
3.0 7.1 4.1
argon 3.1 8.0 5.9
2.6 15.0 12.4
3.2 17.2 14.0
hydrogen c.-__Nride 3.2 10.0 6.8
(gaseous) 3.0 8.5 5.5
argon saturated with 2.7 9.0 6.3
carbon tetrachloride 2.9 5.0 2.1
3.1 9.7 6.6

Table 3
~

1 metal (11)
method of purification evaporating vessel content in
mole/molo
I 2.1
1 x distilled in a copper
apparatus
quartz I 3.1
4.5
~~

2.7
1 x distilled in a rasotherm
1.6
apparatus 1.9
0.4
Dowex 50 +Dowex I 0.45
ion exchange 0.48
0.24
0.1
2 x distilled 0.27
a ) in a rasotherm
b) in a quartz apparatus 0.12
toflon 0.18
0.13
0.09
quartz 0.07
Dowex A-1 0.06
~~ ~~
~ ~

ion exchange 0.03

teflon 0.02
0.02
230 R.I ,VOSZKA
K. R A K S ~ N Y

was treated subsequently with 3 ml of dilute hydrochloric acid, and the concen-
trations were determined by the usual way. The data show the amount of
impurities which would be present in the sodium chloride if a 5 M solution of
the latter made with the above mentioned amount of water have been evapo-
rated (Table 3). From the data we can draw the conclusion that the impurity
content of bidistilled water does not play an important part in such experiments,
where the final impurity content of the substance does not exceed the value of
mole/mole. If the impurity content is to be reduced farther, the water
has t o be purified with Dowex A-1 ion exchange resin.

4. Purification by extraction
Three methods of ion extraction were investigated :
a ) The system described by HEALY,i.e. 0.18 M thenoyltrifluoracetone
+
(TTA) 0.014 M trioctylphosphine-oxide (TOPO) in cyclohexane a t p H 7
b) The system, described by UMLAND and MECKENSTOCK, i.e. 2% 8-hydroxy-
quinoline +2y0 n-buthylamine in chloroform a t p H 11, and finally
c) The system described by WILLIAMSand WILSON,i.e. 0.1% glyoxal-bis-
hydroxyanile (GHBA) in chloroform a t pH 12.5. The experiments were carried
out on a 5 M solution of sodium chloride (reagent grade, Reanal, Budapest)
subjected t o a previous purification process with dithizone in carbon tetra-
chloride for the removal of the interfering heavy metal cations. The solution
was shaken three times with an equal volume of extracting agent in polye-
thylene equipment. The pH values were adjusted by isothermally distilled
ammonium hydroxide.
4.1 TTA in cyclohexane forms complexes with the alkaline earth ions a t p H
values above 7. Having in mind the synergistic effect of TOP0 on these extrac-
tions, we could achieve a favourable extraction ratio a t pH 7 too, where the
reagent alone is not so effective, and above which the reagent decomposes. We
were able to produce a substance with an impurity content below the detection
limit. but the possible introduction of other impurities (e.g. fluorine) rules out
this method.
4.2 n-Buthylamine has also a synergistic effect on the extraction of alkaline
earth ions with 8-hydroxychinoline. However we could not decrease here the
impurity content below 1 x lo6 mole/mole. At a certain elevated pH value
the sodium hydroxyquinolinate precipitates, and thus i t is not feasible t o
increase the p H to the optimal level. The distribution ratio for barium (11)
ions is also unfavourable. We have attempted to mix an alkaline solution of 8-
hydroxyquinoline a t elevated pH with a 5 M sodium chloride solution, and
after a subsequent filtration of the coprecipitated sodium calcium hydroxy-
quinolinate we tried to remove the remained free hydroxyquinoline by adsorp-
tion on activated carbon. Although this simple method is applicable for attaining
an impurity level of about 1 x mole/mole, we have not succeeded in
exceeding this limit.
4.3 By the use of the GHBA system an impurity level of 5 x mole/mole
could be attained. The barium (11) and strontium (11)complexes of this reagent
are not enough stable as it is revealed by the insolubilitv of their carbonates in
Further Data on the Preparation of Extremly Pure Sodium Chloride Crystals 231

GHBA a t elevated pH values. This fact seems to be responsible for the limita-
tions of this method. Nevertheless it was found, that this system is applicable
for spectrophotometric or direct titrimetric determination of calcium (11)ions
in concentrated sodium chloride solution too. We have used a M EDTA
solution as titrating agent. Here it is advisable t o work with a tenth of the
amount of indicator given in the literature (WILLIAMS and WILSON).By using
n-butanol instead of chloroform this method can be employed also for the
determination of relatively high quantities of calcium (11)in conc. sodium chlo-
ride solution because of the greater solubility of the calcium GHBA complex in
higher aliphatic alcohols as compared to chloroform.

5. Precipitation
An attempt was made to purify the reagent grade sodium chloride (containing
5-7 x 10-6 mole/mole of bivalent impurities) by the addition of a hundredfold
excess of calcium (11)ions as compared t o its initial value, and by a subsequent
removal of the calcium (11)ions in the form of carbonate, we have got a sub-
stance containing 1-2 x mole/mole of calcium (11). The p H increase
does not influence the results. Although this method seems not to be an effec-
tive one, yet it is applicable for the removal of other bivalent impurities origi-
nally present in the sample, including strontium (11)and barium (11).
This coprecipitation method is widely used in the isotope technique. Using
barium (11)ions instead of calcium (11)we tried to remove the calcium (11)ions
as a coprecipitate. It was assumed, that by means of a subsequent application
of the zone-melting technique, a substance of sufficiently low impurity content
can be obtained due to the relatively low value (0.6) of the distribution coeffi-
cient of barium (11)ions in the melt of sodium chloride (GRUNDIG, WASSERMANN).
We carried out measurements to justify this assumption. The substance ob-
tained had an impurity content of 4.4 x 10-7mole/mole following a zone-
melting in 20runs. At the end of the crucible we measured a value of 2 x
mole/mole. This indicates a remarkable purification, although the ob-
tainable purity depends here again on the quality of the quartz crucible.

6. Ion-exchange
The ion-exchange resin Dowex A- 1, containing built-in iminodiacetic groups is
capable to exchange the alkaline earth ions in the presence of excess alkali ions
too (SIDES,KENNER).This was observed to be valid also in conc. sodium chlo-
ride solutions, as follows: 1.3 g of the wet resin (having a dry substance content
of 0.3 g) in sodium form was shaken for 6 hours with 16 nil of 4 M sodium chlo-
ride solution doped to 0.01 M with one of the alkaline earth ions in question a t
different pH values (adjusted with an ammonium hydroxide - ammonium
chloride buffer solution in such a manner that the final total concentration of
ammonia be aqua1 to 0.25 M). The alkaline earth ion concentration of the super-
natant was measured in the usual way. The dependence of the logarithms of
distribution coefficients on the pH is given graphically in Figure 1.

s = metal (11)mmole/g of dry substance

~~~

metal (11) rnmole/ml of solution

 K. R A K S ~ N Y
R.I ,VOSZKA

Fig. 1. The dcpendenoe of the logarithms of the

distribution coefficicnts(log S ) on the pH for calcium
(II),(x),magncsium (11) (O), strontium (11) ( 0)and
barium (11) (A) ions during the ion cxchange in a
4 11 sodium chloridc solution

3 4 5 6 7 8 9 I0
PH

From the plotted values it follows, that a t elevated p H values the conc. sodium
chloride solution can be effectively freed from the alkaline earth ions. The
effect,iveiiess of this process increases in the series calcium (11) magnesium >
> >
(11) strontium (11) barium (11).
The flow rate influences the quantity of retained ions in a great extent. For
example the flow rate should be less than 0.7 cm min-l in a solution containing
less than 1 x mole/mole of the ions in question.
The method can be exploited also for analytical purposes as a powerful tool
for preconcentration, as mentioned in section 1. Making use of this method the
impurity content in sodium chloride can be reduced easily t o about 1 x lo-'
mole/mole. The method applied t o a pre-purified substance with a subsequent
crystallization in teflon vessel seems to be the most simple and effective way
for the purification of sodium chloride.

7. Crystallization
We investigated the effectiveness of crystallization for purification of sodium
chloride. We were engaged in this method immediately after the beginning of
our experiments. It happened exactly in the course of these experiments that
our attention was focussed on the great importance of the used laboratory equip-
ment. The experiments were carried out as follows : a 5 M solution of sodium
chloride purified by ion exchange and doped with a known amount of alkaline
earth ions was heated. One part of the vapour condensed on the wall of the
beaker and dissolved during its reflux the little crystals accumulated on the
wall by capillary forces. It is of extreme importance, because a crystal layer
formed by too rapid evaporation has nearly the same impurity concentration,
as that of the mother liquor. I n the just described way 90% of the total amount
of sodium chloride crystallized in 5-6 fractions and the alkaline earth content
of each fraction was measured, The data are plotted in Figure 2 . On the hori-
Further Data on the Preparation of Extremly Pure Sodium Chloride Crystals 233

I 1 I L
0 50 ?OO 750
(LL mole/mole)
Fig. 2. The dependence of the apparent distribution ooofficionts of the cdcinin (11).
magnesium (11) and barium (11)ions on their concentration in solution citlculated for
the amount of dissolved sodium chlorido a t a crystallization rate of
0.8 g/min/1000 ml ( - - - ) and 0.15 -0.2 g/min/1000 ml (-) in a quartz wsscl

zontal axis there are the concentrations of the solution expressed as mole alka-
line earth ions/mole sodium chloride, on the vertical one the distribution coeffi-
cients in per cent, i.e. the ratios of the concentrations of the alkaline earth ions
in the solid and the liquid phase respectively. The distribution coefficients fall
in the range of 4-S% for all mentioned ions a t a crystallization rate of 0.8 t o
1.9 g/min/1000 ml solution. At a lower rate, (0.15-0.25 g/min/1000 nil) the
distribution coefficients have a lower value although a t lower concentrations
show a marked increase. This can be interpreted perhaps as impurities arising
from the quartz wall. To ascertain this, experiments were carried out in teflon
equipment. Here we applied a teflon mixer with 100r.p.m. and allowed t'o
crystallize only a small part of the sodium chloride in order t o achieve a suitably
low concentration. The data are plotted in Figure 3. It can be seen, that the
distribution coefficients are considerably lower, although their increase can be
observed from lower concentrations. This increase can not be attributed to the
impurities arising from the wall, because the error of measurements does not
234 K. R A K S ~ N YR.
I , VOSZKA

exceed 10% a t these concentrations, as it is apparent fromt the data in Table 3.

The data of Table 1. point to the small amount of impurities if a teflon equip-
ment is employed, which does not reach the limit of detection. I n order to
throw light on this phenomenon we can assume, that either the mechanism of
crystallization process undergoes a change a t lower concentrations, or other
cations may be present in very small concentrations, which get enriched in the
solid phase (e.g. lead (11)) oppositely t o the alkaline earths, and are measured
together with the latter. Nevertheless the fact remains, that this rise in bivalent
cation content makes it impossible to apply this method for the production of
extremely pure sodium chloride.
Fig. 3. The dcpendencc of the apparent dia-
tribution coefficients of calcium (11) ions on
it,s concontration in solution calculated as bc-
fore a t a. crystallization rate of 0.6-0.7
g/min/lOOO ml whilc stirring in a teflon
resscl

8. Vaauum distillation
I n the following we have investigated the vacuum distillation of sodium chloride
in order t o study the effect of basic additives on the gas-liquid distribution coeffi-
cient expressed by the concentration ratios in mole/mole. I n the experiments
some interesting results were obtained which are worth of mentioning.

8.1 Preparation of starting material

The sodium chloride (reagent grade) was purified according to the scheme
described in 6., doped with a known amount of calcium chloride obtained from
the reaction of calcium carbonate (reagent grade) and isothermally distilled
hydrochloric acid. The sodium carbonate was prepared by the recrystallization
of the reagent grade material. The amount used in a run was 60 g.

8.2 Equipment
The distillation was performed in a device described in our previous paper as
“first apparatus” where the cooling system was equipped with a nickel plate
of a diameter equal to that of the crucible. Instead of a quartz crucible a plati-
num one was used owing to its greater resistance to basic additives. Blank tests
were also carried out with a specially purified starting material having an impuri-
t y content less than mole/mole, in order to eliminate the error due to the
Further Data on the Preparation of Extremly Pure Sodium Chloride Crystals 235

impurities arising from the crucible wall. The degree of the removal of hydroxide
ions plays an important part in the procedure. This was achieved by the intro-
duction of dried hydrogen chloride gas into the melt. Subsequently the distillate
was collected in portions of 5-10 g. The material remaining in the crucible
proved to be neutral measured after dissolving in water.

8.3 Results
The effect of different amounts of added sodium carbonate on the gas-liquid
distribution coefficient was studied. The data are given in Table 4. The table
shows, that in the used apparatus the effectiveness of distillation goes parallel
with the concentra)tion of sodium carbonate in the melt. Nevertheless if we

Table 4
.______. .~

additives pretreatment with distribution

hydrogen chloside coefficient
__.-
- *) 0.08
- 0.22
0.1 yo sodium carbonate 0.08
0.50/6 sodium carbonate 0.04
1 yo sodium carbonate 0.01

*) In that case the hydroxide content of the substance

remaining in the crucible was 2 x lo-* mole/mole

calculate the distribution coefficient from the data of BREWER,BRACKETT

and HILDENBRAND and assume an ideal behaviour in both phases, a value of
0.002 can be obtained. The great difference between the measured and calcu-
lated data is apparent, and requires a further elucidation. It is obvious here t o
assume either a non-molecular irreversible process or interactions between ions
or molecules in the vapour or in the melt. I n the first case small liquid particles
may be carried away from the melt via the formation of macroscopic bubbles of
dissolved gases freed by the vacuum. I n basic medium this macroscopic bubble
formation was not observable, yet the existence of microscopic bubbles can not
be excluded : A temperature decrease may take place on the gas liquid interface
and this involves the presence of ordered structures acting as thermal insulators
towards the melt, where in the layers lying beneath the interface an overheating
may take place with subsequent eruptions tearing away several liquid particles
towards the nickel plate. The non molecular character of such a vacuum distilla-
tion in a system where the cooled surface is placed directly opposite t o the melt
(ROSENBERGER) is evidenced by the data obtained with a system where a quartz
plate is placed between the two surfaces. I n the latter case we obtained con-
siderably lower results, although the effect of added sodium carbonate was
detectable here too. This can be explained by the reaction of the calcium (11)
with the carbonate or hydroxide ions in the melt. The corresponding compounds
are known t o have a remarkably lower vapour pressure. It is of practical inter.
est, that in a purification process via distillation the direct vapour path between
the melt and cooling plate is to be avoided.
236 R. R A K S ~ N YRI., VOSZKA

9. Conclusions
It appears to us, that the most effective and economic way of the purification
of sodium chloride is the ion exchange one applied to a pre-purified substance
and including several subsequent crystallizations from solution in a teflon
equipment. Further purification o f the sodium chloride is limited by the purity
of water and by the chemical resistance o f used equipment. But in the crystal
growth from the melt the substance is more exposed t o impurity sources, and
the lower limit of attainable concentration is about 1 x mole/mole. This
value was obtained even when growing a zone-refined potassium chloride of
mole/mole purity (calculated from conduction measurements) by the
Bridgman method in a quartz crucible.
Finally we express our gratitude to Prof. Dr. I . Tarjkn for his keen interest
in our work and to Mr. L. &xi for his helpful assistance in performing the
experiments.

Literature
BREWER, L., BRACKETT, E.: Chem. Rev. 61, 425 (1961)
GRUNDIG, H., WASSERMANN, E.: Z. Physik 176, 293 (1963)
HEALY, T. V.: J. inorg. nuclear Chem. 19. 324 (1961)
HILDENBRAND, J. H.: J. phys. Chem. 67, 2232 (1965)
PARRY, E. P., DOLLMANN, G. W.: Analytic. Chem. 36, 1783 (1964)
ROSENBERCER, F.: Mat. Res. Bull. 1, 55 (1966)
SIDES,J. L., KENNER, Ch. T.: Analytic. Chem. 38, 707 (1966)
UMLAND, F., MECKENSTOOK, K. U.: Z. analyt. Chem. 165, 161 (1959)
VOSZKA, I.,
R., T A R J ~ N , BERKES, L., KRAJSOVSZKY, J.: Kristall u. Technik 1, 423
(1966)
WILLIAMS, K. T., WILSON, J. R. : Analytic. Chem. 33, 244 (1961)

(Received August 15, 1968)

Authors' address:
K. R A X S ~ N Y I
Dr. R. VOSZKA
Orvosi Fizikai Inthzet
Budapest V I I I
Puskin U.. 9