\
PIST-CW 66-0556
THE JOUR",:\L OF CHEMICAL PHYSICS
VOLUME 45, NUMBER 10
Effect of Pressure on the Melting Points of the Sodium Halides
general
-
15 NOVEMBER 1966
CARL W. F. T. P1STORIUS
(65)
ere N (s)
upon a
th others.
lute gas),
enological
thod proquivalent
were to
btain the
erage in a
ied in our
ume that
interval
e.g., the
d binary
restricted
h spheres
e torques
difficulty
ntervalof
C/ltlllical Physics Group of tlte )."Tatiol/aJ Physical al/d Natiollal Chemical Research Laboratories, South AfricaIJ Council for &imtiflC
and Indllstrial Research, P. O. Bo .. ,; 395, Pretoria, Sou/II Africa
(Received 5 July 1966)
The melting curves of the sodium halides have been detemlined to 40 kbar. The present curves for
NaCl and r aF are iu agreement with Clark's corrected curves to 17 kbar, bUl the agreement for the other
sodium halides is less good. There are no triple points on the melting curves to 40 kbar and it is concluded
that the transitions found previously in ~aF and NaCl near ",20 kb:u, if real, are metastable and shear
induced.
I•
INTRODUCTION
HE high-pressure melting properties of the potas-
T
shun halides l and the rubidium halides2 have
recently been reported. The present investigation continues this study and is specifically concerned with the
determination of the melting curves of the sodium
halides at pressures up to 40 kbar. The sodium halides
have been studied before by Clark,2 but his pressure
mnge was limited to ,.....,20 kbar, and in some cases to
",10 kbar.
There are indications that the sodium halides may exhibit polymorphism at,.....,I0-20 kbar and 25°-200°C.4-9
However, the stable transition from the ~aCI-type
structure at low pressures to the CsCI-type structure
at higher pressures would seem to occur near ,.....,250
kbar.HI The probable nature of the 10-20-kbar transition is discussed later.
EXPERIMENTAL
The salts were obtained from Merck and from
Hopkin and Williams, and were of reagent grade. They
were dried by heating to """'SOO°C for a few hours, or
by melting in a platinum capsule. The 1-atm melting
points agreed to within 2°- 3°C with the best values
from the literature, viz., 992°C for NaF,3 800.5°C for
~hCl"2 i·HoC for NaBr,3 and 655°e for NaI.3
Pressures up to 40 kbar were generated in a pistonC. W. F. T. Pistorius, J. Phys. Chern. Solids 26,1543 (1965).
C. W. F. T. Pistorius, J . Chern. l'hys. 43, 1557 (1965).
IS. P. Clark, Jr., J . Chern. Phys. 31,1526 (1959).
• V. V. E\'d<lkimova and L. F. Vereshchagin, Fiz. Tvcrd. Tela
4, 1965 ( IQ62) [English trans\.: Soviet Phys.-Solid State 4,
1438 (1963)).
'C. W. F. T. Pi.-torius and H. C. Snyrnan, Z. Physik. Chern.
(Frankfurt' 43, 1 (19CH) .
• C. W. F. T. Pistorius, J. Phys. Chern. Solids 25, 1477 (19CH).
T C. W. F. T. Pistorius, j. Ph),;;. Chern. Solids 26, 1003 (1965).
• D. B. Larson, in Physics of Sn/ids at HiJl.h PrtSSllres, C. T.
Tomizuka and }{, ~L Emrick, Ells. (Acarlemic Press Inc., New
York, 196-), p. 459.
I D. B. Larson, R. N. Keeler, A. Kusubov, and n. L. Hord,
J. }'h),s. Chern. Solids 27,476 (1966).
10 J . C. jarnie;:on, in Physics 11 SIJ/ids at lIigh Pus.fllrts, C. T.
TOmizuka anrl R. ~L Emrick, Etls. (Academic j're<s [nc., New
York, 19(5), p. 4·H.
II G. E. Hauver and A. Melani (unpuhlished work).
II
S. Roberts, Phys. Rev. 23,386 (1924).
I
I
the rate of
hydrodyd directly
<»). These
Chemical
r Charles
n.
cylinder device previously described. 13 •14 :Melting and
freezing were detected by means of differential thermal
analysis, using platinum/ platinum-10% rhodium thermocouples. Corrections were made for the effect of
pressure on the thermocouples, using the data obtained
by Getting and Kennedy!S and making approximate
allowance for the crushing strength of the ceramic
thermocouple tubing and the temperature of the
pressure seal. Typical values of this correction amounted
to +5°e at 10 kbar and +12°C at 40 kbar for a. temperature difference of 800°C between the thermocouple
junction and the pressure seal.
The salts were contained in nickel capsules which
usually incorporated thermocouple wells.! There was
no evidence of any contamination of the sampJe by the
capsule even after many meltings. Sample containers
of stainless sleel were also tried, but the molten sodium
halides reacted strongly with the capsule material.
Each melting curve is based on at least three separate
runs. The results of different runs, even though carried
out using sample capsules of widely differing design,
were consistent to ,.....,3°e or better. Heating and
cooling rates were usually in the range 2°-6°C/sec.
However, beating rates of up to 20 o e / sec gave identical
results. Temperatures could be determined to ±3°-6°e.
Freezing points were usually within a few degrees of
the melting points after the proper corrections had
been made for friction. Extreme care was taken in
making reliable friction corrections and the final pressures are believed accurate to better than ±0.5 kbar.
EXPERIMENTAL RESULTS
The data are shown in Figs. I--t Clark3 corrected
his results for the effect of pressure on the thermocouples by using a linear extrapolation of Birch's
measurements. 16 In the range of interest the resulting
13 G. C. Kennedy and P. N. La~fori, J. Geophys. Res. 67, 851
(1962) .
U G. C. Kennedy and R. C. Newton, in Solids under Pressure,
W. Paul and D. Warschauer, Eds. (McGraw·Hill Book Co., Inc.,
New York, 1962).
16 I. Getting and G. C. Kennedy (unpublishcrl work).
10 F. Dircll, Rev. Sci. Instr. 10, 137 (1939).
3513
•
3514
CARL W.
F. T.
PISTORIUS
1200
u
0
I
FIG.!. The melting curve of NaF to
40 kbar. Solid line, present work;
dashed line, Clark (1959).
w
a:
::>
~
a:
w
0..
1100
~
W
t-
~·L---~5~--~I~O---'~15'---~2~O----~25O--'~3~O----~35O---~4~O--~
PRESSURE - KILOBARS
TABLE I. Parameters of the melting curves.
Substance
Source of data
To(OK)
NaF
Clark (1959)
1265
NaF
Clark (corrected)
1265
NaF
Present work
1265
NaCl
Clark (1959)
1073.5
NaCI
Clark (corrected)
1073.5
NaCl
Present work
1073.5
NaBr
Clark (1959)
1014
NaBr
Clark (corrected)
NaBr
A (kbar)
14.3
c
Standard
deviation
(OC)
b.
b.X103
Standard
deviation
(OC)
I =to+bIP+~p2+baP3
I.
ITJ ·
5.5
1.3
4.886
1.6
992
14.58
-0.205
-0.453
1.7
12.2
5.762
4.2
992
15.47
-0.219
1 .429
3.4
16.7
2.7
3.3
2.810
4.2
799.6
23.44
-0.383
4.928
4.3
2.969
3.7
802.5
21.61
-0 . 201
0.576
2.7
2.9
5.6
1014
3.076
6.4
741
27 . 57
-0.521
6.362
6.5
Present work
1014
3.356
4.0
742
26.41
-0.529
6 .095
3.4
NaI
Clark (1959)
928
2.8
6.1
NaI
Clark (corrected)
928
2.942
6.8
656
31.53
-0.676
9 .350
7.0
NaI
Present work
928
3.649
5.0
655
27 . 66
-0.388
2.064
2.4
C0
~
5.0
12.2
~I
E L TIN G
r 0 I N T S 0 F SOD I U M II A LID E S
3515
1400r-----r-----.-----.-----.-----.-----.-----.---~
/0/
/0
/0
1300
1200
lTve of NaF to
>resent work;
l) .
U
FIG. 2. The melting curve of NaCI . 0
to 40 kbar. Solid line, present work;
dashed line, Clark (1959).
w
a::
,0
1100
/
o
,/
:::J
~
a::
w
.,
,,«
I
I
/'
0..
::E
w
I
/
-
1000
I-
900
/
I
I
,,,I
,
I
I
I
-
/
I
I
I, ___~____~____J -____~'____~__-,~____L -_ _~
800a10
20
30
40
PRESSURE - KILOBARS
correction is too large by a factor of "-'2-3, and the
dashed lines in Figs. 1-4 represent Clark's raw data
corrected according to the results of Getting and
Kennedy.15 It can be seen that the agreement between
the present results and Clark's results is good for XaF
and NaCI, but less good for NaBr and XaI. The
agreement between Clark's data below "-'10 kbar and
the extrapolation of the present melting tines is good
in all cases, but above ,....,,10 kb:u Clark's curves differ
from the present curves by amounts increasing with
pressure. For Nal" our cun'e lies above Clark's curve,
but within the maximum combined experimental error,
while for NaBr and Nal our curves fall somewhat
below Clark's curves. As has been stated above, the
present cun'es are reproducible to within "-'3°e even
when using differing sample geometries, different drying
methods, and salts from different origins. Clark stated
that "for some unknown reason the melting curves of
NaBr and Nal proved to be more difficult to locate
in this apparatus, and the determination of the melting
points of these salts are uncertain by ±lO°C."
A tentative extrapolation of the melting curves!6
indicates that abovC"-'6G-90 kbar the normal order
of melting of the sodium halides is reversed, with
NaF having the lowest and Nal the highest melting point. Similar behavior was found previously
for the potassium! and rubidium2 halides, but at much
lower pressures.
The curves in the figures were calculated from the
simple equation
fitted to the experimental points by least squares.
P was the pressure in kilobars. The parameters are
given in Table I. The experimental points were also
CARL W.
3516
F.
T.
PISTORIUS
/
./.
I~OO
1200
/
!
- '71
..
/
"-
/- :l
1100
/./
/1/
1.1 .
1/
0
0
1000
~
~
I
a.
~
w
~
FIG. 3. The melting curve of NaBr
to 40 kbar. Solid line, present work i
dashed line, Clark (1959).
40
•i
1
das
,
'I"
W
0:::
:l
0:::
W
/'
I
,1
"
I.
900
I.
I.
I
'.
,
'.
700L-----L-----~----~----~~--~25~---,3~0----'3~5----~40
PRESSURE - KILOBARS
fitted to the Simon equation!7
P-P.o=A [(T/ To) c-l],
where T is the melting point (in degrees Kelvin) at
P (in kilobars), and A and c are adjustable constants.
p. and To are the coordinates of the triple point of
the phase in question. For the sodium halides P~O.
A and c were determined by means of Babb's
method. ls All calculations were carried out on the IBM
704 of the National Research Institute for 1\b,thematical Sciences. The parameters are given in Table 1.
~o uncertainties are given for A and c, since it was
previouslyl~ found that very small changes in experimental data can affect the Simon parameters by
large amounts.
1&
F. E. Simon and G. Glaue1, Z. Anorg. Allgern. Chern. 78, 309
(1929) .
17
S. E. Babb, Jr., Rev. Mod. Phys. 35, 400 (1963).
I' C. W. F. T. Pistorius, M. C. Pistorius, J. P. Blakey, and L. J.
Arlmiraal,
J. Chern. Phys. 38, 600
(1963).
iJ
I
111
\'
r
I\
MELTING
POINTS OF SODIUM
HALIDES
3517
I~r-----r---~r---~----~-----.-----'-----'-----'
1000
e of NaDr
sent work;
u
0
FIG. 4. The melting curve of NaI to
4.0 kbar. Solid line. present work;
dashed line. Clark (1959).
W
0::
::>
too
<
0::
900
W
a..
:::E .
w
too
6ooL---~5~--~1~0----~15-----2~0~~~
25~--~3~0----~3~5----~40
PRESSURE - KILOBARS
of Babb's
n the IB~[
IathematiTable J.
nce it was
in experimeters by
~Y. and L. J.
THERMODYNAMICAL CALCULATIONS
The initial slope of the melting curve, (d Tl dP)p-o,
is Tol Ac by differentiation of the Simon equation. It
is a.1so equal to ill/fillS" where llV, and tiS, are the
molar changes of volume and entropy on melting.
Values of the initial slopes calculated from the Clapeyron-Clausius equatiori are compared in Table II with
those observed by Clark,' as corrected for the effect of
pressure on the thermocouples,15 and with the present
values, as obtained from the Simon equation and
from the power series fit. It is obvious that the various
fits yield a fairly wide scatter of initial slopes. This
suggests that the initial slopes obtained from leastsquares tits of melting curves are not as reliable as has
been thought previously. It is particularly dangerous
to obtain the initial slope from a power series fit,
since we have often found .that the mathematically
best fit of the melting points is obtained by a power
series which has a small maximum or minimum near
zero pressure. This is usually found when the experimental data are especially good, with the exception of
•
3518
CAR 1_
W. F. T. PISTORIUS
TABLE II. Melting parameters at zero pressure.
(dT/dP)p=oob.
(deg/bar)
t:.S,
(cal/rnole·deg)
t:.V,
(crna/rnole)
t:.Vt!t:.S,
(deg/bar)
NaF
S.S"-6 .2b
4.15 0 -4.64"
NaCl
6.3 h -6.7"
NaBr
S.9h -6.()&
Compound
NaI
5 .6s .b
To/At;
b,
0 .016-0.020
0 .0151 d
0.0180
0.014'6 d
0.0155
7.55"
0.027-0.029
0.0236d
0.0241
0 .0234d
0.0216
8 .07"
0.032-0.033
0 .0289 d
0.0272
0.0276d
0.0261
0 .037
0.0332 d
0.0315 d
0.0357
0.0277
8.58"
" H . Schinke and F. Sauerwald, Z. Anorg. Allgem. Cbem. 287, 313 (1956).
b F. D. Rossini, D. D. Wagman, W. H. Evans, S. Levine, and I. JaJIe, NaU.
Bur. Std . (U.S.) Cire.. No. 500 (1952).
incorrect friction corrections at low pressures, or when
the room-pressure melting point is incorrect. In such
cases ' the initial slope obtained from the fit will be
totally incorrecL
The agreement between observed and calculated
slopes is fair, with the present values showing better
agreement than Clark's corrected values. The worst
agreement is for NaBr and it would seem certain that
either AS, or A V.t for NaBr is incorrect.
The change of volume upon melting is considerably
smaller for NaF than for the other sodium halides,
and, while the chloride, bromide, and iodide have
melting curves which are remarkably similar in curvature and slope, NaF has a considerably different
melting curve. The explanation for this behavior can
be found in data obtained by Frank and Foster20 for
NaF. They found that molten NaF is strongly dimerized, with the dimers completely ionized to NaF2and Na+ and the monomers completely ionized to
Na+ and F-. At lOOO°C the degree of dimerization
was 32%. Dimerization is not known to occur in
molten KaCl, TaBr, or Nal. The dimer obviously
is denser than the monomer and causes the small
volume change. It is to be e}.:-pected that, at a given
temperature, the degree of dimerization \yill increase
strongly with pressure. It is probable that molten
KF and RbF are also partly dimerized.2
POLYMORPHISM
Evdokimova and Vereshchagin4 reported a very
sluggish phase transition in NaCl near 18 kbar by
means of high-pressure x-ray techniques. This transition was subsequently observed by Pistorius,6 using
volume methods, and by Larson and his co-workers,8.9
using shock-wave techniques. However, several other
., W. B. Frank. and 1.. M. Foster, J. Phys. Chern. 61, 1531
(1957) .
• G. J. Landon and A. R. Ubbelohde, Trans. Faraday Soc. 52, 647 (1956).
s. P. Clark:, Jr., J. Cbem. Pbys. 31,1526 (1959) (data corrected).
d
workersl0.21.~ failed to detect this transition. Similar
phase transitions were observed for N aF,5 and possibly
for N aBr and N aV According to Evdokimova and
Vereshchagin4 the transition is the expected one from
the NaCl-type structure to the CsCI-type structure.
However, Jamieson's work1o on KCI-NaCl solid solutions under pressure indicates convincingly that the
stable NaCl-type to CsCI-type transition in NaCl
takes .place near ,.....,250 kbar, and this is confinned by
shock-wave measurements,u·23 Furthermore, the present
melting curves show no trace of any possible triple
points involving stable low-pressure phases.
The reality of the 18-kbar transition in NaCI is
still open to doubt, although it has been found by
three different laboratories using completely different
techniques. What appears to be the case is that this
transition is metastable and probably shear induced.
The kinetics of this phase change are unusual. One
gets the impression that it occurs mainly while pressure
is being changed and there is a resultant shear component present. In this respect it is very similar to the
martensite reaction. Although a martensite transition
is unusual in a nonmetal, it is known that NaCN,
possessing the NaCl-type structure 'a bove 288°K,
undergoes a martensitic transition to a metastable
rhomboh edral form upon being heated above 200°C
and quenched to 20°C. 24 We suggest therefore that the
,....,1D-20-kbar transitions in NaF and NaCl, if real ,
are metastable martensitic-type transitions from the
NaCl type to probably the CsCI-type structure. If the
0. Johnson, Science 153, 419 (1966).
E. A. Perez·Albueme and H. G. Drickamer, J. Chern. Phys.
43. 1381 (1965).
"A. Kusubov, University of California Lawrence Radiation
Laboratory Rept. UCRL-12473, 61, 1965.
2. n . L. Averbach , quoted by ~r. Cohen, in "The Martensite
Transformation," Pha!~ Trall s/ormatiolls in S olids, R. Smolu·
chowski, J. E. Mayer, and W. A. WeyI, Ells. (John Wiley &
Sons, Inc., New York, 1951).
21
22
•
,,' #
•
I
l
I
I
I
t
;2, 647
ted).
(1956).
NaCI is
found by
y different
; that this
.r induced.
Llsual. One
Ie pressure
;hear commar to the
• transition
tat NaCN,
ve 288°K,
metastable
ove 200°C
,re that the
CI, if real.
s from the
lure. II the
POIXTS OF SODIUM
low-pressure transitions encountered in NaBr and
NaI are real, they are probably similar in nature.
Recently a paper appeared 25 suggesting the use
of NaCI as a standard for calibration of static highpressure high-temperature apparatus. This paper was
criticized9 on the basis that NaCI is a substance whose
phase behavior is extremely complicated. If the
above suggestions are correct, however, NaCI would
still be a suitable standard to "-'200-250 kbar and
0-1500°C, since the small amount of the metastable
phase formed should not interfere with the x-ray
measurement of the lattice spacing of the NaCl-type
phase. However, this presupposes that the original
compression data for NaCl upon which the calibration
curve is based be free of shear effects, i.e., that the data
refer solely to the compression of the NaCI-type phase.
This mayor may not be the case for Bridgman's
results,26 but shock-wave data would not be suitable,
D. L. Decker, J. Appl. Phys. 36, 157 (1965).
P. W. Bridgman, Proc. Am. Acad. Arts Sci. 74, 21 (1940);
76, 1 (1945).
25
Similar
d possibly
nova and
one from
structure.
solid solu. that the
in NaCl
lfirmed by
:hepresent
iible triple
il.
il
MELTIXG
2tI
THE JOURNAL OF CHEMICAL PHYSICS
ce Radiation
e Martensite
s, R. Smolu·
hn Wiley &
since considerable shear is thought to be present in a
shock front. The occurrence near ",,250 kbar of the
stable NaCI-type to CsCI-type transition would in any
case make it impossible to use Decker's equation of
state2' above this pressure.
It must be conceded, however, that the bulk of
present evidence indicates th at the l0-20-kbar transitions in the sodium halides are probably not real,
expecially since Johnson21 showed that the positive
x-ray evidence4 may be in error.
ACKNOWLEDGMENTS
The author would like to thank Mrs. Martha
C. Pistorius for writing the computer programs,
Dr. E. Rapoport of this Institute, and Professor
L. F. Vereshchagin of the Academy of Sciences of the
USSR for valuable discussions, and Mr. B. Clark for
able assistance with some of the experimental work.
Mr. J. Erasmus and his staff kept the apparatus in
good order.
VOLUME 45, NUMBER 10
15 NOVEMBER 1966
EPR in Selenium
P. I. SAMPATH
F1mdamental Researc!: Laboratory, Xerox Corporation, Roc/zester, New York
(Received 27 June 1966)
l(
EPR studies at X-band frequencies have been carried out between 3000 and 77°K on amorphous and
crystalline selenium. Dominating the spectra of the amorphous material were a large number of paramagnetic centers ""1019/cc which contribute to an unusually broad and inhomogeneous line of width,
AD I'p,....,3Kg and g=2.3 to1.8±0.01. The linewldtbs and .g values were very sensitive to heat treatments.
This broad line is attributed to electrons at chain ends in a highly disordered region of the solid. A different
species was also present in much lower concentrations ",,1013 /CC, as observed by a sharp resonance with
peak-to-peak linewidth, tlH 1'"",6 G and g=2.0039±O.0006, which is close to the g value of 2.0023 for a free
electron. Heat treatments altered the magnitude of the signal in an erratic manner and up to """lOI6/cc spins
were observed with no changes in tlil liP or g value. The identification is rather uncertain. Theoretical
considerations of the g value and linewidth favor a center associated with an impurity (Le., oxygen) at
the end of the selenium chain but have shown that a chain end interacting with a neighboring chain cannot
be ruled out. Experimental results tending to minimize the role of oxygen were that no conclusive differences were observed when samples were heat treated in air or vacuum or prepared by evaporation, and
the appearance of a virtually identical resonance when single crystals of hexagonal selenium were damaged
(mechanically and by electron bombardment).
No resonances were found in pure "as grown" monoclinic or hexagonal selenium. This result was unexpected for the latter material, and chemisorption of oxygen on the surface is thought to be the explanation.
,
1
I. INTRODUCTION
Chern. Phys.
3519
H.-\LIDES
ELENlmr c"xists in several allotropic formshexagonal, monoclinic, and amorphous. Grey
hexagonal selenium consists of long helical chains,1
as shown schematically in Fig. 1. The chemical struc-
S
1
A. Von Hippel, ]. Chern. Phys. 16,372 (1948).
ture of each chain is believed to be ·Se-(Se) n-Se·. 2
Thus, one would expect to see electron spin resonance in
hexagonal selenium, due to the unpaired electron at
each end o(the chains. Accordingly, in .a perfect crystal,
one might find that the number of spins observed would
t T. Shirai, S. Hamada, and K. Kobayashi,
]apan84,968 (1963).
J.
Chern. Soc.