Home Search Collections Journals About Contact us My IOPscience

Determination of the Triple-Point Temperature of Gallium

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

1979 Metrologia 15 201

(http://iopscience.iop.org/0026-1394/15/4/005)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 200.10.161.5
The article was downloaded on 28/04/2011 at 19:46

Please note that terms and conditions apply.

 -~ __
Metrologia 15, 201-215 (1979)
~~

metrologia
@ b y Springer-Verlag 1979

Determination of the Triple-Point Temperature of Gallium

B.W. Mangum and D.D. Thornton

Institute for Basic Standards, National Bureau of Standards, Washington, D.C. 20234

Rzceived: March 16, 1978 and in revised form April 10, 1979

Abstract temperature fixed points in the very important range

We have determined the triple-point temperature of between 0.01OC and 122OC, a range of crucial
high-purity gallium to be 29.77406OC using three importance to the biomedical, oceanographic and
standard platinum resistance thermometers (SPRTs) industrial communities.
calibrated on the IPTS-68 and using samples of gallium We report here an investigation of the suitability of
from three commercial sources. All data obtained on the triple-point and melting-point of pure gallium as
the highest purity sample have a standard deviation of secondary temperature fixed points or, in the case of
+0.0001 1OC. The overall systematic uncertainty is the triple-point, as a defining temperature fixed point of
estimated to be !y:f m°C. The day-to-day irrepro- the International Practical Temperature Scale. Because
ducibility of a single melt of each gallium cell was of its widespread use by the semiconductor industry,
found to be < 2 1 5 p 0 C , while over a period of 7 gallium is now commercially available in very pure
months the irreproducibility among melts of each form and is relatively inexpensive, making it readily
gallium cell was < ~O.OO01loC.These limits on the available to anyone to use as a fixed point.
reproducibility appear to be largely instrumental since The abundance of gallium in the earth's crust is
every time an SPRT was left in a gallium cell for about 5 to 15 parts per million but, unlike the more
several days there was no change in the indicated common metals such as lead and tin, it is widely
temperature of the triple-point to within the instrumen- dispersed. It was first isolated [31 in 1875 and the first
+
tal resolution of 1.5 p°C. The pressure dependence determination of its melting temperature [41 was made
by Lecoq de Boisbaudran in 1876. Several other
was measured and determined to be dTldP =
+
-2.01 1 0.001 m°C/atm. melting-point temperature determinations were subse-
quently made [ 5 , 61 but the first really accurate
determination was made in 1934 by Roeser and
I. Introduction Hoffman [71. The samples used by Roeser and
Apart from the boiling-point of water at 100°C and the Hoffman [71 were purified by Hoffman 181, and
melting-point of phenoxybenzene at 26.86OC, both of according to tests made by him, the gallium was
which are extremely difficult to realize at .the milli- 99.999% pure. The temperature obtained by Roeser
degree level, there are no internationally-accepted fixed and Hoffman for the melting-point was 29.780
points on the International Practical Temperature Scale +0.005°C on the ITS-27. In terms of the IPTS-68 this
of 1968 (IPTS-68) [ l , 21 between the triple-point of becomes 29.771 f 0.005°C.
water at 0.01OC and the melting-point of benzoic acid Since the metal has become commercially available
at 122.37OC. Phenoxybenzene is an organic compound in large quantities at high purity, it has been used in
rather than a metallic element and, as such, is relatively many studies of melting behaviour [9-141. Also there
difficult to purify and has a low thermal conductivity. have been several recent reports of determinations of
Consequently, its usefulness as a temperature fixed melting-point or freezing-point temperatures [ 13, 1 5 -
point is severely limited. Practically speaking, then, 181, but none of those adequately documented all of the
there are no accurately-realizable and convenient variables affecting the melting temperature, i.e., purity,
pressure, methodology, etc.
Certain commercial equipment, instruments, or materials are On the basis of our own work and in consideration
identified in this paper in order to adequately specify the of earlier work, it was thought that the liquid-solid
experimental procedure. In no case does such identification imply
recommendation or endorsement by the National Bureau of
equilibrium of gallium should constitute an excellent
Standards, nor does it imply that the material or equipment fixed point. The study reported here was undertaken (a)
identified is necessarily the best available for the purpose to determine both the day-to-day and the sample-to-
0026- 1394/79/0015/Q2Q 1/%03.00
Table 1. Specifications of gallium samples

Source Lot number Nominal purity Comments

Alusuisse F 17/220 99.9999+% Recrystallized six times from semiconductor-grade
(microwave grade) material which had been analysed by mass spectro-
scopy as 6Ns
Alcoa 3809B 99.9999+% Recrystallization. Details unavailable
Eaglepicher J-57-76 99.99999% Recrystallized until no impurities are detected by
emission spectroscopy or until the level of detected
impurities is below 0.1 ppm. Manufacturer certified
that no impurities were detected in this lot

sample reproducibilities, and (b) to measure the outer case to relieve stresses induced by the volume
equilibrium temperature to the highest possible ac- increase incurred upon the freezing of the gallium
curacy. Samples from three commercial sources were sample. The re-entrant thermometer well was con-
investigated. The temperatures were measured with structed of nylon. This choice was made because we
long-stem standard platinum resistance thermometers had found in some preliminary studies that machined
(SPRTs) [191, three such thermometers being used in
the investigation.
In Section 11, we give a description of the samples,
the apparatus and the measurement procedures. In t
Section 111, we discuss the results and their analysis. A ml
summary is given in Section IV.
A
11. Experimental
B
A . Gallium Samples
C
Gallium is available from a number of commercial
sources at a nominal purity of 99.9999% (6Ns) or D
better. Such purities are difficult, if not impossible, to
verify. The limits of detection of most of the heavy E
metals expected to affect the melting-point temperature
of gallium are about 0.1 ppm for emission spectro-
scopy and about 0.01 to 0.05 ppm for mass spectro- F
scopy; neutron activation analysis is not feasible
because of the large neutron absorption cross-section of G
gallium. Thus, we can only be reasonably certain that if
any metals were present in our samples, their con-
centrations were less than 0.1 ppm. The sources, lot H
numbers, and nominal purity of the three samples are
listed in Table 1 along with comments regarding
methods of purification, etc. I

B. Gallium and Water Triple-Point Cells

i. Gallium Triple-Point Cells: A drawing of the gallium
triple-point cells which were used in this study is shown
in Figure 1. The all-plastic construction of these cells
was meant to circumvent potential problems associated F
with the 3.1% expansion of the metal upon freezing, as J
well as to avoid direct contact with free metal. The cell
was made to be used in a constant-temperature oil bath.
The outer case of the cells was made of nylon whereas Fig. 1. Cross-section of gallium triple-point cell. A -
the inner container, the container actually holding the Valve (Zytel), B - bath lid (plexiglass), C - support
gallium, was Teflon. Teflon was selected because it is (nylon), D - pumping tube (polyethylene), E - cap
relatively soft. The design of the inner container, (nylon), F - container (TFE), G - case (nylon), H
incorporating several ridges for mechanical strength - re-entrant well (nylon), I - gallium, J - base
and a loose fit, leaves an excess volume'inside the nylon (nylon)
202
nylon provided a surface that readily nucleated The containers were sufficiently translucent that it was
crystallization of the gallium, which is notorious for its possible to determine when the gallium was completely
ability to supercool. solidified by observing its top surface through the cell
After the components of the triple-point cells were wall.
machined, they were cleaned in the following manner.
The Teflon components were soaked overnight in aqua ii. Water Triple-Point Cells: Four water triple-point
regia, thoroughly rinsed in distilled water, and air dried. cells were used in this investigation. All were obtained
The nylon parts were washed in a gentle detergent from Jarrett, Inc. (USA). One of them was a type B
solution, rinsed thoroughly with distilled water, and air and the other three were type A. The types of cells,
dried. Then they were partially assembled, with epoxy their serial numbers, the dates they were used, and the
being liberally applied to the threaded sections. The cap SPRTs with which they were used are given in Table 2.
assembly was not joined to the base assembly until
after the cells had been filled with gallium.
The gallium samples were received in polyethylene C . Apparatus
squeeze bottles sealed under argon. All handling of the i. Constant Temperature Bath: The constant tem-
gallium, including the filling and final assembly of the perature bath used in this investigation was a
cells, was conducted in a glove box under a dry argon Rosemount Engineering Company Variable Tem-
atmosphere. The gallium was liquefied by raising the perature Oil Bath, Model 913A, with a capacity of 10
temperature inside the glove box above 3OOC. Each cell litres. The fluid was clear mineral oil for which the
was filled with approximately 900 grams of gallium viscosity was 17 centistokes' at 4OoC and 3.6
poured directly from the shipping bottles into the centistokes at 100OC. A lower viscosity fluid would
Teflon container. The cap assembly, which included the have been preferable from the point of view of
thermometer well, the pumping tube and valve, and the temperature stability, but this oil was known to have
support assembly, was joined to the base by a threaded very little, if any, effect on the gallium cell components.
section liberally doped with epoxy. The epoxy was Temperature control was not a problem. Cooling was
allowed to set and the gallium was solidified before the obtained with pressurized air which had been chilled by
valve of each cell was sealed and the cell removed from passage through a copper coil immersed in an ice bath
the glove box. before it was passed through the oil bath cooling coils.
In all cases, the gallium was solidified by lowering The heater was controlled by a Tronac Temperature
the glove box temperature below 29OC and inserting a Controller, Model No. PTC-40, which used a thermis-
liquid-nitrogen-chilled copper rod into the nylon well. tor as the sensing element. Through the use of this
system, we were able to maintain temperatures uniform
and constant to k 1 m°C, or better.
Table 2. Water triple-point cells used in measurements,
dates of measurements, and SPRTs involved ii. Resistance Bridges: Two resistance-ratio bridges
were used in making SPRT resistance measurements in
Date of Water this investigation. One employed ac and the other dc
measurement triple-point cell SPRT techniques. The ac bridge, designed by Cutkosky [201
and built by R.S. Kaeser at NBS, is a resistance-ratio
12 February 1977 B-11-536 374,375 bridge operating at 400 Hz and is stated to be in error
13 February 1977 B-11-536 089,374,375 by no more than 3 p51 when used with a 100 R
17, 18, 19 March B-11-536 089,374,375 standard resistor. It was used in making measurements
1977 of the temperature reproducibility of each of the
16 May 1977 A-13-587 375 gallium cells and for the comparison of the three
17 May 1977 A- 13-587 089,374,375 samples, but not for determinations of the triple-point
18 May 1977 A- 13-587 089,374 temperature. The alignment of the bridge was checked
19 May 1977 B-11-536 089,314,375 prior to the measurements reported here. Its resolution
20 May 1977 B-11-536 089,374,375 when used with a 100 C2 standard resistor and at
23 May 1977 A-1 3-587 089,374 measuring currents of 0.71, 1, and 1.4 mA was kO.15
23 May 1977 B-11-536 375 p51. Since the measured current ratios were found to
24 May 1977 B-11-536 089,374,375 vary a few percent from the nominal factor of d 2 , we
16 August 1977 A-13-506 089,374,375 used current values of 0.71, 1 and 1.41 mA in obtaining
17 August 1977 A-1 3-506 089,375 the extrapolated zero-current value of resistance bridge
17 August 1977 B-11-536 089,374 ratio. The uncertainty resulting from this procedure
18, 19, 23, 24, 25, B-11-536 089,374,375 was estimated to be k0.5 p,R.
26 August 1977 In making SPRT resistance measurements, it is
30,3 1 August 1977 A-13.723 089,374,375 desirable to duplicate calibration conditions as closely
1, 2 September as possible, Since the SPRTs were calibrated using a dc
1977 Mueller bridge, the instrument used in making SPRT
12, 13. 15, 16 A-13-723 375
September 1977 I 1 stokes = lo-' m2/s (unit of kinematic viscosity)

203
resistance measurements for the determination of the obtained in 1976. They were calibrated by the NBS
triple-point temperature was the Guildline Current Platinum Resistance Thermometer Calibration
Comparator, Model 9975. This is a dc instrument with Laboratory at measuring currents of 1 and 2 mA, with
automatic current reversal, the period of which was extrapolations to zero current. The calibrations for zero
adjustable to either 4, 8, or 16 seconds. Unless current were then used in all calculations of tem-
otherwise stated, all measurements were made with a 4 perature, thereby eliminating self-heating effects in all
second reversal period. The inaccuracy of the current of the measurements.
comparator was stated by the manufacturer to be 2 <
parts in lo7plus 1 digit in the last dial (8th). Since we iv. Standard Resistors: The standard resistor used
did not have a temperature-controlled 10 0 standard with each of the above bridges was a nominal 100 Cl
resistor to use with the bridge, we used a nominal 100 precision resistor, Model No. HA518 (which we have
0 standard resistor which was kept in a temperature- designated as H17) manufactured by Vishay. It was
controlled environment. This meant that the last (8th) enclosed by and thermally shielded from a large copper
dial represented 10 pa steps. Using a 3 second time enclosure whose temperature was regulated at 27.8
contant and averaging the strip-chart recorder tracing fO.l°C using a mercury thermostat. On the basis of
of the bridge output for 5 to 10 minutes permitted a
resolution, at measuring currents of 1 and \/2 mA, of
f1.5 p 0 (k 15 p°C).
-
measurements on similar enclosures, it is estimated that
the temperature variation of the resistor was f 3 mK
over a period of a few days. Based on the manu-
facturer’s specifications, the temperature coefficient of
resistance is estimated to be about 0.1 to 0.3
iii. Thermometers: Three SPRTs purchased from ppm/OC at the control temperature, which means that
Leeds and Northrup were used in this investigation. the resistance would vary by no more than kO.1 pa.
One of them had a platinum element that consisted of a The resistor was calibrated at both 400 Hz and at dc by
coiled helix mounted on a mica cross (L&N 8163 the Absolute Electrical Measurements Section of the
Series), while the other two had elements that consisted NBS on 9 February 1977, and the results of that
of a single layer helix of bifilarly wound fine platinum calibration are given in Table 4.
wire on a mica cross (N&N 8167-25 Series). In Table
3, we have listed the serial numbers of the three thermo- U . Chart Recorder: A Leeds and Northrup Speedomax
meters used, the NBS identification number (I,D,#), XL, Model 602, strip-chart recorder with an adjustable
the type of thermometer, and the dates they were zero balance was used with each bridge in making
calibrated. resistance-ratio measurements. The true zero was
The SPRTs were supplied by the manufacturer with determined for the complete measurement system by
standard external copper leads designed for dc setting the measuring current flowing through the
measurements. The four leads from the helical element SPRT to zero and, in the case of the Guideline Current
were platinum wires brought straight up the stem in a Comparator, also setting.the sensitivity to the minimum
square array and through a hermetic seal at the head. position. The same zero was obtained for other
Although the current leads and the potential leads were sensitivity settings of the current comparator but the
normally formed from the diagonal pairs, one SPRT noise was much smaller at the minimum sensitivity
was connected such that the current leads were setting, making the zero easier to determine.
adjacent, as were the potential leads. To correct this
situation and in order to reduce coupling to stray
electromagnetic fields and to reduce capacitive coupling vi. Vacuum System: The vacuum system contained
between the leads, the SPRTs were modified for ac a mechanical rotary pump, a mercury manometer, an
measurements by removing the external cables supplied oil manometer, two alcohol-solid-carbon dioxide
by the manufacturer and installing BNC connectors mixture cold traps, manometer bypasses, and
some valves, as shown schematically in Figure
mounted in bakelite caps such that one coaxial cable
2. With this pumping system, we were able to obtain
would serve as the current leads and the other would
pressures of the order of a few tenths of a pascal.
serve as the potential leads. Using this arrangement of
leads, the SPRTs were insensitive to lead positions.
The SPRT of the 8 163 Series was obtained in 1973,
Although the vapour pressure of gallium is very low,
even at elevated temperatures (< 10 mPa at T
1000°C), this was adequate since the pressure depen-
-
and the two SPRTs of the 8167-25 Series were
dence of the melting-point is relatively small. The
manometer system was used only during the measure-
Table 3. SPRTs used in the study

SPRT NBS SPRT Table 4. Calibration data of standard resistor H 17

S/N I.D.# type Date calibrated dc 400 H z
Resistor Date resistance resistance
1808364 089 8163 January 1977 Manufacturer type calibrated value value
1846617 374 8 167-25 January 1977
Vishay HA518 9 February 99.99346 99.993592
1846729 375 8167-25 January 1977 1917

204
 VACUUM
GAUGE

9
U
GALLIUM
CELL
Hg
MANOMETER
OIL
MANOMETER U
ARGON
PUMP
Fig. 2. Schematic of vacuum and
CYLINDER gas-handling system

ments of the pressure dependence of the melting-point this stage of the preparation, it appeared that the best
of gallium and during the filling of the cells with an practice was to avoid overcharging the heat pipe with
atmosphere of argon at the conclusion of the experi- dry ice. This appeared to regulate the heat removal
ments on a given cell. At all other times the mano- rate, permitting a slow, steady growth of the mantle.
meters were closed off from the system and bypassed. Cracking of the mantle could be avoided by terminat-
A cold trap was located between the gallium cell and ing the freeze after the mantle reached about 1 cm in
the remainder of the system so that even when the thickness as viewed through the water. The heat pipe
manometers were open to the pumping system the and the alcohol were then removed from the well and
gallium cells were protected from contamination. All of the triple-point cell packed in ice. In the early stages of
the argon gas that entered the gallium cells passed this work, the ice mantle was permitted to grow
through the cold trap. overnight under these conditions. The thermometer well
was then filled with water, the initial temperature of
which was approximately 2OOC. This would melt a
D. Measurements and Procedures little of the ice mantle in contact with the thermometer
well, producing a liquid-solid interface near the well.
In this section we shall describe the preparation and
This is essential to obtain an equilibrium temperature.
maintenance of the gallium and water triple-points, the
A small piece of foam rubber was placed in the bottom
procedure for determining the gallium triple-point tem-
of the thermometer well to cushion the SPRT from
perature and its stability, th’e technique used for
shock when it was being inserted into the well. An
establishing thermal contact between the SPRTs and
aluminium bushing, 5 cm long, fitted snugly in the
the triple-points, the power dependences of the SPRTs,
thermometer well and was also a relatively snug fit to
the pressure dependence of the gallium melting-point,
the SPRT. Usually the bushing was prechilled in ice
and the effects of immersion of the SPRTs in the
water before inserting it into the thermometer well.
gallium triple-point cell.
Thermal contact between the inner liquid-solid inter-
face and the SPRT was thus provided by the thin layer
i. Preparation and Maintenance of Water Triple-Point of water and the bushing, the latter playing the major
Cells: Four water triple-point cells were used in this role. Without a bushing present, self-heating effects of
investigation. They were prepared for triple-point the SPRTs were much more pronounced.
measurements in the following way. First, the cell was Once the water triple-point cells had been prepared
cooled in an ice bath for 3 to 4 hours. Then, by means as described above, they were kept packed in ice in
of a liquid-nitrogen-cooled copper tube being alternately either a dewar or a home-made Styrofoam container.
and repeatedly dipped into liquid nitrogen and then into With the use of either of these, a water triple-point cell
0.5 cm3 of ethyl alcohol in the bottom of the thermo- could be kept in good working condition for several
meter well, a ,spherically-shaped mantle of ice was weeks, even though it was being used more or less con-
formed around the bottom end of the well. After tinuously every day. During the measurement period,
approximately 30 minutes, the mantle appeared to be the water triple-point cell was checked at least every
about 1 cm thick. Then a heat pipe, cooled by a solid morning, and normally several times a day, to ensure
carbon dioxide and ethyl alcohol mixture, was placed in that the ice mantle was free to rotate, that is, that there
the thermometer well and the well slowly filled to was no freezing either to the outer wall or to the thermo-
within approximately 1 cm of the top of the water meter well. If there were ice needles or an ice crust ex-
column in the annular space of the cell with cold ethyl tending from the mantle to the outer wall of the cell,
alcohol. After approximately 60 minutes, the ice mantle these were melted either by the heat from one’s hands
was 1 cm or more thick. As experience was gained in or by placing the cell in a container of water at ambient

205
temperature. In any event, the mantle itself was tested end of May, we obtained readings at three currents,
in a container of water such that one could clearly 0.3d2, 1, and d 2 mA, using a 1 second time constant
see the outer boundary of the ice and see that it was and a 5 minute integration time. From these, we esti-
not touching the outer wall of the cell. mated an uncertainty of k5 ps2 in the extrapolated zero-
current value of resistance. In August and September,
ii. Preparation and Maintenance of Gallium Triple- we used only 1 and ,/2 mA (a check had shown that
Point Cells: The filling of the gallium cells and the the current values reproduced precisely) but, as stated
subsequent solidification of the gallium were described earlier, a time constant of 3 seconds and an integration
in Section 1I.B.i. In preparation for each series of time of 10 minutes resulted in a reduction of the
measurements, e.g., May or August-September, each estimated uncertainty to k1.5 ,us2 in the extrapolated
cell containing solid gallium was placed in an auxiliary zero-current value.
oil bath at -60°C. Hot oil was more or less continu-
ously pumped into the thermometer well, maintaining iv. Stability of Gallium Triple-Point Temperature: The
an average temperature of approximately 40° C, to stability of the gallium triple-point temperature of a
ensure that there was always an inner and outer given melt of a given cell was determined by con-
liquid-solid interface the full length of the column of tinuously monitoring the resistance of an SPRT in the
gallium. After 15 to 20 minutes, during which time gallium cell for several days at a time using the
25 to 50% of the sample was melted, the cell was Cutkosky ac bridge. For this procedure, the cell was
placed in the Rosemount bath, the temperature of prepared and maintained as described in Section 1I.D.ii.
which was normally kept approximately 10 m°C above At least three melts of each cell were investigated by
the gallium triple-point temperature. During this this procedure. Only changes of resistance were of
operation, the cell contained argon gas at a pressure interest for this study and the bridge resolution of
of -1 atmosphere. The cell was then connected to the k0.15 ,us2 (k1.5 p"C) at 1 mA measuring current
vacuum system and evacuated. For the pressure enabled us to make an accurate study of this.
dependence experiments, the cell was pressurized with
argon gas, but during the triple-point measurements, v. Repeatability of Gallium Triple-Point Temperature
the cell was pumped continuously. Measurements and Intercomparison of Gallium Cells:
Upon completion of each series of measurements, After preparation of the gallium cells as described in
the samples were solidified. Whenever the sample had Section II.D.ii, the repeatability of the triple-point
been completely melted, it was cooled to -29OC and temperature for a given mantle in each cell was
then a freeze was initiated by repeated insertions into determined using the Cutkosky ac bridge. At the same
the thermometer well of a liquid-nitrogen-cooled copper time, a comparison of the three cells was made using
tube. The cell was then put into a dewar containing a two gallium cells at a time and SPRTs 374 and 375.
small quantity of ice at the bottom and then covered. Before the SPRTs were put into the gallium triple-point
Following those experiments when the sample was not cells, they were preheated in the oil bath for a minimum
completely melted, the cell was placed in a dewar as of 30 minutes. After being in the cells for at least 30
just described and the sample permitted to freeze. minutes, their power dependence was measured. The
Thermal contact to the SPRTs was provided by the SPRTs were then interchanged in the two gallium cells
oil in the thermometer wells. Since the SPRTs fitted and after 30 minutes had elapsed, their resistances were
snugly into the wells, no bushings were required. measured again. Several interchanges were made per
day.
iii. Temperature Measurements: The gallium triple-
point temperature was determined by making
resistance measurements of the three SPRTs in the vi. Gallium Melting Experiments: In order to obtain
gallium triple-point cells and in a water triple-point cell. melting curves, each cell containing solid gallium at
The SPRTs were normally cycled through the gallium room temperature was placed in the Rosemount bath
and water triple-point cells in the following measure- maintained at -4OOC. An SPRT was placed in the
ment sequence each day: the SPRT was preheated in well while the cell was at room temperature and then
the oil bath containing the gallium cell for at least 30 monitored continuously until the sample was totally
minutes, then placed in the gallium cell. After being melted and in equilibrium with the bath. From a con-
there for at least 30 minutes, measurement of its sideration of the cell as shown in Figure 1, it is clear
resistance was begun with the Guildline Current that the principal point of thermal contact between the
Comparator. The SPRT was then removed from the bath and the gallium would be at the bottom of the cell.
cell, stored temporarily at room temperature, and then In order to reduce heat flow at that point and prevent
precooled in an ice bath for at least 30 minutes. The the premature melting of the gallium at the bottom, we
resistance of the SPRT in the water triple-point cell was fitted a Styrofoam jacket to the bottom half of the cell.
then measured in the same way as outlined above for
the gallium cell. vii. Pressure Dependence of Gallium Melting-Point
The power dependence of each SPRT was deter- Temperature: The measurements of the pressure
mined each time its resistance was measured either in dependence began with the SPRT in the gallium cell
the gallium cell or in the water triple-point cell. Until the under a vacuum. The sample used was that obtained

206
from Alusuisse. The variation in the gallium melting
temperature as a function of argon gas pressure over
the melt was then determined from measurements made
with the Cutkosky bridge. Although phase diagrams
indicate that mercury and gallium are completely
insoluble in both liquid and solid phases [211, we took
care to have no mercury vapour from the manometers
backstreaming into the gallium cell during these
measurements.

viii. Thermometer Immersion and Hydrostatic Head

Effects: The immersion characteristics of the cells were
studied using SPRT 375, the resistance of which was
measured with the Cutkosky bridge at 1 mA measuring
current. The cell was prepared as described in Section
1I.D.ii. At the start of the experiment, the tip of the
SPRT was placed at the bottom of the thermometer
well. At that position the centre of the platinum sensing
element was approximately 13 cm below the top of the
gallium. In order to determine the effect of the bath
temperature on the SPRT resistance, it was cycled by
-0.1 " C about the triple-point temperature with a
period of 4 hours. Measurements were made as the 50 52 54 56 58 60 62
thermometer was withdrawn in increments of approxi- (28583o--l
mately 1 centimetre. With the thermometer tip about 7 BRIDGE DIAL SElTlNG
cm above the bottom of the well, the effect of cycling
the bath temperature was pronounced. Consequently,
the cycling was discontinued and the bath temperature Fig. 4. Measurements of the hydrostatic head effect
was set 10m°C above the triple-point temperature. using SPRT 375 and the Alusuisse gallium sample, 0
and H represent data obtained on two different runs.
The straight line represents the hydrostatic head effect
calculated from the density and the results of the
experiment on the pressure vs. melting temperature of
gallium

2950-

2925 -
29101 I I I I I I I I I 1 I I
10 a 6 4 2 0
DISTANCE FROM BOlTOM OF WELL (cml P llml

Fig. 3. Immersion characteristics of SPRT 375 in the Fig. 5 . Pressure dependence of the gallium melting-
triple-point cell containing Alusuisse gallium. The data point temperature. The straight line through the data
have been corrected for hydrostatic head effects has a slope of -19.85 nK/Pa (-2.01 1 m°C/atm.)

207
 Results of these measurements are shown in Figures SPRT in the Alusuisse gallium cell is shown in Figure
3 and 4. These results show that under the conditions of 5 . The uncertainty in the pressure measurement was
our experiments the effects of the temperature outside about f 100 Pa (- & 1 mmHg) and, hence, does not
the cell are no longer discernible when the tip of the show on this scale. The f 1 5 p0C uncertainty in the
thermometer is within about 6 cm of the bottom of the temperature measurement is indicated by the size of
well. As shown in Figure 4, the change of the melting the circle representing the data. The straight line
temperature with the distance of the SPRT from the fitted through the data points has a slope of dTldP =
bottom of the well (hydrostatic head effect) is in agree- -2.011 k 0.001 m°C/atm (or -19.85 nK1Pa). Using
ment with the results of the melting temperature vs. thermodynamic data [221, the Clausius-Clapeyron
pressure experiments presented in Section 1II.A. Both equation (dTldP = T AVIL, where T = temperature, P
the results shown in Figure 3 and the few points = pressure, AV = volume change during phase change,
obtained when the effects of cycling the bath tem- and L = latent heat of fusion) yields the value dT/dP =
perature were observable indicate that the relative -2.032 m°C/atm (or -20.06 nK/Pa). In Table 5 , we
temperature difference decreases by about a decade for list these values as well as those determined by other
every 2 centimetres of immersion. These results are investigations [17,23, 241.
further supported by the observation that varying the
room temperature by as much as 10°C had no effect
on the SPRT reading when its tip was at the bottom of B. Water Triple-Point Results
the well. As stated in Section II.B.ii, we used four water triple-
point cells during the course of this work. The ice
111. Results and Analysis mantles of the triple-point cells were annealed for
several days [19, 251 before they were used for actual
A . Pressure Dependence measurements. With the exception of A- 13-506, the
The results of the pressure dependence of the melting- SPRT resistances in the cells agreed to within the
point temperature of gallium as determined by an scatter observed for a given cell (i.e., f0.07 m°C). The

a
a
a-
a 1
a

1.11815500 I

Fig. 6. Resistance ratio data for gallium triple-point temperatures taken with SPRT 375. Values of R,, and R,, are
Guildline bridge ratios (times lo8) of the resistance of SPRT 375 to that of standard resistor H17

208
 1.11816500
. i
28564900 I I
t i

.
z
1

28564800

1
25547200

-: .
25547100
t' ...
+- --++
I I I I I I I I

+- SEPT. --
1 15 1 15 1 15 1 15 1 15 1 15 1 15 1 15 1
-+FEB.-! MAR. APR. MAY JUN. +JUL. +AUG.
ign
Fig. 7. Resistance ratio data for gallium triple-point temperatures taken with SPRT 374. Values of R,, and RH20are
Guildline bridge ratios (times 10') of the resistances of SPRT 374 to that of standard resistor H17

triple-point temperature of A- 13-506, however, was The resistances of each of the three SPRTs
consistently about 0.2 m ° C below that of the other obtained from measurements in one or the other of the
cells. It is unlikely that the isotopic composition of the three water triple-point cells are shown in Figures 6-8.
water of the various triple-point cells would have been The water triple-point cell used for measurements on
sufficiently different to account for the 0.2 m°C lower any of the various days shown in the figures is given in
temperature observed for A- 13-506 since the manufac- Table 2.
turer obtained the water for the cells from the same The data obtained in March 1977 were not used in
source. Since cell A-,13-506 was an old cell, it calculating the gallium triple-point temperature since
apparently had become contaminated, and consequen- there were problems associated with the ice mantle of
tly, it was not used in the determination of the gallium the water triple-point cell and with the values realized
triple-point or melting-point temperatures. for the water triple-point (but apparently not for the
gallium triple-point). Some of the effects of the
problems can be seen in the bottom portions of Figures
Table 5 . Experimentally determined values of the
pressure dependence of the melting-point temperature 6, 7, and 8. Some of the problems arose because the ice
mantle was permitted to freeze to the outside wall.
of pure gallium
Although the mantle was melted so that it was free to
dT/dP value Reference rotate, the ice would grow back and stick to the wall
after several hours. This was detected during attempted
-2.01 1 mK/atm. This work measurements by a pronounced drift in SPRT
-2.12 mK/atm. Bridgman [231 resistance. By melting the mantle substantially away
-2.0 mK/atm. Jayaraman et al. [241 from the wall, the refreezing problem and the drift in
-1.5 mK/atm. Sostman [ 171 SPRT resistance were solved although for some reason,
-2.032 mK/atm. Calculated perhaps stresses still present in the ice mantle, the
resistance values of the SPRTs were still low.

209
 28545450 I I

28545350 -1
25530250 I I

I I I I I I I I I

Fig. 8. Resistance ratio data for gallium triple-point temperatures taken with SPRT 089. Values of R,, and RH20are
Guildline bridge ratios (times lo*) of the resistances of SPRT 089 to that of standard resistor H17

C. Results of Gallium Melting Experiments increase on the uppermost curve, the other curves show
Melting curves shown in Figures 9 through 1 1 were a very abrupt increase in temperature. We attribute this
obtained, as described in Section II.D.vi, for each cell behaviour to penetration of the mantle by liquid in the
after the gallium had been frozen in the usual way. The vicinity of the thermometer sensor. Subsequent melting
melting curve shown in Figure 12 was obtained for the of the remaining solid accounts for the gentle increase
Eagle-Picher sample after it had been rapidly frozen in temperature over the next 2 to 3 hours, as shown by
from a totally liquid state. The data shown in Figure 13 the uppermost curve. Upon complete melting, the
were obtained for the Eagle-Picher sample when it had sample temperature approaches the bath temperature
been prepared according to Section 1I.D.ii for a triple- exponentially.
point temperature determination before being placed in The details of the melting curves at the highest
a bath at 4OOC. Although data points are shown on the resolution are strongly characteristic of each sample,
curves, the thermometer was monitored continuously reflecting the influence of at least two factors. Firstly,
using a strip chart recorder. The points indicated in the the distribution of impurities in a sample, as determined
figures were used for transferring the data from the by its prior thermal treatment, plays a major role in
strip charts to the graphs. Each set of melting data is determining the shape of the melting curve. Secondly,
presented at three levels of resolution to display the the establishment of the hydrostatic head, which varies
main, features of the melting behaviour. There are three from cell to cell, will affect the initial portion of the
salient features in the lowest resolution curve for each melting curve. Although it would be dimcult to account
sample. The temperature rapidly attains a constant for the details of the highest resolution curve for each
value at which it remains for about 7 hours. It then sample, the curves do show that the samples, when
gradually increases by one to two degrees over the next prepared in the same way, have different melting
2 to 3 hours before increasing rapidly and then ranges.
approaching the bath temperature. The effect of different freezing methods on the
Approximately 7 hours after melting began and shape of the melting curve can be seen by comparing
when the temperature begins to show a gradual Figures 1 1 and 12. The results of Figure 13 are
2 10
 i I I I I
' 2-1-
c

J
i

35 OV

I°Fc I! i
'cmofF
250 &-a I l m 'T
ll0p C
105T

i
-I
-1
428682960

--28682ooo

I
200-1

I
0 1 w m 3 0 0 4 0 0 5 L m e Q l 7 r m e Q l
Time Iminuted

Fig. 9. Melting curve at three levels of resolution Fig. 11. Melting curve at three levels of resolution
obtained with SPRT 375 for the Alusuisse gallium obtained with SPRT 375 for the Eagle-Picher sample
sample in a 40°C bath following a standard freeze. The in a 40°C bath following a standard freeze. The scales
temperature scale on the left-hand side of the figure is are the same as in Figure 9
associated with the upper curve; the scale running from
28581000 to 28585000 on the right-hand side of the
figure is associated with the middle curve; the other
scale on the right-hand side is associated with the
lowest curve

r 36 OC

I
r--

10 P 'C

- 28582950

+28581wo I 1
0 1 " ~ u y I u x ) em 7w 800
Time lminutesl

Fig. 12. Melting curve at three levels of resolution

Fig. 10. Melting curve at three levels of resolution obtained with SPRT 375 for the Eagle-Picher sample
obtained with SPRT 375 for the Alcoa gallium sample in a 40°C bath following a fast freeze. The scales are
in a 40°C bath following a standard freeze. The scales the same as in Figure 9. At 160 minutes the SPRT was
are the same as in Figure 9 removed briefly from the cell and reinserted

211
 D . Gallium Triple-Point Results

c
The gallium triple-point temperatures, tabulated in

'6.0 I! i I Table 6, were calculated from the zero-power resistance

ratios shown in Figures 6-8, using the standard
interpolating formula specified by the IPTS-68, the
SPRT calibration constants provided by the NBS
Platinum Resistance Thermometer Calibration
Laboratory, and the appropriate corrections described
below. In Figures 6-8, R,20 (0) represents the ice-point
resistance of an SPRT calculated from the value,
R , (t.p.), obtained at the triple-point. The resistances
of the SPRTs in the gallium and water triple-point cells
were corrected for the depression of the true triple-
.
.I point temperatures due to the hydrostatic pressures
exerted by the columns of liquid. The size of the
20.0 depression in the water triple-point cell is 7 p°C per cm
of water above the point of measurement [191. For
B-11-536, the approximate height of the water column
was 25 cm and for A-13-587 and A-13-723, the heights
15.0 I I I I I I I of the water columns were approximately 28 cm. The
variations in the amount of ice comprising the mantle
could cause an uncertainty of 50.5 cm in these
numbers. The depression of the water triple-point
Fig. 13. Melting curve for the Eagle-Picher sample temperature was calculated for the appropriate water
which had been prepared for the determination of the triple-point cell used in the measurements, i.e., 175 p°C
triple-point temperature and then placed in a 4OoC for cell B-11-536 or 196 p°C for cells A-13-587 and
bath. The scales are the same as in Figure 9 A-13-723. The zero-power resistance of the SPRT
determined in the triple-point of water cell and experi-
encing a temperature of 0.009825°C for cell B-11-536
or 0.009804°C for cells A-13-587 and A-13-723 was
significant in that when the cell is prepared according then corrected to O°C by dividing that resistance by
to Section II.D.ii, the temperature, as indicated by the the appropriate resistance ratio (either 1.000039 16 or
highest resolution curve, is independent of time and is 1.00003908, respectively).
independent of the fraction of gallium melted. The depression of the temperature due to the

Table 6 . Gallium triple-point temperature as determined for each sample by each of the SPRTs used in the
investigation

Alusuisse Alcoa Eagle-Picher

Aug-Sept Aug-Sept Aug-Sept

SPRT I.D.# May 1977 1977 May 1977 1977 February 1977 May 1977 1977

3 75 29.77403 29.77402 29.77420 29.77393 29.77388 29.77400 29.77392

29.77406 29.77399 29.77410 29.77396 29.77392 29.77392 29.77395
29.77400 29.77393 29.77397
29.77405 29.77395 29.77389
29.77402 29.77399
29.77392
29.77398
29.77395
3 74 29.77409 29.77395 29.77393 29.77386 29.77363 29.77389 29.77388
29.77390 29.77390 29.71402 29.77395 29.77409 29.7741 1 29.77384
29.17393 29.77396 29.77386
29.77387
089 29.77420 29.77427 29.77432 29.77413 29.7742 1 29.77416 29.77419
29.77409 29.77427 29.77421 29.77415 29.77414 29.77414
29.774 19 29.774 15
29.7742 1 29.774 15

212
column of liquid gallium was calculated from the The average temperature and its standard deviation
results of the pressure dependence of the melting-point derived for each SPRT in each gallium triple-point cell
temperature, which was shown to be -2.001 m°C/atm, are given in Table 8. The systematic differences among
and is 12 p°C per cm of gallium (at the triple-point the thermometers, as shown there, may arise from three
temperature). Since the height of the liquid gallium sources. For measurements of a given SPRT, the
columns was approximately 13 cm, this results in a uncertainty arising from calibration errors (calibrated
depression of 156 p0C. This was added to the tem- against fixed points of the IPTS-68 as maintained at
perature actually measured in the gallium triple-point NBS) may be as large as i-0.21 m°C [191. The presence
cell as calculated above for the appropriate resistance of moisture on the thermometer insulation is another
ratio. source of error and this, of course, varies from thermo-
In addition to the hydrostatic head corrections, it meter to thermometer. It has been shown [261 that
was also necessary to make corrections for the effects shunting of an SPRT resistor with moisture can cause
of the measuring current reversal frequency. Most of an error at the triple-point temperature of water
the SPRT resistance values were obtained for a 4 equivalent to !::!m°C without the thermometer ex-
second reversal period for the measuring current but hibiting a detectable "wet kick". Apart from any cali-
changes in resistance of each SPRT in both gallium and bration errors, different SPRTs may indicate discrepant
water triple-point cells were measured when the current temperatures at fixed points intermediate to the calibra-
reversal period was changed from 4 to 16 seconds. tion points due to an inherent source of error resulting
Small effects were observed only at the triple-point of from the use of real materials. Although we have no
water and for all SPRTs except 375. This effect was documentation, we would expect the error from this
probably due to the presence of moisture on the source to be small compared to 0.1 m°C near 30°C.
thermometer insulation since it was not observed at the Overall then, different SPRTs may indicate tem-
gallium point. Wet insulation is known [261 to give rise peratures of a given intermediate fixed point near 30°C
to polarization effects. Assuming that the 16 second which differ by approximately 0.9 m°C. Consequently,
period of current reversal of the Guildline bridge more the differences in the temperatures calculated for the
nearly reproduces the situation regarding measure- different SPRTs for a given gallium triple-point cell are
ments with the Mueller bridge in the Calibration not inconsistent with these error estimates.
Laboratory [271, we have corrected the 4 second period The results in Table 8 indicate that there is a
current reversal data to those at a 16 second period. systematic difference in the temperatures of the
These corrections also have been included in the values different gallium cells. The Alusuisse sample has the
given in Table 6. highest triple-point temperature, the Alcoa sample is
In order to get some indication of the long-term lower by 0.01 m°C and the Eagle-Picher sample is
reproducibility of the three gallium triple-point cells, we lowest, being 0.06 m°C below the Alusuisse sample.
have averaged the temperatures of each set of measure- Since these temperature differences are of the same
ments made in February, May and August-September. order of magnitude as the experimental uncertainty, it
These are given in Table 7. It is difficult to draw any was decided to compare the three gallium cells more
definite conclusions regarding a possible change with precisely. The results of this comparison (using SPRTs
time of the gallium triple-point temperature for each 374 and 375 and the ac bridge), shown in Figure 14,
cell, since only two temperatures per cell were clearly confirm the dc results and demonstrate that the
measured with each SPRT in May. We observe that the Alusuisse sample triple-point temperature is indeed
differences between the May and August-September 0.02 m"C higher than that of the Alcoa sample and
averages are of the order of the differences in the May approximately 0.07 m°C higher than that of the Eagle-
values and are consistent with the i-5 pUn instrumental Picher sample. It is equally clear that the day-to-day
uncertainty estimated in 1I.D.iii. Moreover, the reproducibility of the triple-point temperature of each
differences are not particularly systematic from SPRT cell is at least as good as the instrumental uncertainty.
to SPRT. Thus, about all that can be concluded from In accounting for the differences among the cells, it
the results in Table 7 is that over the long term the should be recalled that the hydrostatic pressure
triple-point temperature of a given gallium cell repro- correction for a centimetre of gallium is 12 p°C. Since
duces to within +O. 1 1 m"C and that this limit largely the height of the column of liquid gallium above the
reflects the measurement errors. sensor varied less than a centimetre, both as a result of

SPRT
1
Table 7. Average gallium triple-point temperature determined as a function of time

May ~lusuisse
Aug-Sept ~
May Alcoa

Aug-Sept February
Eagle-Picher

May Aug-Sept
1
3 75 29.77404 29.77402 29.77415 29.77394 29.77390 29.77396 29.77395
3 74 29.77400 29.77393 29.77398 29.77392 29.77386 29.77400 29.77386
089 29.774 14 29.77427 29.77426 29.77417 29.77421 29.774 15 29.77416
 Table 8. Average temperature and standard deviation for each SPRT in each gallium triple-point cell
SPRT Alusuisse Alcoa Eagle-Picher

375 29.77402 f 0.000025 29.77401 f 0.0001 12 29.77394 f 0.000039

3 14 29.77395 f 0.000079 29.77394 f 0.000058 29.77390 f O.OO015 1
089 29.77421 f 0.000085 29.77420 f 0.000066 29.77416 f 0.000027
Average 29.77406 & 0.0001 1 29.77405 k 0.00013 29.77400 f 0.00014

the volume change on freezing and from cell to cell, the sample could have been as high as 0.1 to 0.5 ppm,
largest possible effect on the equilibrium temperature i s depressing the melting temperature by as much as 0.0 1
of order 0.01 m°C, and hence, is at or below our level of to 0.25 m°C. The effect of impurities in the Alusuisse
detection. Possible differences in isotopic composition sample would be to depress the melting temperature by
of the starting material and isotopic fractionation les than 0.01 m°C. It follows, then, that the triple-point
during purification and freezing could conceivably be temperature of the Alusuisse sample most closely repre-
factors in the sample-to-sample differences, since the sents the triple-point temperature of pure gallium. That
Alusuisse sample is refined from European bauxite, value is 29.77406 with a standard deviation of all data
while the other samples are from North or South points of f0.0001I0C. The standard deviation of the
American bauxite. The differences observed among the data for individual SPRTs, however, range from
triple-point temperatures of the three gallium samples, f0.000025 to f0.000085°C.
however, most likely arise primarily from differences in Potential sources of systematic errors which contri-
purity. This is supported by the differences in melting bute to the overall uncertainty of the measured triple-
range as shown in Figures 9-11. (See Sections 1I.D.vi point temperature of gallium on the IPTS-68 arise from
and 1II.C). A 0.1 ppm difference in concentration of calibration, “wet” insulation, bridge inaccuracy,
alloying impurity, such as lead, silver, copper, indium, variations in isotopic composition, and impurities, as
etc., would cause a change of about 0.01 to 0.05 m°C in indicated above in the discussion of differences among
the melting temperature [211. Estimates of the purity of the thermometers and among the gallium cells. The un-
the Alusuisse sample put concentrations of each such certainties in realizing the fixed points of the IPTS-68
impurity at less than 0.0 1 ppm, perhaps as low as 0.001 used in the calibration of the thermometers are
ppm, whereas the Eaglepicher sample seems more f0.1 m°C for the triple-point of water [ l , 191 and
likely to have concentractions of 0.01 to 0.05 ppm of k 1 m°C for the freezing points of both tin and zinc [ 191.
each. These estimates are based on information These result in an uncertainty from calibration of
provided by the manufacturers. Thus, the total con- 20.4 m°C. The three thermometers used in the deter-
centration of alloying impurities in the Eagle-Picher mination of the gallium triple-point temperature
showed no “wet kick”, but, even so, we attribute an
uncertainty of !: m°C to the realization of the
216641116
L I I I” I triple-point temperature of water to allow for the possi-
bility of moisture on the insulation [261. The uncertainty
at the triple-point temperature of gallium due to such
“wet” insulation would be negligible. The estimated
inaccuracy of the bridge, as discussed in Section II.C.ii,
5 corresponds to an uncertainty in the temperature
21614111

1 measurements of f0.16 m°C. As stated above, the

uncertainty resulting from impurities in the Alusuisse
1 sample is estimated to be less than fO.O1 m°C.
Variations in isotopic composition of the gallium
samples would affect the gallium triple-point tem-
perature but we have no information regarding the
magnitude of such an effect. Therefore, it is not possible
for us to make a realistic estimate of the uncertainty
arising from this source. The triple-point temperature of
SEPlEMBEllO7l water is known to vary with the isotopic composition
[ll and this, in fact, is the source of the uncertainty in
Fig. 14. Direct comparison of gallium triple-point cells that temperature discussed above under systematic
using SPRTs 374 and 375. Bridge dial settings errors due to calibrations. One could normally include
represent Cutkosky bridge ratios (times lo8) of the this uncertainty again in temperature measurements
resistances of the SPRTs to that of standard resistor involving calibrated SPRTs, but since the triple-point of
H17. 0 represent data for Alusuisse sample; I water cells used in our measurements came from the
represent data for Alcoa sample; 0 represent data for same source as those used in the calibration of our
Eagle-Picher sample thermometers, it is inappropriate to include this

2 14
uncertainty a second time. Thus, the total uncertainty Id. Comptes rendus des seances de la Quinzieme
from all of the sources from which we can form Conference Generale des Poids et Mesures,
estimates is approximately !y:;m"C. Resolution 7, p. 105, Annexe 2, p. A 1 (1975)
The total uncertainty is the sum of both systematic le. Also Metrologia 12, 7 (1976)
and random uncertainties and for our measurements is 2. The corrected value of 2636°C for the melting-
+0.7
-,.lm"C. point of phenoxybenzene is to be published in
Metrologia
3. Lecoq de Boisbaudran, P.E.: Comp. Rend. 81,493
(1875)
IV. Summary 4. Lecoq de Boisbaudran, P.E.: Comp. Rend. 82,
We have determined the triple-point temperature of 1036 (1876); Comp. Rend. 83,611 (1876)
high-purity gallium to be 29.77406OC using three 5. Richards, T.W., Boyer, S.: J. Am. Chem.'Soc. 43,
standard platinum resistance thermometers calibrated 274 (1921)
on the IPTS-68 and using samples from three commer- 6. Craig, W.M., Drake, G.W.: J. Am. Chem. Soc. 56,
cial sources. An assessment of the relative purity of our 584 (1934)
samples was confirmed by the results of an inter- 7. Roeser, Wm.F., Hoffman, J.I.: J. Res. NBS 13,
comparison of the triple-point temperatures of the 673 (1934)
samples and by their melting curves. All data obtained 8. Hoffman, J.I.: J. Res. NBS 13,665 (1934)
with the three SPRTs on the highest-purity sample have 9. Jach, J., Sebba, F.: Trans. Faraday Soc. 50, 226
a standard deviation of 20.0001 1"C, although the (1 954)
standard deviations of the data for individual SPRTs 10. Boedtker, O.A., La Force, R.C., Kendall, W.B.,
range from +0.000025 to 20.000085°C. The overall Ravitz, S.F.: Trans. Faraday Soc. 61,665 (1965)
systematic uncertainty is estimated to be !:: m"C. 11. Pennington, P.R., Ravitz, S.F., Abbaschian, G.J.:
We have established that over a 7 month period the Acta Met. 18, 943 (1 970)
irreproducibility was 5 +0.00011 "C while the day-to- 12. Pashaev, B.P.: Fiz. Tverd. Tela 3,416 (1961)
day irreproducibility of a single melt of each gallium 13. Wenzl, H., Mair, G.: Z. Phys. B21,95 (1975)
cell was found to be < 215 p°C. These limits on the 14. Abbaschian, G.J., Ravitz, S.F.: J. Cryst. Growth
reproducibility appear to be largely instrumental since 28, 16 (1975)
every time an SPRT was left in a gallium cell for 15. Thornton, D.D., Mangum, B.W.: Proces Verbaux
several days, there was no change in the indicated du CCT, Document No. CCT/76-13, June, 1976
temperature of the triple-point to within the instru- 16. Lau, S . , Schlott, P.: Proces Verbaux du CCT,
mental resolution of k1.5 p°C. The pressure depen- Document No. CCT/76-23, June, 1976
dence was measured and determined to be dT/dP = 17. Sostman, H.E.: Rev. Sci. Instrum. 48, 127 (1977)
-2.011 C 0.001 m°C/atm (or -19.85 nK/Pa). This 18. Thornton, D.D.: Clin. Chem. 23, 719 (1977)
results in a value of the normal melting-point tem- 19. Riddle, J.L., Furukawa, G.T., Plumb, H.H.:
perature of 29.7 7205 " C . Platinum resistance thermometry, NBS Mono-
graph 126, 1973, Supt. of Documents, U.S.
Government Printing Office, Washington, D.C.
Acknowledgememts: We wish to acknowledge helpful 20402
discussions with Dr. G.T. Furukawa and with Mr. 20. Cutkosky, R.D.: J. Res. NBS 74C, 15 (1970)
W.R. Bigge. 21. Vol, A.E.: Handbook of binary metallic systems,
Vol. 11, Israel Program for Scientific Translations,
Jerusalem, Israel, 1967
22. Adams, Jr., G.B., Johnston, H.L., Kerr, E.C.: J.
References Am. Chem. Soc. 74,4784 (1952)
la. Comptes rendus des seances de la Treizieme 23. Bridgman, P.W.: Phys. Rev. 48,893 (1935)
Conference Generale des Poids et Mesures, An- 24. Jayaraman, A., Klement, Jr., W., Newton, R.C.,
nexe 2, p. A1 (1967-1968) Kennedy, G.C.: J. Phys. Chem. Solids 24,7 (1963)
lb. Comite Consultatif de Thermometrie, 8' Session, 25. Berry, R.J.: Can. J. Phys. 37, 1230 (1959)
1967, Annexe 18 26. Berry, R.J.: Metrologia 2, 80 (1966)
IC. Metrologia 5 , 35 (1969) 27. Bigge, W.R.: private communication

215