890
HOMER FAY
The Electrical Conductivity of Liquid A1,0, (Molten Corundum and Ruby) *
by Homer Fay Union Carbide Corporation, Gnde Division. Speedway Laboratories, Indianapolis, Indiana (Received October 99, 1965)
The electrical conductivity of pure liquid Al203 has been measured a t ca. 2400°K and found to be 384 mhos m-’ 5%. This value is one-fourth that usually quoted. The measurements were made in an iridium crucible with a coaxial iridium rod as the second electrode. By measuring the resistance as a function of the immersion depth of the iridium rod, both the conductivity and the series circuit resistance were determined. The cell was calibrated with aqueous electrolyte solutions after correcting for polarization effects. The possibility that the electrical conductivity is dependent on the oxidation-reduction properties of the atmosphere is considered.
*
Introduction
Experimental Section
Data on the electrical conductivity of liquid metal oxides are still quite scarce. The conductivities of the more refractive oxides in particular, have in most cases only been estimated from electric furnace measurements. Mackenzie‘ classifies oxide melts as either nonconducting “network liquids” or as ionic or electronic conductors. Liquid A I 2 0 3 is considered to be an ionic conductor. The most comprehensive study of the conductivities of molten oxides is probably still that of van Arkel, Flood, and Bright.2 They measured the conductivities of several of the lower melting oxides and compared them with halides. They also tabulated values for MgO, CaO, TiOz, ZrOz, ThOz, Cr203, and Ah03; these, however, were not measured directly but were estirnated from electric furnace operations. The conductivity of Al1O3at the melting point is given as 15 X 10%mhos m-l. Recently, we have been experimentally investigating the possibility of growing crystals from melts maintained by inductive coupling of radiofrequency power directly to the melt. I n this process, the melt acts as its own susceptor, and its conductance is very significant. Although not quantitative, these experiments have rather consistently indicated that the conductivity of A1203 was considerably lower than the values given by van Arkel. We, therefore, decided to attempt a direct measurement of the conductivity of liquid (molten corundum or white sapphire) and of this liquid doped with Crz03(molten ruby).
The scarcity of high-temperature conductance data is due to the difficulty of finding sufficiently inert and refractory materials to use for crucibles and electrodes. After some experimentation, an iridium crucible as one of the electrodes and a small-diameter iridium rod as the other were found to be satisfactory for this purpose. A diagram of the electrode and crucible configuration is shown in Figure 1. The crucible was ca. 1.25 in. in diameter and 1.2 in. deep but was slightly rounded on the bottom. A 2.25-in. diameter rim was welded to the top edge of the crucible, and an iridium lead wire was welded to this rim. The crucible was supported by its rim inside a special oxy-hydrogen combustion furnace, the combustion zone being directly outside the crucible. The central iridium electrode was mounted in a Jacobs chuck of a drill press and was adjusted to be approximately coaxial with the crucible. The drill press permitted vertical movement of this electrode, and measurements were made at various immersion depths as described below. Resistance measurements were made with a General Radio Type 650A bridge driven by an internal oscillator at 1000 cps. This bridge has but a single balancing
The Journal of Physical Chemistry
*This research is part of project Defender under joint sponsorship of the Advanced Research Projects Agency, the Office of Naval Research, and the Department of Defense. (1) J. D. Mackenzie, Advan. Inorg. Chem. Radiochem., 4, 293 (1962). (2) A. E. van Arkel, E. A. Flood, and N. F. H. Bright, Can. J. Chem., 31, 1009 (1953).
ELECTRICAL CONDUCTIVITY O F LIQUIDAlp03
891
9
1
LIQUID LEVEL
-
no indication of any appreciable dispersion or change in resistance with frequency.) The audiofrequency measurements yielded resistance values from 0.15 to 0.38 ohm, after subtracting the lead resistance. A small but arbitrary quantity of Crz03was then added to the charge, and the measurements were repeated while lowering the central electrode. After the crucible was cooled, the crystallized ruby was removed, and an estimate was made of the liquid level. The electrical assembly was then reconstructed, without the furnace, and the cell was filled to the same level with nearly saturated NaCl solution. The electrical conductivity of this solution was measured with a calibrated dip cell to be 10.3 mhos m-l. Resistance measurements were then made as a function of immersion depth of the electrode. However, in this case the zero resistance reference point was obtained by directly contacting the crucible with the central electrode. It was found that the cell polarized quite badly when this electrolyte was used. A similar polarization had not occurred with the fused AlzOa. In fact, the reactance corrections were negligibly small in the high-temperature measurements: and the external capacitor was used only to sharpen the balance. Large values of external capacitance, up to a few tenths of a microfarad, were required to obtain a satisfactory balance with the electrolyte. It was, therefore, necessary to correct the bridge readings. It was assumed that the capacitive reactance was in series with the solution resistance, dielectric effects being negligibly small for this solution. The equivalent series resistance was calculated, and from this the cell constant, K , was determined as a function of immersion depth. This calibration curve is shown in Figure 2. The cell constant was never “constant” but always decreased with immersion. However, the variation, over a range of depths, was smooth enough to allow reasonably accurate measurements. The measured resistances in the A1203 experiments have been plotted as a function of the cell constant, K , as shown in Figure 3. The points should lie on a straight line, since R = Ro K / a . When the data taken with the electrode descending and ascending are treated separately, good linear fits are obtained. The two straight lines in Figure 3 have been fit by the method of least squares. The values obtained for Roand u are
1 r -WELO
-0
-1/8
-IN
-3/8
I12 5/8 3N
P---
-
I /2”-
Figure 1. Iridium crucible and electrode configuration (scale shows immersion depth in inches).
control when used for ac resistance, and thus can measure only relatively pure resistances. A General Radio 1000-pf variable capacitor was, therefore, inserted in the bridge in parallel with either the arm adjacent to or opposite to the unknown, as required to sharpen the balance. The bridge contained a tuned amplifier and was balanced using a Tektronix oscilloscope as the final detector. The cell was connected to the bridge through a coaxial cable, with extended lengths of bare copper wires and clips to attach to the iridium wire and to the chuck holding the iridium rod. The crucible was connected to the shield of the cable, which was grounded at the bridge. A measure of the lead resistance could be made by shorting out the leads where they connected to the iridium. This did not, however, correct the readings for the resistance of the iridium, which was not negligible. The crucible was charged with fragments of pure crystalline A1203 and heated until the charge was melted. The temperature was maintained at ca. 2400°K as measured by an optical pyrometer uncorrected for emissivity. The central iridium electrode was then lowered until it just touched the melt surface. This position was noted and identified as the “zero” position. The electrode was then lowered in ’/&. intervals to a depth of 3/4 in. below the zero point. A series of resistance measurements was made with the electrode descending and later repeated as the electrode was withdrawn. (When the electrode was at a depth of 3/4 in., a few measurements were attempted with a radiofrequency bridge. Quantitative data were not obtained, but the radiofrequency measurements gave
+
BO, ohm
Electrode descending Electrode ascending
0.0702 0.0535
U,
mhos
m-1
385 383
The variation in Ro may represent real changes in contact resistance but is more likely an effect of the Volume 70, Number 3 March 1966
892
HOMERFAY
13C I20
1 IO
0.4 100
I
90
7
80
E Y
70
t L + *
g
”
60
0 _t
d
50
90 CELL CONSTANT,
30
\
\
0
\
I
I
I
I
I
I
.5
I H M E R S I O N DEPTH,
I
I .IO
I
AI
I
Figure 3. Measured resistance us. cell constant for pure Al103 and ruby a t 2400’K: straight lines, least-squares fits to the equation R = RO K / u ; 0 , pure A1203,electrode descending; 8, pure A1203, electrode ascending; X, ruby (A1203 Crr03), electrode descending only. Ruby points at 1/8 and in. are off scale.
INCHES
Figure 2. Cell constant calibration curve as determined with aqueous NaCl solutions.
meniscus a t the electrode. The mean conductivity a t 2400°K is u = 384 f 8 mhos/m. The range limits have been calculated from an analysis of variance and represent a probable error of 2%. However, the absolute value of the cell constant is in doubt by more than this amount, and we thus estimate the conductivity to be accurate to 5%. The addition of Crz03to the charge caused a definite initial decrease in conductivity. As the measurements progressed, however, the resistance values for ruby approached those for pure A1203, and the last two points measured fit the average slope very well. Apparently, the conductivity was not constant during these measurements but was drifting toward an equilibrium value identical with that of pure A1203. Finally, measurements were made of the resistance of the charge as the furnace was cooled. The values of resistance increased smoothly but very rapidly as the melt solidified. The high-temperature conductivity of the solid is apparently not negligible, but it is still much less than that of the liquid.
Discussion The method used for measuring the conductivity of high-melting oxides in metallic crucibles appears to The Journal of Physical Chemistry
m-l
+ +
20
IO
K,
be capable of considerable accuracy, despite the fact that the “cell constants” are rather small and the cell must be calibrated for various immersion depths. The conductivity of pure liquid A1203 near its melting point is, according to the present study, 384 mhos m-’. This is almost a factor of 4 lower than the value of 15 X lo2 mhos m-’ quoted by van ArkeL2 The conductivity is, however, still sufficient to consider liquid Ah03 to be an ionic conductor. The chargecarrying species are apparently small and mobile. No further structural inferences can be made from these few measurements. It is unfortunate that the atmospheric composition could not be controlled during these measurements. The fusion was made in the open air but the local atmosphere could have been somewhat reducing from the flame gases. Other observations indicate that the appearance of the melt, its viscosity, and its conductivity may all depend on the atmosphere, reducing conditions favoring a higher conductivity. One possible mechanism that could account for such behavior is the establishment of an equilibrium between the oxygen in the atmosphere and oxygen “vacancies” in the liquid. We hope eventually to test this hypot,hesis by further experiments in controlled reducing and oxidizing atmospheres. The addition of small amounts of Crz03 to the melt definitely decreases the conductivity, but the effect is apparently only temporary. This further indicates
STRUCTURESOF C,H,O COMPOUNDS ADSORBED ON Fe
that some process of equilibrium with the atmosphere is taking place.
Acknowledgment. The author thanks B. J. Corbitt for his assistance in making the measurements and R. &I.Youmans and P. V. Vittorio for making avail-
893
able the furnace. He also is indebted to Dr. M. N. Plooster for valuable discussions on the properties of A1203 melts and to the Crystal Products Department of the Linde Division for permission to publish these results.
Structures of Some C,H,O Compounds Adsorbed on Iron'
by G. Blyholder and L. D. Neff Department of Chemistry, University of Arkansas, Fayetteville, Arkansas
(Receined October 22, 1966)
The infrared spectra over the range 4000 to 300 cm-' of methyl, ethyl, n-propyl, n-butyl, isopropyl, isobutyl, and t-butyl alcohols, acetaldehyde, ethylene oxide, acetone, methyl ethyl ketone, diethyl ether, methyl vinyl ether, and tetrahydrofuran adsorbed on Fe have been obtained. All except diethyl ether and tetrahydrofuran, which gave no infrared evidence of adsorption, adsorb at 25" to give an alkoxide structure as the main stable surface species. The alkoxide structure is shown to be in accord with the main stream of thought on the stability of organometallic compounds, thus linking surface chemistry and organometallic chemistry firmly on a structural basis. An alkoxide structure is proposed for intermediates in the Fischer-Tropsch synthesis reaction.
Introduction Carbon-, hydrogen-, and oxygen-containing molecules have been found to undergo many interesting and occasionally useful reactions on metal surfaces. As well as hydrogenation and hydrogenolysis of many kinds of compounds, dehydration of many compounds including alcohols, which yield aldehydes and esters, have been observed. CO and Hzinteract on metal surfaces under appropriate conditions to produce hydrocarbons and alcohols. Although there is a voluminous literature on reactions occurring on metal surfaces, there is relatively little direct experimental evidence about the structure of species adsorbed on metal surfaces. Infrared spectroscopy has proved to be one of the most effective ways of obtaining structural information about adsorbed species. Most of the spectral work has concerned CO and hydrocarbons with carbon-, hydrogen-, and oxygen-containing molecules being somewhat neglected. The interaction of
methyl and ethyl alcohols2 and a variety of alcohols and other oxygen-containing molecules3 adsorbed on Ni have been studied by infrared spectroscopy in this laboratory. On Ni these studies indicated that at room temperature chemisorbed CO and acyl structures are the most stable species. When the carbon atom to which the oxygen is bonded is bonded to only one other carbon atom, that C-C bond is readily broken to produce chemisorbed CO. This decomposition to CO and acyl structures indicates that hydrogen atoms are fairly readily removed by the surface interaction. In this paper the infrared spectra of a variety of C,H,O compounds adsorbed on Fe are examined. Since Fe is generally found to be not so good a dehy(1) This paper is taken in part from the Ph.D. dissertation of L. D. Neff, University of Arkansas, 1964. (2) G. Blyholder and L. D. Neff, J. Catalysis, 2 , 138 (1963). (3) G. Blyholder and L. D. Neff, in preparation.
Volume YO, Sumber 3 March 1066
