Hig
rature an NEW METHOD FOR MEASURING PRESSURE EARL A. GULBRkYSEN AND KENNETH F. ANDREW Westinghouse Research Laboratories, East Pittsburgh, Pa.
O n e of the trends in chemical and metallurgical processing of metals and other materials is the use of high vacuum at elevated temperatures. This paper describes the design, construction, and testing of refractory porcelain furnace units which are of double-walled construetion and which can be sealed direct to a Pyrex glass vacuum system. Pressures of mm. of mercury or better can he obtained with these units at 1175" C. A new method for measuring the performance of a furnace tube is suggested. This method utilizes the measurement of the rate of reaction of zirconium with gases present in the furnace tube.
HE problem of producing and maintaining high vacuums of mm. of mercury and better in ceramic furnace tubes a t temperatures above 900" C. is one of incrcasing technical and scientific importance. This problem was discussed a t the Louisville meeting of the Electrochemical Society in 1947 as part of a symposium on "Refractory hlaterials." Following this meeting the authors decided to design, construct, and test several types of furnace assemblies using the ceramic materials which would most closely meet their specifications. These furnace assemblies are designed for t\To purposes: First, the furnace unit should extend the temperature iange of the vacuum microbalance apparatus to 1200' C. and higher; and, secondly, the reaction chamber should make possible high vacuums for the carrying out of metallurgical preparations and heat treatments of metals a t temperatures up to 1200" C. and higher. GENERAL DISCUSSION
I n the design of high temperature and high vacuum furnace systems it is necessary to evaluate carefully the sources of gas and the methods for removing the gases. For purposes of classification, the authors divide the several materials concerned in a vacuum system into three groups. These are the materials used in the several parts of the vacuum system, the materials used in connecting one part to another, and the specimen and its support. For a vacuum system consisting primarily of glass and high temperature ceramic materials, the following mechanisms of gas formation must be considered: (a)physical and chemical adsorption of gases on the glass and ceramic surfaccs; ( h ) permeation of gascs into the surface of the glass and ceramic surfaces; (c) permeation of gases through the walls of the glass and ceramic tubing; and ( d ) gase- formed by thermal decomposltion and by chemical reaction of the several component in the glass and ceramic tubing at the elevated temperature. Factors (a)and ( b ) can be minimized by long periods of pumping and by heating of the glass and ceramic parts. Factor (c) may be minimized
by the use of protective v a c u u m , by the choice of ceramic tubing, and by careful construction of the connecting seals. Factor ( d ) may be minimized by the manufacturer's choice of material in preparation of the ceramic tubing. The sources for the inaterials used in Connecting one part t o another can be minimized by eliminating wax, metal, rubber, and grease connections while the sources for the materials used in the specimen and its support may be minimized by careful selection and pretreatment of the support material. Owing to t8hesmall quantities of gas available it is difficult ta distinguish one type of gas source from another. Thus, it is difficult to distinguish surface permeation from direct permeation and gas sources from thc glass tubing from gas sources due t o the ceramic material. The removal of gases forming in a vacuum system can be accomplished iii several ways. These methods arc vacuum pumps, cold traps, and getters, The speed of a n-ell-designed vacuum pumping system can be increased within limits by increasing the size of the pumps and by increasing the diameter and decreasing the length of the connecting tubing. If the gases forming in the vacuum system are condensable, the effective speed of the system can be increased by the use of cold traps. The speed of the cold trap will depend on its effective area and the size of thc connecting tubing. The temperature of the trap will determine which of the gases will be condensed. The use of getters in the form of filaments a t elevated temperature is very effective in removing those gases which will react with the getter. Zirconium metal filaments, for example, are very effective in removing many of the gases normally present in vacuum systems. The use of suoh devices is increasing rapidly (2, 6 , 7 ) . If one design :and constructs a vacuum system so as to combine a fast pumping system with the complete elimination of all rubber, metal, and grease seals, then the furnace tube itself becomes the major sources of gas. In choosing the ceramic material for the furnacc tube it is necessary to consider the following properties: a melting point of 1800" C. or higher; an expansion coefficient of 35 X lo-' to 50 X lo-' em. per cm. per C., which will allow a vacuum-tight glass seal to be made to the ceramic material; a nonporous sintered ceramic body which is vacuum tight, in the range of mm. of mercury or leas a t teniperatures greater than 1000 C.;a loiv vapor pressure a t temperatures of 1000' C. and higher; a low decomposition pressure for the oxide; and a high degree of stability to cryst,al structure transformations for temperatures up to 1500' C. Two ceramic materials meet these specifications. Those arc: synthetic mullite having a melting point of about 1835O C . and an expansion coefficient of 45 X 10-7 em. per em. per C. between 20" to 1320" C., and synthetic zircon writh a melting point of 1775" C. and an expansion coefficient of 42 X lo-' between 20' to 1550" C.
2'162
O
. INDUSTRIAL AND ENGINEERING CHEMISTRY
December 1949
LITERATURE DATA
*
GLASSTO PORCELAIN SEALS.Few papers are to be found in the published literature on the direct sealing of ceramic materials to glass although the technique is well known here and abroad. Vatter (11) has published a paper reviewing the practice but the paper is not available except in the abstract form. Sue (9) states in a recent paper that Berlin porcelain can be sealed directly to Pyrex, Sibor, and Jena glass by ordinary procedures. PERMEABILITY DATA. A number of papers have dealt with the permeability of quartz and various glasses to helium, hydrogen, and other gases. Many of these are reviewed by Barrer (1) who has contributed extensively in this field. There appears to be no question about the diffusion of the lighter weight gases. However, the diffusion of air, nitrogen, and higher molecular or atomic weight gases in silica is difficult to measure because of the low values for the permeability constants. Roeser (8)in a study of the permeability of refractory porcelain tubes to air has observed a great variation in tubes from several manufacturers. A comparison of Roeser's data on the permeal5lity of porcelain to air with that of silica to nitrogen indicates that porcelain is the superior material for furnace tubes. In addition porcelain does not devitrify a t high temperatures as does silica.
2763
and thermocouple are covered with a zircon slip and dried before the seal at g is made. The thermocouple and heater leads are taken out of the glass envelope by means of tungsten press seals. Figure 2 shows a schematic diagram of the main components of the authors' vacuum system. The furnace tube shown in Figure 1 is attached to a tube leading to the balance housing and by a second tube t o the remainder of the vacuum system which includes a liquid nitrogen trap, a mercury ahutoff for isolating the system from the pumps, and the several gages for measuring the pressure. McLEOO GAGE AND MANOMETER
Hg SHUT OFF TO GAS SOURCES
nn TO LIQUID NZ TRAP
Figure 2.
Vacuum System
..
EXPERIMENTAL DATA
GLASSTO PORCELAIN SEALS. Strain-free seals can be made between synthetic mullite and zircon tubing with Pyrex 70740 having an expansion coefficient of 33 X om. per om. per C. It is preferable t h a t the glass have a slightly lower expansion coefficient than the procelain. The seals are made by the regular procedures using a freshly cut and smooth porcelain surface. These seals permit the construction of complex glass-ceramic apparatus including doublewalled ceramic furnace tubes. I n the latter unit the annular space is evacuated or used to introduce test gases for permeability studies. DOUBLE-WALLED FURNACE TUBEAND VACUUMSYSTEM. Figure 1 shows a design of a double-walled furnace tube for use with Gulbransen's vacuum microbalance system (3, 4). Other designs for larger units have also been made and built. Figure 1 will be described because i t is the basis of the tests. Either mullite or zircon procelain tubing is used. Zircon is harder to cut using the cutting wheels regularly available to a glass blower.
,
NGSTEN LEA0 SEALtI)
INNER PUMPING TUIULATION
Figure 1. Internal Heater for High Temperature DoubleWalled High Vacuum Tube '4
The outer tube is shown e t a, the inner tube a t b, and the glass porcelain seals a t c. The inner tube is 2.57 cm. in (inner) dlameter and 30 cm. long. The wall thickness is 3 mm. ; the inner and outer tubes are joined a t g by a glass seal. This construction together with the refractory spacers allows free movement of one tube relative to the others when thermal gradients are present in the furnace unit. The spacers may be provided for in the manufacture of the tubing or made from Alundum brick as shown in the figure. The space between the two tubes is evacuated by a mercury diffusion pump and a Hyvac forepump. The inner tubulation is attached to the all-glass vacuum system. A molybdenum heater, h, is wound on the inner tube and clamped a t both ends. The details of the winding depend upon the temperature distribution desired. A heavy molybdenum rod carries the current to the heater proper. This rod is connected to a tungsten lead wire by means of a flexible stranded nickel wire. A platinum-platinum 10% rhodium therm* couple is mounted between the heater wires. The heater wires
+
The main vacuum pumping system consists of a high speed, singlejet, Illinois-type mercury diffusion pump; a two-jet Prince. ton-type mercury diffusion pump; and a Hyvac forepump. MEASUREMENT O F PERFORMANCE O F VACUUM SYSTEM. From a chemical and metallurgical point of view the performance of a given system is related to the reactivity of the given metal in the vacuum system. A measurement of the pressure a t some point away from the furnace unit by a McLeod or ion-type gage is of value, However, in a system which includes an effective cold trap a t liquid nitrogen temperatures many of the gases given 081 in the furnace unit afe not measured. The measurement of the reactivity should be carried out in the furnace unit itself. I n a well baked-out system the composition of the gases found in an all-glass and ceramic vacuum system depends largely upon the gases t o which the system has recently been exposed after baking out. The glass parts of the system, a t room temperature, may adsorb small quantities of water vapor even if dried gases are used. The furnace tube operating a t high temperature will not preferentially adsorb water vapor over other gases. The authors will show later that if their system consisting of a furnace tube a t 900" C. and the glass parts of room temperature is exposed t o hydrogen or helium, these gases will constitute the chief components of the gases forming in their vacuum system for 24 to 48 hours after exposure. If the same system were exposed to air they would expect to find as the gases forming in the vacuum system nitrogen, oxygen, water vapor, carbon dioxide, carbon monoxide, etc. MEASUREMENT OF PRESSURE. The McLeod gage is used in two ways to measure the pressure. The f i s t method is the direct measurement of the pressurk. A large McLeod gage is mm. of mercury can be estimated used and pressures of 1 X However, pressure readings in this range are only qualitative. The second method consists in isolating the vacuum system and furnace from the pumps by means of a mercury shutoff. The g u e s forming in the system are allowed to accumulate for 30 minutes a t which time the pressure is noted on the McLeod gage. Since the system includes a cold trap, water vapor is remoyed. It could not be measured accurately on the McLeod gage in any case. The volume of the vacuum system and furnace unit is 3 liters and since the pressure can be estimated to 1 x 10-6 mm. of mercury, the sensitivity of the method is 3 x 10-6 liter-mm. of mercury in 30 minutes. The authors call this vacuum test the apparent leak rate method since they have not established the sources of the gases and the system behaves as if a leak were present. REACTION OF ZIRCONIUM.The reactivity of the gases in the vacuum of the furnace unit can be memured directly by suspending a specimen of zirconium metal in the vacuum system and measuring the rate of reaction with the gases present. This can be readily achieved by the use of the vacuum microbalance technique which has been described (9, 4 ) . Zirconium has been found to react quantitatively a t elevated temperatures and a t very low pressures with most gases except
2764
INDUSTRIAL AND ENGINEERING CHEMISTRY
hydrogen and the inert gases (6,6). For low reaction rates the reaction products will dissolve in the metal as fast as formed and ieave the surface in a film-free condition. These reactions are the basis of the use of zirconium as getters a t high temperatures in vacuum tube technology The thermodynamic equilibria for the reactions of zirconium with oxygen, water vapor, carbon monoxide, carbon dioxide, and nitrogen have been considered in a previous work ( 5 ) . A11 of these reactions are feasible in the temperature range of 800" to 1200" C. and down to pressures of 10-8 mm. of mercury. I n this temperature range the rate of solution of the compound is sufficient to maintain the surface in a film-free condition providing the reaction rate is maintained below the rate of solution, For very low pressures the reaction rate probably is proportional to the pressure of the gases present. The critical conditions for thc reactions are the pressuIe and temperature a t which the rate of formation of the compound equals the rate of solution in the metal. Although the authors have not determined these precise conditions, experience has shown that the metal is probably in a film-free condition a t 800" to 1200" C. for pressures of the order of 1 x 10-6mm. of mercury and less. Because it is impossible to measure the APPARENT PRESSURE. pressure in the middle of a furnace tube by conventional ion and McLeod gages the authors propose to use the reactivity of zirconium as a reasonable approximation to the true pressure They call this method the measurement of the apparent pressure Kinetic theory allows one to formulate an expression for the weight gain of a given surface of zirconium as a function of the temperature, pressure, molecular weight of the reacting gases, and the efficiency of the surface reaction. It is necessary to make several assumptions in applying kinetic. theory to this problem. Since zirconium acts as a getter it m a j aid the pumps and cold trap to lower the pressure. The assumption is made that this will not affect the reactivity measurement under conditions in which the mean free path is large compared to the dimensions of the vessel. From kinetic theory the basic equations are stated. Consider a metal sample of 1 sq. em. area intercepting a column of gas of average velocity fi, mass m per molecule, and concentration. n molecules per cc. The weight intercepted IS given by
w = -noma: 4
Vol. 41, No. 12
the following experiments: to compare the perforniance of two types of furnace tubes and associated vacuum systems as a function of the temperature, to measure the reaction rate in one furnace tube a t several high temperatures and from these data evaluate the apparent pressures; and to study the effect of long periods of pumping on the rate of reaction. A 5-mil strip of ductile zirconium is used for each experiment. The weights of the specimens are 0.6840 gram and the measured areas are about 15 sq. em. After placing the specimen on the microbalance (3,4 ) the system is sealed and evacuated to a pressure of mm. of mercury. An auxiliary furnace is raised around the furnace tube and the specimen and furnace tube heated to the test temperature. The position of the balance beam is read with a micrometer microscope as a function of time. In this manner the rate of reaction in vacuum is studied. COMP~RISON O B Two FURNACE TUBES AND ASSOCIATED VacUUM SYSTEMS.The first system consists of a single-walled furnace tube of Vycor glass 7900 and the authors' older pumping system (3, 4). The second system consists of a zircon doublewalled furnace unit with the new vacuuni pumping system. The speed of the pumping system has been increased in the new system by the addition of a high speed Illinois-type mcrcury diffusion pump and by the enlargement of the connecting tubing The results are shown in Figure 3 . The weight gain in divisions (1 microgram = 0.8 division) is plotted against the time in minutes and the temperature is noted. A rate of reaction of 0.7 division or 0.056 microgram per sq. em. per minute is observed for the zirconium specimen in the Vycor glass 7900 furnace tube a t 900' C . while a rate of reaction of 0.070 division, or 0.0056 microgram per sq. cin. per minute is observed in the zircon tube at 900 C. It has been suggested by a reader of this paper that Vvcor glass 7911 would give a better prrforrnancc than type 7900.
Here a is the fraction of the molecules which react on striking tbe surface. Since the value of CY is not known, it is assumed to be 1. This awears reasonable for a film-free surface and for a low reaction rate (pressures of the order of 1 x 10-6 mm. of mercury). This formula also assumes that the mean free path is large compared to the dimensions of the vessel and that the area of the specimen is small compared to the walls of the furnace tube. I n the authors' microbalance system the specimen area is about 5% of the wall area of the furnace tube. The assumption is made that the average molecular weight of the gases corresponds to oxygen; therefore, the apparent press r e , A . P . , in millimeters of mercury can now be related to the weight gain, W , in micrograms per square centimeter per minute by the following equation for a temperature of 900' C. A . P . = 1.5 x 10-0
w
0
RESULTS AND DISCUSSION
Preliminary. Furnace tubes made from several mullite and zircon compositions were used. I n one of the compositions the sodium content of the alumina is reduced and in another composition anhydrous alumina is used. The firing temperature of the zircon tube is better than 3000' F. All of the furnace tubes gave excellent Pyrex 7740 to porcelain seals. A number of the seals have been in use for a year and have been subjected to a large number of thermal shocks. The seals are free from gas pockets and are of good strength. A number of single- and double-walled furnace tubes have been rested by sealing to the glass vacuum system. At a temperature of 900" C. all of the tubes tested gave direct McLeod gage pressure readings of 10-6 mm. of mercury or less. Several of the tubes have been tested at 1000 C. and one double-walled assembly at 1175" C. Again pressures of 10-6 mm. of mercury or less are observed by direct measurement. Vacuum Microbalance Study of Reaction of Zirconium in Vacuo and the Calculation of the Apparent Pressure. The measurement of the reaction rate of zirconium in vacuo is used in
TIME (M1N.I
Figure 3.
Comparison Reaction, Zirconium in Vacuo at 900' C.
Studies on the rate of reaction of zirconiurn in a furnace tube connected to the vacuum syst,em and then isolated from the pumping system show that the reaction rate is not appreciably affected by closure. The authors conclude that the improvement, shown in Figure 3 is due primarily to Lhe new Purnacc unit and not t o improvemenk in the pumping system. EFFECTO F TEMPERATERR ON REACTION RATE IN ZIRCOX FURNACE UNIT .4ND CALCULATION O F -kPPAREXT PRESSURE. Figure 4 shows the results of the rate of reaction of zirconium in a zircon double-walled unit with t8heinterniediate space evacuated. A reaction rate of 0.0053 microgram per sq. em. per minute is observed a t 900" C. and 0.10 microgram per sy. cm. per minute a t 1100" C. Apparent pressures of 0 3 X mm. of mercury and 1.5 X ' 0 1 mm. of mercury are calculat,ed for temperatures of 900" and 1100" C., respectively. I n both experiments a McLeod gage direct pressure reading of 10--6 mm. of mercury i s
December 1949
-
observed. The furnace tube is closed from the pumps and the rate of reaation measured. No appreciable effect is noticed. These experiments show that a n excellent vacuum can be obtained at temperatures of 900" t o l l O O o C. and that the rate of reaction of zirconium gives a reasonable method for studying the performance of a vacuum system. The sample weight is 0.6840 gram and the area is 15 sq. om. Using these figures the reaction rate at 900" C. corresponds to a n addition of 1.17 X 10-6yo of the sample weight per minute at 900" C. and 2.2 x lo-'% per minute a t 1100' C. These contaminations would not appreciably affect the mechanical properties of the metal unless the temperature treatment is extended over a long period of time.
10
8
5
"=
6
: 4
s 2
0
0 TIME (MINI
Figure 4. Reaction of Zirconium in Vacuo at 900' and 1100' C. in Zircon DoubleWalled Furnace Tub?
second per square centimeter per millimeter per centimeter of mercury.
TESTSAT 900' C. The long period or equilibrium annarent leak rates of both the mullite and zircon double-walled furnace units are of the order of the limit of error of the authors' measuring unit, or 1.7 t o 3.4 X liters-mm. of mercury per second, or 2 to 4 x 10-9 cc. a t N.T.P. per second. These values are observed after 15 to 24 hours of pumping and after exposure to a dry gas atmosphere with the zirconium specimen removed. Control experiments are made without the furnace tube present. T h e above limiting values are achieved after several hours of pumping. Both mullite and zircon furnace tubes have been tested and found t o give similar equilibrium apparent leak rates. Two effects should be noted concerning these equilibrium leak rates. First, the constant use of these furnace tubes a t temperatures of 900" C. and higher for several months leads to a deterioration in the apparent leak rate. Owing t o the long periods of time necessary to make the tests the authors are not able to state statisticall t h a t this is the general behavior. Secondly, some variation c a s been noted in tubes received from the manufacturer Again statistical information is not available. TESTSAT 1000" C. After 1 hour of pumping the apparent leak rate of a mullite double-walled vessel is 1.7 X 10-8 litermm. of mercury per second or 2 X lo-* cc. (N.T.P.) per second. In terms of permeability rates this value is equivalent to 3 X cc. per sq. cm. per second per mm. per cm. of mercury. Roeser (8) has studied a number of refractory porcelain tubes from several manufacturers. His permeability values vary from 8.3 X 10-10 to 5 X 10-8 cc. per sq. om. per second per mm. per cm. of mercury. The authors have studied two double-walled vessels. These two tubes give nearly identical apparent leak rates although one is constructed of mullite and the other from zircon. It may be possible that more sensitive tests would show up differences in apparent leak rates.
rn
0
B 2
*
2765
INDUSTRIAL AND ENGINEERING CHEMISTRY
EFFECTOF PUMPING ON REACTIONRATES AND APPARENT Figure 5 shows the results a t 900' C. The weight PRESSURES. gain in divisions is plotted against the time. Curve A shows a reaction rate after 1 hour of pumping of 0.2 division or 0.016 microgram per sq. cm. per minute. Curve B shows a reaction rate measurement after 15 hours of pumping of 0.04 division per minute or 0.0033 microgram per sq. cm. per minute. The apparent pressures are calculated t o be 2.4 X 10-8 mm. of mercury after 1hour of pumping and 5 X mm. of mercury after 15 hours of pumping. The improvement in performance as a result of a long period of pumping is shown. The next section will attempt to explain the nature of the improvement in the vacuum as a result of pumping. Apparent Leak Rate Studies. A large number of observations have been made on the apparent leak rate of severaI furnace tubes in a n effort t o understand the sources of gas in a furnace system. The experiments are classified in the following manner: the apparent leak rate at 900 ' C. and the effect of long periods of usage; the apparent leak rate a t 1000" C. and comparison with other data in the literature; the apparent leak rate as a function of the intermediate chamber pressure; the permeability of zircon porcelain to helium; the time variation of the apparent leak rate after exposing to various gases; the effect of gas pressure on the surface permeation; and the effect of time of gas reaction on the surface permeation. For these measurements the authors' system consists of a furnace tube 2.57 cm. in diameter (inside) and 3-mm. wall. The length of the tube in the hot zone is about 25 cm. and the effective area of this tube is about 200 sq. cm. The vacuum system has a total volume of about 3 liters and a surface area of about 2000 sq. cm. The leak rate data are given in units of liter-millimeter of mercury per second and as cubic centimeters of gas at N.T.P. per second. I n experiments where the permeability is to be calculated the data will be given in the units of cubic centimeter (N.T.P.) per
A
1310
-+,
-'
0 2 DIVISIONS/MIH 7
-CLOSED$ISTEM FROM PUMPS
-
J
T h e comparison with Roeser's data is only valid if the authors interpret their data in terms of permeation through the walls. They do not think t h a t this type of permeation is the source of gas since a protective vacuum is used. However, regardless of the source of the gas, the performance of the tubes studied here is superior to the tubes studied by Roeser (8). The difficulties inherent in measuring and interpreting apparent leak rate data will be discussed in the next several sections. APPARENTLEAK RATE AS A FUNCTION OF INTERMEDIATE CHAMBERPRESSURE. With the intermediate chamber evacuated the apparent leak rate of a mullite double-walled furnace tube after 24 hours of pumping a t 900" C. is 3.4 X 10-9 litermm. per second or 4 X CC. (N.T.P.) per second. Air is now added to the intermediate chamber t o 1 atmosphere and the apparent leak rate measured six times over a period of 50 hours. No change in the leak rate is noticed. One must conclude t h a t gas permeation through the walls is not important, in a system with air surrounding t h e tube. If a protective vacuum is used, gas permeation through the tube wall is negligible. The maximum rate of permeation can be calculated from t h e sensitivity of the authors' method and they find that the permeability of mullite at 900" C. is not more than 8.0 X 10-18 cc. per sq. cm per second per mm. per cm. of mercury. This value may be compared t o the permeability of silica to nitrogen (1)-namely, 9.5 X cc. per sq. cm. per second per mm: per om. of mercury. Care must be exercised in comparing permeability data because of the difficulty of separating t h e permeability from other gas sources. PERMEABILITY OF ZIRCONPORCELAIN TO HELIUM.The results are shown in Table I. In this table the McLeod gage pressures,
INDUSTRIAL A N D ENGINEERING CHEMISTRY
2766 TAl3LE
LEAKRATEAND 1. APPARENT
PERJIEABILITY O F I N ZIRFONPORCELAIS AT 900" C.
Condition Vacuum in chamber E e (88 om. Hg), 10 min. ,He(88crn.Hg),90 min. He (88 ,om. Hg), 180 min. Vacuum in chamber, 37 min. Vaauum in chamber, 1080 min.
(30-minute t e s t ) Apparent Pressure, Leak Rate, Mm. of Hg, Cc. (N.T.P.)/ 30 Min. Sec.
HELIVM
Permeability
Cc. (K.T.P.),
Sec./Sq. Cm. Mm./Cm. H y 6 . 8 x 10-19 1.35 x 10-9
4 X
4 x 7.9
2 . 8 X 10-'2
5 . 5 X 10-5
9.3
x
10~-*
3 . 0 X IO.->
5.9
x
1.0
x
10
1.6 X 1 0 - 9
3 . 2 X 10-8
5 ~ x 4
6 . 0 X 10-5
1 . 2 x 10-7
2.0
2 X 10-5
10-8
x
10-0
10-5
x
8
lo--' 10
13
after 30 minutes of closing off t h e vacuum pumps, are tabulated, From these values the apparent leak rate and the permeabilities &recalculated. Six conditions are studied. These are with a vacuum in the intermediate chamber; with helium in the chamber for 10 minutes; with helium in the chamber for 90 minutes; with helium in the chamber for 180 minutes; after evacuating for 37 minutes; and after evacuating for 1080 minutes. T h e helium is added t o the intermediate chamber to a pressure of 88 cm. of mercury. T h e apparent leak rate increases very rapidly with time until after 180 minutes an apparent leak rate of 5.9 X 1 0 - 6 cc. (N.T.P.) per second is observed. T h e result of this experiment is in marked contrast t o what is observed on adding: air t o the chamber. T h e authors feel t h a t in this case they are dealing with the phenomenon of permeation through t h e walls of the tube. T h e permeability is calculated t o be 1 X 10-8 cc. per sq. cm. per second per mm. per cm. of mercury. This value can be compared t o a value of 0.67 X 10-8 given by Barrer ( 1 ) and a value of 3.6 X 10-8 given by Tsai and Hogness (IO)for helium in silica at 900' C. T h e results show t h a t t h e permeability of silica and porcelain tot helium differs only to a minor extent. The deleterious effect of addine helium is amarent. T h e data in Table I also show that, after evacuating t& helium, t h e vacuum system is slow t o recover. Even after 1080 minutes t h e vacuum system has not re covered its initial performance. TIMEVARIATIONOF APPARENTLEAKRATEAFTER EXPOSURE TO VARIOUS GASES. I n these experiments t h e surface permeation of the furnace tube and the Pyrex 7740 glass system is studied. Mere, t h e intermecliate chamber is maintained under vacuum conditions and one of the gases, hydrogen, helium, nitrogen, oxygen, or argon, is added to the furnace tube and vacuum system after raising the mercury shutoff. Experimentally the removal of gas From the surface is measured by studying t h e apparent leak rate as a function of time. TABLE11. EFFECT OF GAS TREATXENT OF TUBEON
APPAREXTLEAKRATE
(goOD C. mullite double-walled tube, 30-minute test) Apparent Leak Rate, Cc. (X.T.P.) per HI at 900' C. t,o 2.0-Cm. Pressure for 2 Hours
Time, Min. 10
90
226
2.0
5.2
x x
10-8
10-7
2 . 2 x 10-7 1 . 6 x lo-? 4 . 0 X 10-Q
305 4203 He a t 900° C. t o 7.6-Cm. Pressure for 2 Hours 8 . 8 x 10-6 10 1 . 4 X 10-6 90 5 . 2 x 10-7 210 8 . 0 x 10-9 1402 N1 a t 900" C . t o 7.6-Cm. Pressure for 2 Hourb 2 . 0 x 10-7 12 102
258 1259
2.0 1.0
x
x
10-8 10-8
x
10-8
x
10-9
x
10-9 0 2 a t 900° C . t,o 7.6-Cm. Pressure for 2 Hours 4 x 10-7 10
4.0
90
6
210
4 2
5 X 10-9 226 Ar a t 900° C. t o 7.6-Cm. Pressure for 2 Houra 2 x 10-7 10 1.4 x 10-8 90
1208
x
10-s
Set
THE
Vol. 41, No. 12
Table I1 shows a summary of the experiments. The gases are ttdmitted to the vacuum system in turn and, after evacuating, the apparent leak rate is studied after definite periods of time. If helium is added for 2 hours a t 900' C. and a 7.6 cm. of pressure, the vacuum system has not recovered after 1402 minutes. Hydrogen behaves in a similar manner, T o test the relative influence of the adsorption and surface permeation of the Pyrex 7740 glass, the authors sealed the furnace tube from the system and made a blank run. Apparent leak rates similar t o those observed for the heavier gases were observed. Since a direct comparison is not possible, however, one must conclude t h a t the surface permeation and adsorption for the Pyrex 7740 glass system is appreciable and must be considered in the interpretation of the leak rate data. It is important to note t h a t although water vapor is undoubtedly given off in the natural decomposition of the furnace tube it is not playing a role in these measurements. Water vapor is not only removed in the liquid nitrogen traps but is in addition not measured by a McLeod gage. EFFECT OF Gas PRESSURE O N ADSORPTIOX ASD SURFACE PERMEATIOV PROCESSES. In order to understand the nature of the processes involved it is necessary t o study the extent of adrorption or permeation as a function of the pressure with the intermediate chamber evacuated. T h e extent of these processes is measured by an apparent leak rate measurement 10 minutes after evacuation of the gas. Table 111 shows the results for oxygen. T h e extent of the adsorption or permeation process is very strongly affected by the pressure. A plot of the apparent leak rate follows roughly a square root of pressure law. This suggests that the adsorption or permeation process follows a square root of pressure law. One explanation of this law is t h a t the adsorption cr snrface diffusion is occurring as atoms.
TABLE
111.
EFFECT O F PRESSURE O F R E A C T I N G GAS O S THE APPAREST LEAKRATEFOR OXYGEN
(R,?act,ion time is 10 min., 900' C. mullite double-walled tube, 3(l-iiiinutr test) Pressure, Apparent Leak Rate, Cc. (K.T.P.) pei' S w Cjm. of H r 10 Min. aft,er E v a , c u a t i n n 7.6 1.7 x IO-' 0.76 0.076 10 -9
1 1.6 4.0
x x x
10-0
10-7
10-8
EFFECTOF T ~ VOF E Gas Rsacrrox ON ADSORPTIOSOR SURPERMEATION PROCESSES. For these experiments the inter-
FACE
mediate chamber is evacuated and either hydrogen or oxygen added to the inner tube at a given pressure for several periods of time. The extent of adsorption or permeation is measured by an apparent leak rate measurement 10 minutes after evacuation of the gas. T h e results for oxygen and hydrogen are shown in Table EV. I n both cases t h e extent of adsorption or permeation is roughly linear with time. T h e rate of adsorption or permeation of oxygen at a pressure of 7.6 cm. of mercury and 900' C. is 1,7 X 10-9 cc. (N.T.P.) per second while the corresponding rate for hydrogen a t 3.2 cm. and 900" C. is 1.55 X 10-8 cc. (N.T.P.) per second. If the furnace tube alone were taking up t h e gas, it would be possible t o give the rates in terms of ~t unit area. Since the nature of the process and the extent of the reacting surface are not known, the authors prefer t o leave the calculation in the units given above.
INTERPRETATION. It is of interest to try t o interpret the data from the knowledge of the several adsovption processes and the
OF Trim TABLE IV. EFFECT
OF REACTIOX o x APPARENT LEAK RATE
(900' C. mullite double-walled tube, 30-minute tert) Apparent Leak Rate, Cc. ( 3 T.P.) ppr Spr I'ime, Min 10 Min. after Evacuation HA,Pressure = 3 2 Cm. of H g 1 28 6 8 xx 110-8 0 : 10 30 2 . 8 x 100 2 , Prraaurr = 7 6 Cm. ot Hp: 1.1 x 10-7 1 . 7 x 10-6
3.0
x
10-6
December 1949
P
w
INDUSTRIAL AND ENGINEERING CHEMISTRY
nature of the permeation of gases in solids. A rough estimate of the total volume of gas given off from the walls of the vacuum system can be made from a consideration of the data in Table 11. With the exception of the hydrogen and helium data the volume of the several gases given up by the surfaces can be accounted for by the removal of a monomolecular layer. The time dependence of the gas evolution rate indicates that an activation energy is involved. For the gases oxygen, nitrogen, and argon the process may be one of chemi-sorption or one of chemi-sorption plus permeation. The pressure effect mentioned above shows that permeation may be playing a more important role than straight ehemi-sorption. Barrer ( 1 ) has shown that one interpretation of the square root of pressure law is that the diatomic molecules are dissociating before diffusing into the ceramic material. A more extensive study is necessary to interpret the details of the process. A study of the hydrogen and helium data indicates that a much larger adsorption surface is involved or that a combination of adsorption and permeation is occurring. This must be true since the volume of gases rel-ased after treatment with hydrogen or helium is 10 to 50 times that observed for the other gases. The permeation of the ceramic material to helium was shown above t o occur rather rapidly. This fact leads us to conclude that the surface of the ceramic material is removing gas and then releasing this gas by a process of chemi-sorption and permeation. SUMMARY
is made of the effect of long periods of pumping on the apparent. pressures. The apparent leak rate of a number of furnace tubes has been studied extensively as a function of a number of variables. After pumping periods of 24 hours apparent leak rates of the order of I t o 3 x lo-@liter-mm. of mercury per second are observed. A study of the apparent leak rate of a zircon double-walled vessel has shown t h a t the apparent leak rate is not due to permeation of gases through the walls of the ceramic tubing. The permeability of the tubing to helium is measured and shown t o bP in agreement with the data in the literature for helium in silica. The source of the evolution of gases in furnace tubes is studied extensively as a function of the exposure of the surface of the tubing to various gases. The adsorption or surface permeation is greatest for helium and hydrogen and least for oxygen and argon. This would indicate t h a t a permeation process is important. The process is further studied as a function of the pressure of the gas and time of exposure t o the gas. The rate of permeation is a function of the square root of the pressure and is linear with time. ACKNOWLEDGMENT
Furnace tubes made from several mullite and zircon compositions were supplied by the McDanel Refractory Porcelain Company, Beaver Falls, Pa. Ductile zirconium was obtained from Foote Mineral Company, Philadelphia, Pa.
h number of synthetic mullite and zircon porcelain to Pyrex
7740 glass seals have been made. A double-walled furnace unit is described for high temperature and high vacuum studies.
Pressures of 10-8 mm. of mercury or better can be obtained at 1175’ C. A new method for measuring the pressure of the gases present m a furnace tube is developed. This method utilizes a sensitive weight gain method for measuring the rate of reaction of 5irconium with the gases present in the furnace tube. The authors call the pressures measured in this manner the apparent pressures. Apparent pressures of the order of 0.8 x 10-8 mm. of mercury and 1.5 X IO-’ mm. of mercury are measured a t 900” C. and I1OQo C. in a n evacuated zircon double-walled vessel. A study
2767
LITERATURE CITED
(1) Barrer, R. M., “Diffusion in and through Solids,” Cambridge,
England, University Press, 1941. (2) Fast, J. D., Foote Prints, 13,22-30 (1940). (3) Gulbransen, E. A.,Rev. Sci. Instruments, 15,201-4 (1944). (4) Gulbransen, E.A., Trans. Electrochem. Soc., 81,187-97 (1942). ( 5 ) Gulbransen, E.A., andAndrew, K. F., J.M e t a k , 1,515-25 (1949). (6) Guldner, W. G., and Wooten, L. A., J.Electrochem. Soc., 93,22335 (1948). (7) Raynor, W. M., FootePrints, 15, No.2,3-10 (1943). (8) Roeser, W. F., B u r . Standards J. Research, 6,485-94 (1931). (9) SUe, P., Bull. SOC. chim. France, 1946,410. (10) T’sai, L.S., and Hogness, T., J.Phule. Chew., 36,2595 (1932). Feinmech. u . Prhzision. 51,181-6 (1943). (11) Vatter, H., RECEIVED January 26, 1949.
m
Effect of Surface lension on Heat Transfer in Boiling 0
A. I. MORGAN, L. A. BROMLEY, AND C. R. WILKE University of Calvornia, Berkeley, Calif. I
Heat transfer coefficients were observed to increase qualitatively with decrease in surface tension of the liquid for temperature differences below the critical value. This is in agreement with other investigations. The critical temperature difference occurs at lower values of the temperature difference as the surface tension is lowered. For solutions of wetting agents in which surface tension is a function of time it is shown that the surface tension of a “young” surface controls the heat transfer coefficients.
N
UMEROUS investigators have been concerned with the
effect of surface tension on the heat transfer coefficients in boiling (1-4, 8). The effect on boiling of the rate of approach of the surface tension of a new surface to its equilibrium surface tension has not been studied. Since the formation of bubbles in boil-
ing involves the continuous formation of new surface, the condition of equilibrium surface tension is never truly attained although it may be closely reached in some cases. APPARATUS
The apparatus consisted essentially of a steam-heated, chromium-plated copper tube from which the boiling occurred and which itself passed through a Pyrex tube containing the boiling liquid. The copper tube was in reality two concentric tubes The annular space between the tubes contained the insulated thermocouple leads and was filled with soft solder. Although an attempt was made to have the thermocouples merely touch just below the surface of the outside tube it is known that the thermocouples contacted points considerably below the surface and
