Determination of Traces of Water Vapor in Gases FRANK E. HARRIS AND LEONARD K. NASH
Harvard University, Cambridge, Mass.
but 15 minutes. The measurement is unaffected by (or readily compensated for) wide and rapid fluctuations in the composition of the matrix gas, and the method may be used for the determination of 0.001 to 0.1 volume % of water vapor. Water determinations formerly accomplished with difficulty can be conducted by this method; and the thermistor bridge technique should find important application in a number of analogous trace determinations based on thermometric measurements.
A general method for the trace determination of water vapor in condensable gases, particularly hydrocarbon gases, has been sought. The novel method here described is based on measurement, with a thermistor bridge, of the temperature rise produced when a stream of the test gas is passed over solid calcium hydride. The apparatus is readily assembled from normally available components, calibration is easily accomplished with a dynamic blending system, and a single determination requires
A
3700" K. If two thermistors (designated by the subscripts and $ ) are maintained a t different temperatures,
GREAT variety of methods for the determination of the water content of gases has been developed. However, there is currently available no completely satisfactory method for the estimation of minute traces of 13 ater vapor in gases in general, and in hydrocarbon gases in particular. Yet the need for such a method has become ever more acute-for example, in connection with the measurement of the aqueous content of a number of gases involved in industrial processes based on the use of catalysts that are readily poisoned by water. The method described herein presents certain advantages in that ( a ) it is easily applied to condensable (as well as noncondensable) gases; ( b ) it is semispecific for water; ( c ) though primarily intended for the analysis of batch samples, it is at least potentially serviceable in semicontinuous automatic analysis of streaming gases; and ( d ) the apparatus is readily assembled from commonly available components. It seemed desirable to base the determination on some specific or semispecific chemical property (reaction) of water that could be followed in terms of an easily measured physical property of the system. Thus, water enters as a reactant into a number of fairly selective and highly exothermic processes that involve the progressive destruction of solid reagents-for example, phosphorus pentoxide and metallic sodium. To be useful in the proposed determination it is essential that the reactivity of the solid reagent is not substantially impaired by its progressive exhaustion, and by the accumulation of solid reaction products. Solid calcium hydride appeared to be the most promising of a small number of possible reagents. At room temperature this material is substantially inert to most readily volatile materials (except for acids, alcohols, etc.), but it reacts rapidly and quantitatively with minute traces of water. Inasmuch as some 25,000 calories are released for each mole of %rater involved, the reaction is sufficiently exothermic to be followed by thermometric measurements. Of particular importance is the fact that the calcium hydroxide formed in the reaction is nonadherent; much of it is blown away by a rapid gas stream, and the activity of the surface is maintained until the hydride is practically exhausted (3). To measure the small temperature differences expected in the proposed determination, multijunction couples and/or fairly elaborate electrical equipment would ordinarily be required. The use of a thermistor bridge appeared to be a more promising possibility. Such a bridge retains the desirable feature that the measurement of the temperature rise is unaffected by small changes in room temperature; it is, potentially, a much more sensitive instrument than a thermocouple, and it involves only the simplest and least expensive electrical equipment. The resistance, R, of a thermistor a t any temperature, T , is calculable from the relation, R = a.eb/T,where a and b are constants ( 2 ) . The constant, b, is characteristic of the material used in the thermistor and, in the authors' specimens, had a common value of
1
From t,his it follows that:
T1T2 In a2Rl T2 - T I = b alR1 In the system proposed herein, T1and T2never differed by more than a degree or so from room temperature which, in turn, commonly fluctuated over a range of not more than a few degrees. Under these conditions, and noting that ul, al, and b are all constants, it , 5 v is plain that the difference of the temperatures prevailing around the therniistors is a simple function of the ratio of the resistances, which can easily he determined by placing the thermistors in a Wheatstone bridge circuit.
-
-
APPARATUS AND PROCEDURE
The heart of the apparatus shown in Figure 1 is based on the design of a related system described by Cohn The greater length of the inner reaction tube, K , was enclosed in a silvered vacuum jacket, J. Tube K was made from a 25-cm. length of 16mm. tubing, with a median constricFigure 1. Remtion of about 9 mm. in diameter. As a tion Cell and protection against drafts, the unTh erm i *t or jacketed projections of tube K were Bridge wound with two layers of sheet asbestos. The calcium hydride was used in the form of fragments with a mean diameter of about 1 mm. These fragments were easily selected from a partially crushed mass of the commercial (Metal Hydrides, Inc.) substance. One gram of the calcium hydride was introduced into the hemispherical cup, HI formed from 50-mesh platinum gauze. On insertion into the reaction tube, this cup was supported by the shoulders of the constriction. RI and R1 were Western Electric 17A thermistors4isks about 1 mm. thick and 5 mm. in diameter, with a nominal resistance of 1000 ohms. The thermistors were supported, a t a distance of about 3 cm. from the calcium hydride mass, by leads of heavy bell wire that passed through the rubber stoppers with which the ends of K were plugged. The rest of the measuring circuit consisted of a dry cell, a dial-decade resistance box, XI a precision wire-wound 1OOO-
736
V O L U M E 23, NO. 5, M A Y 1 9 5 1 ohm fixed resistance, Y , and a Rubicon box-type galvanometer, G , with a resistance of 10.27 ohms and a sensitivity of 0.002 pa per mm. scale deflection. The stream of the test gas was passed through the reaction cell a t a rate of 1.3 liters per minute, and the variable resistance, X, was set (usually only t o the nearest ohm) to bring the bridge into balance.
Figure 2.
737 Table I. Trials with Determinate Samples Volume Percentage of Water 0.0000 0,0028 0.0070 0.0084 0.0140 0.0393 0.060 0.097
Matrix Gas Oxygen
CALIBRATION
The operating characteristics of this system w r e investigated in a series of trials with a number of gases containing various determinate concentrations of water vapor. These test mixtures were prepared with t,he aid of a dynamic blending system, shown in Figure 2 . which is related to that described by Kalker and Ernat ( 4 ) . The gas enter5 a t the left, under a positive pressure that can Le cont,rolled clo~elywith the aid of the variable leak, XI. The gas stream passes through vessel D , where it is dried by a packing of Dehydrite folloned by calcium hydride. The stream then bifurcates: the major proportion of the gas is by-passed through tap X,, whereas the rest is routed through tap to a capillary flowmeter, F,, and thence t o the water saturator, S . The two gas strea.ms are then reunited, and the mixture flows successively through anot)her flowmeter, F P ; through a tube, T , containing a thermometer; through the reaction cell, R (shon-n in detail in Figure 1); and through the desiccant contained in the guard tube. G. to the vent. The aqueouc content oi the miFture delivered a t R depends primarily on the aqueous tension in the saturator, which was maintaincd constant. and on the proportion of wet and dry gas entering thr final mixture. 1vhic.h iyas cont,rolled by suitable manipulaThe saturator consisted of a long bead tion of tape S:and ?io. ton-er, partially filled with n-ater! and immersed in a n ice bath. The gas rntered at the bottom of the tower, and was precooled during itP passage through the ice bath. The column of beads above the wat>erlevel in the saturator served as an effective spray trap. \Talker and Ernst found ( 4 ) that the gas delivered from their saturator wm only 93 to 96% saturated. However, the authors' saturator was more than tn-ice as long as theirs and the masimuni flow rate through the saturator described herein was approximately one fifth of their maximum rate. It seemed safe to conclude that the authors' saturating efficiency was close t o 100%. K i t h the saturator a t 0" C. this corresponds t o a concentration of about 0.60% water vapor in an effluent gas at a pressure close to 1 atmosphere.
X,
Ohms 1080.0 1078.5 1076.5 1076.0 1072.5 1060 1045 1022
Temperature Rise 0
c.'
0.000 0,033 0.077 0.088 0.166 0,447 0.79 1.32
Butane
0.0000 0.0130 0.0286 0,0324 0,0467 0.061
1080 1073 1066 1063 1058 1051
0.000 0.155 0.312 0.381 0.49 0.65
Carbon dioxide
0,0000 0.0169 0,0345 0.063
1079 1072 1063 1050
0.000 0.155 0.359 0.65
Ethylene
0.0000 0.0156 0.063 0.104
1080 1073 1046 1023
0.000 0.155 0.77 1.30
Hydrogen
0.0000 0.0355 0.0587
1080 1062 1049
0.000 0.403 0.695
Calibration T r a i n
The circuit design then requires that Rl/Rl = S / Y = X/lOOO The temperature rise must be a simple function of R2/R1; hence the temperature rise must also be a simple function of the value of X which, after calibration, can be taken as an indication of the water content of the test gas. The observations are essentially unaffected by the thermal effects of the thermistor bridge current. Under the authors' experimental conditions the electrical heat output is barely of the order of the heating e y e c t e d from a gas containing mole yoof water; this concentration is ell below the authors' present range of investigation.
Resistance,
The proportion in which the wet and dry gas streams were mixed, to form final mixtures of determinate low water concentrations, was measured with the capillary flowmeters, F 1 and F2. Preliminary trials verified that, other things being equal, the reading of each flowmeter T ~ a Slinear function of the rate of gas flow. Flowmeter Fz was accurately calibrated with oxygen in a series of trials in which the gaseous throughput in unit time was measured volumetrically. FloJymeter F , was then calibrated in terms of the readings of FP by passing a stream of gas through the system with tap X3 completely shut, so that the same quantity of gas flowed through Fl and F2. hfter this calibration it was possible to determine, from the relative readings of the two flowmeters, what proportion of any final mixture consisted of gas that had been routed through the saturator. Thc final mixture was blended from wet and dry gas streams a t pressures equal to each other and to the pressure (calculable from the reading of the small, open-end manometer, AI) a t n-hich the wet gas left the saturator. The aqueous concentration of the wet gas was determined from the equilibrium vapor pressure of water at 0' C. and the total gas pressure a t the exit from the saturator. Hence, knowing the proportion in which the wet and dry gases were mixed, the concentration of water vapor in the final mixture could be calculated All the test materials used were C.P. grade tanked gases. With each new gas the same general procedure was followed, except that the total flow was adjusted until the ratio between the pressure drop indicated by the second flowmeter and the pressure drop that had been produced there by oxygen, flowing a t a rate of 1.3 liters per minute, was equal t o the ratio of the viscosity of the test gas to the viscosity of oxygen. The results obtained are shown in Figure 3 and Table I. RESULTS
The temperature rises listed in the last column of Table I were calculated with the aid of Equation 2. When a perfectly dry gas is passed through the system there should be no rise in temperature-that is, T2 = T,-so that a 2 / a l = R2lR1. But &/RI = X / Y ; thus, in this system, a2/a1 = lOSO/lOOO = 1.08. For temperatures only slightly different from room temperature,
(-In
1.08.
1000
=
55.2 (3.0334 - log X) (3)
ANALYTICAL CHEMISTRY
738 In the range of the authors' experimental data the difference ( 2 ' 2 - TI)is an almost perfectly linear function of the resistance, X , as may be seen from the double graduation of the horizontal axis of Figure 3. This happy circumstance permits the preparation of undistorted calibration curves of water concentration as a function of the measured resistance, X,and there is no necessity to calculate the actual temperature rises corresponding to these measurements.
'
Figure 3.
0
Hydrogen
0O w m
n-Butane
Graph of Calibration Data A Carbon dioxide V Ethylene
The equipment mas fairly rapid in its response. After a large change in water concentration a reading within 0.002'% water of the final value could be obtained in 10 minutes, and the final value in 15 minutes; with smaller changes of water concentration the corresponding times were 5 and 10 minutes. The final readings were steadily maintained for periods of hours. Thus the progressive exhaustion of the calcium hydride charge, provided i t did not proceed too far, was without effect on the experimental results. The water content of the gas under test determines the frequency Mith which the charge has to be replaced. For concentrations of the order of O.l%, the maximum studied, the useful lifetime of the charge exceeded 1hour-that is, the exhaustion of 15% of the hydride. The lifetime of the charge was greatly extended when lower concentrations of water vapor were involved; the same charge has been used for as long as 48 hours. The approach of effective exhaustion of the charge is unmistakably signalized by a pronounced upward drift of the resistance readings and a failure to settle down to any steady equilibrium value, as in normal operation. \Then a gas of invariant water content was passed through the train while an exhausted charge was replaced with a fresh one, the last steady reading obtained with the old charge and the first steady reading obtained with the nen one
were always the same. The results seemed to be unaffected by a change of as much as 10% in the total flow rate. The accuracy of a calibrated method of analysis can best be judged from its reproducibility and sensitivity. This is particularly true when, as in the present case, it seems probable that the instrumental readings are better defined than the compositions of some of the test samples. The scatter of the points secured Kith a given matrix gas, observable in Figure 3, probably reflects an indeterminacy in the latter sense. Given occasionaI checks of the zero point, it appears that the sensitivity of this method approximates 0.0005 volume per cent of water, and that the reproducibility is of the same order of magnitude. The large scale plot of the results obtained with low concentrations of water vapor in an oxygen matrix, displayed as an inset in Figure 3, countenances such an evaluation. The nature of the measurements is such that the accuracy should be fairly uniform throughout the accessible range of concentrations. Calculation of the thermal efficiency of the system, based on a comparison of the theoretical and observed temperature rises, yielded results which, although they varied with the nature of the matrix gas, were of the order of 35%. A thorough examination of the origin(s) of this relatively meager thermal recovery indicated that a major part of the heat leakage was due to a reverse flow of gas, partially cooled by contact with the upper, unjacketed portion of the reaction cell over the hot thermistor. When the upper (hot) thermistor was inserted more deeply, so that it was almost in contact with the calcium hydride mass, the thermal efficiency mounted from 35% to close to SO%, a reasonable approach to the 80% thermal efficiency reported by Cohn in a related application of similar apparatus. In his work the hot junction of a thermocouple was imbedded in the downstream end of the catalyst mass in which the heat was liberated, whereas in the authors' system the hot thermistor was normally some 3 em. distant from the bed of calcium hydride. The calibration data, plotted in Figure 3, indicate a relatively weak dependence of the experimental findings on the nature of the matrix gas. However, with a given water concentration, the various temperature rises should be inversely proportional to the heat capacities of the various matrix gases; Cohn observed a t least a qualitative approach to such proportionality (1). However, the agreement was far from exact, and he attributed this failure to heat leakage from the reaction cell. The apparent heat leakage from the equipment described herein was much greater than that noted by Cohn. Furthermore, because of the character and magnitude of the heat losses in this system, the measured temperature rise with any matrix gas cannot be a simple (inverse) function of its specific heat but, rather, a complex (and undetermined) function of its specific heat, heat conductivity, viscosity, etc. On the not unreasonable assumption that this function is such that all the test gases behave much the same, the close similarity of the calibration curves obtained with different matrix gases is readily understood. The thermal economy of the iystem could easily be improved by using small thermistors that could be imbedded in the calcium hydride, by extending the vacuum jacket to cover the upper part of the reaction cell, and/or by using a top-to-bottom direction of gas flow, the plan adopted by Cohn, so that natural convection would not tend, as it did in the authors' system, to cause a recirculation over the hot thermistor. These changes might be advantageous if it were desired to extend this method to lower concentrations of water vapor, if a matrix gas of perfectly stable composition were involved, or if the autocompensatory system described below were used. However, these changes seemed decidedly disadvantageous in connection with the batch analyzer already described, because it seemed plain that the existent heat losses were chiefly responsible for the similarity of all the calibration curves, Thus, in accepting a relatively poor thermal economy, there is obtained a favorable situation in which only limited calibration, especially in the range of low water concentra-
V O L U M E 23, NO. 5, M A Y 1 9 5 1
139
tion ranges, is required; and in which, in all accessible concentration ranges, the results are relatively insensitive to fluctuations in the composition of the matrix gas. For routine semicontinuous analysis of the water content of streaming industrial gases, a relatively simple apparatus can be used. 9 linear train of the following components promises to furnish the desired resulk
1. A flow controller, consisting of a diaphragm valve, to control the gas pressure at the inlet, and an orifice or capillary to govern the rate of flow under this pressure. 2. A small flowmeter to show that the flow is of the correct order of magnitude. The flow need not be closely controlled. 3. A reaction cell and measuring bridge, as shown in Figure 1. 4. A saturator, consisting of an intimate mixture of a salt hydrate and its anhydride, held in a constant temperature (ice) bath. The chosen hydrate should furnish an aqueous tension of the same order of magnitude as that to be measured. The output of the saturator must be accurately reproducible, but it need not be a perfect equilibrium mixture. The actual percentage of water in the hydrated gas can be determined &ally by an absolute (gravimetric) method. 5 . A second cell exactly like the first. (3. A guard tube containing a desiccant and leading to the vent. The null readings of both cells ought first to be determined by passing a perfectly dry gas through the train with the saturator
temporarily by-passed. The train can then be reconstituted and used without any further calibration; the data supplied by the second cell serve as a reference standard in terms of which the readings of the first cell can be interpreted. The calibration curves are essentially linear. Therefore the reading of the second cell with the gas of known water content delivered from th? saturator, taken together with the null reading of that cell, serves to define the calibration line appropriate to the composition and flow rate of the gas passing through the system. With this calibration line the original water content of the test gas can be easily and uniquely determined from the reading of the first cell. ACKNOWLEDGMENT
Much of the preliminary work in the development of this analytical system was carried out by J. A. and Peter Kafalas. LITERATURE CITED
(1) Cohn, Gunther, ANAL.CHEM.,19,832 (1947).
( 2 ) Greenhill, E. B., and Whitehead, J. R., J . Sci. Insfrunients, 26,
92 (1949).
(3) Metal Hydrides, Inc., Beverly, Mass., Pamphlet, "Calcium Hy(4)
dride.'' Walker, A. C., and Emst, E. J., Jr., IND.ENG.C H m f . , .IN.u.. ED., 2, 134 (1931).
RECEIVED September 7,
1950.
Precision Multiple Sorption-Desorption Apparatus W. 0. MILLIGAN, WARREN C. SIMPSON', GORDON L. BUSHEY2, HENRY H. RACHFOKD, JR.3, AYD ARTHUR L. DRAPER The Rice Institute, Hotiston, Tex.
A precision multiple apparatus has been constructed which is capqble of determining fifteen sorptiondesorption isotherms or isobars simultaneously, thus permitting the rapid accumulation of accurate sorption data. The apparatus is based on the gravimetric principle. Weight changes are followed by silica spring balances. A micrometer slide and microscope are moved into successive spring balance positions by means of a hand-operated or motordriven screw. The reproducibility of the readings is independent of the screw mechanism. Extensive evacuation and degassing ensure removal of air from
T
HE determination of the amount of adsorption of gases or
vapors on solids has usually been carried out by volumetric methods. Gravimetric methods possess the advantage of permitting the simultaneous examination of several samples, and do not require a determination of the volume of the adsorbent. Since the classic experiments of McBain and Bakr (8) who first employed fused silica spring balances, the gravi. metric method has come into widespread use. A multiple apparatus consisting of six McBain-Bakr balances has been described by Stamm and Woodruff (1.9). In this device, each of six silica spring balances may be rotated in turn into position before a common measuring microscope. Harris, Ott, and Arnold (2) have discussed a similar apparatus. In this laboratory there has been constructed a multiple sorption-desorption apparatus capable of following fifteen samples Present address, Shell Development Co., Emeryvilla, Calif. Present address, Department of Chemistry, University of Illinois, Urbana, Ill. 8 Present address, Humble Oil and Refining Co., Houston, Tex 1
2
the samples prior to the introduction of the sorbate. Equilibrium points are attained at a temperature constant to better than 0.001' C. and at prwures determined to within 0.007 mm. of mercury. Isotherms and isobars are usually obtained in the temperature range -20' to 60" C. and at pressures from 10-6 to 400 mm. of mercury. The apparatus is especially designed for the examination of hydrous oxidewater systems but is not limited to these materials. Adsorption and desorption isobars and isotherms have been obtained for over 500 samples, representing 15,000 equilibrium point readings.
simultaneously. Silica spring balances are used, and the measuring microscope is moved from balance to balance by means of a translatory motion, employing a screw. The apparatus is designed to yield accurate data in quantity with speed and convenience in over-all operation. APPARATUS
Fifteen fused silica springs are mounted in vertical positions in individual borosilicate glass tubes suspended in a thermostatically controlled bath. The springs employed vary in sensitivity from 0.3 to 40 mm. per mg. Each sample tube is approximately 1 meter in length, in order t o allow the use of the high sensitivity springs, and also to allow the springs to operate a t room temperature. This latter feature is essential to prevent the continuous sensitivity variations observed when silica springs are exposed to high pressures of water vapor a t elevated temperatures. Also, the springs exhibit a small negative temperature coefficient; in this apparatus, temperature corrections are employed. Figures 1
