INDUSTRIAL AND ENGINEERING CHEMISTRY
358
Assume that air enters a dryer at 80" F. dry bulb and 68% relative humidity (0.7 inch of mercury, 68"F. dew point,). It is desired to dry the air to a maximum dew point of 15" F. (0.1 inch of mercury). This corresponds to the removal of 6.5 grains of moisture per cubic foot. At the end of the adsorption cycle, the gel at the inlet of the dryer i d 1 be in equilibrium with the entering air, and will therefore contain 32.6% adsorbed water (read from 0.7 inch of mercury isopiestic at 80" F.), The gel at t,he exit, end will be in moisture equilibrium with the dried air, but, a t a temperature calculated to be 80" plus (10" X 6.5), or 145' F., since silica gel is heated about 10' F. for every grain of moisture adsorbed per cubic foot of air at atmospheric pressure. On the 0.1 inch of mercury isopiestic. this corresponds to 1.6% adsorbed water. The useful capacit'y will then be (32.6 1.6) / 2 , or 17.1%. If, say, t-he air is to be passed at 100 cubic feet per minute for 8
+
Vol. 46, No. 2
hours, the quantity of wat,er t o be removed is 48.000/7000 X 6.5, or 43.9 pounds. Dividing this by 0.1711we obtain 257 pounds of gel as the required quantity. For long-term use (see above) t'he quantity of gel would be arbitrarilj- increased by about lo%, t o give 283 pounds as t'he amount required. LITERATL-RE CITED
(1) Dehler, F. C . , Ciieni. and M e t . Eng., 47, 305 (1940). (2) Patrick. W, A,, Colloid S~ymposiu7nAnnual, 7, 192 (1930). (3) Perry, J. H., "Chemical Engineers' Handbook,'' 3rd ed., p. !I1 2, Yew York. h1cGraw-Hill Book Co., 1950. (4) Taylor, R. K., IND.ENG.CHmr., 37, 649 (1945). RECEIVED for review June 21, 1953.
ACCEPTED October :i, I!).%.
J TT, T. GRANQtTIST, E'. A. MITCH, .IND C. H. EDV-ARDS Flnridin Go., Warren, Pa.
T
HIS paper deals with the adsorption of C1, Co, Ca, and n-C1 saturated hydrocarbons on Florida-Georgia fuller's earth ( floridin, an aluminum magnesium silicate mineral, average mineral analysis: 75% attapulgite, 10% montmorillonite, 5% free silica, 10% calcite, dolomite, etc.) found extensively in Gadsden County, Florida, and Decatur County, Georgia. An understanding of the availability of the adsorbent, surface to molecules of increasing size was obtained. This information could then be used t o determine t,he applicability of the Methanite g s process (4, IO), an above-ground lowtemperature adsorption storage system, to hydrocarbons other than methane. Earlier papers (3,4 ) described the adsorption of nit,rogen and methane on the same adsorbent. The same sample WE used for all the adsorption work: but a s h d y of the results listed below. together with those in t,he previous publications, shows that the two sets of results are not' directly comparable. This seems t'o be due to a loss in surface area of the sample studied, with essentially no accompanying change in gas saturation volume. A nen- nitrogen adsorption isotherm was determined. and the area and saturation volume were calculated, Ivith these results: 86.4 square meters per gram and 353 ml. (standard temperature and pressure). The earlier work on t,he same sample gave a surface area of 128 square meters per gram, and a saturation volume of 352 mi. The apparatus and operator techniques were checked by determining the area of a sample of Spheron furnished by P. H. Emmett, on which he reported a surface area of 120 square met,ers per gram (7); the authors' result on this sample was 119 square meters per grain. For this reason, the resuks discussed include nitrogen adsorpt,ion, t o present a picture of the comparative adsorpt'ion of all five gases on the same surface. Gas to gas comparisons should be valid, and the clay characteristics determined should be proportional to the values for a fuller's earth of higher surfacc area. EXPERIMENT -iL
The gas-adsorption apparatus, essentially that described by Emmett (8), and techniques employed have been adequately described in the earlier publications (3,4 ) . The only modification concerned the temperature-control system; this consisted of a bath of boiling adsorbate at atmospheric pressure in the case of nitrogen and methane, and of slush baths for the other adsorbates studied. Because these slush baths could result in serious temperature variations in the sample, the sample tube was fitted into one hole of a bronze block, containing two holes symmetrically located. The other hole contained the bulb of the vapor pressure thermometer. Care was taken to obtain a close fit between the glass bulbs and the wall of the block, and any
aiinular space was filling with mercury. Agitation was carried out by a reciprocating stirrer operating close t o the outer ivall of the block. Thus, the temperature noted on the vapor pressure thermometer should be that of the sample. Compounds iuscd for the baths TTere met,hyl ethyl ketone (melting point, -86" Cy,)! monochlorobenzene (melting point - 45' C.)>and water (melting point 0" C.), for ethane, propane, and n-butane, respect,ively. In many cases, the adsorption was cont,inued until duplicat,c points m r e obtained a t PO. The gases used were obtained from the following sources: high purity nitrogen, Linde Air Products Co.; methane, research grade, Phillips Petroleum Co., reported purity, 99.31%; ethane, research grade, Phillips Petroleum Co., reported purity, 99.78%; propane, research grade, Phillips Petroleum Co., reported purity, 100%; n-but,ane, c. P., Ohio Chemical and Manufacturing Co.! reported purity, 99%. No attempt \vas made to incrraso thc purity of these gases prior to their use. The sample of fuller's earth was the same as that used in t>hc xork xith nitrogen and methane (3,4): It had been extrudccl (a process involving auger-driven plastic flow a t high prcwurc through a suit>able die plate to disrupt the natural clay st>rurture), screened to 30/60 mesh, and thermally activated in air at, 649" C. (1200' F.) for 1 hour prior to use in the adsorption program. The sample was outgassed a t ea. 10-5 mm. of mercury at 200" C. ininiediiitely before an actual adsorption run. 1)1 YCL SYION
Table I and Figures 1 to Ipresent the experimental results OC this investigation. The five adsorption isotherms shown in ihc curves are representative of Type I1 isotherms, and all exctcpt nitrogen possess narrow hyPteresi3 loops. These isotherms n'crc used to determine the monolayer volume, T,i and t,he saturation volume, V,. The value of 8, was obtained by t'he use of the Bruiiauer-Emmet,t,-Tellel equation; V , mas determined by not,ing the point at which PIP0 = 1 . 111 gas volumes were expre as milliliters at standard temperature and pressure. Molecular areas for the different adsorbates must be knowti for any calculation of surface arc&. For these, t>heequat,jon tirea = (4) (0.866) (.1[/4 ~ / % A D ~ ) Z / ~ vias used, where V is molecular might. A is Avogadro's number, and D is adsorbate density at the boiling point. The molecular areas obtained in this way xere then used in conjunction with the monolayer volumes to obtain B E T surface areas. The surface areas for the hydrocarbon gas adsorbates were fairly constant and consistently lower than the nitrogen area. This was considered as being indicative of incomplete surface coverage on the
INDUSTRIAL AND ENGINEERING CHEMISTRY
February 1954
359
I 40
/I:
1-
.-
8
lf
40
10080 -
60-40-
20-
I
0 0
01
I
02
03
05
04
07
06
08
10
09
P/ Pn
Figure 1 .
P/Po
Adsorption o€ Nitrogen
011
Fuller’s Earth
0 Adsorption points
9 Adsorptjon and
desorption point*
meters/e. . ...,
Nz 0.804
CHI 0.424
(6)
(1)
C?Ha 0.456 (1)
CaH8 0.582
n-C1Hlo 0.579
(1)
(1)
37.7 17.2
30 54.9 22.1
44 75.6 27.1
58 100.1 36.2
. .
21.3
26.8
35.8
47.0
19.7 353 0.548 17.9 86.4
13.0
I 1 .9
322 0.542 21.5 69.8
217 0.533 18.2 71.0
8.92 174 0.587 19.5 66.0
113 0.505 16.6 66.2
155
150
150
153
28 34.8 16.2 ,
16
6.80
E
R W . (S .~A . X I O ~ )127 , A. % coverage b y mono- 100 layer Hysteresis ? N O Experimental Pa 768.5
80.8
Yes 702.6
82.1
Yes 794
0 Desorption points
0 Adsorption points
TABLE I. SGMMARY OF EXPERIMENTAL RESULTS Density a t boiling point, g./ml. Mol. wt. Mol. vol., ml. Mol. area, calcd. from density, sq. A. Mol, area, method of h-ay and Morrison (9),sq. A. V m (gas), ml. STP Vr (gas), ml. STP V S (lis.), ml. Va (gas)/Vm (gas) Surfacearea, BET,sq.
Figure 2. Adsorption of Methane and Ethane on Fuller’s Earth
76.4
Yes 655
76.4
Yes 763.7
first layer, and so per cent surface coverage was calculated for each gas from the ratio of hydrocarbon gas area to the nitrogen area. Kay and Morrison (9) suggested that values equivalent to nitrogen areas could be obtained from other gases by using a molecular area calculated by
’
( N z ) X 16.2 sq. A., where 16.2 Vm (gas)
sq. A. is the commonly accepted cross-sectional area of the nitrogen molecule ( 6 ) in low-temperature adsorption work. As a matter of interest, such values were determined. The pore volume, V 8 (liq.), was computed by correcting the saturation gas volume to the corresponding volume of liquid nitrogen a t the boiling point. The density values used were obtained from the literature (1, 6). The average pore radius was
2zq) s.
X lo4,$.,for the adsorpA. tion of each hydrocarbon gas. The number of monolayers of adsorbate, the ratio of the saturation volume to the monolayer volume, were also calculated for each gas. Finally, the molar volume of each adsorbate a t its boiling point was obtained by dividing the molecular weight by the density a t the boiling point.
calculated from the relationship
360
INDUSTRIAL A N D E N G I N E E R I N G C H E M I S T R Y
c t-
4 0
L
I
13 \ W 3
m a:
-1 -I
~
i
230
>
Vol. 46, No. 2
200 --
i
-1
I-
80
I
rp
ki
I
e?
40L
p0
B
20
2
a
t Lo
4-
Q
(3
i 40k20,-
PROW
P/ Po
F/ Po
Figure 3. Adsorption of Propane on Fuller's Earth 0 Adsorption points
Figure 4. .idsorption of n-Butane tan Fuller's I k t r i h C Adcnrption point5
@ Desorption points
Desorption points
The loss of surface mentioned earlier was a cause of considerable perplexity, but should not great,ly affect the results, since saturation volumes were not affected and clay characteristics related to the monolayer volume should be proportional to the value- for a fuller's earth of higher area. The earlier xork on the same sample ( 4 ) gave a surface area of 128 square meters per gram arid a saturation volume of 352 nil. for nitrogen adsorption; the present results are 86.4 square meters per gram and 383 nil. The earlier paper on methane adsorption (3) list,ed a surface area of 74.0 square meters per gram, based on the m e a of the methane molecule calculated from the liquid density, and a satupation volume of 298 ml.; current values are 69.8 square meters per gram and 322 nil. Close correspondence of Ti, (liq.) for all gase9, toget'her with the drop in nitrogen area, means a greater effective pore "radius"-calculated
from ___ Va(liq.)
s.A.
x
101, as 85 A. for
original nitrogen data and 127 *I,for the present work. f ~ ' . in ~~ t,he present paper is approximately constant, a t 150 A. for rhe hydrocarbons. Thus, greater effective surface coverage at V,, should have been effected for methane than in the earlier xork, and this is confirmed experimentally, since the first methane coverage value was 65%, as compared with a fairly const,ant value of ca. 80% for all the hydrocarbon g a s . Because the extrusion process disrupts the well-defined floridin crystallites into a randomly distributed mass of broken fragments of crystallite, the loss of area may have been due to crystallite growth and r e x rangement during the long period of time since the original sample preparation. High temperature damage of the surface is a logical explanation, but careful survey of experimental proccdure does not support this view. Figures 5 and 6 indicate the variat,ion of monolayer volume and saturation volume with molecular area and molar volume, respectively. For the variation of V,, the areas calculated both from the density and by the method of Nay and Morrison have been used. The density area cuive represents the 80% surface coverage, while the Nay and Morrison curve represents a coniplete monolayer coverage, since the molecular areas have becn
increased to a value nerePssry for this condition. Using thr: density areas and a V , corrected to complete surface coveragc as described below ~ ~ o uresult ld in a curve coincidental with t h e Nay and Morrison curve: but, terminating c\t the density area for butane. At any rate, it ie obvious that the monolayer volume clecreases regularly with increase in molecular area of adsorbate, as would be expected. Hnbuitively, it is espect,ed that some 1 ' ~ ' lntionship should exist between the saturation volume and the molar volume. Actually, a very real relationship does exist, as v (gas) X -- where can be shown by the follominy: V , (1iq.i = 22,400 DL JI is t8hemolecular weight md D L is the density of the adsorbate at' t'he boiling point. But, M,'DL is the molar volume (Vmo,ar) of the particular adsorbate, and 22,400 is a constant, ,so that l y e
x
Lr)
(lis.) = ___ X. T-,,,. 9'
However, in the series of gases
studied, the values of T7* (liq.), or pore volume, are e,sentially constant, indicating that the pore volunie of the sample is conpletdy accessiblc to the different, gasmolccules. Thus, 8, (liq.) =:
K
X V m o i a r , or -= Ti, (gas), where k, VXIlOlB, IC = Til k,. A plot of T', (gas) against l/Vmaiai should tre linear. with the slope equal to t'he product of the pore volume and molar gas volume. Figure 6 sho\\-e good linearity, with the exception of the propane point, and s n evaluation of the pore volume lrom the slope of the curve give,! a value of 0.572, compared with a11 average of 0,843 ml. for the five adsorbates studietl. It is interesting to note the approximate constancy of t,he surface arca values for the hydrocarbon gases and of the per ccnt cowxage based on nitrogen as 100%. It is probable that thc filling of 'the monolayer is riot complete at V, obtained from BET equation, and that the .arface does become covered at some V/Vm>1. The use of niolecular areas calculated by the Nay and Morrieon method would give 100% coverage, but these larger values for molecular area, m e hard to associate with the €act that Tr,/V, is :w large BE or slightly larger than V J V , for nitrogen. h corrected value of T',n was calculated by dividing the
lis, and k2 =
INDUSTRIAL AND ENGINEERING CHEMISTRY
February 1954
361
2 0 s The presence of hysteresis loops in the adsorption isotherms for the four hydrocarbon gases, but no hysteresis in the case of nitrogen adsorption, is difficult to understand. It is realized that few desorption points were taken in the nitrogen work, hut such points extend to a low enough relative pressure to eliminate the possibility of hysteresis. Furthermore, the absence of hysteresis in nitrogen adsorption checks the results of the earlier work ( 4 ) . KO explanation is offered for the difference in behavior. However, F. L. Shea, Jr., has noted that the pressure difference across the hysteresis loop increases with molecular size, a point deserving further investigation. For maximum capacity in an adsorption storage system such as the Methanite process, the total capacity is not the point at which the isotherm reaches PO. Rather, the storage would continue at Po, until the gross pore volume of the adsorbent was conipletely saturated and the particles appeared wet. Actually, commercial practice would be at constant pressure and variable temperature, and the final soaking operation would occur at this normal pressure. The amount of physically adsorbed hydrocarbon corresponds to approximately 60% of the total capacity (established in pilot plant operation, IO), and the additional amount ia simply the condensation of adsorbate into the gross pore space of the adsorbent. Therefore, saturation volumes by the method described should provide valuable information relating to the eventual behavior of the adsorbent in actual storage operation, and the data for the various hydrocarbons should also he of value in predicting the behavior of gas mixtures. Table I11 lists values calculated from Table I for the storage capacity of the adsorbent, and estimates total storage capacity. A suitable equation describing the adsorption of these gases on the fuller's earth surface would be of great utility in design operations. Hill (2) has discussed an equation deduced by Halsey
14
i 0
A
- MOL
AREA FROK DENSITY RELATIONSHIP
- MOL AREA
CALC EQ NAY h MORRISON
2
In N = -a/ra 1
1
/
I
,
/
/
I
!
where z = P / P o , r is the surface concentration, and a and s are constants. This isotherm has been developed to cover the adsorption of spherical, symmetrical molecules on a nonporous, nonpolar adsorbent for 8>2, where e = V/V,. The adeorption of gases such as Cz, C3, and C4 on the fuller's earth surface violatgs
experimental V , by the per cent coverage, and using the corrected value, V , (corr.) to obtain V J V , (corr.). Table I1 shows these results. Thus, by assuming that values of molecular area from density are correct, and that surface coverage is complete only a t V/V,>l, a V , corrected to give complete coverage yields V J V , values that are in line with the increase, over the nitrogen molecular area, of the molecular areas of the other adsorbates.
I I
__
80-
-1
601-
,
P 4 0 F
20--
300-
OF V , ASD V,/V,, TABLE11. CORRECTIOK COVERAGE Nl c1 cz
v,, V,
(gas)
O& coverage Vm (oorr.) V./Vm (oorr.)
19.7 353 100 19.7 17.9
15.0 322 80.8 18.6 17.3
11.9 217 82.1 14.5 14.9
TO
TOTAL SURFACE CS
8.92
174
76.4 11.7 14.7
c4 6.80 113 76.4 8.90 12.7
TABLE111. STORAQE CAPACITY OF FULLER'S EARTHFOR HYDROCARBON GASES
c1
Cl
cs CC
Standard Cu. Feet Gas/ Cu. Foot Adsorbent 165 111 89 57
Est. M a x . Standard Cu. Feet/ Cu. Foot Adsorbent 315 211 169 108
Gross Heat Value Stored Est. Max. B.t.u./Cu. FoAt B.t.u./Cu. Foot Adsorbent Adsorbent 336,000 176,500 392,000 206,500 236,000 448,000 196,000 372,000
+
6
20
c
~402 ' ' 20 -
'
/
0
Jl
I
l
'
'
l
'
' ' ' ' 02
03
INVERSE OF MOLAR VOLUME
Figure 6. Variation of Gas Saturation Volume with Molar Volume
I N D U S T R I A L A N D E N G I N E E R I N G CHEMISTRY
362 T.4BLE
Iv. EVALUATION OF k AND S
FOR
V, = volume o f gas at STP adsorbed at saturation HYDROCARBON V, (liq.) =: volume of liquid adsorbate a t boiling point.
A4.DSoRPTIoN D A T A
In
5
= -k/Rs where z = P/Po, B = V / V m
k
Vol. 46, No. 2
s 1.8
1.6 1.3 1.3 1.6
these conditions, and the value of 8 at saturation is of the order of 10, demanding agreement over a wide range of 8. The above equation in the form In z = - X ’eq where 8 is V / V , as above, and k and s are constants, has been applied to the authors’ data, and k and s have been evaluated for the different adsorbates. The results are listed in Table IV. A close fit to the experimental data was obtained up to 0.975 relative pressure. Hill (2) indicated that values of k S 4 should be of the right order of magnitude; this waq not true in this instance. No significance of the values of k and s obtained is suggested. ACKBOW‘LEDGMENT
The authors wish to acknowledge the valuable experimental assistance given by H. J. Streich and Anthony Rock, of the Floridin Co L. B. Christie, also of this laboratory, prepared the curves. The authors are also grateful to P. H. Emmett of Mellon Institute for his review of the early phases of this study. N0;VENCLATURE
Vm = volume
of gas a t S T P adsorbed at monolayer coverage V, (corr.) = monolayer volume corrected to give coverage equal to nitrogen
Corresponds to pore volume Vmraalar = volume of I mole of liquid a t its boiling point and 760 mm. mercury pressure P = equilibrium gas pressure PO = pressure a t saturation, corresponding t o vapor pressure of adsorbate at temperature employed DL = density of adsorbate a t boiling point X = molecular weight = geometrical pore radius. A 2v ( h . 9 R,, s.A . A = Avogadro’s number 0 = V/V,, number of molecular layers a t any P r = P/Po F = surface concentration k , s, a = constants S.A. = surface area LITERATURE CITED
(1) Egloff, G., “Physical Constants of Hydrocarbons,” Vol. I, New York, Reinhold Publishing Corp., 1939. (2) Frankenburg, W. G.. Komarewaky, V. I., and Rideal, E. K.,
eds., “‘Advances in Catalysis and Related Subjects,” Vol. IV, New York, Academic Press. Inc., 1952. (3) Granquist, W. T., ISD. ENG.CHEM..42, 2572 (1950). (4) Granquist, W. T., and Amero, R. C., J. Am. C h e m Soc., 70, 3265 (1948). (5) Harris, B. L., and Emmett. P. H., J. Phys. R. CoZZoid Chem., 53, 811 (1949). (6) Hodgman, C . D., ed.. “Handbook of Chemistry and Physics.” 31st ed., Cleveland, Ohio, Chemical Rubber Publishing Co.. 1949. (7) Joyner. L. G., and Emmett, P. E.,J. Am. Chern. Soc.. 70, 2353 (1948). (8) Kraemer, E. D., ed., “ddvances in Colloid Science,” Vol. I, New York, Interscience Publishers, 1942. ,and Morrison, J. L., Can. J , Reseaich, 27B,205 (1949). (10) Bpangler, C. V.,Bodle, W. W., and Granquist, W. T., Oil Gas J., 48, 170 (April 20, 1950). RWBIVEDfor review August 13, 1953.
ACCEPTED December 10, 1953.
ressibility Factors ne Mixtures i GEORGE BI. \TATSOAT, A . B. STEVENS, R . B. E V A 3 3 111, AND DOS HODGES, JR.] Agricultural and Mechanical College of Texas, College Station, Tex.
ITROGEN is an important and common component of industrial gaseous mixtures. Adequate tabulations of experimental data on the pressure-volume-temperature behavior of pure nitrogen (1, 6. 8, 16) and of pure propane (4,12, 16, 16) are available in the literature. Literature references for fluid mixtures with propane as one of the components are given by Sage (14), but no references are found for nitrogen-propane mixtures. Xtrogen-methane mixtures have been investigated by Keyes and Burks (11), and the Beattie-Bridgeman equation of state has been tested for mixtures from these measurements ( 2 ) . Nitrogen-ethane data have been presented by Reamer et al. (19). Pressure-volume-temperature relations for nitrogen-ethylene mixtures have been reported by Hagenbach and Comings (9). To date, the authors are unaware of any previous experimental measurements on nitrogen-propane mixtures. The present investigation is the first of a series of pressurevolume-temperature studies on nitrogen-hydrocarbon mixtures contemplated in this laboratory. 1
Preeent address, The Texas Co , Bellaiie, Tex.
IIATERI h L S
The 99.9901, C.P. grade propane used was obtained from the Mathieson Co., East Rutherford, K.J. The nitrogen was acquired from the Linde Air Products Co., East Chicago, Ind., and it was, according to the manufacturer’s specification, 99.99% pure. The purity of these gases Tyas checked by means of the mass spectrometer and found to be pure within the experimental uncertainty of the analytical procedure. APP4RATUS AND PROCEDURE
A modified Burnett system ( 7 , 1 7 ) was used in this investigation. Briefly the apparatus consists of an Amagat-type dcadweight pressure gage, expansion bomb and diaphragm, thermostat, injection system, and sampling system. The dead-weight gage is similar to that described by I l e p s (10). The piston-cylinder combination for the gage was built by the Pratt and Whitney Co., WePt Hartford, Conn., and meets the specifications enumerated by Beattie ( 3 ) . The gage constant was determined by calibrations against the vapor pressure of liquid carbon dioxide a t 0’ C. This method is described by Bridgeman (6). The value of the vapor pressure of liquid car-
