I
JAMES R. WHITE and ALBERT T. FELLOWS Research and Development Laboratory, Socony Mobil Oil Co., Inc., Paulsboro, N. J.
Thermal Diffusion Efficiency and Separation of Liquid Petroleum Fractions bb
Molecular configuration and not equipment design may govern heat requirements for separating liquid petroleum fractions by thermal diffusion
G R O W I N G INTEREST in commercial separation of liquid mixtures has resulted in several recent studies of thermal diffusion (2, 70-72, 7 4 , most of which have neglected efficiency. Components differing primarily in molecular shape rather than characteristics such as molecular weight, volatility, or polarity are separable by this process, and hydrocarbon mixtures, especially petroleum, are promising material for yielding to such shape selection. The process requires dissipation of large amounts of heat, which limits its commercial application. T o determine if this requirement results from the molecular nature of the process or from inadequate apparatus design, measured thermal efficiencies have been compared with limiting thermodynamic efficiencies. Two vertical diffusion columns were constructed to measure heat requirements for separating two binary test mixtures. Independent cell measurements of the ordinary diffusion coefficients and thermal diffusion constants made possible estimates of limiting thermodynamic efficiencies. Thermal efficiencies for the column operations compared favorably with the estimated limiting thermodynamic efficiencies. These columns were then used to measure heat dissipation required to separate several complex petroleum fractions, including distillate and re1 Middle Eastern lubricating oil ons, naphthenic Coastal distillate oil, a Coastal residual oil, and paraffin scale wax. To interpret some of the work and facilitate extrapolating from laboratory to commercial scale, several simple cascades were computed, illustrating the principle that partitioning or baffling alone is incapable of altering thermal efficiencyin thermal diffusion columns.
Efficiency Relationships The mathematical theory of thermal diffusion, particularly in vertical columns, has been developed by Furry,
Jones, and Onsager (7), and summarized by Jones and Furry (73). Ideal Binary Mixtures. The energy that must be supplied to separate 1 mole of an ideal binary mixture into two new mixtures each of an equal number of moles and differing in mole fraction composition by A is -XI
In
xi
- x2
In xz
+
where XI and x2 are mole fractions in the unseparated mixture, and T is a mean temperature. When A < ~ 1 x 2 Equation , 1 can be accurately approximated by 1 - ha F = - RT8
XlXZ
Equation 2 is a reasonable approximation, even u p to A = X I X Z when 0.5 mole fraction mixtures are separated. Hydrocarbon mixtures of the type studied here are nonideal and generally exhibit positive deviations from Raoult's law. Separation of typical mixtures such as cetane-naphthalene, when computed by the Hildebrand-Scatchard method, will require a n energy supply, roughly 75% of Equation 2. If the mixture is separated at a rate of u grams per second, the rate a t which energy must be supplied' is
where li;i is a mean molecular weight. When liquid mixtures are subjected to thermal diffusion, operating conditions usually imposed assure that virtually all heat is transported between hot and cold surfaces by conduction. Mass flow a t the ends of the apparatus and heat lost to effluent liquid streams accounts for only a small fraction of the
heat that must be supplied. The temperature gradient and dimensions of the apparatus, together with the heat conduction coefficient of the liquid being separated, then determine the heat transport, Q, thus,
where L and B are length and width of the apparatus, and 2w is distance between hot and cold surfaces at temperatures Tz and T I . For most hydrocarbon systems the heat conduction coefficient, A, varies in a narrow range around 3 X loF4 cal. per sec.-cm.-deg. Dividing Equation 3 by Equation 4 leads to an energy efficiency,
Onsager (75) has shown that when a small volume of gas mixture is subject to a small temperature gradient which achieves a small separation, the entropy inequality applies
where 8, is rate of entropy decrease during separation; s'x, rate of entropy change due to heat conduction; a, thermal diffusion constant; D12,ordinary diffusion constant; and P,pressure. Expression 6 has been rewritten as an energy efficiency. This requires substitution for the rate of entropy decrease during separation, - l / T x dF/dt, and for the rate of entropy production in heat flow, ( 7 ' 2 - T l ) / T 2X dQ/dt, ( 3 ) . Also, PRT/;i? has been included in a gas-law substitution for pressure. Thus,
where 5 is a mean density. This expression, originally derived for gases, is also applicable to liquids. Equality in Expressions G and 7 refers to the particular condition in the thermal diffusion process when the flux VOL. 49, NO. 9
SEPTEMBER 1957
1409
FLOW IN
due to thermal diffusion is twice the ordinary back-diffusion. The existence of an upper efficiency limit at this condition is a consequence of imposing continuity conditions on the equation of transport (75), Flux (1 ) = - ~ D Igrad. Z XI LYPDIZ XIXZ grad In T (8)
UO;
CONCN.
Xg
t
+
This upper efficiency limit is a thermodynamic limit, independent of mechanical design for separation apparatus. When operated under the named restrictions, various designs such as thermogravitational columns, horizontal columns, and packed columns may approach but not exceed this limit. Use of Expression 7 for liquids implies that this equation of transport describes liquid behavior. Equation 5 is an experimental efficiency requiring no knowledge of irreversible process thermodynamics. Expression 7 is a limiting thermodynamic efficiency when temperature differences and separations are small and is determined by the diffusion coefficients. The adequacy of an experimental apparatus and operating procedure has been judged by comparing the best obtainable efficiency, Equation 5, with the limiting efficiency, Expression 7. Making this comparison requires independent measurement of the thermal diffusion constant and the ordinary diffusion coefficient.
Simple Cascades The most efficient operation of a particular thermal diffusion apparatus, measured by a maximum in the quantity, aA2 of Equation 5, does not necessarily permit the desired separation to be made in that apparatus. If a larger separation is needed, a higher column may be constructed, or clearance bemeen plates may be reduced, or a cascade of columns may be employed. This latter may prove the more attractive approach. Subject to discussed restrictions, that apparatus which can be operated to yield the highest Equation 5 efficiency, will also when constructed into a cascade of similar units, yield the highest over-all process efficiency. This is a principle implied in the equations of Furry and Jones and is also derivable directly from entropy of mixing considerations. The simple “constant A’’ cascade described below illustrates the principle. I n a constant A cascade, each stage effects the same separation, 4, between equally divided streams leaving the stage. Binary systems in Tvhich the mixture and product are not far removed from 0.5 mole fraction can be separated in cascades approximatelysatisfying the constant A restriction. The inflowing stream of the constant 4 cascade, uo, is divided equally into product streams of volume ao/2 exiting
1410
w-‘
P
FLOW O U T
3
FLOW OUT
L
7
In this constant 4 cascade, thermal efficiency of a single unit is identical to that of the cascade; n stages; [(n 1)/2] units; per stage separation 4; over-all separation [ ( n l )/2]A
+
+
from final stages. For unequal division of product streams either branch of the cascade may be extended unsymmetrically. Because the volume that must be handled by each stage of the cascade varies, units comprising each stage are paralleled as shown. If A is the separation made in a single stage, the separation made by the cascade \vi11 be [(n 1)/2] A, where n is the number of stages. The number of units in the cascade, each of which dissipates the same heat, Q5 is [ ( n 1)/212. Since
+
+
thermal efficiency of a single unit and that of the cascade are identical. Because the cascade adopted avoids mixing of streams of different composition, no other assembly of similar units can achieve higher efficiency. Thus, the cascade can be regarded as an ideal assembly of smaller units. Conversely, the cascade may be regarded as ideal partitioning of a large unit into smaller units. I n each case, assembly or subdivision does not of itself alter thermal efficiency. An interesting illustration of this principle appears in the vertical column theory of Furry, Jones, and Onsager. Vertical columns with continuous transport of material through the column are characterized by a single, best combination of forward transport, a: and separation, A, for each height and heat dis-
INDUSTRIAL AND ENGINEERING CHEMISTRY
sipation. However, when height and transport alone are varied and concentrations are close to 0.5 mole fraction, the best combination satisfies the condition that crA2/height is constant.
Apparatus
Two vertical diffusion columns were employed. In each column, liquids were contained between parallel plates ~vhich served as heat reservoirs. The smaller of the columns possessed an operating area 22.9 cm. wide and 58.4 cm. high. Heavy annealed brass plates comprised the hot and cold surfaces. These plates were hand-scraped to a flatness of better than 0.001 cm. The plates were separated from each other by a peripheral neoprene gasket, which served the dual purpose of spacing the plates and confining the liquid contained between the plates in the diffusion space. Specially purchased neoprene made it possible to space the plates with a uniformity of better than 0.0025 cm. By choosing the neoprene thickness, spacing between the plates could be varied at will; gage points to determine this spacing were provided around the periphery of the plates. The hot plate was heated by a 4000watt bank of contact electric heaters clamped to its back surface; this was sufficient to achieve a maximum plate temperature of 315’ C. while maintaining a temperature gradient of 1000” C. per cm. Heat was removed from the cold plate by circulating a high-velocity
LIQUID MIXTURE SEPARATION
In
0
*
> V z
2.0
w
0
LL LL
1.0
Ga LLI
Z w 0 0
.I
.2
.3
.5
.6 FEED, 0' G./SEC. .4
.7
.8
.9
1.0
Figure 1. Cetane-] -methylnaphthalene separation efficiency Column operation; T = 355' K., Tz = 65' K., 1 = 58.4 cm., B = 22.9 cm., 2 w = 0.0838 cm.
uniformly distributed stream of water or glycerol over its back surface. Automatic controls maintained plate temperatures at any desired level. Temperatures were measured by inserting thermocouples into a number of wells, 1 mm. in diameter, drilled in from the plate edges and terminating at various points directly behind the working surfaces. The temperature profile in both lateral and vertical directions of both surfaces was thus measured. Under typical conditions, temperatures varied in the lateral direction by no more than 13' C. and in the vertical direction by no more than =!=7O C. The temperature maximum was at the center of the working surfaces. The liquid mixture to be separated was pumped into the working space through a distribution manifold and slot port opening into the center of the hot surface. Separated streams were withdrawn through identical slot ports at the top of the hot surface and the bottom of the cold surface. The larger apparatus possessed an operating area 11.4 cm. wide and 294 cm. high. Hot and cold plates were constructed from aluminum channels 15.3 cm. wide, planed on the outer web surfaces to a flatness of about 0.0025 cm. in the lateral direction. Back cover plates bolted to the channel flanges converted the channels into conduits of rectangular cross section. Heat-transfer liquids, water and glycerol, were pumped in high-velocity turbulent flow through these conduits. The flat outer surfaces of the channels were clamped together with an intervening peripheral neoprene spacing gasket. The liquids to be separated were brought into and withdrawn from the working space so
formed, through a manifold and slot port arrangement similar to that of the smaller apparatus. Heat was supplied to the glycerol heat-transfer liquid by pumping it in turbulent flow over a bank of electrical heaters of 12-kw. capacity. Heat was withdrawn from the cold side of the apparatus by water circulating in turbulent flow which in turn was heat exchanged against tap water. Heated glycerol was pumped from top to bottom behind the hot surface: cooling water was pumped from bottom to top behind the cold surface. Thermocouples inserted into 1-mm. diameter holes in the channel webs enabled measurement of the temperature profile of the operating surfaces. No significant temperature variation was observed in the lateral direction. With maximum heat transfer, temperatures increased smoothly from bottom to top, by 20' C. on the hot surface and by 10' C. on the cold surface. Reduced heat dissipation resulted in a smaller temperature variation. Spacing between surfaces was altered by selection of the neoprene peripheral gasket thickness and pressure adjustment of the C-clamps holding the assembly. Gage points were provided around the periphery., Spacing errors in the lateral direction were no greater than the flatness uncertainty of 0.0025 em. I n the vertical direction variations in spacing were generally less than 0.005 em.
Materials The petroleum stocks selected for study were of a conventional kind described by tabulated physical and chem-
ical properties and source. The cetane used was a material synthesized for Diesel-index reference fuel. I t possessed a specific gravity of 0.771 (20)/4 and a refractive index of n2: = 1.4341; accepted values of specific gravity and refractive index are 0.77344 (20)/4 and nzz = 1.43453. respectively (7). The 1,I ,2,2-tetrachloroethane was Eastman Kodak Co.'s white label grade not further purified. Its specific gravity of 1.575 (20)/4 and refractive index of n2: = 1.4940 compared with accepted values of 1.574 (20)/4 and 1.475 ( 9 ) . 1-Methylnaphthalene of a practical grade was fractionally distilled to yield a fraction possessing a refractive index of n2,0 = 1.6136 and specific gravity of 1.0181 (20)/4. Comparable accepted values are n2,0 = 1.6174 and 1.0202 (20)/4, respectively ( 7 ) . The cetane-1,1,2,2-tetrachloroethane mixtures were analyzed by comparing pycnometric densities with a similarly determined density composition curve. Cetane-1-methylnaphthalene mixtures were analyzed both by this procedure and by comparing refractive indices with a composition refractive index curve. Ordinary Diffusion Coefficients and Thermal Diffusion Constants. Two binary test mixtures, cetane with 1,1,2,2tetrachloroethane and l-methylnaphthalene, were employed to test the performance of the vertical thermal diffusion columns. Ordinary diffusion coefficients and thermal diffusion constants have not heretofore been measured for these mixtures. Density-mixed diaphragm cells of a type recently described (20) were employed to measure the ordinary diffusion coefficients. Measurements were made at temperature and cell concentrations corresponding approximately with the mean temperatures and concentrations which prevailed in the continuous-flow vertical column experiments. For cetane-l,1,2,2-tetrachloroethane, 0 1 2
( 6 5 " C . ) = 1.29 z!z 0.03 X 10-5 set.-*
for solutions initially 2.4M (-0.5 mole fraction) in tetrachloroethane diffusing into cetane. For cetane-1-methylnaphthalene 0 1 2
( 6 5 " C . ) = 1.15 =!= 0.02 X
10 - 6 sec. -*
for solutions initially 2.3M (-0.5 mole fraction) in 1-methylnaphthalene diffusing into cetane. The 0.03 and 0.02 are standard deviations of the mean. When these coefficients were used, small corrections from 65O C. to the mean temperature of the column experiments were made by assuming Dl2 directly proportional to temperature and inversely proportional to viscosity i n VOL. 49, NO. 9
SEPTEMBER 1957
141 1
accordance with the Stokes-Einstein relation. Measurements of the viscosity temperature behavior of the mixtures were made to obtain these corrections. Thermal diffusion constants, a, for both systems have been computed and measured. While these constants have not been obtained with precision, their magnitudes have been sufficiently well established to permit rough estimation of limiting thermodynamic efficiencies. Following the theoretical approach of Dougherty and Drickamer ( 4 ) , LY for the system cetane-l,l,2,2-tetrachloroethane has been computed from excess thermodynamic properties estimated from available heats of vaporization in the Hildebrand-Scatchard (8,79) manner. a has also been estimated from measured activation energies for viscous flow in the manner suggested by Dougherty and Drickamer (5). The former procedure results in an estimate of 2.6 (25' C.); the tatter leads to a value of 3.3 (25' C.). Saxton and Drickamer (18) have measured a as a function of temperature between 25' and 55' C. for binary mixtures of normal paraffin hydrocarbons (nC6, nC7, nCs, nC9, nCle. nC18) with 1,I ,2.2-tetrachloroethane. Reasonable interpolation for the binary mixture, cetane (nC,J-tetrachloroethane, result3 in an CY of 3.25 (25' (2.). Values for these binary systems changed only slowly with temperature. Thus, from the slope of the curve of a us. temperature for nC1a-tetrachloroethane, a for cetanetetrachloroethane has been estimated as 2.5 at 65' C. The thermal diffusion constant has been measured for the system cetane1,1,2,2-tetrachloroethaneusing a vaportight stainless steel diaphragm cell similar in design to that described by Saxton, Dougherty, and Drickamer (77). Steady-state concentrations on each side of the porous glass diaphragm were established both from conditions of initial concentration equality on both sides of the diaphragm and from conditions of initial concentration inequality. Measurements were made at a mean concentration of 0.5 mole fraction and a t a mean temperature of 65' C. with a temperature difference across the diaphragm of about 20' C. The a values obtained varied between 2.0 and 2.2. These values, however, are believed subject to greater uncertainty than indicated by their repeatability. When the stirrers were operated at speeds low enough to avoid cavitation, isothermal conditions on each side of the diaphragm were not fully realized; systematic errors were thus introduced. Accepting this uncertainty, the range of estimated and measured values of CY for the equimole cetane-l,1,2,2-tetrachloroethane mixture lies befiveen 2.0 and 2.6 at 65' C. Physical data to estimate excess ther-
1 41 2
modynamic properties of the cetane-lmethylnaphthalene system were not available. Activation energies for viscous flow were computed from viscosity measurements and used to estimate an a of 1.5 at 25' C. Direct measurements of the equimole mixture in the diaphragm cell at 65' C. yielded an a of 1.9. Systematic errors as previously described are possibly contained in this value. Its magnitude, however, is believed sufficiently well established for rough estimation of limiring thermodynamic efficiencies.
Column Efficiencies Operation data for both vertical columns with the two test mixtures are given in Figures 1 through 4 where the energy efficiency computed by Equation 5 has been plotted as a function of the continuous rate at which the equimole test mixture has been fed into the column. Characteristic of each operation is a single maximum in the efficiency curve corresponding to a feed rate, cr, such that separation between product streams, A, is roughly one half that at zero feed rate. These results are in good accord with the so-called "two-log-q" rule proposed by Furry and Sones. Limiting thermodvnamic efficiencies have been estimated by Expression 7 and are shown as shaded areas on each of the graphs. These areas embrace the range of values that may be computed from Expression 7 corresponding to a range of values for a where a correct mean value is estimated to lie. The assumptions necessary in interpreting Expression 7 for practical column experiments are such that computed thermodynamic efficiencies must be viewed as approximate. In addition to uncertainties in applying a relationship which refers to small temperature and concentration differences to experiments in which both differences are appreciable, averaging in both temperature and concentration coordinates has been necessary. Thus, the ordinary diffusion coefficient is a strong nonlinear function of temperature and an unknown but probably small function of concentration. This coefficient has been referred to the mean temperature and concentration of each column experiment in an arbitrary manner. Similarly, mean values of molecular weight, density, and mole-fraction product have been chosen. Allowing for such uncertainties, Figures 1 through 4 suggest that welldesigned vertical columns are capable of thermal efficiencies of the same magnitude as thermodynamic limiting efficiencies. As a consequence, improvements in column design should not result in large thermal efficiency gains. Modest efficiency gains and engineering design improvements are of course
INDUSTRIAL AND ENGINEERING CHEMISTRY
possible. It may be added that columns obeying the equations of the Furry, Jones, and Onsager (7) operating theory achieve an efficiency of 60 to 70% of the thermodynamic limit. Failure to reach the limit is a consequence of failure to satisfy the condition for equality in Expression 7 throughout the column. These columns and similar operating procedures have been employed for separating complex petroleum fractions subsequently described. Underlying the presentation of heat-dissipation data for these systems is the assumption that. the columns operated with an efficiency which approaches the limiting efficiency. The heat requirements for the separations described are believed to be determined primarily by the molecular configuration of the molecules in the mixtures and are only secondarily the consequence of the column design.
Petrsleu m Fractions Naphthenic Coastal Distillate Lubricating-Oil Stock. Thermal diffusion separation of a naphthenic Coastal lubricating oil stock of mid-boiling range, has been extensively studied. This stock lends irsell well to study because it contains no wax or asphalt fraction, is more easily separated by thermal diffusion than any other stock studied, and extreme physical property changes are not required. Table I lists properties of the raw stock, the properties after conventional furfural refining: and the properties after thermal diffusion separation to a target viscosity index equal to that of the furfural refined product. The gross breakdown into aromatic and nonaromatic Iraccions shown has been obtained by chromatographic separation. To follow the separation process quantitatively, a single numerical parameter which could be treated as a concentration was desired. Ramser's (76) proposed index, V T R is a suitable choice. VTR equals -2.09 log [y(100' F.)/y(210O F.)] where y is a viscosity in centistokes. Ail fractions separated from a single stock by thermal diffusion will fit a linear relationship between the logarithm of the viscosity at any temperature and the VTR. Also, a convenient, approximately linear additivity in volume fractions and VTR exists. In complex petroleum fractions: to a first approximation, mole and volume fraction may be regarded interchangeable. Accordingly, the mole fraction notation, x , has been retained to represent either. Thus, VTTR,i,.
=
XI
(V'TRL)
+ (1 -
XI)
(VTR2)
Exhaustive analytical separation of this stock in a narrow-annulus column similar to that described by Jones and Milberger (12) has provided a VTR, volume-fraction relationship from which a rough volume fraction, binary composition
LIQUID M I X T U R E S E P A R A T I O N scale can be contrived. Thus, the raw stock of -2.53 V T R can be regarded as composed of &?yoby volume of oil averaging -2.10 V T R and 38yo by volume of oil averaging -3.23 VTR. The binary concentration scale resulting, while lacking physical reality, nevertheless provides a useful approximation for viewing this complex system. For ideal binary systems not too far removed from mole fraction 0.5, the most efficient operation of a thermal diffusion column is that operation which maximizes the quantity, uA2 divided by the heat transported, Q (Equation 5). An ideal constant A cascade of columns operates with an efficiency equal to that of any single unit comprising the cascade. These relationships permit estimation of the heat requirements of a cascade from that of a single prototype column, provided the constant A restriction is satisfied. At least within the practical limits shown in Table I, the naphthenic Coastal oil separation meets this restriction. Several measurements have established the (AVTR) measured from top to bottom of an operating column to be nearly independent of the VTR for the entering stream. This result is consistent with that expected in separating a binary mixture in which initial mixture and separated products are not too far removed from 0.5 mole fraction. Both thermal diffusion columns described have been operated with several spacings between plates, at a variety of mean temperatures. Temperature differences have also been varied. For each operation, the flow rate has been adjusted to maximize the quantity, o(AVTR)~/Q. Regardless of the actual small separation made, that column, whose dimensions and operating conditions yield the maximum in this quantity, has been regarded as a prototype for a cascade of similar columns arranged to make a desired greater separation with maximum efficiency. For any fixed throughput, heat transported in the cascade will increase as the square of the ratio of the separation desired to the separation achieved in a single prototype unit. The cascade diagrammed provides an equal division of stream volumes exiting from top and bottom of each unit of each stage. This cascade also provides equally divided streams exiting from the cascade. I t is evident that unequal division of streams exiting from a cascade of unit stages, each stage dividing stream volumes equally, can be arranged by extending either branch of the cascade unsymmetrically. Such an arrangement will modify heat dissipation requirements. Therdore, in order to specify thermal requirements for thermal
0 Figure 2.
.I
.2 FEED, O'GAEC.
Cetane-1 -methylnaphthalene separation efficiency
- TI = 56'
Column operation; 7 = 340' K., Tz
diffusion separation of a lubricating oil, it is also necessary to specify the product yield. This has been taken as 500j0. Such a raffinate yield approximates that obtained in conventional furfural refining. Columns can, of course, be operated with unequal volume streams exiting from the top and bottom. Ideal cascades "of such columns can be constructed. When the descriptive equations of Furry, Jones, and Onsager are applied to the fractionating section and scrubber placed above each other and the whole is treated as a single column with a reservoir at the feed point, unequal
I
0
.4
.3
.I
K., I = 294 cm., 6 = 11.4 cm., 2w = 0,0838
cm.
division of stream volumes does not lead to greater efficiency than realizable with equal division and can achieve the efficiency of the latter only when the feed-port location between the exit ports is displaced from the mid-position toward that exit port carrying the larger stream. The columns described have been operated with a variety of feed-port locations and stream divisions. I n no instance has a higher efficiency been achieved than that for equal division of product streams and mid-position location of the feed port. Figures 5 through 8 illustratezthe
I
I
I
I
.2
.3
z
.4
FEED, 0' G./SEC. Figure 3.
Cetane-1 ,I,2,2-tetrachloroethane separation efficiency
Column operation; f = 340° K.,Tz
- TI = 56'
K.,L = 294 cm., 6 = 1 1.4 cm., 2w = 0.0838 cm. VOL. 49, NO. 9
SEPTEMBER 1957
141 3
I
I
I
I
I
I
I
I
I
I
-I 0
4.
Figure
I
I
I
.I
.2
.3
I
Cetane-1 ,I ,2,2-tetrachloroethane
Column operation; T = 355' K., Tr - TI = 68'
I
I
1
.8
.9
1.0
I
.5 .6 .7 FEED, 0'G./SEC. .4
separation
K., 1 = 58.4
cm.,
I
B = 22.9
efficiency cm., 2 w =
0.0838 cm.
Table 1.
Coastal Distillate Thermal
Conventionally Refined
Raw Distillate
Vis., cs. 37.7' C. (looo F.I 98.8OC. (210°F.)' Viscosity index
189.0 11.6 22 -2.53
VTR Specific gravity,
20 20
nn 2O
s, %
N, % Aromatics, % Nonaromatics, %
113.5 9.7 61 -2.22
Diffusion Product 107.5 9.4 60 -2.22
0.9285
0.8984
0.9100
1.5104 0.27 0.05 40.3 59.7
1.4917 0.14
0.26
1,5012 0.04 35.6 64.4
0.01
21.4 78.6
ul
0 2.0 X
-I
(u
L1:
0
.I
I
1
.2
.3
FEED, 0' G./SEC. Figure 5.
141 4
Naphthenic Coastal distillate
INDUSTRIAL AND ENGINEERING CHEMISTRY
.4
principal features of naphthenic Coastal oil separation and the ranges of dimensional and operating variables explored. The quantity, O ( A V T R ) ~divided by rate of heat conduction between the plate surfaces, has been plotted against thc rate, o (grams per second), at which raw oil stock has been fed into the columns. Operating temperatures and column dimensions are noted on each plot. Each point plotted represents an operational steady state---Le., the columns have been operated under steady conditions with continuous transport of oil until composition of effluent streams remains constant in time. This has required operating times from a minimum of about 2 hours for large values of o, to a maximum of about 16 hours for small values. Each curve exhibits a single maximurn, roughly established by the point distribution; the ordinate chosen magnifies scattering. For this stock, the maximum increases with increasing average column temperature-at least up to temperatures where oil cracks at the hot plate. The maximum, obtained at a mean temperature of 188' C. with a hot plate temperature of 224' C. (Figure 5) is not far removed from a practical upper limit. A reduction In the temperature difference between the plates reduces the observed maximum in the chosen efficiency function (Figure 6). Figure 7 demonstrates the effect 01 a fivefold increase in column height and a twofold decrease in column width. The maximum shifts toward lower values of g. A larger separation offsets both the lower flow and the increase in heat dissipation; as a consequence the efficiency itself is substantially unchanged. Reducing the spacing between plates displaces the maximum toward lower values of 0 (Figure 8). Accompanying increases in the separation achieved again offset the lower flow and increase heat transport and result in little change of efficiency maximum. Experiments have been performed at both smaller and larger spacings than those noted. At much smaller spacings, the times necessary to establish steady concentrations in product streams become long, thus introducing appreciable uncertainty. At larger spacings, the measured separations become so small that they are obscured by errors in viscosity measurement. A cascade of columns, each operating with the maximum efficiency of Figure 5, arranged to make a product of the quality indicated in Table 1. in 50% yield, will conduct from hot to cold surfaces a total of 1.48 X lo4 calories per gram (2.04 X 105 B.t.u. per gallon) of raw naphthenic Coastal distillate fed into the cascade. This, therefore, determines the minimum fuel requirement for any thermal diffusion system to
LIQUID M I X T U R E S E P A R A T I O N achieve such a separation, within limits of the assumptions made. Heat recovery may, of course, be used to improve utilization.
2.0 0 -
Middle Eastern Residual Lubricating-Oil Stock. The propane-deasphalted and dewaxed Middle Eastern lubricating oil stock chosen for separation, possessed the properties shown in Table 111. After conventional furfural refining this oil possesses approximately the properties tabulated. Table I11 lists also the properties of this residual oil separated by thermal diffusion to a target viscosity index about equal to that conventionally obtained. Studies of this oil were limited. A linear V T R relationship with volumefraction blending is approximately obeyed. Studies were not undertaken to ascertain whether the AVTR achieved
X (u
[I:
I-
\
_'
T-T=86"C. 2 \ T = 121°C.
\
>
Q +
'b
'-*p \
P
*U
1
L z 5 8 . 4 crn. B = 22.9 cm. 2w=O.O86crn.
In
Middle Eastern Distillate Lubricating-Oil Stock. This raw distillate contains an appreciable wax fraction which concentrates in the desired product stream during thermal diffusion. T o simplify handling, the raw distillate was dewaxed. Properties of the dewaxed raw distillate, this distillate after furfural refining and dewaxing, and a thermal diffusion product separated to a target viscosity index equal to that of furfural refined product are listed in Table 11. VTR is a less satisfactory concentration parameter for this stock than for naphthenic Coastal oil. A volume-fraction additivity in VTR is more approximate and in practice was replaced with an experimentally determined volume fraction, V T R scale derived from blending experiments with separated fractions. I n addition, within the separation range of Table 11, the (AVTR) obtained in a given thermal diffusion column experiment decreased as the VTR of the oil fed to the column approached the target VTR. As a consequence; the efficiency expression, c(AVTR)~/Q, which is applicable to a constant A cascade, requires correction. The efficiency of a cascade of columns will be less than that computed by assuming that v(AVTR)~/Q for the cascade and for each individual column making up the cascade are alike. This assumption, however, provides an upper limit for cascade efficiency. Two typical efficiency curves for the columns are shown in Figure 9. The most efficient operation of a cascade arranged to make the product described in Table I1 in 50y0 yield will require that somewhat greater than 5.0 x lo4 calories of heat be conducted per gram (7.0 x 106 B.t.u. per gallon) of dewaxed Middle Eastern distillate fed into the cascade.
I
\
\\
\
T2-T,= 4 5 ° C .
x'
\ \\
T= 121°C.
; i
0
.05 .I FEED, O'G./SEC.
0 Figure 6.
.I5
Naphthenic Coastal distillate
in a single stage could be regarded as independent of the entering oil composition within the range of Table 111. Assumption of this independence, how-
ever, is implied in using the U(AVTR)~/Q expression to estimate the performance of a cascade from that of a single stage. The two sets of experiments in Figure
Table II. Middle Eastern Distillate Dewaxed Raw Distillate
Vis., cs. 37.7O C. (100' F.) 98.8' C. (210'F.) Viscosity index ' VTR 20 Specific gravity, 20 n as
Conventionally Refined
43.0 5.6 63 -1.85
s, %
N, % Aromatics, % Nonaromatics, % '
Thermal Diffusion Product 22.2 4.3 113 -1.49
27.2 4.9 115
-1.56
0.9200
0.8618
0.8735
1.5117 2.89 0.08 54.5 45.5
1.4748 0.73 0.01 21.7 78.3
1.4850 1.77 0.02 36.9 63.1
1
1
I
1
2.0
-.
'\
- __.. cm.
B = 22.9 cm. \ \ 2 w = 0.006 cm. \ \ \
\
1.0
111
L = 294 cm. B = 11.4 cm.
\
\
\
-
\ \ L
2w=0.086cm.
0 I
0
I
I
I
I
.05
.I
.I5
.2
FEED, 0 G./SEC. Figure 7.
Naphthenic Coastal distillate VOL. 49, NO. 9
SEPTEMBER 1957
1415
L = 2 9 4 crn. B = 11.4 cm.
ul
2
2.0
X
2w=O.I39 c m .
I
\
? = 132°C. crn.
1
20=0.086
0
.05
0
Figure 8.
.I5
.I FEED, CY G./SEC
.2
Naphthenic Coastal distillate
10 are representative of the efficiencies obtained in separating this oil. At the smaller plate spacing, the maximum occurs at low flow rates and could not be satisfactorily determined.
Table 111.
Urhen the thermal diffusion product listed in Table 111 is prepared in 5070 yield from the raw stock in the most efficient operation shown in Figure 10, 7.2 X lo4 calories of heat per gram
Middle Eastern Residual Thermal
Dewaxed,
Vis., cs. 3 7 . V C. (l0OOF.) 98.8O C. (210' F.)
Visc. index
VTR n
Deasphalted Ran, Stock
Conventionally
Diffusion
Refined
Product
1013 42.8 78 -2.95
525 31.7 97 -2.55 1.4949 1.11 0.02 41.9 58.1
436 27.6 95 -2.51 1.5059 2.52 0.05 64.0 36.0
20
s, %
3.04
... ...
N, % Aromatics, % Nonaromatics, %
...
T2-T, = 80°C. f = 135°C
//
/+o---
/
a 0.5
/"
A ' \
/HY
//O
--
\
T2-
/a.'
'ip/
L/
' \
= 83"C , '--\ f=l35"C. L = 2 9 4 cm.
B = 11.4 crn. 2w=0.139crn.
I/ Or
141 6
I
1
INDUSTRIAL AND ENGINEERING CHEMISTRY
I
I
' 0 -
of raw stock (9.9 X 106 B.t.u. per gallon) must be transported between hot and cold surfaces. Coastal Residual Lubricating-Oil Stock. After propane-deasphaking and dewaxing, a Coastal residual stock chosen for study possessed the properties shown in the first column of Table IV. After this oil has been conventionally solvent-refined, its properties approximate those tabulated in the second column. After thermal diffusion separation to a target viscosity index of slightly less than that of the conventionally refined product, properties of the residual oil become those indicated in the third column, Studies of this oil were restricted to expression for operatthe CT(AVTR)~/Q ing conditions shown in Figure 11. The approximation that a separation measured by AVTR be independent of the entering stream composition was not experimentally examined; prediction of the performance for a cascade from values of the ordinate of Figure 11, however, includes this approximation. When the thermal diffusion product of Table IV is prepared in 50% yield (conventional solvent-refining yields range up to 7570) from the deasphalted, dewaxed Coastal residual. approximately 2.3 X IO4 calories of heat per gram of raw stock charged (3.2 x IO5 B.t.u. per gallon) must be transported, Paraffin Scale Wax. The wax product obtained from a distillate Middle Eastern or Mid-Continent lubricaiingoil stock may contain up to 570 of occluded oil at that stage in the refining where it has been removed from the filters and stripped of dewaxing solvent. SubsequenL de-oiling of the wax by sweating or solvent precipitation is conventionally employed to reduce the oil content of the wax to a few tenths of Oil contents are usually specified in terms of ASTM procedure D 721-44 which distinguishes the oil phase as those hydrocarbons soluble in methyl ethyl ketone at -25' F. The wax itself varies M ith source and distillation conditions; a typical paraffin scale wax may contain about 94y0 of iso- and normal paraffins, 57, of cycloparaffins, and a few tenths per cent of aromatic compounds distributing in a range of 20 to 36 carbon atoms per molecule. The oil. phase comprises molecules of roughly the same molecular weight but which are cyclic, aromatic, or very highly branched. There is, thus, primarily a molecular shape distinction between the phases. This shape difference makes it possible to employ thermal diffusion to de-oil the wax. In the thermal diffusion process, liquid wax of lowered oil content separates from the top of the column; oil-rich liquid wax is withdrawn from the bottom. For descriptive purposes the wax-oil system has been regarded as a two-component
LIQUID MIXTURE SEPARATION binary system with concentrations given by the ASTM method. Entropy approximations and the constant A cascades used to describe the oil systems are inadequate for this system because all concentrations distribute at one extreme of the binary concentration scale. Each stage of a cascade used to separate the wax-oil system will make a separation, which should here be defined as, q = ( ~ 1 / ~ 2 ) ~ ~ ~ d ~ ~feed, ~ / ( ~and 1 / ~ *which ) is roughly independent of entering stream composition. In order to extrapolate from a single column making a separation less than that desired, it is sufficient to develop this column into a cascade of the same q in each stage, making the desired over-all separation, and compute the process heat dissipation for the cascade. Such "constant q" cascades can be generally constructed to avoid mixing streams of different concentration only as a limit when a large number of stages are employed. The desired process qo for the paraffin scale wax deoiling is 17-i.e.,
I .o X
0.5
'bl
0
I
0
.I FEED, O'G./SEC.
.05 Figure 10.
Coastal Residual
Dewaxed, Deasphalted Raw Stock
Viscosity, cs. 37.7O C. (loooF.) 98.8OC. (210"F.) Visc. index VTR .
Conventionally Refined
1162 41.1 68 -3.03
N, %
Aromatics, % Nonaromatics, %
Thermal Diffusion Product
681 33.0 82 -2.75 0.57 0.02 38 62
... ... ... ...
~~~
s, %
.I5
Middle Eastern residual
Table IV.
0997 0.950 = (0%)/(0;05)
A number of cascades employing a large number of stages realizing this q. and yielding product wax (taken as 99.7% wax) in 50% yield have been computed. Each stage in each cascade produces a separation, q, less than 17. The computed ratio of the heat dissipation of each single stage, Q, to that of its corresponding cascade, Q., is shown on Figure 12 as a function of the separation made in each single stage. The Furry, Jones, and Onsager operating theory for columns in which height and separation alone are varied yields a relationship between best values of transport, height, and separation. Near extremes of the binary concentration scale, parametric solutions of this relationship are analogous to these cascade computations. Figure 12 provides the basis for estimating the most efficient operation of thermal diffusion columns used to de-oil the wax. The rate at which crude wax is fed into a column divided by the heat transported in the column and multiplied by the ratio read from Figure 12, corresponding to the separation made, will be proportional to the efficiency. This efficiency expression is employed in Figure 33 showing typical paraffin scale wax de-oiling experiments. The most efficient operation, in which product wax in 50% yield and 99.7% purity has been prepared from crude wax of
1
I
I
682 32.1 79 -2.77 0.61 0.07 45 55
0
, O ' ' / \
/ /O O/
I I
9 I
\ \
T2-T,=8OoC.
\ \
i = 130°C. L = 294 cm. B=
\
I
I
I
11.4 cm.
\ 2w=O.I32cm. \ \ \
\
0 0
.05
.I FEED, 0 G./SEC.
Figure 1 1 ,
-15
.2
Coastal residual VOL. 49, NO. 9
SEPTEMBER 1957
1417
c I
2o
I
I
efficiencies. This was established by separating two test mixtures of cetane with 1,1,2,2-tetrachloroethane and 1methylnaphthalene. Thermal diffusion constants, a, and ordinary diffusion coefficients, D12, for these two mixtures have been independently determined. These quantities have made possible an estimate of limiting thermodynamic efficiencies with which column operating efficiencies have been compared. Five complex petroleum fractions have been separated using the same columns and operating procedures employed for the test mixtures. T h e minimum heat requirements for these separations are believed to be determined primarily by configuration of molecules in the mixtures rather than by apparatus design. Heat requirements for separations and yields, comparable with present refining practice, range from a minimum of 2.0 X 105 B.t.u. per gallon for a naphthenic Coastal distillate lubricating-oil fraction to 2.8 X 106 B.t.u. per liquid gallon for de-oiling of paraffin scale wax.
References
0 .I
.5
( 1 ) Am. Petroleum Inst., Project 44, “Selected Values of Physical and Thermodynamic Properties of Hy-
1.0
Q/Qo (2)
Figure 12. Computed ratio of heat dissipation for each stage to that of its corresponding cascade a5 a function of separation in each stage
(3) (4)
fins and away from branched or cyclic structures.
95% purity, has required that 2.1 X 105 calories of heat per gram (2.8 X
(6)
106 l3.t.u. per liquid gallon) of crude wax fed into the cascade be transported between hot and cold surfaces. The wax so prepared differs slightly from conventionally purified wax ; mass spectrometric analysis reveals some shift toivard higher concentrations of normal paraf-
summary
I
Q
0 -
\ \
\
70” C 145°C L = 2 9 4 cm. B = 11.4 cm. 2 w = 0.080 cm.
\
0 \
/
\
\
\
0
0
.01
I
I
I
I
I
.02 .03 .04 .05 .06 .07 .08 .09 FEED, s‘ G./SEC. Figure 13.
141 8
I
Paraffin scale wax
INDUSTRIAL AND ENGINEERING CHEMISTRY
(1923). 17) Furrv. W.H.. Jones. R. C.. Onsarer. L.; Phvr Rkv. 55, i o 8 3 (f939). ( 8 ) Hildebr H., Scott, R. L., “Soll. >
The two, plane-parallel, vertical, thermal diffusion columns described appear to operate with efficiencies of the same magnitude as thermodynamic limiting
61
I
(5)
drocarbons and Related Compounds,” Carnegie Press, 1953. Begeman, C. R., Cramer, P. L., IND. EXG. CHEM.47, 202 (1955). Bridgman, P. W., “The Nature of Thermodynamics,” Harvard Univ. Press, Cambridge, 1941. Dougherty, E. I>.,Drickamer, H. G., J. Chem. Phys. 23, 295 (1955). Dougherty E. L., Drickamer, H. G., J . Phis. Chem., 59, 443 (1955). Eckart, H., Brennstoff-Chem. 4, 24 V
I
I
Keinl
1 9 ) International Critical Tables, vol. 3, p, 28, McGraw-Ilill, New’York, 1928. (10) Joncs, A. L., Petroleum Processing 6, 132 (1951). (11) Jones, A. L., XND. ENG. CHEM.47, 212(1955). (12) Jones, .4.L., Milberger, E. C., Ibid., 45. 2689 (1953). (13) Jones, R. C., Furry, W. H., Reo. Mod. Phys. 18, 151 (1946). (14) Melpolder, P. W., Brown, R. .k9 >
,
Washall, T. A., Doherty, W., Young, W. S., Anal. Chem. 26,
1904 (1954). (15) Onsager, L., Phys. Rev. 55, 11371% (1939). (16) Ramser, J. H., XXD. ENG. CHEM” 41,2053 (1949). (17) Saxton, R. L., Dougherty, E. L., Drickamer, H. G., J . Chem. Phys. 22, 1166 (1.954). (18) Saxton, R. L., Drickamer, H. G., Zbid., 22,1287 (1954). (19) Scatchard, G., Chem. Rev. 8, 321 (1931). (20) White, J. R., J . Chem. Phys. 23, 2247 (1955); 24, 470 (1956).
RECEIVED for review August 23, 1956 ACCEPTED Feb. 26, 1957
