Radial Chromatography in Industry J
PETER P. HOPF R'arrl, Blenkirisop & Company, L t d . , London, England
Chroniatographic adsorplion can be etfectecl if d solirtiori is allowed to flow radially through a disk or c ' j liiider from the axis to the perimeter. -idsorption taLes place in much the same fashiou as in a coluiiiri. The f l o w ma? be assisted by centrifugal force- to gi\e greater speed and vlearer zone definition. i n apparatus was con-triirtetl for this purpose, and the term "rhroniatofuge" is proposed for this type of plant. [t allows some l a r g e wale separation* Mith a sa\iirp of time, labor, and floor s p a ' e .
disk thcii becomes a zone o r part of a zone mid may b i b urariged to contain one constituent only. This technique, although very elegant, is suitable only for valuable products, which are manufactured in comparatively small bulk, as it, is cxpcnsivc \virh regard to control and floor space. T l i it~ appears that the principal objections to technique. xiid plant either in use or under consideration a t present are expense of construction, large requirement of floor space, and difficulty of control. TI1 E CHROL\I.ATOFUGE
T
HE separation and pui,ificat,iori of suhstaiicc..:
solutioii h!. chromatographic adsorption have btwi an object of research for Rome time. T h e techniqrie most commonly employed makes use of a column fitted a t the bottom with a fritted glass disk and connected t o a suction pump. The original Tswett column has been modified from time to time. This colunin is packed n i t h an adsorbent, and t h e solution passed through with the aid of suction or by gravity alone. A vast nuniber of substances have thus been isolated or purified from other constituents. Pure materials are thus obtained in a solvent which may be evaporated. This technique has lately been adapted for industrial purl)oscs, arid plants have been erected on a pilot as 1 ~ 1 1as a full conimercia1 scale with substitution of the column by adsoqitioii to\\ (lrs. T h e cost of this plant arid its operation is, however, too high to allow cheaper products to be isolated or purified in this fashioii. Various improvements and techniqucs have been cvolvcd to this end, notably in the production of penicillin. The adaptioii of existing plants and the recognition of the principlcs i i i v o l r d in known techniques have beeii of considerable help in thi.: advaiicc. iti
L 4 R G E SCALK ADSOKPTIOV
T h e practice of decolorizing solutions or organic liquid.- t)y passage through a filter press packed with a n adsorbent, usually animal charcoal, has been in usv for a long time. Similarly sand filters packed with adsorbents, usually coal ashes or silica gel, have served the same purpose. In several examined cases aiialysis of the adsorbent along t h e dirclction of flow has shown that t h r impurities are adsorbed \Tithin more or. lese rleai,ly defined zones. the exact position of which is dependent 011 the rate of flon, thc nature of the adsorbent, and the total amount of impui,itic,,present. I n the case of a fi1tt.r press the first plate invariably contains the colloidal impurities, such as carbon and rwinous matter. As these have to he regarded as filtered off rattier. than adsorbed, they have to be ignored in this consideration. Thrs remainder of the impurities, hon-ever, will be found to tw atlsorbed singly in zones and to be in chromatographic serics. A method of chromatographic adsorption used in industry. but hitherto for very special purposes only, is the chromatographic disk, the action of \vliich is analogous to that of a siiiglc. plate or chamber in a filter press. This has so far beeii used only for the separation of single substances, such as penicilliii. I solution containing several coustituents, the isolation of Tvhich is desired, might conceivably pass through a succession of such disks, each mounted separately oil a suction filter. Every single
938
.A uc~w type of cwristructiori proposed t,o facilitaic. tlic ust: chromatography 011 industrial scales and for bulk products eniploys radial rather than longitudinal adsorption. The plaiit (Figure 1) is capabk~of \wing modified to suit any particular. purpose. It consists iri esscnre of a cylindrical drum into which t h t , liquid is fed from t h r asis atid passes t o the outer pei-imetci-. The cyliiidric.al roiitairicr (or series of contaiiicrs~of cuiivt~riic~rit diameter arid height li perforated outer wall and is I i i i c d inside w i t h filtrr cloth. The axis is in the form of a hullo\v tub(.: its wall is also prrforated and laid out ivith filter cloth. This central tube serves as feed pipe. T h e whole is c~nclostdi r i a suitable vessel fitted lvith tin out1r.t iiear tho bottom. T t w t)ottom of the feed pipe may be s e a l d or. open, \vhcri a cwntiriuous rcturn flow may tie arranged. For the purpose of this investigation a lahoratory scale apparatus arid a works scale plant were built. T h e first \vas mad(, of copper, the perforated walls being substituted by 6-mm. t)r,ass \virc netting. The lincs of design laid down in the sketch \vi'ri; folloncd. The dianicter of the feed pipe was 2.5 mi.. a d thr. radius of adsorbent container 'A, 30 cm. Outer container B \vas a copper boivl of 70-em. diameter, and the height of a d s o r k i l t disk C was 10 cm. ,111 joints were brazed. T h e apparatus was drivcri by a I-horsepowel, motor: because it was not sufficieitt 1y sparkproof, it n as iv(*llremovcd from the apparatus and protected by a tiri screen through r\hich the belt was passed t o allow the sinallest aperture permissible. T h e larger scale plant was adapted from a n available hydro or ~ o r k scentrifuge such as is cominon1)- used for the drying of crystals. Tlic outer bowl was made of cast iron with direct drainage iiito a glass-lined sump. T h e inner container was of galvanized iron with perforated walk t o which a lid of the same material was fitted. The feed pipe, Jvhich was welded onto this lid, was also madyadjusting the tap of a drum placed directly over the funnel. This construction was also belt-drivcri by a gear affixed t o the bottom of the Construction, the motor beiiig some distance away, and completely nonsparking and vaporproof. These precautions are necessary since, in several cases, solutions containing ethtsr. were used, and nonaqueous flammable solveiits are coniniiori in chromatography. T h e outer case M'&S also fitted with a lid of galvanized iron, consisted of two well fitting halves, and was fixed t o the outer container by means of hinges. T o make this lid as tight fitting as possible and to avoid frictioiial sparking, the feed pipe n-as fitted n-ith a gland packed
August 1947
INDUSTRIAL AND ENGINEERING CHEMISTRY
with graphite and asbestos a t D. The diameter of the feed pipe was 15 cm.; the radius of adsorbent container A was 125 cm., and its height C, 80 cm. Although a stationary design is possible, ivhere the passage of liquid through the adsorbent is assisted by suction or pressure, far greater allowances would have to be made for gravitational forces. The full benefit of this design is obtained only when i t is in the form of a hydro or centrifuge as outlined, when the disk or cylinder is made to revolve round its axis-that is. its feedpipe. The centrifugal forces assist, not oiily the flow but also the distribution of the adsorbent and of the liquid with regard to the adsorbent. This has been found to give good definition of the adsorption zones and embodie- the maximum advantage over existing designs, as far as speed, floor space, and adsorptive efficiency are concerned. The name proposed here for such a device is “chromatofuge.” The rate of flow may be adjusted within narroJv limits by adjusting the speed of the revolving disk to give maximum separation. This speed must not, however, drop belox the minimum necessary t o ensure even distribution of pressure. If the cylinder is packed unevenly with adsorbent material, rotation before the passage of solution will cause the adsorbent to rearrange itself, and a n even packing is automatically obtained. I n some cases, where the rate of adsorption is very slow, it was found t h a t the stationary deaign is preferable. This applies generally to cases where no great affinity exists between the adsorbent and the adsorbrd substance-that is, where the rate of reaction of the adsorption process is less than the rate of flow of the solution under rotation. In order to adapt the plant to stationary use, the liquid is fed ILS before under convenient pressure. The bottom of the fced pipe is not sealed off but contains a bean valve which keeps the desired pressure constant and allows for the liquors to be returned through the top. POSSIBLE USES
11‘ the chromatofuge is used for separation by chemical reaction-that is, by a change in the chemical composition of the constituent which react wit,h the adsorbent-the process is not, strictly speaking, one of chromatographic adsorption. Such reactions, however, may take place a t different levels, and the zones are then to be regarded as st>aticrather than fluid. I n these cases the centrifugal adaptation is preferable to any other design. The disk is packed n i t h a basic material in the case of acids and with a n acid in the case of bases. I n all examined eases of separation by chemical reaction, the solution could be passed at maximum speed, and clearly defined, circular, static zones were obtained for each constituent. The zones thus formed cannot be eluted and do not travel on addition of solvent. Such a design is remarkably adaptable to a variety of problems and may lead to a wide range of new separations and purifications being attempted on a n industrial scale. The low cost of construction and the high working speed permit cheaper bulk products to be inanufactured in this \yay. As the sectional area through which the solution is to pass varies directly with the squaw of the distance from the center’clear zones, morc even distribution over the disk may be obtained by making the preva1~ntconstituents stay in the outer parts of the disk. This will hc possible in many instances either by connection of two chroinatofuges in series or careful prcvious adjustment of the solution. Collection of more data than have so far been obtained may lead to a theoretical foundation which would correlate adsorption affinity and rate of flow and thus allow the position of each zone to be calculated in advance. It is safe to say that, as the zones approach the perimeter, the rate of flax is slowest; to some extent this counterbalances the fact t h a t constituents with the smallest affinity for the adsorbent will be found there. Gravitational forcrs would also come into play and would have to be allowed for. I t was hoped that a clearer understanding of the theory as
Figure 1.
939
Chroniatofupe
Reell as a balancing oi the gravltattonal anti rentiiiugal torces would be derived by replacing the cylindrical adsorbent containei with one of the shape of a n inverted cone. This might have given clearer zones and, in the rase of fluid chromatogram, might have allowed these zones to be continuciusly removed along the sloping perforated side a t various distanoes from th? center Experiments so far, however, have not shown any advantage in such a design. ELUTIO3
The final separation of the single pure vonstituents ha6 to follow known practice. Usually this ail1 be done by elution with a suitable solvent, which may be the original solvent adulterated with water or a lower aliphatic alcohol. If the top of the adsorbent container is removed, however, the single zones may be dug out by hand or with a n adjustable stencil. For industrial products the only feasible method available a t present is elution. The eluent will be passed through the chromatofuge in the same manner as is the solution-that is, under centrifuging-or statically-that i., under pressure whilc the apparatus remainq stationary. PRACTICAL APPLICATION
.4 variety of separations known and described in the literature werc c a r r i d out on the con tructions outlined, and i t was found in each case that separations which have been carried out on a column gave exactly analogous results on t,he chromatofuge, t’hr single zones being much clearer in definition. The sequence of these zones was identical with that found in columns, and, in the case of inorganic solutions, tho chromatographic series rrniained unaltered. The procedure in each case involvtd lining of the outer and inner walls with filter cloth and close packing of the container with adsorbent. The container mas then allowed to revolve at maximum speed, after Tvhich more adsorbent had to be added to fill the available space. I t was then spun dry for a few more minutes before the solution was passed through the feed pipe and the effluent collected. I n this manner on the smallcr construeticin a solution of 100 grams of quinidine and 50 grams of quinine in 2 liters of ether gave two concentric zones with a blank space between them. These zones were eluted without rotation by passing the solvent under pressure and collected individually. This operation was carried out with 2.5 liters of et,her containing 57, methanol. The adsorbent used was activated alumina.
INDUSTRIAL AND ENGINEERING CHEMISTRY
940
On the large construction, packed with barium carbonate, a mixture of 25 liters of oleic acid and 60 liters of ricinoleic acid in 20 liters of ether were similarly separated, both compounds being obtained in highly purified form b y elution with moist ether. Recovery of all constituents except the ether was nearly quantitative. Also un the large construction 100 liters of a t h y alculiolic solution containing 25 liters of octan-2-01 and 18 liters of methyl hexyl ketone m r e passed through activated charcoal and e l u t d with moist alcohol under rotation, when the two constituents could be collected in separate fractions. These contained 90c; of the higher alcohol and 85Yc of the ketone. h mixed interniedia r y fraction containing the remainder of both constituents \vas also obtained and would, in production, h a r e to hc rt;tui,ned to t h e separator after drying. On t h e laboratory scale a disk was packed \vitli a mixture ol 75% alumina and 25% 8-hydroxyquinoline. .4 dilute sulfuric acid solution of vanadium, iron, nickel, and zinc was pasbed through under rotation, Clear zones with blank interspaces were obtained for vanadium, iron, and nickel, whereas the ziti? zone could be made visible under t h r mcrcury vapor lamp.
Vol. 39, No. 8
A packing of activated alumina 1r-a~ u . d fur the purification of castor oil on tlie works scale. A dark oil (82yo glycerol triricinoleate) containing ricinolvic acid, dihydroxystearic acid, and colloidal impurities was thoroughly dried and passed through t h e chromatofuge at a rate of 500 liters in 1.5 hours. A little dry ether \vas added t o reduce the viscosity a n d assist passage. The effluent was found to consist of ether and pure, colorless, neutral castor oil in what could be regarded as quantitative yield. On examining the opened container under ultraviolet light, and tiy taking analyticalsamples along a given diameter, clear zones for cach c-onstituent were identified. T h e colloidal impurities \vc>readsorbed or, more likely, filtered off in a small zonc round the crnti,al tube, the other constituents being separated by blank z o n r s of c l i w dcfinition. Castor oil was found loosely adsorbed in thr outnniost zone bordering on the perimeter. Elution with (+her cwntaining 5"; methanol under rotation allowed the zones to bc ciillected individually. It is hoped that these expc~i~iments and tlirir publication v,ill \videti the scope and application in industry of chromatography, and provide a means for using the wealth of experimental results of wsearch norliers in this field in works practice.
ECONOMIC PROCESS OPERATION Method for Determination of Optimum Operating Conditions
A
mathematical method is yreseriled for the poaiti\e clrtermination of the process conditions that giTe maxim~rrn profit for a plant operation. The complete interrelations of the process variables and the costs of additional unitc of each of the variables are necessary for the calculatiou. T w o illustrative examples are given, and the epecnial requirements of the method are discussed.
0
YE: of t h r most important arid, at tht. s m i e tiin(,, m o ~ t
difficult problems of a process engineer is the determiriatioil of the optimum operating conditions of a plant. The decisions as t o what temperatures, pressures, flows, etc., lead to inaxiriiuin plant monetary return can easily mean the difference betwwn profit and loss for a n entire organization. T h e operating conditions chosen are usually based on the engineer's judgment of the economic and engineering relations involved and ordinarily have little or no mathematical background. In the present article a mathematical systthni is tlevc.lopcd for organizing engineering and economic data to make possible a complete, positive solution for the optimum operating condition, of a plant on the single basis of maximum monetary return. The mathematics of the method is exact, and the rrsults obtaiiiod r a n be made as accurate as the data ujed. RIATHER14TICS OF METHOD
A simple Hiid familiar example of economic design i,< tlit, calculation of the optimum thickness of insulation or lagging for a steam line. Figure I gives a typical set of data for avcbragetl weather conditions. The cost of the steam lost, the co>t of the lagging, and the total cost, all ill dollars per year, arc plotted against t h e t'hickness of lagging i n inches. The steam loss in M
p o u u d ~ih tit35igiitlttd as u, and tiit, thickness of lagging in inches is t l t 4 g n a t d as b. Also the cost of a n additional M pounds of ,steam is -1, and the cost per year of an additional inch of lagging i,s H. The total cost per year of maintaining the steam flow is 7'. \\-it11 these definitions B is the slope of t h e cost-of-lagging vurvt, of Figure 1. Also d a / d b is the change in the steam loss with change in thickness of lagging. Then A(&/&) is the change in the cost of the steam loss with change in thickness of laggingthat is, A ( d a / d b ) is the slope of the cost-of-steam-loss curve of t h e figure. Plots of these slopes, B and -A(&/&), are given by t,he broken curves of Figure 1. At the optimum design the total cost curve goes through a rnimimum cost point. But a n additional property of the optimum is that the slopes of the cost lines, A ( d u / d b ) and B, are equal and opposite in sign. This is shown graphically in Figure 1 n-here the broken line plots of - 9 ( d a / d b ) a n d B cross at the minimum cost point. This same property can be stated algebraically as follows: The slope of the total cost curve, d T / d b , is the sum of the slopes of the cost-of-steam loss and cost-of-lagging curves, and a t the minimum cost point the slope of the total cost curve, d?'/dh, is xei'o. Then the equation
A(da/db)
+u
=
dT,flb
=
0
(1)
is true a t the minimum cost point. This equation can also be ivritten as A da B d b = 0. This simple relation can be used to define the economic optimum for any system of two variables. I t states t h a t a t the economic optimum an additional dollar outlay yields a n additional dollar return. If the return is more than a dollar, then the outlay should he increased, and if the return is less than a dollar t1ic.n the outlay should be decreased. This assumes that there arc no discontinuities in the immediate vivinity of the optimum values.
+
