This ion exchange process removes toxic species from waste waters and forms a concentrated waste; subsequent treatment of the regenerant is necessary for ultimate disposal. T h e process offers the potential advantages of cartridge ion exchange beds and concentrated wastes, but some economic and further technical evaluation is recommended. Addenda for Ect,nornics
T h e conditions described in these technical feasibility tests are probably not econoinically competitive \vithout change, as no attempt was made to minimize costs. To provide a basis for cost estimation, further work should include tests designed to reduce the volume of regenerant and resin required because they are the primary influence on costs. Such tests should include use of unbalanced resin, lower regeneration time, and lower regeneration flow rates. With reduced regenerant volumes, higher concentration ratios would be obtained. T h e economics of this process is probably more favorable for the small manufacturer for whom conventional treatment is quite costly than for the large manufacturer.
Ac knowledgrnent
T h e support of the Connecticut Water Resources Commission and the analytical assistance of Joseph Masselli and Nickolas Masselli are gratefully acknoivledged. T h e author appreciates the experimental assistance of Kenneth Hansen in the actual waste water tests, James Colthart in the repeated cycle tests: and Tony Chen and Christopher Eik. T h e author is especially indebted to Joshua Dranoff for his guidance, discussion, and experimental contributions. Literature Cited
Goldblatt, E., Znd. Eng. Chem. 48, 2107 (1956). Masselli, J., Masselli, N., TVesleyan University, Middletown, Conn., personal communication, February 1958. Ruder, J., M. Engr. thesis, Department of Chemical Engineering, Yale Unikersity, New Haven, Conn., 1959. Smith, G. F., Tt’ilkins, D. H., Anal. Chem. 25, 510 (1953). Tt’alker, C. A., Zabban, TV., Plating 40, 165, 269 (1953). TVills, E. G., J . Water Pollutzon Control Fed. 23, 1288 (1951). RECEIVED for review December 21, 1966 ACCEPTED June 8, 1967 Division of Tt’ater 8L Tt’aste Chemistry, 148th Meeting, ACS, Chicago, Ill., September 1964.
CONCENTRATION AND SEPARATION OF IONS BY DONINAN MEMBRANE EQUILIBRIUM R I C H A R D M . WALLACE Savannah River Laboratory, E . I . du Pont de .\‘emours
&3 Go., rliken, S.C. 29801
Continuous cciuntercurrent processes were developed on a laboratory scale using Donnan membrane equilibrium across permselective membranes to concentrate and separate ions in solution. This technique was used to concentrate uranyl ions from dilute solutions of uranyl nitrate, remove excess acid from a typical feed solution, concentrate trivalent ions such as lanthanum, remove strontium from sodium nitrate solutions, and separate silver and copper ions with a process based on differences in their ionic charge, and with processes baised on differences in the stability of their complexes with different ligands.
membrane equilibrium across permselective (ion exchange) membranes has been used to measure charges on ions in solution (Wallace, 1964), dissociation constants of acids (Wallace, 1966), ,and stability constants of inorganic complexes (Wallace, 1967). This paper extends these studies to the development of processes for the continuous concentration and separation of ions in solution. These processes are called “Donnan dialysis” because they are based on the approach to Donnan membrane equilibrium. Donnan membrane equilibrium is any thermodynamic equilibrium subject to the restraint that some of the species are excluded from a portio i of the system (Donnan, 1924). I n the processes described here, the restraint is imposed by permselective membranes which ideally pass cations or anions but not both simultaneously. I n the following theoretical discussion, the behavior of cations in a system containing a cationpermeable membrane is considered ; a completely analogous treatment of the behavior of anions in a system containing a n anion-permeable membrane is also valid. If solutions of electrolytes are placed on opposite sides of a membrane permeable to cations but not to anions or water, the anion composition of the two solutions must remain constant. T h e cations, however, will redistribute between the ONNAN
solutions until the following condition of equilibrium is established (Wallace, 1964; Donnan, 1924; Helfferich, 1962)
= K
(%+)l’Z
where C i Z + is the activity (approximately the concentration) of the ith cation of charges Z; R and L refer to the solutions on opposite sides of the membrane; and K is a constant for all cations in the system. If a n electrolyte solution containing a univalent cation N and a noncomplexing anion is placed on one side of the membrane, while a different concentration of a second salt containing the same anion but a different cation C Z + is placed on the other side, Equation 1 shows that, a t equilibrium,
Thus, all cations tend to concentrate in solutions containing the higher anion concentration; cations of higher charge tend to concentrate preferentially over those of lower charge. If different anions-one of which complexes a specific cation while the other does not-are placed on opposite sides of a VOL. 6
NO.
4 OCTOBER 1 9 6 7
423
cation-permeable membrane, the cation will concentrate in the solution containing the complexing anion, because Equations 1 and 2 refer to free ions in solution, and the formation of the complexed ion reduces the free ion concentration in that solution. If two different cations are introduced into the system, only one of which is complexed by the complexing anion, that cation will concentrate preferentially in the complexing solution and the two cations can be separated. This distribution of cations between two solutions may be regarded as a form of solvent extraction in which both phases are aqueous. The distribution coefficient of the cations between the two aqueous “solvents” can be varied by adjusting the composition of the two solutions. A wide variety of concentration and separation processes is therefore possible for two countercurrent solutions flowing over opposite surfaces of a permselective membrane.
Feed-
Product
--Strip
+-
-Raffii l a t e
Figure 1 . Schematic representation of experimental membrane separator
Figure 2. Schematic diagram of membrane spacer of experimental membrane separator
Experimental
T h e membrane separator (Figure 1) consists of two end plates, shown as shaded areas, and a series of membrane spacers (flow compartments), each separated from an adjacent one by a permselective membrane. Slots and holes cut near the edge of each spacer act as ducts to direct the flow of the solutions. The ducts are arranged so that solution entering the bottom of one spacer leaves at the top, bypasses the second spacer, and enters the top of the third spacer. Thus, different solutions flow alternately cocurrent and countercurrent to each other in adjacent spacers; however, the over-all flow is countercurrent. The end plates Ivere 3 X 3 inch squares of 3/8-inch stainless steel. Each end plate contained two ports to which adapters were welded to accommodate external piping for the inlet and exit streams. Each plate also contained one hole at each corner to accommodate */?-inch bolts, which were used to clamp the separator together. Figure 2 is a schematic of a membrane spacer. The solution enters the spacer from the preceding stage through a hole and slot at the lower left. It then proceeds through the spacer as shown by the arrows and leaves through the slot at the top right, where it is conducted b y means of the hole into the next appropriate spacer. Baffles within the spacer direct the flow to expose all of the membrane. T h e upper left hole is a bypass port that conducts the second solution from the preceding spacer to the succeeding one. All of the membrane spacers in the countercurrent separator are alike. but they are orientated differently to provide the floiv pattern. If Figure 2 represents the first spacer after the end plate, the next three spacers are oriented by successively rotating Figure 2 through 180°, first about the vertical axis in the plane of the paper and then about the horizontal axis. Since there are only four possible orientations of the spacer, before the original configuration is restored, a membrane separator of any size can be made of a series of identical units each containing four spacers. The membranes between the spacers are punched with two holes a t opposite sides of the same edge to accommodate the flow between the spacers. T h e appropriate orientation for successive membranes is obtained by 90’ rotations about a n axis normal to the membrane surface. The membrane spacers were constructed as a sandwich of polyethylene sheet cut as in Figure 3A, between two stainless steel sheets cut as in Figure 3B. When assembled in this fashion. the spacer defined the flow pattern shown in Figure 2. When the membrane spacer was a simple design as in Figure 2, the membrane collapsed into the slots and allowed the solutions to flow into the wrong spacer. However, the stainless steel outer sheath in the final sandwich design bridges the slot in the polyethylene center and prevents membrane collapse. The outer sheaths (Figure 3B) were of 10-mil 316 stainless steel shim stock, while the inner part (Figure 3A) was of 20-mil polyethylene. Each spacer was a 3 X 3 inch square with a 2 X 2 inch area in the middle comprising the flow compartment. T h e ribs in each component were 0.10 inch wide, and the flow channels through the spacers were 0.32 inch wide. T h e total membrane area exposed to the solutions was 3.2 sq. inches per membrane. 424
I&EC PROCESS D E S I G N A N D DEVELOPMENT
0
0 A. Polyethylene
Figure
01 8. Stainless Steel
3. Patterns for membrane spacers
Thicker spacers, desired in some tests, were made by adding 20-mil polyethylene components of the Figure 3B design to the middle of the sand\vich. The membranes \vere AMFion C-103C cation- and AMFion A-104B anion-permeable membranes (trade-mark of the American Machine & Foundry Co.). Membrane separators of different numbers of stages were used in the early development. The final separator contained 23 membranes and 24 spacers; the total exposed membrane surface was 73.5 sq. inches. A separator containing only 17 membranes and 18 spacers was used in studies with anion membranes. Solutions were fed into the membrane separator with either Beckman solution metering pumps (trade-mark of Beckman Instruments, Inc.) or Lapp Pulsafeeders (trade-mark of the Lapp Insulator Co.). Normal feed flows for the most rapidly moving streams were 1 to 10 ml. per minute. For very slow flows (0.02 to 0.20 ml. per minute), a Pulsafeeder was geared down to about one-half stroke per minute. Agitation within the membrane separators was provided by pulsing some of the feed streams with a pulse pump (Lapp Pulsafeeder from which the check valves had been removed). T h e pulse pump \vas geared to deliver about 120 pulses per minute with a pulse size of about 0.3 ml. per pulse in each direction. T h e pulse pumps were in series with the feed streams, between the feed pumps and the membrane separator. T h e flow of solution through the pulse pump helped purge them of air that was generated by cavitation. Air traps, placed between all pumps and the membrane separator, prevented air bubbles from entering the membrane separator. Needle valves, between the Pulsafeeder metering pumps and the pulse pump, prevented solution flow through the ball-check valves of the metering pumps.
Input stream flows were determined by calibration of the metering pumps. Exit stream flows were measured by collecting the effluent in burets; the time for delivery of a known volume was measured periodically. After assembly. water was pumped through the membrane separator units to remove the air; however, troublesome amounts remained trapped. This trapped air was removed as a foam by pumping a solution of a surface-active agent through the separator while pulsing; the system was then flushed with water.
recovery of the cation, the charge balance across the membrane separator requires
ZaAsgs
cy=-
Mass balance across the membrane separator shows the cation concentration in the product stream to be
(3) where C is the concentration of the cation, g is the volumetric flow, and subscripts P, F , and R refer to the product, feed, and raffinate, respectkvely. The flows of the raffinate and feed streams are assumed to be the same in Equation 3. If the cation concentration iri the raffinate, CR, were zero, the concentration factor would be the flow ratio gF/gP. I t might appear that C R could be made as small as desired simply by decreasing the flow to allow more time for transfer through the membranes. However, other factors must be considered if the cation is to be recovered completely. If the feed stream flows too rapidly or the strip stream too sloirly, not enough anions are introduced into the strip to satisfy the valence of the cations from the feed. When this happens, the product becomes saturated with the cation and some loss to the raffina1:e must occur. Therefore, for complete
Table I.
(4)
where A is the anion concentration of charge Z A , ZC is the charge of the cation, and subscript Srefers to the strip. (Other symbols have their usual meaning.) I n cation concentration studies it is convenient to define a n operating parameter
Concentration of Ions
A continuous membrane concentrator can be treated ideally as t\ro long, thin channels separated by a cation- (or anion-) permeable membrane. The feed, a dilute solution of a salt of the ion to be concentrated, enters one of the channels, flows rapidly in one direction, and emerges as the raffinate. T h e strip, usually a concentrated acid solution (although a salt of a univalent cation may also be used), enters the other channel, flO\YS slonly in the opposite direction, and emerges as the product stream. As feed flows through the channels, the cation originally in the feed exchanges across the membrane with the hydrogen ion from the strip in the approach to the equilibrium represented by Equation 2 . T h e raffinate emerges as a dilute acid solution; the product stream emerges as a roncentrated solution of the cation originally in the feed.
> ZCCFgF
ZAAsg s
(5)
ZCCFgF
Condition 4 is satisfied for all values of cy 2 1. When cy = 1, conditions for complete recovery are just met and considerable losses to the raffinate are expected. However, these losses should decrease as a increases, since the excess of anion in the strip, occurring with larger values of cy, will provide a larger driving force for the transfer of the cation into the strip stream. I n the previous discussion, the membranes were assumed to be ideal and not to transport water. Unfortunately, this assumption is not valid for real membranes. As the strip stream progresses through the separator, water is transferred by osmosis from the feed stream into the strip stream, thereby decreasing the concentration in the strip and increasing its flow. Although the rate of osmosis is slow, the strip stream flow is also slow, so that considerable dilution can occur. Osmosis, therefore, limits the concentration of cation in the product. Concentration of Uranyl Ion. NITRICACIDSTRIP. Uranyl ion (UO?+) was concentrated from a 0.01M uranyl nitrate [UO,(NO,),] feed with a 2M nitric acid ("OB) strip in tests with the membrane separator (Table I). The feed flow was kept constant within separate groups of tests but cy was varied. Each test was run for 7 hours. Steady state was attained in all tests, except the first two a t feed flows of 5 ml. per minute and a a t 0.8 and 1.0. T h e attainment of steady state was demonstrated by measuring the uranium concentration in the raffinate as a function of time. Uranium losses to the raffinate decreased with increasing values of cy a t constant feed flow; the uranium concentration in the product remained fairly constant u p to cy = 1.5, then decreased with increasing cy. Uranium losses decreased with decreasing feed flow a t constant a, as expected. T h e concen-
Concentration of Uranyl Nitrate with Nitric Acid
UO2+ -
Feed, qF 5.0
3.0
.lTot
Flow, __. ?vfI./Min. Strifi,
Rajinate
Product,
g.3
gP
a
0.04 0.05 0.06 0.075 0.100 0.200
0,0995 0.112 0.142 0.176 0.230 0.339
0.80" 1. o a 1.2 1.5 2.0 4.0
0.111 0.141 0.181 0.344
1.2 1.5 2.0 5.30
0.036 0.045 0.060 0.1'60 at steady state.
Product concn., M 0.282 0.274 0.273 0.268 0.210 0.148 0.220 0.189 0.124 0.189 0.185 0.155 0.084
H i Concn., M
concn.,
7% of feed
Product 0.0126 0.0297 0.0466 0.109 0.269 0.551 0.0941 0.218 0.534 0.0207 0.0726 0.188 0.596
24.6 17.8 7.46 3.34 1.33 0.67 1.41 0.56 0.18 1.76 0.15 0.09