IE61021B1 - Separation of citric acid from fermentation broth - Google Patents

Description

The field of art to which this invention pertains is the solid feed adsorptive separation of dtric add from fermentation broths containing citric add, carbohydrates, amino adds, proteins and salts. More specifically^ the invention relates to a process for separating dtric add from fermentation broths containing same which process employs a non-zeolite polymeric adsorbent, which selectively adsorb dtric add, and is selected from the group consisting of a neutral, crosslinked polystyrene polymer, a nonionic hydrophobic polyacrylic ester polymer, a weakly basic anionic exchange resin possessing tertiary amine or pyridine functional groups, and a strongly basic anionic exchange resin possessing quaternary amine functional groups and mixtures thereof .

Citric add is used as a food addiilant, and in pharmaceutical, industrial and detergent formulations. The increased popularity of liquid detergents formulated with dtric add has been primarily responsible for growth of worldwide production of dtric add to about 320 million Kg per year which is expected to continue in the future.

Citric acid is produced by a submerged culture fermentation process which employs molasses as feed and the microorganism, Aspergillus Niger. The fermentation product will contain carbohydrates, amino adds, proteins and salts as well as dtric add, which must be separated from the fermentation broth.

There are two technologies currently employed for the separation of dtric add. The first involves caldum salt predpitation of dtric add. The resulting caldum dtrate is acidified with sulfuric add. in the second process, dtric add is extracted from the fermentation broth with a mixture of trilausyl-amine, n-octanol and a Cjq or Cj n isoparaffin. Citric add is reextracted from the solvent phase into water with the addition of heat. Both techniques, however, are complex, expensive and they generate a substantia] amount of waste for disposal Use patent literature has suggested a possible third method for separating citric acid from the fermentation broth, which involves membrane filtration to remove raw materials or high molecular weight impurities and then adsorption of contaminants onto a nonionic resin based on polystyrene or polyaerylic resins and collection of the citric add in the rejected phase or raffinate and crystallization of the dtric add after concentrating the solution, or by precipitating the dtric add as the caldum salts then acidifying with H2SO4, separating the CaS0*4 and contacting cation- and anionHgxdrangers. Has method, disdosed in European Published Application No. 151,470, August 14,1985, is also a rather complex and lengthy method for separating the dtric add. In contrast, the present method makes it possible to separate the citric add in a single adsorption step and to recover the dtric add from the adsorbent to obtain the purified dtric acid using an easily separated desorbent.

This invention relates to a process for adsorbing dtric acid from a fermentation broth onto a polymeric adsorbent selected from the group consisting of a neutral, crosslinked polystyrene polymer, a nonionic hydrophobic polyaerylic ester polymer, a weakly basic anionic exchange resin possessing tertiary amine or pyridine functional groups, and a strongly basic anionic exchange resin possessing quaternary amine functional groups and mixtures thereof aad thereafter recovering the dtric add by desorption thereof with a suitable desorbent under desorption conditions. One aspect of tire invention is in the discovery that highly selective separation of dtric add from salts and carbohydrates is only achieved by adjusting and maintaining the pH of the feed solution lower than the first ionization constant (pKaof dtric seid (3.15). The degree to which the pH must be lowered to maintain adequate selectivity appears to be interdependent on the concentration of dtric add in the feed maxtore; the pH is inversely dependent on the concentration. As concentrations are decreased below 13% to very low concentrations, the pH snay be near the pKa^ of dtric add of 3.13; at 13%, the pH may range from 0.9 to 1.7; however, at 40% dtric add feed concentration, the pH must be lowered to at least about 12 or lower. At higher concentrations, the pH must be even lower; for example, at 50% dtric add, the pH must be at or below 1.0. It is thus preferred to maintain the pH of the feed mixture in the range of 0 J to 25 with a range of 0-5 to 22 giving best results. Another aspect of the invention is the discovery that the temperature of adsorption can be reduced for the polymeric adsorbent used herein by the addition of acetone, or other low molecular weight ketone, to the desorbent; the higher temperatures associated with adsorbent breakdown can thus be avoided.

The invention also relates t© a pro-cess for separating citric add from a feed mixture conprishog a fermentation broth containing same, which process employs a polymeric adsorbent selected from the group consisting of a neutral, crosslinked polystyrene polymer, a nonionic hydrophobic polyacrylic ester polymer, lo ε weakly basic anionic exchange resin possessing tertiary amine er pyridine functional groups, and a strongly basic anionic exchange resin possessing quaternary amine functional groups and mixtures thereof which comprises the steps of: (a) maintaining net fluid flow thorugh a column of said adsorbent in a single direction, which column contains at least three nones having separate operational functions occurring therein and being serially interconnected with the terminal zones of said column connected to provide a continuous connection of said nones; (b) maintaining an adsorption zone in said column, said zone defined fey the adsorbent located between a feed input stream at an upstream boundary of said zone and a raffinate output stream at a downstream boundary of said zone; (c) maintaining a purification zone immediately upstream tai said adsorption zone, said purification zone defined by the adsorbent located between an extract output stream at an upstream boundary of said purification zone and said feed input stream ata downstream boundary of said purification zone; (d) maintaining a desorption zone immediately upstream ta said puriScafion zone, said desorption zone defined by the adsorbent located between a feesbeni input stream at an upstream bonndary of said zone and said extract output stream at a downstream boundary of said zone; (e) passing said feed mixture into said adsorption zone at adsorption conditions to effect the selective adsorption of said citric arid by said adsorbent in said adsorption zone and withdrawing a raffinate output stream comprising the nonadsorbed components of said fermentation broth from said adsorption sooeae; (f) passing a desorbent material mt© said desorption sons at desorption conditions to effect the displacement of said citric acid from the adsorbent in said desorption zone; (g) withdrawal an extract output stream comprising said citric add and desorbent material from said desorption sone; (h) passing at feast a portion of said extract ©Wet stream to a separation means and therein separating at separation conditions at least a portion oi said desorbeat material; as.d, (i) periodically advancing through said column of adsorbent in a downstream direction with respect to fluid flow in said adsorption zone the feed input stream, raffinate output stream, desorbent input stream, and extract output ' stream to effect the shifting of zones through said adsorbent asd the production ef extract output and raffinate output streams to produce a raffinate product having a reduced concentration of desorbent material Further, a buffer zone may fee maintained immediately upstream from said desorption zone, said buffer zone defined as the adsorbent located between the desorbent input stream at a downstream boundary of said buffer zone and the raffinate output stream at an upstream boundary of said buffer zone.

Other aspects of the invention encompass details of feed mixtures, adsorbents, desorbents and operating conditions which are hereinafter disclosed.

Figure 1 is a plot of concentration of various citric add species versus the pH of citric add dissociation which shows the shifting of the equfiibriura point of ths citric add dissociation by varying the concentration of citric add, citrate anions and the hydrogen ion.

Figure 2 is a static plot to determine the effect oi pH on amount of citric add that ca® be adsorbed by the adsorbent.

Figures 3A, SB and 3C are the plots of the pulse tests in Brasagde I using XAD-4 to separate citric add from a feed containing 13% citric add at pHfc rf 2.4, L7 and 0.9, re^ectivri^.

Figures 4A, 4B, 4C, 4D and 4E are plots of the pulse tests of Exanpfe H at pH’s of 2.4,17,0.9, X8 and 14, respestivdy, run on different adsmfrent samples.

Figures 5Α aad 5B are plots of the pulse tests of Example 3Π at pH's of 2.8 aad 14, respectively, aad temperatures ©f 93° C.

Figures ©A, 6B aad 6C are plots of the pulse tests of Example IV at pH’s of L94,1J3 aad 05, respectively.

Figures 7A 7B aad 7C are plots of the pulse tests of Example V at pH's of 1.82,0.5 aad 03, respective^.

Figures SA aad SB are plots ©f the pulse tests of Example VI at pH's ©f 1.5 aad 1.0, respectively.

Figure 9 is a plot ©f the pulse test ia Example VH showing the adsorption achieved at lower temperatures (93°C versus 45®C) through the incorporation of 10% acetoae in the desorbent water.

Figure 10 is the plot of the pulse test in Example VX0 using a weakly basic anionic exchange resin having a tertiary amine functionality in a cross-linked acrylic resin matrix to separate citric add from a feed containing 40% citric arid at a pH of 1.6, desorbed with water.

Figures IIA, HB and MC are plots of the pulse tests of Exaxople DC, at pH's of 7.0,35 and 2.4, respectively.

Figures 32,33A aad 33B are the plots of the pulse test of Example X at a pH of 1.6 run on several different adsorbent samples of weakly basic anionic exchange resin possessing pyridine functionality in a cross-linked polystyrene resin matrix. The citric arid is desorbed with 0.05N sulfuric arid or water.

Figure 34 is a plot of the pulse test of Example XB5 at a pH of 1.6.

Figure 15 is the plot of the pulse test of Example XIV at a pH of 22 run on a different adsorbent sample ©f a less strongly basic anionic exchange resin possessing quaternary ammonium functionality in a cross-linked potystyreme resin matrix, desorbed with dilute sulfuric arid. * At the outset the definitions ©(various tenns used throughout the verification will be useful in making dear the operation, objects and advantages of tfee instant process.

A Teed mixture* is a mixture containing one or more extract components and one or more raffinate components to be separated by the present process. The term Teed stream indicates a stream of a feed mixture which passes to the adsorbent used in the process.

An extract component is a compound or type of compound that is more selectively adsorbed hy the adsorbent while a raffinate component is a compound or type of compound that is less selectively adsorbed, in this process, citric add is an extract component and salts and carbohydrates are raffinate components. The term desorbent material shall mean generally a material capable of desorbing an extract component The term desorbent stream* or desorbent input stream indicates the stream through which desorbent materiel passes to the adsorbent Tne term raffinate stream or raffinate output stream means a stream through which a raffinate component is removed from the adsorbent The composition of the raffinate stream can vary from essentially 100% desorbent material to essentially 100% raffinate components. The term extract stream or extract output stream shall mean a stream through which am extract material which has been desorbed by a desorbent material is removed from the adsorbent The composition of the extract stream, likewise, can vary from essentially 100% desorbent material to essentially 100% extract components. At least a portion of the extract stream and preferably at least a portion of the raffinate stream from the separation process are passed to separation means, typically fractionators, where at least a portion of desorbent material is separated to produce an extract product and a raffinate product ’The terms extract product amd raffinate product mean products produced fey the process containing, respectively, am extract component and a raffinate component is higher concentrations than those found in the extract stream and the raffinate stream. Although it is possible by the process ©f this invention to produce a high purity, citric add product at high recoveries, it will be appreciated that an extract component is never completely adsorbed by the adsorbent likewise, a raffinate component is completely nonadsorbed Ity the adsorbent Therefore, varying amounts of a raffinate component can appear in the extract stream and, likewise, varying amounts of an extract component can appear in the raffinate stream. The extract and raffinate streams then are further distinguished from each other and from the feed mixture ity the ratio of the concentrations of an extract component and a raffinate component appearing in the particular stream. More spedfieaEy, the ratio of the concentration of dtric add to that of the fess selectively adsorbed ©eaaponests will be lowest in the raffinate stream, next highest in the feed mixture, and the highest in the extract stream. Likewise» the ratio of the concentration of the less selectively adsorbed components io that of the more selectively adsorbed citric add will be highest in the . raffinate stream, next highest in the feed mixture, and the lowest in the extract stream.

« The term selective pore volume of the adsorbent is defined as the volume of the adsorbent which selectively adsorbs an extract component from the feed mixture. The term nonselective void volume of the adsorbent is the volume of the adsorbent which does not selectively retain an extract component from the io feed mixture. This volume includes the cavities of the adsorbent which contain no adsorptive sites and the interstitial void spaces between adsorbent particles. The selective pore volume and the nonselective void volume are generally expressed in volumetric quantities and are of importance in determining the proper flow rates of fluid required to be passed into an operational zone for efficient operations to take place for a given quantity of adsorbent When adsorbent passes” into an operational zone (hereinafter defined and described) employed in one embodiment ©f this process its nonselective void volume* together with its selective pore volume carries fluid into that zone. The nonselective void volume is utilized in determining the amount of fluid which should pass into she same zone in a countercurrent 2o direction to she adsorbent so displace the fluid present in the nonselective void volume. If the fluid flow rate passing into a zone is smaller than the nonselective void volume rate of adsorbent material passing into that zone, there is a net entrainment of liquid into the zone by the adsorbent Since this net entrainment is a fluid present in nonselective void volume of the adsorbent, it in most instances comprises less selectively retained feed components. The selective pore volume of as adsorbent can in certain instances adsorb portions of raffinate material from the fluid surrounding the adsorbent since in certain instances there is competition between extract material and raffinate material for adsorptive sites within the selective pore volume. K a large quantity of raffinate material with respect to extract material surrounds the adsorbent, raffinate material can be competitive enough to be adsorbed by the adsorbent The feed material contenmlated in this invention is the fermentation product obtained from the submerged culture fermentation ©x molasses by the microorganism, Aspergillus Niger. The fermentation product will have a composition exemplified fey the following: ' Citric add Salts Carbohydrates (sugars) Others (proteins and amino adds) 22.9% + 3% 6,000 ppm 1% % The salts will be K, Na, Ca, Mg and Fe. The carbohydrates are sugars including glucose, xylose, mannose, oligosaccharides of DP2 aad DP3 plus as many as 12 or more unidentified saccharides. The composition of the feedstock may vary from that given above and still be used in the invention. However, juices such as dtrus fruit juices, are not acceptable or contemplated because other materials contained therein will be adsorbed at the same time rather than dtric add alone. Johnson, X MJE&2d£gri£» Vol 33 (3) pp 287-93.

It has now been discovered that the separation of dtric add can be enhanced significantly by adjusting the pH of the feed to a level below the first ionization constant of dtric add. The Sist ionization constant (pKaj) of dtric add is 3.13„H®jidtok^^^ 53rd Edition, 1972-3, CRC Press, and therefore, the pH of the dtric add feed should be below 3.13. When the pH for a 13% concentrated solution of dtric add is 2.4 or greater, for example, as in Figure 3A (Example 1), dtric add breaks through (is desorbed) with the salts and carbohydrates at the beginning of the cycle, indicating that all the dtric add is not adsorbed. In contrast, less break through of dtric add is observed when the pH is L7 aad ao break through when the pH is 0.9 at the 13% level, for example as ia Figures 3E and 3C, respectively. fe aqueous solution, unionised dtric add exists in equilibrium with the several ci tratsanions and bydrqgen ions. Inis is shown fe fee following equations, where fee add dissociation constants, pKaj, pKa^ and pKa^ of dtric add at 25®C axe 3 J3,4.74 aad 5.4¾ respectively: Equation 1 B.CA pKaj * 3.13 χ · --—-T p?(a^ « 4.74 _ —-y hca' S.40 The equilibrium poiat of dtric add dissociation can be shifted by vantiag tbs concentrations of dtric acid, the dtrate anion or the hydrogen ion. This fe demonstrated in Figure ζ for She concentration of the several dtric add spedes in solution versus pH at 90eC The result shows a higher percent of nonionized dtric add (H3GA) at a higher hydrogen ion concentration (lower pH). Decreasing the pH (raising the HT ion concentration) will introduce more nonionized dtric add while reducing tine citrate anionic spedes (HgCA', HGAZ and ΟΑή$ΐη rise sofctiesBL Based on the citric add equilibrium and the resin properties mentioned above, nonionized dins add wiO be separated from other Ionic series (including dtrate anions) ia the fermentation broths using the resin adsorbents described. However, for a higher dtric add recovery, a lower pH solution fe required. The static adsorption isotherm of a particular resin falling within the invention, Amberlite XAD-4, for dtric add was carried out at room temperature about 25°C> as ε function of feed pH. Figure 2 shows the results of the study. The results show adsoiption of the monionic dtric add as the pH is lowered. Without the intention of feeing limited by this explanation, it appears that the nonionic dtric add species in the solution fe preferentially adsorbed on the adsorbents of the present invention either through an add-base interaction mechanism or a hydrogen bonding mechanism or a mechanism based on a strong affinity for relatively hydrophobic spedes er a combination ef these medsanfem.

Desorbent materials used ia various prior ast adsorptive separation processes vaiy depending upon such factors as the type of operation employed. In the swing bed system, fo which the selectively adsorbed feed component fe removed from the adsorbent by a purge stream, desorbent selection fe not as critics! and desorbent materials comprising gaseous hydrocarbons such as methane, ethane, etc., or ©Sher types of gases such as nitrogen ©r hydrogen say be used as elevated temperatures or reduced pressures or both to effectively purge the adsorbed feed component from the adsorbent. However, in adsorptive separation processes which are generally operated continuously at substantially constant pressures and temperatures to ensure liquid phase, the desorbent material must be judiciously selected to satisfy many criteria. First, the desorbent material should displace an extract component from the adsorbent with reasonable mass flow rates without itself being so strongly adsorbed as to unduly prevent an extract component from displacing the desorbent material in a following adsorption cycle. Expressed in terms of the selectivity (hereinafter discussed in more detail), it is preferred that the ID adsorbent be more selective for all of the extract components with respect to a raffinate component than it is for the desorbent material with respect to a raffinate component Secondly, desorbent materials must be compatible with the particular adsorbent and the particular feed mixture. More specifically, they must not reduce or destroy the critical selectivity ef the adsorbent for an extract component with respect xo a raffinate component. Desorbent materials should additionally be substances which are easily separable from the feed mixture that is passed into the process. Both the raffinate stream and the extract stream are removed from the adsorbent in admixture with desorbent material and without a method of separating at least a portion of the desorbent material the purity of the extract product and tise raffinate product would not be veiy high, nor would the desorbent material be available for reuse in the process, ft is therefore contemplated that any desoibent material used in this process will preferably have a substantially different average telling point than that of the feed mixture to allow separation of at least a portion of the desorbent material from feed components in the extract and raffinate streams by simple fractional distillation thereby permitting reuse of desorbent material in the process. The term '’substantially different as used herein shall mean that the difference between the average boiling points between the desorbent material and the feed mixture shall te at least about 5°C The boiling range of the desorbent material may te higher or lower than that of the feed mixture. Finally, desoibent materials should also fee materials which are readily available and therefore reasonable an cost, fra the preferred isothermal, isobaric, liquid phase operation of tine process of the present invention, it has been found that water is a particularly effective desorbent material Also, it has been determined that acetone and ©ther tow molecular weight ketones, such as methylethyl ketone amd diethyl ketone to be effective in admixture with water ia small amounts, ap to 15 wl%. The key to their usefulness lies in their solubility in'water. Their advantage, however, lies in their ability to reduce the tenperature at which the desorption can take place. With some adsorbates and water as desorbent, the temperature must be raised to aid the desorption step. Increased temperatures can cause premature deactivation of the adsorbent A solution to that problem in this particular separation is to add acetone ia the amount of 1 to 15 wt.% of the desorbent, preferably, ί to 10 wt.% with the most preferred range of 5 to 30 wt.%. The low molecular weight ketone may also affect the adsorbent stability in possibly two ways, by removing solubilizing components which cause deactivation or by effecting regeneration, Le., by removing the deactivating agent or reversing its effect For example, a reduction of the desorption tenperature of this separation by approximately 50°C has been achieved by adding 10 wt.% acetone to the desorbent A reduction of from about 5°C to about 70®C can be achieved by the addition of 1 to 15 wi.% acetone to the water desorbent Dilute inorganic adds have also been found to give good results when used as desorbents. Aqueous solutions of sulfuric add, nitric add, hydrochloric add, phosphoric add and mixtures thereof can be used in amounts corresponding to 0.01 to ION (normal), with best results obtained with dilute sulfuric add at 0.01 to LON.

The prior art has also recognized that certain characteristics of adsorbents are highly desirable, if mot absolutely necessary, to the successful operation of a selective adsorption process. Such characteristics are equally important to this process. Among such characteristics are: (1) adsorptive capacity for some volume ©f an extract component per volume of adsorbent; (2) the selective adsorption of an extract component with respect to a raffinate component and the desorbent material; and (3) sufficiently fast rates of adsorption and desorption ©f an extract component to and from tfee adsorbent Capacity cf the adsorbent for adsorbing a specific volume of an extract component is, cf course, a necessity; without such capacity the adsorbent is useless for adsorptive separation. Furthermore, tfee higher the adsorbent's capacity for an extract conponent the better fe the adsorbent increased eapadiy tf a particular adsorbent makes it possible to reduce the amount of adsorbent needed t© separate an extract omnponent of known concentration contained in a particular charge rate of teed mixture. A reduction in the amount of adsorbent required for a specific adsorptive separation reduces die cost of the separation process. It fe important that the good initial capacity of the adsorbent 'be maintained during actual use in the separation process over some economically desirable life. The second necessary adsorbent characteristic is the ability of the adsorbent to separate components of the feed; or, in other words, that the adsorbent possess adsorptive selectivity, (B), for one component as compared to another component Relative selectivity can be expressed not only for one feed ©osoponemi as compared to another but caa also be expressed between any feed mixture component and the desorbent material Tse selectivity, (B), ss used throughout this ^stdfication fe defined as the rati© of the two components of the adsorbed phase over the ratio of the same two components ta the unadsorbed phase at equilibrium conditions. Relative selectivity fe shown as Equation 2 below; Selectivity = (3) = Equation 2 . £vok percent C/vol. percent £vol. percent C/vok percent Djy where C and D are two components of the feed represented in volume percent and the subscripts A and U represent the adsorbed and unadsorbed phases respectively. The equilibrium conditions were determined when the feed passing over a bed of adsorbent did not change composition after contacting the bed of adsorbent lh other words, there was no set transfer of material occurring between the unadsofbed and adsorbed phases. Where selectivity of two components approaches 1.0 there fe no preferential adsorption of one component by the adsorbent with respect to the other; they are both adsorbed (or nonadsorbed) to about the same degree with respect to each ©ther. As the (B) becomes less than oc greater than 1.0 there fe a preferential adsorption by the adsorbent for ©ae component with respect to the other. When comparing the selectivity by the adsorbent of one component C over component £>, a (B) larger than 1.0 indicates preferential adsorption of component C within the adsorbent A (B) less than 1.0 would indicate that component D fe preferentially adsorbed leaving an unadsofbed phase richer in component C and an adsorbed phase richer in component D. Ideally desorbent materials should have a selectivity equal to about 1 or slightly less than 1 with respect t© all extract components so that all of the extract components can be desorbed as a dass with reasonable flew rates of desorbeni material and se that extract components can displace desoibent material in a subsequent adsorption step. While separation of an extract component from a raffinate component is theoretically possible when the selectivity of the adsorbent for the extract component with respect to the raffinate component is greater than 1, it is preferred that such selectivity approach ε value of 2. like relative volatility, the higher the selectivity, the caster the separation is to perform. Higher selectivities permit a smaller amount of adsorbent to be esed. The third important characteristic is the rate of exchange of the extract component of the feed mixture material or, in other words, the relative rate of desorption of the extract component This characteristic relates directly to the amount of desorbent material that must be employed in the process to recover the extract component from the adsorbent; faster rates of exchange reduce the amount ©f desorbent material needed to remove the extract component aad therefore permit a reduction in the operating cost of the process. With fester rates of exchange, less desorbent material has to be pumped through the process and separated from the extract stream fer reuse in the pro-osss.

Resolution is a measure of the degree ef separation of a two* component tystem, and eaa assist in quantifying the effectiveness of a particular combination ©f adsorbent, desorbent, conditions, etc. for a particular separation. Resolution for purposes ef this application is defined as the distance between the two peak centers divided by the average width of the peaks at 1/2 the peak height as determined by the pulse tests described hereinafter. The equation for calculating resolution is tibsis: Lg ~ M “ΤΤΟνπςι— where Lj and are the distance, in mi, respectively, from a reference point, e.g, zero to the centers of the peaks aad Wj and W2 are the widths of the peaks at 1/2 the height of the peaks.

A dynamic testing apparatus fe employed to test various adsorbents with a particular feed mixture and desorbent material fe measure the adsorbent la characteristics of adsorptive capacity, selectivity and exchange rate. The apparatus consists of an adsorbent chamber comprising a helical column of approximately 70 cc volume having inlet and outlet portions at opposite ends of the chamber. The chamber is contained within a temperature control means and, in addition, pressure control equipment is used to operate the chamber at a constant predetermined pressure. Quantitative and qualitative analytical equipment such as refiractometers, polaximeters and chromatographs can be attached to the outlet line of the chamber and used to detect quantitatively or determine qualitatively one or more components in the effluent stream leaving the adsorbent chamber. A pulse test, performed using this apparatus and the following general procedure, is used to determine selectivities and other data for various adsorbent systems. He adsorbent is filled to equilibrium with a particular desorbent material hy passing the desorbent material through the adsorbent chamber. At a convenient time, a pulse of Seed containing known concentrations of a tracer and of a particular extract component er of a raffinate component or both, all diluted in desorbent, is injected for a duration of several minutes. Desorbent flow is resumed, and the tracer and the extract component or the raffinate component (or both) are eluted as in a liquidsolid chromatographic operation. The effluent can be analyzed onstream or, alternatively, effluent samples can be collected periodically and later analysed separately by analytical equipment aad traces of the envelopes of corresponding component peaks developed.

From information derived from the test adsorbent, performance can be in terms of void volume, retention volume for an extract or a raffinate component, selectivity for one component with respect to the other, and the rate of desorption of an extract component by the desorbent. Tse retention volume of an extract or a raffinate component may be characterized by the distance between the center of the peak envelope of an extract or a raffinate component and the peak envelope of the tracer component or some other known reference point. It is expressed in terms of the volume in cubic centimeters of desorbent pumped during this time interval represented by the distance between the peak ewelopes. Selectivity, (B), for an extract component with respect to a raffinate component may be characterized by the ratio of the distance between the center of the extract component peak envelope and the tracer peak envelope (or other reference point) to the corresponding distance between the center ©f the raffinate component peak envelope and the tracer peak envelope. The rate of exchange of an extract component with the desorbent can generally be characterized by the width of the peak envelopes at half intensity. The narrower the peak width, the faster the desorption rate. The desorption rate can also be characterized by the distance between the center of the tracer peak envelope and the disappearance of an extract component which has just been desorbed. This distance is again the volume of desorbent pumped during this time interval.

T© further evaluate promising adsorbent systems and to translate this type of data into a practical separation process requires actual testing of the best system in a continuous countercurrent liquid-solid contacting device. The general operating principles of such a device have been previously described and are found in Broughton U.S. Patent 2,985,589. A specific laboratory size apparatus utilizing these principles is described in deRosset et ah, U.S. 3,706,812. The equipment comprises multiple adsorbent beds with a number of access lines attached to distributors within the beds and terminating at a rotary distributing valve. At a given valve position, feed and desorbent are being introduced through two of the lines and the raffinate and extract streams are being withdrawn through two more.

All remaining access lines are inactive and when the position of the distributing valve fe advanced by one index, all active positions will be advanced by one bed.

This simulates a condition in which the adsorbent physically moves in a direction counteroirrent to the liquid Sow. Additional details on the above-mentioned nonionic adsorbent testing apparatus and adsorbent evaluation techniques may be found in the paper Separation of Cg Aromatics by Adsorption by A, J. deRosset, R. W. Neuzil, D. J. Korous, and D. H. Rosback presented at the American Chemical Society, Los Angeles, California, March 28 through April 2,1971 One class of adsorbents to be used in the process of this invention will comprise nonionqgenic, hydrophobic, water-insoluble, crosslinked styrenepoly(vi23yl)benzene copolymers and copolymers thereof with monoethylesticalty unsaturated eompounds or poJ^etaylenicaMy unsaturated monomer ©ther tihan poly(vmyl)benzenes, including the acrylic esters, such as those described in Gustafson U.S. Patent Nos- 3,531,463 and 3,663,467, although not limited thereto. As stated in U.S. Patent No. 3,531,463, the polymers may be made by techniques disclosed in U.S. Serial No. 749,526, filed July %3958, now Patent Nos. 4,221,872; 4,224,415; <256,840; <297,220; <382,124 and 4,501,826 to Meitzner et al.

Adsorbents sueh as just described are manufactured by the Rohm and Haas Company, and sold under the trade name Amberlite. The types of Amberlite polymers known to be effective for use by this invention are referred to in Rohm and Haas Company literature as Amberlite adsorbents XAD-1, XAD-2, XAD-4, XAD-7 and XAD-8, and described in the literature as hard, insoluble spheres of high surface, porous polymer. The various types of Amberlite polymeric adsorbents differ somewhat in physical properties such as porosity volume percent, skeletal density and nominal mesh sizes, but snore so in surface area, average pore io diameter and dipole moment. Ihe preferred adsorbents will have a surface area of MP20G0 square meters per gram and preferably from 100-100C mx/g. These properties are listed in the following table: TABLE 1 Properties of Amberlite Polymeric Adsoroeni Chemical Nature XAJM Polystyrene ? XAD-2 Olystyrene VMh Λ ΛΜΙί’»*·® Polystyrene V ft « Acrylic : Ester «ei ίδ# Acryl' Estei Porosity folamg % 0*i*7' (%£> 42 IC,’a *4 ab 55 IE 5» ie’i Tra® Het Density graas/cc 1.02 1.02 *5 fiLi> <4 ο» V £» H iftE, rib «Β W 1.09 10 Surface Area H”/gram 100 300 780 450 160 Average Pore ΙΙΗ&βοβ An g s> t ter 200 §0 50 90. / f Si 15 Skeletal Etensity graas/cc 1.07 1.07 1.03 1.24 1.23 tomal Mesh Size 20-50 20-50 20-50 20-50 25-50 Dipole Moment ©f Functional Groups 0.3 0.3 0.3 1.8 1.8 Applications for Amberlite polymeric adsorbents suggested in the Rohm and Haas Company literature include decolorizing piilp mill bleaching effluent, decolorizing dye wastes and removing pesticides from waste effluent There is, of course, no hint in the literature of my surprising discovery of the effectiveness of Amberlite polymeric adsorbents in the separation of citric acid from Aspergillus-Niger fermentation broths.

A second class ©f adsorbents to be used in the process of this invention will comprise weakly basic anion exchange resins possessing tertiary amine or pyridine functionality fa a cross-linked polymeric matrix, e.g., aesy&c or styrene. They are especially suitable when produced in bead form, have a high degree of uniform polymeric porosity, exhibit chemical and physical stability and good resistance to attrition (not common to macroreticular resins).

Adsorbents such as just described are manufactured by the Rohm and Haas Company, and sold under the trade name Amberlite. The types of Amberlite polymers known to be effective for use by this invention are referred to in Rohm and Haas Company literature as Amberlite adsorbents XE.-275 (IRA-35), IRA-68, and described in the literature as insoluble in all common solvents and having open structure for effective adsorption and desorption of large molecules without loss of capacity, due to organic fouling.” Also, suitable are AG3-X4A and AG4-X4 manufactured by Bio Rad and comparable resins sold fey Dow Chemical Co., such as Dowex 66, and Dow experimental resins made fa accordance with U.S. Patents 4,031,038 and 4,098,,867.

The various types of polymeric adsorbents of these classes available, will differ somewhat in physical properties such as porosity volume percent, skeletal density and nominal mesh sizes, and perhaps more so in surface area, average pore diameter and dipole moment. The preferred adsorbents will have a surface area of 10-2000 square meters per gram and preferably from 100-1000 nrf/g. Specific properties of the materials listed above can be found in company literature and technical brochures, such as those in the following Table 2 which sre incorporated herein by reference. Others of the general class are also available.

IAB1E2 .Masderi&Ee AG3-4A (Bio Rad) Polystyrene Chromatography Electrophoresis Immunochemistry Molecular Biology-HPLC - Price List M April 1987 (Bio-Rad) AG4-X4 Acrylic Chromatography Electrophoresis Inununochemistrv Molecular Biology - HPLC - Price List M April 1987 (Bio-Rad) ftw Experimental Resins Polystyrene U.S. Patent Nos. 4.031,038 and 4,098,867 Dower 66 Polystyrene Material Safetv Data Sheet Printed 2/17/^7 (Dow Chemical USA) IRA-35 (XE-275) Acrylic Amberlite Ion Exchange Resins (XE-275) Rohm & Haas Co. 1975 IRA-68 Acrylic Amberlite ion Exchange Resins Amberlite IRA-68 Rohm & Haas Co. April 1977 Applications for Amberlite polymeric adsorbents suggested in tiie Rohm and Haas Company literature include decolorizing pulp mill bleaching effluent, decolorizing dy© wastes and removing pesticides from waste effluent There is, of course, no hint in the literature of any surprising discovery of the effectiveness of Amberlite polymeric adsorbents in the separation of citric add from AspcigQlus-Higer fermentation broths.

A third class of adsorbents to be used in the process of this invention will comprise strongly basic anion exchange resins possessing quaternary ammonium finactiosality in a cross-linked polymeric matrix, divarylbenzene cross-linked acrylic or styrene resins. They are espedaJly suitable when produced in bead form and have a high degree erf uniform polymeric porosity and exhibit chemical aad physical stability. In the instant case, the resins can be gelular (or gel-type) or maororeticulaf' as the term is used in some recent literature, namely Eunin aad Hetherington, l^acrorgik?jlar>jL^chmigeSejiflSe paper presented at the International Water Conference, Pittsburg, PA, October 1969, reprinted by Rohm & Haas Co. In recent adsorption technology, the term nucroreticular refers to the gel stmcture per se, size ©f the pores which are of atomic dimensions and depend upon the swelling properties of the gel while macroreticular pores and trae porosity refer to structures in which the pores are larger than atomic distances and are not part of the gel structure. Their size and shape are not greatly influenced by changes in the environmental conditions such as those that result in osmotic pressure variations” while the dimensions ©f gel structure are markedly dependent upon the environmental conditions. In classical adsoiption the terms microporous and macroporous normally refer to those pores, less than 20 A and greater than 200 A, respectively. Pores of diameters between 20 A and 200 A are referred to as transitional pores. The authors selected the term macroreticular”, instead, to apply to the new ion exchange resins used in this invention, which have both a microreticular as well as a macroreticular pore structure. He former refers to the distances between the chains and crosslinks of the swollen gel structure and the latter to the pores that are not part of the actual chemical structure. The macroreticular portion of structure may actually consist of micro-, macro-» and transitional-pores depending upon the pore size distribution. (Quotes are from page 1 of the Kunin et ah artide). The macroreticular structured adsorbents also have good resistance to attrition (not common to conventional masTwetiosfer resins). In this application, therefore, all reference to maOToreticuliar indicates adsorbent of the types described above having the dual porosity defined by Kunin and Hethesing. GeF and gel-type are used in their conventional sense.

Looking at both the quaternary ammonium function-containing ion exchange resins of the invention, the quaternary amine has a positive charge and can form an sortie bond with the sulfate ioa The sulfate form of quaternary anMnomum anion exchange resin has a weakly basic property, which in turn, can adsorb citric add through an add-base interaction.

CR)2 (c2h4®) ?' where P * resinous aoiety E * Tow@r C«,e,s C.A,., * cUrate io>o Adsorbents such as just described are manufactured by the Rohm and Haas Company, and sold under the trade name Amberlite. The types ©f Amberlite polymers known to be effective for use by this invention are referred to in Rohm and Haas Company literature as Amberlite IRA 400 and 900 series adsorbents described in the literature as insoluble in all common solvents, qxn structure fer effective adsorption and desorption of large snoleoiles without fess of capacity, due to organic feulisg? Also suitable are AG1, AG2 and AGMP4 resins manufactured by Bio Rad and comparable resins sold by Dow Chemical Co, such as Dowex 1,2, II, MSA-1 and MSA-2, etc. Also useful in this invention are the socalled intermediate base ion exchange which are mixtures of strong and weak base exchange resins. Among these are the following commercial^ available resins: BioRex 5 (Bio-Rad 1); Amberlite IRA-47 and Duolite A-340 (both Rohm & Haas).

For example, they may be useful where a basic ion exchange resin is needed which is not as basic as the strong base resins, or one which is more basic than She weakly task resins.

The various types of polymeric adsoxbents of these dasses available will differ somewhat in physical properties such as porosity volume peroent, skeletal density asd nominal twssb sizes, asd perhaps more so fo siariace area, average pore diameter and dipole moment The preferred adsorbents will have a surface area of -2000 square meters per gram and preferably from 100-1000 mA/& SpedSc properties of the materials listed above can be found in company literature and technical brochures, such as those mentioned fo the following Table 3 .

TABLE 1 PROPERTIES OF ADSORBENTS Adsorbent Matrix Resin Type IRA 453 (Rohm & Haas5 Acrylic gel-type IRA 953 Acrylic macroporous IRA 900 Polystyrene macroporous IRA 904 Polystyrene macroporous IRA 910 Polystyrene macroporous IRA 400, 402 Polystyrene macroporas IRA 410 Polystyrene gel-type AG 1 (Bi® Rad) Polystyrene gel-type AS 2 Polystyrene gel-type AS-HFM Polystyrene sacroporc'us Bi© Rex E (Bi© Rad) Mixture of strong base and weak base resins (e.g. AG-2 and A6-3 or AS-4 Reference to Company Literature Amberlite Ion Exchange Resins 1986 & Technical Bulletin IE-207-74 34 Technical Bulletin and Material Safety Data Sheet ar© available Technical Sullstin is available and Amber]it® Ion Exchange Resins,, IE-3,00-66.

Technical Bulletin,, 1979 and IE-208/74, Jan. 1974 Technical Bulletin,, 1979 and IE-101-66,, May 1972 Amberlite Ion Exchange Resins,, Oct., Seot. 1976, Aon! X972 aod IE-S9-S2, October 1975 Amberlite Ion Exchange Resins IE-72-53,, August 1970 Chromatography Electrophoresi s Irnrnunochemistry Molecular Biology HPLC, Price List M April 1987 Chromatography El ectrophoresi s Immunochemistry Hoiecu]ar Biology HPLC,, Price List M April 1987 Chrcroatography Electrophoresls Immunochemi stry Hoiecu]ar Biology HPLC, Price List H April 1987 Chromatography El eet whores 1 s Immunochemi stry Mol ecu! ar Biology HPLC, Price List M April 1987 The adsorbent may be employed in the form of a dense compact feed feed which is alternatively contacted with the feed mixture and desorbent materials, in the simplest embodiment of the invention the adsorbent is employed in ih® fonn of a single static bed in which case the process is only semicontinuous. ϊη another embodiment a set of two or more static beds may fee employed in fixed bed contacting with appropriate valving so that the feed mixture is passed through one or more adsorbent beds while the desorbent materials can be passed through one or more of the other beds in the set Tse flow of feed mixture and desorbent materials may be either up or down through the desorbent Any of the conventional apparatus employed in static bed fluid-solid contacting may be used.

Countercurrent moving bed or simulated moving bed countercurrent flow ^sterns, however, have a much greater separation efficiency than fixed adsorbent bed systems and are therefore preferred. In the moving bed or simulated moving bed processes the adsorption and desorption operations are continuouriy taking place which allows both continuous production of an extract and a raffinate stream and the continual use of feed and desorbent streams. One preferred embodiment of this process utilizes what is known in the art as the simulated moving bed countercurrent flow system. The operating principles aad sequence of such a flow system are described in U.S. Patent 2,985,589.

In such a system it is the progressive movanent of multiple ligui access points down an adsorbent chamber that simulates the upward movement of adsorbent contained in the chamber. Only four of the access lines aw active at any one time; the feed input stream, desorbent inlet stream, raffinate outlet stream, and extract outlet stream access lines. Coincident with this simulated upward movement of th®. solid adsorbent is the movement of the liquid occupying the void volume of the packed bed of adsorbent So that countercurrent contact is maintained, a liquid flow down the adsorbent chamber may be provided by ε pump. As an active liquid access point moves through a cycle, that is, from the top of the chamber to the bottom, the chamber circulation pump moves through different sones which require different Slow sates. A programmed Slow controller may be provided to set and regulate these flow sates.

The active liquid access points effectively divided the adsorbent chamber into separate zones, each of which has a different function. in. this embodiment of my process it Is generally necessary that three separate operational zones be present in order for the process to take place although in some instances an optional fourth zone may fee used.

The adsorption zone, zone 1 is defined as the adsorbent located between the feed inlet stream and the raffinate outlet stream. In this zone, the feedstock contacts the adsorbent, extract component is adsorbed, and a raffinate stream is withdrawn. Since the general flow through zone 1 is from the feed stream which passes into the zone to the raffinate stream which passes out of the zone, the flow in this zone is considered to be a downstream direction when proceeding Sran the feed inlet to the raffinate outlet streams.

Immediately upstream with respect to fluid flow in zone 1 is the purification zone, zone 2. The purification zone is defined as the adsorbent between the extract outlet stream and ths feed inlet stream. The basic operations taking place in zone 2 are the displacement from the nonselective void volume of the adsoibent of any raffinate material carried into zone 2 by shifting of adsorbent into this sone and the desorption of any raffinate material adsorbed within the selective pore volume of the adsorbent or adsorbed on the surfaces of the adsorbent particles. Purification is achieved by passing a portion of extract stream material leaving zone 3 into zone 2 at zone 2’s upstream boundary, the extract outlet stream, to effect the displacement of raffinate material The flow of material in zone 2 is in a downstream direction from the extract outlet stream to the feed inlet stream. immediately upstream of zone 2 with respect to the fluid flowing in zone 2 is the desorption zone or zone 3. The desorption zone is defined as the adsoibent between the desorbent inlet and the attract outlet stream. The function of the desorption zone is to allow a desorbent material which passes into this zone to displace the attract component which was adsorbed upon the adsorbent during a previous contact with feed in zone 1 in a prior cycle of operation. The flow of fluid in zone 3 is essentially in the same direction as that of zones I and 2.

In some instances an optional buffer zone, zone 4, may be utilized. Ibis zone, defined as the adsorbent between the raffinate outlet stream and the desorbent inlet stream, if used, is located immediatety upstream with respect to· the fluid flow to zone 3. Zone 4 would be utilized to conserve the amount of desorbent utilized in the desorption step since a portion of the raffinate stream which is removed from zone ϊ can be passed into zone 4 to displace desorbent material present in that zone out oi that sone into the desorption zone. Zone 4 will contain enough adsorbent so that raffinate material present in the raffinate stream passing out of zone 1 and into zone 4 can be prevented from passing into zone 3 thereby contaminating extract stream removed from zone 3. in the instances which the fourth operational zone is not utilized the raffinate stream passed from zone 1 to zone 4 must be carefully monitored in order that the flow directly from zone 1 to zone 3 can be stopped when there is an appreciable quantity of raffinate material present in the raffinate stream passing from zone Ϊ into some 3 so that the extract outlet stream fe sot contaminated.

A cyclic advancement of the input and output streams through the io fixed bed of adsoibent can be accomplished by utilizing a manifold system ia which the valves in the manifold are operated in a sequential manner to effect the shifting ©f the input and output streams thereby allowing a flow of fluid with respect to solid adsorbent in a countercurrent manner. Another mode of operation which can effect the countercurrent flow of solid adsorbent with respect to fluid involves the use of a rotating disc valve in which tbe input and output streams are connected to the valve and the lines through which feed input, extract output, desorbeat input and raffinate output streams pass are advanced in the same direction through the adsorbent feed. Both the manifold arrangement and disc valve are known im the art. Specifically rotary disc valves which can be utilized in this operation can be found im UJS.

Patents 3,040,777 and 3,422,848. Both of the aforementioned patents disclose a rotary type connection valve in which the suitable advancement of the various input and output streams from fixed sources can be achieved without difficulty.

Is many instances, one operational zone will contain a much larger quantity of adsoibent than some other operational zone. For instance, in some operations the buffer zone can contain a minor amount of adsorbent as compared to the adsorbent required for the adsorption and purification zones. It cam also be seen that in instances in which desorbent fe used which cam easily desorb extract material from the adsorbent that a relatively small amount of adsorbent will be needed in a desorption zone as compared to the adsorbent needed im the buffer zone or adsorption zone or purification zone or ah of them. Since it is not required that the adsorbent be located ia a single column, the use ef multiple chambers or a scries of columns fe within the scope of the invention.

It is not necessary that all of the input or output streams be simultaneously used, amd in feet, ia many instances ome of the streams cam be shut off while others effect an input or output of material. The apparatus which can be utilized to effect the process of this invention can also contain a series of individual beds connected by connecting conduits upon which are placed input or output ws to which the various input or output streams can be attached and alternately and periodically shifted to effect continuous operation. In some instances, the connecting conduits can be connected to transfer taps which during the normal operations do not function as a conduit through which material passes into or out of the pro-cess.

It is contemplated that at least a portion of the extract output stream will pass into a separation means wherein at least a portion of the desorbent material can be separated to produce an extract product containing a reduced concentration of desorbent material· Preferably, but not necessary to the operation ©f the process, at least a portion of the raffinate output stream will also be passed to ε separation means wherein at least a portion of the desorbent material can be separated to produce a desorbent stream which can be reused in the process and a raffinate product containing a reduced concentration of desorbent material· The separation means will typically be a fractionation column, the design and operation of which is well-known to the separation art Reference can be made to D. B. Broughton U.5, Patent 2,98538¾ and to a paper entitled Continuous Adsorptive Frocessing~A New Separation Technique by D. B. Broughton presented at the 34th Annual Meeting of the Society of Chemical. Engineers at Tokyo, Japan on April 2, 1969, for further explanation of the simulated moving bed countercurrent process flow scheme.

Although both liquid and vapor phase operations can be used in many adsorptive separation processes, liquid-phase operation is preferred tor this process because ©f the lower temperature requirements and because of the higher yields of extract product than can be obtained with liquid-phase operation over those obtained with vapor-phase operation. Adsorption conditions will indude a temperature range of from. 20°C to ’ '2O0°C with ' 65°C to . 1O0°C being more preferred and a pressure range ef from atmospheric to .: 500 psig (3450 kPa gauge) being more preferred to ensure liquid phase.

Desosptioa conditions will include the same range of temperatures and pressures as used for adsoiption conditions.

The size of the waits which can utilize the precess of this invention can vary anywhere from those of pilot plant scale (see for example our assignee’s UJS.

Patent 3,706,812, to those of commercial scale and can range in flow rates from as little as a few cc an hour up to many thousands of gallons per hour.

The following examples are presented to illustrate the selectivity relationship that makes the process of my invention possible. The example is not intended to unduly restrict the scope and spirit of claims attached hereto. ϊη this example, three pulse tests were run with a neutral styrene divinylbenzene polymeric adsorbent (XAD-4 made by Rohm & Haas 'Company) to determine the ability of the adsorbent to separate citric add, at different pH's, from its fermentation mixture of carbohydrates (DPI, DP2, DPS, including glucose, xylose, arabinose and raffinose) and ions oi salts, including Na’, W, Mg* +, Oa+ * *, CF, SO4=, PO/5 and NCXf, amino adds and proteins. The first test was nm at a pH of 2.4 aad 45°C Two further tests were ran at a pH of 1.7 and 0.9. Citric add was desorbed with water. ‘The fermentation feed mixture had the foOosring eonmssMoa: .Feed .CMwestw Joanm Gtac Add 125% Salts Na\ Ca*Mg* + Fe+ + *> 0.60% (6Ό ppm) Carbohydrates (Sugars) 1% Others (SO^, CT, P04s, NOf, proteins and amino ados) . 5% Water 815% Retention volumes and resolution were obtained using the pulse test apparatus and procedure previously described. Specifically, th® adsorbent was tested ia a 70 cc straight column using the following sequence of operations for the pulse test Desorbent material was continuously run upwardly through the column containing the adsorbent at a nominal liquid hourly space velocity (LHSV) of about 1.0. A void volume was determined by observing the volume of desorbent required to fill the packed dry column. Ai a convenient time the flow of desorbent material was stepped, and a 10 cc sample of feed mixture was injected into the column via a sample loop and the flow of desorbent material was resumed. Samples of the effluent were automatically collected in an automatic sample collector and later analyzed for salts and citric add by chromatographic analysis. Some later samples were also analyzed for carbohydrates, but since they were eluted at approximately the same rate as the carbohydrates, they were not analyzed in these examples nor io were other minor ingredients, amino acids and proteins. From the analysis of these samples, peak envelope concentrations were developed for the feed mixture components. The retention volume for the citric acid was calculated by measuring tine distance from the midpoint of the net retention volume of the salt envelope as the reference point to the midpoint of the citric add envelope. The resolution, R, is calculated from Equation 3, given earlier. The separation factor, B, was calculated as previously indicated.

He results for these pulse tests are shown in the following Table 4. ,il Test A - pH - 2.4 Het retention Component Volume Salts 0 Citric add 44.4 Test B - pH - 1.7 Het retention Component Volume Salts 0 Citric add 4.2.2 Test C » pH 0.9 Het retention Component Volume Salts 0 Citric acid 40.,9 Peak Width itesoH'twn at 0.5 Height (0.5 Height) 14.4 2.39 49.5 Reference Pes.k Width at 0.5 Height (0.5 Height) 11.,6 2.49 45.2 Reference Peak Width at 0.5 Height Resolytion (0.5 Height) 23.3 1.,4 45.1 Reference The results are also show® in Figure 3A fc which il is dear that whale citric add fe more stropgty adsorbed thaa the other ©pmponeats, there fe a substaatial loss of citric add which is unadsorbed aad removed with the salts and 2o carbohydrates (not shows). Citric acid fe satisfactorily separated fc the process ia Figure 3B where the results are judged good aad fc Figure 3C where the results are judged excellent. The process clearly will have commercial feasibility ai a pH of 1.7 aad lower. At a pH of X4 (Figure 3A), however, it fc noted that a substantial amount of the citric add will be recovered fc the form of the citrate, HjCA^, in the raffinate with the salts and carbohydrates. From this, it fe evident that the ionized, 3ί soluble species should be reduced, as explained previous^, by maintaining & lower pH in the feed, thereby driving the equilibrium in Equation 1 to «fee left.

JEXAWLOi This example presents the results of using a neutral crosslinked styrene diviitylbenzene (XAD-4) and a neutral crosslinked polyacrylic ester copotymer (XAD-8) with the same separation feed mixture as Example ί at different pHs to demonstrate the poor separation when the pH is 2.4 or higher, or above the first ionization constant, pKa j - 3.13, of citric add. The same procedure and apparatus previously described in Example I were used in the separation, except the temperature was 60^C and 5 ml of feed mixture was used.

Figures 4A, 4B and 4C are, respectively, graphical presentations of the results of the pulse tests using XAD-4 at pHs, respectively, of 2.4,1.7 and 0.9.

Figure 4A shows that citric add breaks through with the salts (and carbohydrates). This problem can be partially alleviated by lowering the pH to 1.7 as in Figure 4B. An excellent separation can be achieved by lowering the pH further to 0.9 as in Figure 4C. This separation, with adjustment of the pH, again, dearly has commercial utility.

Figures 4D and 4E are, respectively, graphical representations of the results of pulse tests, rum under the same conditions as above, using XAD-8 at pHs of 2.8 and 1.4 and temperatures of 65°C Figure 4D, which was made ax a pH of 2.8S shows no separation, but rather the salts, carbohydrates and dtric add during together MtiaSiy. After about 67 ml, after most of fee carbohydrates and salts and some of the dtric add have been recovered, some relatively pure dtric add can be obtained, but recovery fe low. Figure 4Ε» which was made at a pH of 1.4, shows a selectivity between dtric add and carbohydrates and salts which results in a satisfactory separation and recovery of the dtric add.

WAWIEIB This example presents the results of using a neutral crosslinked styrene divinylbenzene copolymer (XAD-4) with the same separation feed mixture as Example 3 at two different pHs to demonstrate the poor separation when the pH is 2.4 or higher. The same procedure and apparatus previously described in Example ί were used, except the temperature was 93°C in Figures 5A and SB and the amount of feed mixture was 10 ml.

Figures 5 A and 5jB are, respectively, graphical presentations of the results ©f pulse tests using XAD-4 at pHs, respectively, of 2.8 and 1.4. Figure 5A shows that citric add breaks through” with the salts and carbohydrates. This problem can be alleviated by lowering the pH to 1.4 as in Figure SB. This separation, with adjustment of the pH, again, clearly has commercial utility.

EXAMgmiv The procedure and apparatus previously described in Example I was used on the samples of this example. The temperature was 60°C and 5 ml of feed mixture was used. The feed composition was similar to that previously used except that citric arid has been concentrated to 40% in the feed mixture. The effect of concentration on the pH will be seen. In. Figure 6A, even with the temperature at 60°C, the pH of 1.9 is too high to separate the ritric arid at 40% concentration. By adjusting the pH downward as in Figures 6B and 6C, the ritric arid is preferentially adsorbed and excellent separation is achieved at a pH of U3 and at a pH of 05. In each of these samples, caibohydrates were not analyzed, but it can be assumed that the carbohydrates riosety followed the salts ia the separation.

.EXAWLBV The procedure and apparatus previously described in Example I was used on the three samples of this example. Tne temperature was 93°C and the amouat of feed mixture was 5 nd. The feed composition was similar to that previously used except that ritric arid has been concentrated to 40% in the feed mixture t© demonstrate the further effect of concentration on the pH. In Figure 7A, even with the tenperature at 93®C, the pH of 1.8 is too high to separate the ritric arid at 40% concentration. By adjusting the pH downward as in Figures 7B and 7C, the ritric arid is preferentially adsorbed and excellent separation is achieved.

Again, carbohydrates were not analyzed, but it can be assumed that the carbohydrates closely followed the salts in the separation.

EXAMPLE VI The pulse test of Example I was repeated on two 50% dtric add samples «sing XAD-4 adsorbent The desoibent in both cases «as water. The composition of the feed used was the same as used in Example Ϊ except that dtric add has been concentrated to 50%. The temperature was 93°C In the first sample, the pH was 15. As shown in Figure 8A, dtric add was not separated. After reducing the pH to 1.0 in the second sample, dtric add was readily separated as seen in Figure 8B. Again, carbohydrates were not analyzed, but assumed to closely follow the salts. The separation in Figure SB was judged good. exampwvb The separation example represented by Figures TB and 7C required high temperatures, e-g., 93°C to achieve the separation of 40% dtric add due to the difficulty in desorbing dtric add from the XAD-4 adsoibent In this example, high temperatures, which adversely affect the adsorbent life and the eost to operate, are eliminated and the separation is readily achieved at 45°C through the use of a desorbent mixture of 10% (by wt) acetone'amd 90% water. Referring to Figure 9, a feed comprising 40% (wt.) dtric add, 4% carbohydrates and 2% salts of the following elements: KT, Na*, Mg* *, Fe* * *, Ca* * plus proteins and amino adds, was introduced into the pulse test apparatus as set forth previously and the test ran as before except that the temperature was 45°C. In this test th© pH was maintained -at 05, but the desorbent contained acetone as mentioned above. The net retention volume for citric add was 10.7 ml, and the resolution was 0.61 and, therefore, the separation was easily made.

In this example, four pulse tests were ran with a weakly basic anion exchange resin having a tertiary amine function hydrogen bonded to a sulfate ion, ia a cross-linked gel-type acrylic resin matrix (AG4-X4 made by Bio Rad Laboratories, Richmond, California) having a tertiary amine function hydrogen bonded to a sulfate ion, in a cross-linked acrylic resin matrix to determine the ability of the adsorbent to separate citric add from its fermentation mixture of carbohydrates (DPI, DP2, DP3, including glucose, xylose, arabinose and raffinose) and ions of salts, iadudingNa+, K+, Mg* +, Ga* * Jfo* * +, Or, SO4e, PO4e and NOSL amino adds and proteins at a pH of 1.6. The first test was run at a temperature of 75°C The remaining tests were run at 60°C. Citric add was desorbed with water in Pulse Test No. il (Figure 10) and sulfuric acid in two concentrations: 0.05 (Pulse Test No. 2) and 0.25N (Pulse Test No. 3). Pulse Test No. 4 was like Pulse Test No. 2 except that it was made after the adsorbent was used with 24 bed volumes of feed. He fermentation feed mixture had the following composition: Feed Qomwifron Per. Cent Citric Add 40% Salts (K*, Na*, Ca* +»Mg* * Fe* + *) 1J% Caibohydrates (Sugars) 4% Others (SO4~\ CC NO39 proteins and amino adds) 5% Water 495% Retention volumes and separation factor @3) were obtained using the pulse test apparatus and procedure previous^ described in Example I except that a 5cc sample was used. Tie net retention volume (NRV) for the citric add was calculated by measuring the distance from the midpoint ©f the salt envelope as the reference point to the midpoint of the citric add envelope. The separation factor, S, is calculated from the ratio of the retention volumes of the components io be separated to the retention w&rne for the first salt componet (Le. Salts 1).

The results for these pulse teste are shown in the following Table No.

. TABLE WO. 5 Pulse Test Resin/Desorbsnt AG4~X4/Water AG4-X4/0.05N H2S04 AG4-X4/0.25N H2S04 AG4-X4/0.05W H2S04 . Feed Component MRV B Salts 1 1.5 34.25 Citric Acid 54.8 Reference Unknowns A 0 Tracer Unknowns B 6.5 8.30 Salts 2 54.6 1.00 Salts 3.2 11.87 Citric Acid 38.0 Reference Unknown A 0 Tracer Unknown B 2,7 14.07 Unknowns A 0 Tracer Citric Acid 26.9 Reference Sal ts 2.3 11.70 Unknowns B 7.6 3.54 Unknowns A 0 Tracer Citric Acid 38.0 Reference Salts 2,4 15.8 Unknowns 3 7.2 5.28 The results of Pulse Tests 2-4 are similar to Figure 10. From Table 5, it is dear that while citric add is satisfactorily separated in the process, in highly purified form, with water, desorption with water is slower than with dilute sulfuric add as evidenced by larger net retention volume. After aging the adsorbent with 24 bed volumes of feed, the adsorbent shows no signs of deactivation, as observed in Figure 13, which is substantially identical to Figure 11 (conducted under identical conditions with fresh adsorbent).

The first pulse test of Example VUI was repeated using the same procedure and apparatus except that the temperature was 65°C The desorbent was water. This example presents the results of using a macroporous weakly basic anionic exchange resin possessing a cross-linked polystyrene matrix (Dowex 66) with the same separation feed mixture as Example VUI (40% citric add) in the first two pulse tests at a pH of 7.0 and 35 (Figures HA and HB, respective^) to demonstrate the failure to accomplish the desired separation when the pH is above the first ionization constant, pKaj « 3.13, of citric arid, and more specifically ia these two samples, where the concentration of citric arid is 40%, when the pH is above 17. In the third part of the example (represented by Tigoxe HC), the feed was diluted to 13% citric arid and the pH reduced to 2.4. While there is evident inrorovement, it is apparent that the pH and/or concentration will have to be reduced further to prevent breakout of the citric arid. For example, at 13% concentration, it is estimated that the pH must be lowered to about 1.6 to 2J2 Figures HA and HB are, respectively, graphical presentation of the results of the pulse test using Dowex 66 at pHs, reqpectivety, of 7.0 and 35. Figures EA and HB show that citric arid “breaks through with the salts (and carbohydrates) at the higher pHs. This problem can be partially alleviated by reducing the concentration to 13% and lowering the pH to 2.4 as in Figure HG, where it fe shown that only a smaH amount of citric arid is not adsorbed and breaks through in the raffinate while most is adsorbed onto the adsorbent resin (but not desorbed in this Figure). This separation, with adjustment of the concentration and pH to opirmum levels, dearly wfll have commercial utility.

EXAMPLE X Three additional pulse tests under the same conditions as Example ¥10, except as sored, were made ©n dtric add samples of the same feed ,composition, but with two different adsorbents. The desorbent in the first W 5 samples was ©.05 N H2SO4 (Figures 12 and 13 A) while water was used in the third sample (Figure 13B). The composition of the feed used was the same as used im Example VBL The temperature was 6 εκ. i 510 P - CH, - li - CH, where P is the polyswrens moiety fening the resin. Tne second adsorbent (#2), used im the second and third samples, is a tertiary amine, also with a pyridine fimctioeal group, having the following fo®& CK^ rso4“ K I CK3 where P fe as defined above. Both resins are crc^s-linked wfefe divinylbenzene. In some cases, whale water is an effective desorbent, with excellent separation, it fe not strong enough fo recover the adsorbed dtric add quickly enough to make the process coramaerdally attractive. See Figure 15B, ia which the o conditions are the same as above, using adsorbent #2, where water fe tbe desorbent. b this case dtric add does not elute until about 95 ml of desorbent have passed through the adsorbent Dilute sulfuric add is, therefore, the preferred desorbent, as wffl be apparent from the results shown in Figures 12 and 13A. Also, from Figures 12» BA and BB, it will be seen that an excellent separation of citric acid is obtained.

Ihe procedure, conditions and apparatus previously described in Example V3H were used to separate four samples ox citric add from the same feed with two different resins of the same class of adsoibent ss Example VHI, (except that in the first and fourth samples, the column temperature was 50®C and the desorbent was 0.05N HjSOa; in the second and third samples the pH was 22 and the desorbent was dilute sulfuric add at 0.15 N concentration). Both resins, XRA-68 lo and IRA-35, obtained from Rohm and Haas, have an amine function and the following structural formula: R” F - CH2 N:H+SO4e R where P is fhe pdyaaylae matrix, and R’andR = Qfy, Amberlite IRA-68 (Sample Nos. L 2 and 3) is a gel-type resin. IRA-35 (Sample No. 4) fe a macroreticular-typc resin. Sample No- 3 was identical to Sample No. 2, except that the adsorbent had previously been used to separate 69 bed columns of the feed. Sample Nos. I and 2 are both excellent adsorbents for separating citric add fron its fermentation broth within the pH rang© of 16 te 22. Sample Ho. 3, after aging the adsorbent with 69 bed volumes cf feed, demonstrates the stability of the resin (little or no deactivation has taken place) in this separation. Net retention volume (NRV) and selectivity ($) are shows in the fofflovring Table 6.

X6SL££ Sample Ho. Resin Component WRV ] Amberlite IRA-68 Salts Citric Acid ’Unknown A Unknown B 5.5 45.3 0 9.3 8.24 Reference Tracer 4.87 2 Amber1ite IRA-68 Salts Citric Acid Unknown A Unknown 3 2.3 29.0 0 6.5 ‘ 12.51 Reference Tracer 4.46 3 ,Amber]1 fe IRA-58 Salts Citric Acid Unknown A Unknown B '2.85 29.4 0 7.0 10.32 Reference Tracer 4.2 4 Amber]i te IRA-35 Salts Citric Acid Unknown A Unknown 3 K3 36.9 0 5.9 27.38 Reference Tracer 5.25 In a further comparison of the daimed adsorbents of Btamples I through VH with Examples VDT through XI, several samples of the extract wens analyzed for readily cmbooizable impurities (RCS) (Food & Qtemica! Codex (FCC) Monograph #3) and potassium level RCS as detennined in the following Enaosier: a 1 gm sample of the extract (acted ojseesiratioa of dtric add fe determined) is carbonized at 90°C with 10 ml of 95% H2SO4. The carbonized substance is ^xtrophatometrically measured at 500 nm using a 2-ean cell with a 05 inch diameter tube and the amount of RCS is calculated for 50% citric add ssWara. The number arrived at can be compared with that obtained by using this procedure on the cobalt standard solution of the FCC test mentioned above. Potassium is determined by atomic adsorption spectroscopy. For comparison, the same analytical determinations were made on a sample of the same feed and RCS calculated for 50% dtric add with XAD-4, AG4-X4, and adsorbents No. 1 and Na, 2 of Example HL The results are shows ia TaHe X JBSE7 RCS ppmK CA. Net RexJyat XAEM h2o 6.86,8.98 59,137 13.0 AG4-X4 .05NJH2SO4 1.77,, L42 24,81 34.8 #2(Ex.X) .05NJH2SO4 3.17,333 . 24,54 30.8 #l(Ex.X) .05N.H2SQ4 2.17 62 31.0 An improvement in both reduction of levels of RCS and K for the weakly basic resins compared to the neutral resins is indicated by this data, in all samples, RCS was reduced by at least 50% and in two samples, K was reduced by over 50%. It is noted from the net retention volume that both classes of adsorbents have good resolution, but the strong base adsorbents suffer somewhat from increased cycle times. The cycle times can be reduced by using higher concentrations of sulfuric add, e.g., up to about 02N. in the preferred rouge of 0.1 to 02N. in another embodiment, citric add adsorbed on the adsorbent map be converted in situ to a citrate before being desorbed, for example, by reaction with an alkaline earth metal or alkali metal hydroxide or ammonium hydroxide and then 2o immediately eluted using a metal hydroxide, ammonium hydroxide or water as the desorbent Deactivation of the adsorbent by the unknown impurities may tabs place in time, but the adsoibent may be regenerated by flushing with a stronger desorbeat, e„g„. a higher concentration of sulfuric acid than the desorbent a& alkali metal hydroxide or or an organic solvent ©«g·» acetone or alcohol ΜΑΙ^ΗΞΧΠ In this «sample, two pulse tests were run with a gel-type strongly basic anion exchange resin (IRA 458 made hy Rohm & Haas Oo.) having the structural formula like (1) on pageJJI above, substituted with three methyl groups, to <1 determine the ability of the adsorbent to separate citric acid from its fermentation mixture ©f carbohydrates (DPI, DK, DPS, including glucose, xylose, arabinose and raf&nose) and ions of salts, including Na*, KT, Mg* *, Ca* *,Fe* * *, CT, SO4, PO4S and NO3-, amino adds and proteins at a pH of 22. P is aaylic cross5 linked with divinylbenzene. Pulse test Sample No. 1 was run at a temperature of 50°C Pulse test Sample No. 2 was nm at 60°C but after the bed had been aged with 33 bed volumes of feed. Further mans to 62 bed volumes have been made with no signs of deactivation of the adsorbent. Citric add was desorbed with 0.1N solution of sulfuric add in both samples. The fermentation feed mixture had the foflowing composition: a» tion QtricAdd Salts (K*, Na*, Ca* +, Mg+ + Fe* + +) Carbohydrates (Sugars) Others (50^, Cl, PQ^ NOj, proteins and amino ados) Water 495 Retention volumes and separation factor were obtained using the pulse test apparatus and procedure previously described in Example L 9 The results for these pulse tests are shown im the following Table No. g.

JAgISKQJS Feed Sample (to. Resin Component HRV _8 . 1 IRA-458 Salts KO 38.9 ’ Citric Acid 38.9 Reference Unknowns A El Tracer Unknowns 3 6.6 5.89 2 IRA-458 Sal ts 0.9 43.3 Citric Acid 39.0 Reference Unknown A 0 Tracer Unknown 3 7J 5.49 It is dear that citric add is satisfactorily separated m the process, aad after aging the adsorbent with 33 bed volumes of feed, the adsoibent shows no signs of deactivation, which is substantially identical to the results under closely identical conditions with fresh adsorbent, ixaMaoai The pulse test of Easmple XH was repeated ©n additional, citric add samples using the same feed, but a different, macroporous, strongly basic anionic exchange adsorbents, BRA-958, possessing quaternary ammonium functions and an acrylic resin matrix cross-linked with dirinylbenzene matrix. The desorbent was 0.05 N H2SO4. Tse composition of the feed used was the same as used in Esample ΧΠ. The temperature was (XfJC and the pH was 1.6. He adsorbent in this test was a resin obtained from Rohm & Haas baring the structure (1) shown on page. 21, where R. fc methyl As shown in Figure 14, citric acid starts eluting after 45 ml of desorbent have passed through the adsorbent and is very effectively separated from the fermentation mixture in high purity with excellent recovery, EXAMPLE X?V The pulse test of Example XS was repeated on an additional citric add sample using the same feed, but a different, strongly basic anionic exchange resin adsorbent, AG2-X8 (obtained from Bio Rad Company) having a structure like formula (2) above, (page4f) where R is methyl, with a cross-linked polystyrene geltype resin matrix having quaternary ammonium functional groups thereon. The desoibent was 0.15N H2SO4. The composition of the feed used was the same as used in Example XI The temperature was 50°C and the pH was 22.

As shown in Figure 15, citric add starts eluting at about 43 seal of desoibent have passed through the adsorbent and is very effectively separated from the fermentation mixture in high purity with excellent recovery.

In a further comparison of adsorbents of Examples I through VII with Examples ΧΠ through XIV, several samples of the extract were analyzed for readily carbonizable impurities (RCS) (Food & Chemical Codes: (FCC) Monograph #3) and potassium level as described above. The results for each of the adsorbents, XAD-4, IRA 458, IRA 959 and AG 2 - X4 with the indicated desorbent are shown in the following Table 9.

RCS TABIM9 EXTRACT QUALITY iRCS /PQTASSRIM1BYHJLSE3SST Adsorbent .Desorbeot (Calculated at50%CA) Rlnits) spmK (Calculated GA Net Retention Volume XAD-4 h2o 6.86 8.58 . 55 137 13Λ ERA 458 OJN H2SO4 15 SO 375 ERA 958 O.O5NH2S04 2.73 82 32 AG2-X4 OJ5NH2SO>4 55 131 43 An improvement in both reduction of levels of RCS and K for the strongly basic resins compared to the neutral resins fe indicated by this data. In all samples, RCS was reduced fey between 40-85% and K was reduced between 0-20%. it is noted fern Examples ΧΠ, ΧΠΙ and XU (Fig. 14),, that both classes of adsorbents have good separation, but the ©reseat adsorbents suffer somewhat from increased cycle times. The cycle times can be reduced by using higher concentrations of sulfuric add, e.g.s up to about 0.2N in the preferred range of 0.1 to 0JN.

In another embodiment, citric add, adsorbed on the adsorbent may be converted in sfea to a dtrate before being desorbed, for example, by reaction with an alkaline earth metal hydroxide, alkali metal hydroxide or ammonium hydroxide and then immediate^ eluted using ε metal hydrosdde, ammonium hydroride or water, as the desorbent Deactivation of the adsorbent fey the unknown impurities may sake place in time, but the adsorbent may fee regenerated fey flushing with a stronger desorbent, e.g., a high concentration ef sulfuric add than the desorbent, an alkali metal hydrosdde or NH4OH, or an organic solvent, e-g., acetone or alcohol·

Claims (14)

CLAIMS :

1. A process for separating citric acid from a fermentation broth feed mixture containing citric acid characterized in that said mixture is contacted with a polymeric adsorbent selected from a neutral, crosslinked polystyrene polymer, a nonionic hydrophobic polyacrylic ester polymer, a weakly basic anionic exchange resin possessing tertiary amine or pyridine functional groups, and a strongly basic anionic exchange resin possessing quaternary amine functional groups, and mixture thereof at adsorption conditions whereby selectively to adsorb said citric acid, and thereafter said citric acid is recovered from said adsorbent with a desorbent at desorption conditions.

2. A process according to claim 1 characterized in that said adsorption and desorption conditions include a temperature of from 20 to 200°C and a pressure of from atmospheric to 500 psig (3450 kPa gauge).

3. A process according to claim 1 or 2 characterized in that said desorption is effected in the liquid phase with water or an aqueous inorganic acid or a ketone or mixtures thereof.

4. A process according to any one of claims 1 to 3 characterized in that said absorbent comprises a tertiary amine functional group supported on a matrix comprising a crosslinked acrylic resin.

5. A process according to any one of claims 1 to 3 characterized in that said adsorbent comprises a pyridine functional group supported on a matrix comprising a crosslinked polystyrene resin. A .?

6. A process according to any one of claims 1 to 3 characterized in that said adsorbent comprises a quaternary amine functional group supported on a matrix comprising a crosslinked acrylic resin.

7. A process according to any one of claims 1 to 3 characterized in that said adsorbent comprises a quaternary ammonium salt of pyridine supported on a matrix comprising a crosslinked polystyrene resin.

8. A process according to any one of claims 1 to 3, 6 and 7 characterised in that the quaternary ammonium group is in the sulphate form.

9. A process according to any one of claims 1 to 3 wherein said adsorbent is in the sulphate form.

10. A process according to any one of claims 1 to 9 characterized in that it comprises the steps of: (a) maintaining net fluid flow in a single direction, through a column of said adsorbent containing at least three zones having separate operational functions occurring therein and being serially interconnected, with the terminal zones of said column connected to provide a continuous connection of said zones; (b) maintaining in said column an adsorption zone, defined by the adsorbent located between a feed input stream at an upstream boundary of said zone and a raffinate output stream at a downstream boundary of said zone; (c) maintaining immediately upstream from said adsorption zone, a purification zone defined by the adsorbent located between an extract output stream at an upstream boundary of said purification zone and said feed input stream at a downstream boundary of said purification zone; (d) maintaining immediately upstream from said purification zone, a desorption zone defined by the adsorbent located between a desorbent input stream at an upstream boundary of said zone and said extract output stream at a downstream boundary of said zone; (e) passing said feed mixture into said adsorption zone at adsorption conditions to effect the selective adsorption of said citric acid by said adsorbent in said adsorption zone and withdrawing a raffinate output stream comprising the nonadsorbent components of said fermentation broth from said adsorption zone; (f) passing the desorbent material into said desorption zone at desorption conditions to effect the displacement of said citric acid from the adsorbent in said desorption zone; (g) withdrawing an extract output stream comprising said citric acid and desorbent material from said desorption zone; (h) passing at least a portion of said extract output stream to a separation means and therein separating at separation conditions at least a portion of said desorbent material; and (i) periodically advancing through said column of adsorbent in a downstream direction with respect to fluid flow in said adsorption zone, the feed input stream, raffinate output stream, desorbent input stream, and extract output stream to effect the shifting of zones through said adsorbent and the production of extract output and. raffinate output streams.

11. A process according to claim 10 characterized in that a buffer zone is maintained immediately upstream from said desorption zone, said buffer zone defined as the adsorbent located between the desorbent input stream at a downstream boundary of said buffer zone and the raffinate output stream at an upstream boundary of said buffer zone.

12. An adsorption process for separating citric acid from a fermentation broth feed mixture containing citric acid characterized in that said mixture is contacted at a pH lower than the first ionization constant PKa-j of citric acid, with a weakly basic anionic exchange resin possessing tertiary amine or pyridine functional groups, at adsorption conditions whereby selectively to adsorb said citric acid, converting the citric acid into a salt by reaction with an alkaline earth metal hydroxide, alkali metal hydroxide or ammonium hydroxide, and eluting the salt with a metal hydroxide, ammonium hydroxide or water as eluant.

13. A process according to claim 1 for separating citric acid from a fermentation broth feed mixture, substantially as hereinbefore described and exemplified.

14. Citric acid whenever obtained by a process claimed in a preceding claim.