Membrane-Based Process for Debittering Citrus Juice

(FIGURES AND TABLES ARE UNAVAILABLE ON THIS WEBSITE.)

I. IDENTIFICATION AND SIGNIFICANCE OF THE PROBLEM

Citriculture is an important segment of world food production and nutrition. Oranges account for 67% of total worldwide citrus production. Oranges are consumed, principally in the form of juice, at an annual rate of approximately 17 kilograms per person (The Almanac of the Canning, 1979).

A major problem in the citrus industry worldwide is the formation of bitterness in citrus juice and products within hours after extraction from the fruit. The problem occurs in certain varieties of oranges (including mandarins), grapefruits, and lemons. The problem is estimated to cause losses for California citriculture of $6 million to $8 million per year (Anon., 1986a).

The primary cause of the "delayed bitterness" in oranges is the presence of an intensely bitter group of compounds called limonoids, principally limonin and nomilin (Figure 1). The extent of bitterness imparted by limonin has been studied by Levi and coworkers (Levi et al., 1974), who obtained the values shown in Table I.

Much research has been focused on preventing the formation of limonin precursors prior to harvest, as well as on finding means to remove limonin after harvesting and processing. An excellent review of the approaches is given by Hasegawa and Maier (Hasegawa and Maier, 1983), and a summary of recent R&D activities is given in Section V.A. of this proposal. However, existing methods have severe economic and technical limitations (Anon., 1985). Additionally, existing methods often impair the stability and quality of the juice by affecting components other than limonin. Thus, despite extensive research efforts, there is no acceptable commercial method for reducing levels of bitter citrus components. The only current option is to blend excessively bitter juice with non-bitter juice.

The proposed program is directed at demonstrating the feasibility of using a novel process for the removal of limonin and nomilin from bitter navel orange juice. The process is based on a membrane that selectively allows the passage of limonin, limonoic acid, nomilin, and nomiloic acid, while retaining desirable flavor and nutritional components. The proposed process is designed so that the juice or concentrate contacts only materials that are accepted for food use or have been shown to be generally recognized as safe. Furthermore, the proposed process promises to be economical on the basis of our preliminary analysis, as is discussed in Section II.C.

II. BACKGROUND, TECHNICAL APPROACH, AND ANTICIPATED BENEFITS

II.A. SOLUTION-DIFFUSION MEMBRANES OFFER A MEANS TO DEBITTER ORANGE JUICE

An attractive approach to the removal of off-flavors (a solute) in fruit juices involves the use of a synthetic membrane that allows the off-flavor to pass through but retains the desired components in the juice. Membranes can be conveniently categorized into two groups (Figure 2): 1) filtration membranes, and 2) solution-diffusion membranes. The more conventional filtration membranes have been studied extensively for food-processing applications, and their limitations are well-known. Those limitations include severe problems with membrane fouling in the application of interest here (Matthiasson, E., 1985). Solution-diffusion membranes, on the other hand, show substantial promise in this type of application.

Filtration membranes contain pores. When a solution is forced through a filtration membrane (convective flow), solute molecules smaller than the pores pass easily; solute molecules larger than the pores are retained. Consequently, the selectivity of this type of membrane is dictated by molecular size. Moreover, a phenomenon known as "concentration polarization"--i.e., the buildup of solutes on the membrane surface-- leads to membrane fouling and drastically reduced performance.

Solution-diffusion membranes do not have pores, and thus there is no convective flow through the membrane. Instead, solute molecules dissolve in the membrane phase and permeate the membrane by molecular diffusion. Molecular solutes that exhibit high solubilities and/or diffusivities in the membrane phase permeate rapidly; molecular solutes that are insoluble in the membrane phase and/or that have low diffusivities are retained. Thus, the selectivity in this type of membrane is closely correlated with solute solubility, and an analysis of this type of membrane parallels liquid-liquid extraction rather than filtration. The key distinction is that, relative to filtration membrane processes, solution-diffusion membrane processes exhibit minimal concentration polarization and consequently resist fouling. This is because solute transport occurs via dissolution in the membrane phase, rather than by convective flow through pores.

In our proposed process, we will use solution-diffusion membranes. Specifically, the membrane consists of a thin layer of a hydrophobic liquid, about 25 um thick, that separates two aqueous solutions. The hydrophobic liquid is supported and stabilized within the pores of a microporous, hydrophobic polymer film. Such supported-liquid membranes have been shown in our laboratory to be stable for up to a year, even when separating aqueous solutions that have acid concentrations that differ by ten or more pH units. That degree of chemical stability is critical in the proposed application.

The operation of this type of membrane for the removal of limonin and limonoic acid ( *Because the solubility and chemical properties of limonin and nomilin are extremely similar, we believe that the membrane process will operate equally well for nomilin and nomiloic acid. We will therefore use limonin and limonoic acid as a generic terms for all four types of liminoids of interest here.) is illustrated in Figure 3 . A feed solution containing orange juice at pH 3.2 flows on the left side of the membrane, and a stripping solution containing aqueous sodium hydroxide at pH 12 to 13 flows on the right side. Limonin from the feed solution dissolves in the hydrophobic supported liquid of the membrane and is transported across the membrane to the strip side by diffusion.

Limonoic acid, produced by the hydrolysis of limonin, is also present in the feed solution. Since the pKa's of limonoic acid are 2.7 and 4.7, at a pH of 3.2--the pH of the juice--one carboxyl group of limonoic acid and a significant fraction of the other carboxyl group are protonated. Thus, limonoic acid, bearing no charges, also dissolves in the hydrophobic supported liquid of the membrane and is transported to the strip side. At the membrane/strip-side solution interface, limonin and limonoic acid dissolve in the strip solution and are rapidly and irreversibly hydrolyzed and deprotonated to limonate under the basic conditions of the strip solution. Since anionic limonate, bearing charges, is insoluble in the supported liquid of the membrane, it cannot back-diffuse; therefore, limonin and liminoic acid are irreversibly transported from the feed solution to the strip solution.

Because limonate is confined to the strip stream, it can be concentrated by using a strip-stream volume that is smaller than the feed-stream volume. The strip-stream volume is reduced by setting the strip-stream flow rate lower than the feed-stream flow rate. From material balance considerations,

Limonin Conc. in Strip Stream = Limonin Conc. in

Feed Stream x (Feed-Stream Flow Rate/ Strip-Stream Flow Rate).

Hence, the proposed supported-liquid membran process makes possible the removal and concentration of limonin from orange juice.

II.B. PRELIMINARY EXPERIMENTS SUPPORT THE PROPOSED APPROACH

We have conducted preliminary experiments that indicate that the proposed supported-liquid membrane process efficiently transports limonin and limonoic acid. In a typical experiment, a supported-liquid membrane consisting of Shell Sol 71 (*Shell Sol is a mixture of aliphatic solvents manufactured by Shell Chemical Company; it is approved for contact with food products.) supported in the pores of a flat-sheet microporous polypropylene membrane (Celgard 2400 manufactured by Celanese Separations, Inc., Charlotte, NC) was clamped between the two compartments of the membrane-permeability apparatus illustrated in Figure 4. The feed compartment was filled with limonin (60 ppm) dissolved in 20 vol% acetic acid (pH 3.2); the strip compartment was filled with 0.01M aqueous sodium hydroxide (pH 12). (**Control experiments indicated that the solubility of limonin in water was raised from about 6 to 8 ppm at pH 3.2 to more than 200 ppm at pH 12, presumably due to the base-catalyzed hydrolysis and ionization of limonin.) The aqueous solutions in the two compartments were stirred and their temperatures maintained at 25 degrees C. The concentration of limonin in the feed compartment was monitored as a function of time by removing aliquots from the feed compartment and assaying the limonin concentration by means of HPLC and UV detection. ( +Limonin was separated from acetic acid on a C18 reverse-phase HPLC column using a 10 vol% to 50 vol% acetonitrile/water gradient and was detected on a diode-array UV detector set at 207 nm. Concentrations were determined from detector response curves obtained using limonin samples of known concentration. ) We determined an initial limonin flux (++Flux is defined as the amount of material that permeates a given area of membrane in a given time.) of 2.5 ug/cm squared -min from the slope of a plot showing limonin concentration in the feed versus time. Dividing the flux by the initial limonin concentration gave a concentration-normalized flux (permeability) of 1.0 cm/hr.

Since limonin in juice exists as a mixture of forms that include limonin and limonoic acid in three possible states of ionization (Chandler and Robertson, 1983), a major concern was whether the membrane could remove all forms of limonin. To address this question we conducted a membrane-based separation experiment on a sample of hydrolyzed limonin. Specifically, limonin (100 ml/L) was suspended in 20 vol% acetic acid and the pH of the suspension was raised to 12 by the addition of sodium hydroxide. Mild heating of the mixture (to 30 degrees C) caused all the limonin to dissolve. The solution was then heated to 65 degrees C to hydrolyze the limonin to limonate. After heating for 30 min, the pH of the solution was adjusted to 3.2 by the addition of concentrated sulfuric acid. The resulting solution (50 ml) was then added to the feed compartment of a membrane-permeability apparatus containing a supported-liquid membrane composed of 30 vol% isohexadecyl alcohol and 70 vol% Shell Sol 71 and a strip solution (168 ml) of pH 12 aqueous sodium hydroxide. The experiment was conducted and analyzed as described previously. The results are shown in Figure 5.

The results shown in Figure 5 make three important statements. One, the membrane process effectively reduces limonin in the feed from about 55 ppm to 11 ppm. Two, the membrane process is effective at removing all forms of limonin, as prior hydrolysis of limonin in the feed has no adverse effect on total limonin removal. And three, the membrane process follows the expected exponential loss of limonin in the feed with a permeability coefficient of 1.1 cm/hr.

The initial results obtained at Bend Research strongly support the validity of the proposed approach. However, the ultimate feasibility on actual orange juice and the favorable economics must be verified via a thorough R&D program designed to identify the optimum membrane materials and operating conditions. The elements of such a program are described in detail in later sections of the proposal.

II.C. PRELIMINARY COST ANALYSIS SUGGESTS THAT THE PROPOSED PROCESS IS ECONOMICAL

II.C.1. Anticipated Technical Results

In Phase I we will develop a membrane-based process that selectively removes and concentrates limonin. We will incorporate this membrane into a test module and use it to demonstrate the removal of limonin from samples of clarified, very bitter navel orange juice. We will carry out studies to measure the effect of operating parameters on limonin removal, including feed composition, temperature, pH, and feed- and srip-stream flow rates. The results of these studies will be used to make engineering calculations to establish the technical and economic feasibility of the membrane-based limonin-removal process.

II.C.2. Economic Impact of Proposed Debittering Process

The economic impact of "delayed bitterness" in citrus can be illustrated by examining the commercial production of orange juice concentrate in California and Arizona. In a typical year, approximately 40 million pounds of navel orange solids are processed as juice concentrate. Of this amount, approximately 80% would benefit from a debittering process. (**Dr. Denny Nelson, Sunkist Research Center, Ontario, California, personal communication.) Juice concentrate that is not bitter sells for approximately $1.30 per pound of solids. Bitter juice, on the other hand, is discounted a minimum of $0.20 per pound of solids. Consequently, "delayed bitterness" in navel oranges alone results in an economic loss of at least $6.4 million in California and Arizona.

The technical objective in Phase I of the proposed program is to develop a process that reduces limonin concentration in orange juice from 40 ppm (a typical value for early-season navel oranges) to 8 ppm. The resulting debittered juice can be mixed with a late-season juice to produce a blend of marketable quality containing less than 6 ppm limonin. A preliminary analysis of the economics of the proposed process is given below.

Given an average concentration gradient across the membrane of 16 ppm limoqin, the analysis uses an average limonin flux of 0.29 lb per ft squared of membrane per year. This value is based on the permeability values obtained in the preliminary studies reported in Section II.A. We assume that freshly pressed juice contains 11% solids. Equipment and operating costs are based on previously reported metal-recovery studies (Babcock et al., 1983) employing liquid membranes. A list of the assumptions is given in Table II.

The results of our preliminary analysis are shown in Table II. For debittering 32 million poounds of navel orange solids, the total process cost is $605,940/yr, or 1.9 cents/lb of orange solids. The projected cost compares favorably with the minimum $0.20-per-lb discount incurred with bitter juice and represents a $6.1 million potential gain for the citrus industry in California and Arizona alone. Furthermore, it is important to note that in the above analysis the limonin flux used in the calculation is a lower limit that was achieved during very preliminary investigations. Process costs can be further reduced by using thinner membranes or optimized operating conditions to increase limonin flux.increase limonin flux. II.C.3. Advantages of the Proposed Debittering Process

The proposed membrane process offers a number of important advantages over existing processes for the removal of limonin:

• The process can be carried out at low temperature without phase change, thus minimizing deterioration of temperature-sensitive nutritional components in orange juice (e.g., ascorbic acid).
• The process is modular and thus can be operated at any desired scale by simply increasing or decreasing the number of membrane modules used.
• The process can remove limonin from dilute solution and produce it in concentrated form.
• The process can be operated in a continuous mode, thereby minimizing processing time while maximizing process reliability.
• The process is relatively immune to fouling problems that are common when more conventional membrane processes are used in food processing.

Other advantages that we expect to derive from the proposed membrane-based limonin removal process are listed below.

• The process is simple, requiring a minimum of operator control and judgment.
• The process is readily adaptable to sanitary operations.
• The process is designed for high throughput--a condition necessary for large-scale utilization.
• The process will not alter desirable flavor and nutritional components in orange juice or significantly remove them.(*Although some flavor components are hydrophobic and might permeate the membrane, they are not expected to be concentrated in the strip stream. Recall that limonin is hydrolyzed to limonate dianion in the strip stream and that the net charge on limonate is what prevents its back-diffusion and allows it to be concentrated. Using a ratio of (feed-stream flow rate)/(strip-stream flow rate) = 20/1, there would be--in the worst case--a 5% loss of those flavor components that rapidly equilibrate across the membrane.)

II.D. STRUCTURE OF THE PHASED PROGRAM

Phase I is directed at demonstrating the feasibility of using a membrane process to debitter early-season orange juice. Once proof of concept has been accomplished in Phase I, the Phase II effort will focus on 1) optimization of the membrane separation, 2) development of more efficient membranes in hollow-fiber form, and 3) fabrication of prototype hollow-fiber membrane modules. During Phase II, we will do long-term testing of membrane modules under realistic field conditions. The goal is to establish that any adverse effects on module performance caused by sustained operation on actual juice can be rectified in a way that is consistent with large-scale plant operation. It should be noted that the field tests will form the basis for a comprehensive technical and economic evaluation.

If Phase II is successful, we anticipate entering into an agreement with a commercial sponsor for Phase III of the program. Phase III would involve construction of a 1,000-gal/day field-test unit and approximately 12 months (elapsed time) for field-testing and evaluation before full-scale commercialization would be initiated. We anticipate that the membrane modules necessary for Phase III will be fabricated by Consep Membranes of Bend, OR, a manufacturing subsidiary of Bend Research, Inc.

III. PHASE I TECHNICAL OBJECTIVES

The overall goal of this program is to demonstrate the feasibility of using a membrane process to remove and concentrate bitter-tasting limonin from orange juice. We aim to reduce the concentration of limonin in early-season juice from 40 ppm down to 8 ppm so that the purified juice can be mixed with non-bitter, late-season juice to produce a blend of marketable quality (6 ppm limonin and acceptable flavor).

In pursuit of this goal, we will address the following technical questions:

1. What type of solution-diffusion membrane offers high limonin fluxes and selectivities?
2. How do operating parameters (e.g., flow rate, pH, temperature) affect membrane performance (selectivity and flux)?
3. What is the most effective way to concentrate limonate on the strip side--e.g., high pH, anion-exchange resins, charcoal adsorbers?
4. What is the projected lifetime of the membrane?
5. Does the membrane process alter the flavor or nutritional value of the juice by removing beneficial components?
6. Does the proposed membrane process release materials into the juice that are not accepted for food use or are not generally recognized as being safe?
7. What are the projected economics of the process?

The tasks necessary to answer the above question are detailed in the next section.

IV. PHASE I RESEARCH PLAN

IV.A. TASK 1: SELECTION OF SUPPORTED-LIQUID MEMBRANE

The objective of this task is to identify candidate membranes that transport limonin and limonoic acid but reject limonate. The best candidate membranes will be fabricated and evaluated as described in Tasks 2 through 5.

In this task we will investigate two types of supported-liquid membranes: immobilized-liquid membranes, and liquid-swollen polymer membranes (Figure 6). Immobilized-liquid membranes consist of a hydrophobic liquid held in the pores of a microporous support. Liquid-swollen polymer membranes consist of dense polymer membranes that have been swollen by hydrophobic liquids.

Our approach to selecting candidate supported liquids is guided by the principle that high limonin and limonoic acid transport is associated with high limonin and limonoic acid solubility in the supported liquid of the membrane. Thus, by employing solubility and distribution-coefficient data, we aim to identify supported liquids that exhibit high limonin and limonoic acid solubilities but low solubilities to desirable flavor and nutritional components. We will use ascorbic acid as a representative desirable nutritional component, linalyl anthranilate (essence of orange) as a representative desirable flavor component, and a O.1M (pH 3.2) citric acid solution containing 12 wt% sucrose as a synthetic orange juice base. We will then measure the distribution coefficient of limonin, ascorbic acid, and linalyl anthranilate between the synthetic orange juice base and a number of hydrophobic liquids. Candidate liquids include high-molecular-weight even-numbered straight-chain alcohols. (* We have chosen even-numbered alcohols because they are approved for some food-separation processes.) For example, 1-decanol has low water miscability and is essentially nonvolatile. Liquids will be ranked on the basis of their tendencies to preferentially partition limonin and liminoic acid.

Limonin and limonate concentrations will be determined by HPLC analysis as described in our summary of preliminary experiments (Section II.A.) and as reported by Shaw and Wilson (Shaw and Wilson, 1984). Ascorbic acid concentrations will be determined by the reduction of Fe(III) by the ascorbic acid and subsequent measurement of the absorbance of Fe(II)-ferrozine chelate at 562 nm as described by Taselskis and Nelapaty (Taselskis and Nelapaty, 1972). Linalyl anthranilate concentrations will be determined by HPLC analysis.

IV.B. TASK 2: MEMBRANE FABRICATION AND TESTING

During the proposed Phase I feasibility study, we will study only flat-sheet membranes, as they are easy to fabricate and they can be readily evaluated in our membrane-permeability apparatus (Figure 4). However, we plan to use hollow-fiber membranes in Phase II development work because of their higher surface-area-to-volume ratio. It is our experience that developments obtained with flat-sheet membranes are readily transferable to hollow-fiber membranes.

Immobilized-liquid membranes will be prepared by sorbing the solvents selected in Task 1 into the pores of a microporous membrane. We intend to examine the feasibility of using Goretex Type S 11003 (polytetrafluoroethylene), Celgard (microporous polyethylene), and BRI's microporous polysulfone and microporous polyvinylidene difluoride as supports for the liquid membranes. The immobilized-liquid membranes will be prepared by immersing the microporous support in the solvent and removing the air bubbles remaining in the pores by repeatedly drawing and releasing a vacuum above the solvent. The organic liquid is retained in the pores by capillarity.

Liquid-swollen polymer membranes are prepared by immersing the polymer membrane in the selected liquid with gentle warming. Polymer membranes will be obtained as commercial membranes or prepared as films supported on porous backings by using one of the following methods: 1) casting a solution of the polymer dissolved in an organic solvent onto a preformed porous support, such as polysulfone, and letting the membrane air-dry to form a thin, nonporous film on the support; 2) forming a supported-polymer membrane by casting the polymer solution onto a glass plate and then precipitating the polymer in water; or 3) laminating a precast polymer membrane onto a preformed porous support by heating, such that the polymer membrane melts just enough to be "heat sealed" onto the support. These membrane-fabrication techniques have been successfully used at Bend Research for a variety of membranes, and thus we expect the probability of success in this task to be very high.

The supported-liquid membranes will be tested to ascertain that they are free of leaks by measuring initial fluxes of ascorbic acid in a membrane-permeability apparatus of the type illustrated in Figure 4. Defect-free supported membranes are expected to display high resistance to ascorbic acid transport. We will also measure initial fluxes of limonin and limonate. Initial fluxes will be determined by imposing a concentration gradient across the membrane and periodically removing and analyzing samples of the solution on the permeate side of the test cell as described in Section II.A. Permeabilities will be calculated using the relationship

Permeability = (Permeate Flux)/ (Concentration Difference).

Selectivities will be calculated relative to ascorbic acid permeability using the relationship

Selectivity = (Limonin Permeability/Ascorbic Acid Permeability).

IV.C. TASK 3: OPTIMIZING STRIPPING CONDITIONS

Limonin fluxes are governed, in part, by the limonin-concentration gradient across the membrane. Thus, effective limonin removal hinges on maintaining low concentrations of membrane-permeable forms of limonin in the strip side. This can be accomplished by transforming all limonin forms to limonate, or adsorption of limonin forms onto adsorbents. In this task we will measure the solubility of limonin as a function of strip-solution pH (in the range of 10 to 13) and temperature (in the range of 25 degrees to 65 degrees C). Additionally, we will study the use of commercially available, strong anion-exchange resins (e.g., Dowex l-X2, Amberlite IR-401, Duolite ES-861) and activated carbon (e.g., Darco) as a means to adsorb and concentrate limonin forms in the strip solution.

IV.D. TASK 4: MEMBRANE PERFORMANCE WITH CLARIFIED ORANGE JUICE

On the basis of limonin flux and selectivity, we will select the two best supported-liquid membranes prepared in Task 2 and the best stripping conditions determined in Task 3 for evaluation on actual orange juice. A test module that holds a 1-ft section of flat-sheet membrane will be fabricated for this task. To eliminate problems associated with handling orange juice pulp, we will carry out our Phase I membrane-performance studies on clarified orange juice. During Phase II we would design a system that would treat all components of orange juice concentrate.

To aid us in this task, we have established a collaboration with Dr. Denny Nelson of the Sunkist Research Center in Ontario, California. Dr. Nelson has provided us with authentic samples of limonin and will provide us with clarified early-season navel orange juice containing 30 to 40 ppm limonin. Furthermore, he has offered to test the purified juice for bitterness removal, as well as nutrient and flavor loss. Additionally, we will test the purified juice for the presence of materials released from the membrane. If any major difference in membrane performance is evident when actual clarified juice is used as compared with performance when a synthetic mixture of limonin, ascorbic acid and linalyl anthranilate is used, an attempt will be made to determine the cause.

IV.E. TASK 5: EFEECT OF OPERATING CONDITIONS ON LIMONIN REMOVAL

The objective of this task is to measure the effect of operating conditions on the efficacy of limonin removal from clarified orange juice. Operating parameters to be considered include temperature, feed- and strip-stream velocity across the membrane, pH of the strip stream, and composition of the feed stream.

Liquid membranes similar to those to be used in the program but applied to inorganic-salt separations are known to perform adequately for 6 months without requiring reloading with fresh organic solution. Liquid-swollen membranes are known to perform adequately for up to 2 years. However, such membranes have not been evaluated for limonin removal from citrus juice. Hence, the additional purpose of this task is to run limited (about 1 month) lifetime studies using clarified orange juice to be certain that there are no short-term adverse effects on membrane life.

IV.F. TASK 6: PRELIMINARY TECHNICAL AND ECONOMIC ANALYSIS AND PREPARATION OF FINAL REPORT

In this task we will make use of the data obtained in Tasks 4 and 5 to perform a technical and economic evaluation. These evaluations will form the basis of the decision to proceed into the Phase II program. The results will appear in our final report.

V. RELATED RESEARCH AND DEVELOPMENT

V.A. R&D ACTIVITIES RELATED TO PROPOSED EFFORT

Previous approaches to bitterness reduction in citrus products fall into three categories: 1) preharvest treatments that inhibit the formation of limonoids in citrus fruits, 2) biotransformation of bitter liminoids into non-bitter metabolites, and 3) liminoid removal by adsorption on polymer resins. However, none of these approaches has been put into commercial practice. A brief discussion of recent R&D activities in these three categories follows.

Shin Hasegawa from the Fruit and Vegetable Chemistry Laboratory of the Department of Agriculture's Agricultural Research Service in Pasadena, California has recently reported a preharvest treatment that claims to inhibit the formation of liminoids in citrus fruit (Anon., 1986b). The treatment consists of applying synthetic auxins (plant-growth regulators) that are potent inhibitors of liminoid biosynthesis. However, this approach has not been demonstrated to be cost-effective.

Several investigators have reported studies directed at using biotransformations as a means to avert bitterness by converting bitter liminoids in citrus juice to non-bitter metabolites. Approaches include the use of imobilized cells (Hasegawa et al.,1983; Hasegawa and Pelton, 1983) and the use of enzymes (Hasegawa, 1975a; Hasegawa, 1975b). A drawback of these approaches is the deterioration of the flavor and color of treated juice.

Most R&D activities have been directed at removing bitter limonin from juice by adsorption or complexation on polymer adsorbents. Approaches include the use of ion-exchange resins (Coca-Cola Co., 1987; Mitchell et al. 1985; Johnson and Chandler, 1985; Purl, 1984), cyclodextrin polymers (Shaw et al., 1984), and cellulose dster gels (Johnson and Chandler, 1981a; Johnson and Chandler, 1981b, and Chandler and Johnson, 1979). These approaches, however, have a number of shortcomings, including unfavorable process economics, excessive process complexity requiring extensive operator control, the need for organic solvents to strip limonin from the adsorbent, and production of difficult-to-manage waste streams.

V.B. BEND RESEARCH ACTIVITIES RELATED TO THE PROPOSED EFFORT

Bend Research is currently involved in several separate but related membrane-based separation projects. Dr. Paul van Eikeren, the principal investigator, is currently directing an NSF SBIR Phase II project entitled, "Optical Resolution of D,L-Phenylalanine in a Membrane Reactor" that is focused on using supported-liquid membranes to selectively remove the product of an enzyme-catalyzed reaction. Additionally, he is directing several proprietary projects for private Bend Research clients directed at using membrane separations to recovery valuable food additives and flavoring agents.

VI. KEY PERSONNEL AND BIBLIOGRAPHY

It is anticipated that Dr. Paul van Eikeren will serve as principal investigator, with key contributions being made by Dr. Harold Lonsdale. Dr. van Eikeren has extensive R&D experience in areas directly related to the proposed work. He is now directing several Phase I and Phase II SBIR programs. Dr. Lonsdale has 25 years of experience in the field of membrane technology and is recognized internationally as a leader in the field. He is the founding and current editor of the Journal of Membrane Sciene. Abbreviated resumes for these individuals follow.

VII. FACILITIES AND EQUIPMENT

Bend Research is a 10,000- square ft laboratory and office complex; an additional 5,000 square ft is under construction. The firm employs a technical staff of 45 and has 70 employees. Work is devoted exclusively to the development of membranes and membrane-based processes. In-house material and equipment relevant to the performance of the proposed program include the following:

a) atomic absorption, UV-visible, and FT-IR spectrophotometers;

b) multiple HPLC and GC instruments;

c) scanning electron microscope;

d) organic-synthesis laboratory and capability;

e) continuous-production equipment for flat-sheet and hollow-fiber support membranes; and

f) membrane-permeation apparatus.

VIII. CONSULTANTS

Dr. Denny Nelson,Director of R&D at Sunkist Research Center, Ontario, California, is an expert in the field of debittering navel orange juice. He has agreed to consult with Bend Research on this project without compensation.

IX. POTENTIAL COMERCIAL APPLICATIONS

Succesful development of a membrane system for the removal of limorin offers a practical procedure for the removal of bitter limonoids from all citrus juices. Furthermore, successful development would provide the basis for fabricating other supported-liquid membranes for the selective removal of off-flavors and off-colors in fruit-juice concentrates. For example, supported-liquid membranes could offer a means of removing polyphenol off-colors in natural sweeteners produced from apple or pear concentrates.

X. CURRENT AND PENDING SUPPORT

No work substantially similar to that proposed here is being conducted at this time, nor is any pending.

XI. EQUIVALENT PROPOSALS

No proposal substantially similar to this has been submitted to any other agency.

REFERENCES

Anon., "New Methods Remove Citrus Bitterness," C&E News, May 27 (1985) 58.

Anon., "Method to Reduce Bitterness in Citrus Found," October 27(1986a)25-26.

Anon., "Method to Reduce Bitterness in Citrus Fruit," C&E News, October 27 (1986b) 25.

Babcock, B.C. et al., Renovation of Printed Circuit Board Etchants with Coupled Transport Membranes. Final report to the National Science Foundation from Bend Research, Inc. on Grant No. DAR-8023179, November 17, 1983.

Chandler, B.V., and G.L. Robertson, "The Solubility of Limonin, the Bitter Principle of Orange Juice," J. Sci. Food Agric., (84(1983)1272-1284)

Chandler, B.V., and R.L. Johnson, "Adsorbent Materials for Debittering Citrus Juices," Australian Patent 498,130, February 15, 1979.

Coca-Cola Co., "Method of Reducing Bitterness in Citrus Fruit Juices," Israeli Patent 70634 AL, January 30, 1987.

Hasegawa, S., and V.P. Maier, "Solutions to the Limonin Bitterness Problem of Citrus Juices," Food Techn., 37(1983)73-77.

Hasegawa, S. et al., "Metabolism of Limonoids by Arthrobacter Globiformis II: Basis for a Practical Means of Reducing Limonin Content of Orange Juice by Immobilized Cells," J. Agric. Food Chem., 31(1983)1002-1004.

Haseqawa, S., and V.A. Pelton, "Reduction of Nomilin Bitterness in Citrus Juice and Juice Serum with Arthrobacter Globiformis Cells lnmbilized in Acrylamide Gel," J. Agric. Food Chem., 31(1983)178-180.

Hasegawa, S., L.C. Brewster, and V.P. Maier, "Limonate:NAD(P) Oxidoreductase and Debittering of Citrus Juices and Other Products," U.S. Patent 3,920,851, November 18, 1975a.

Hasegawa, S., L.C. Brewster, and V.P. Maier, "Limonate:NAD(P) Oxidoreductase Manufacture and Use in Prevention of Bitterness in Citrus Juice," U.S. Patent 3,917,512, November 4, 1975b.

Johnson, R.L., and B.V. Chandler, "Ion Exchange and Adsorbent Resins for Removal of Acids and Bitter Principles from Citrus Juices," J. SCI. Food Agr., 36(1985)480-484.

Johnson, R.L., and B.V. Chandler, "Reactivation of Cellulose Acetate Gel Bead Column Used for the Removal of Limonin from Citrus Juices," j. Sci. Food Agr., 32(1981a)1191-1196.

Johnson, R.L., and B.V. Chandler, "A Pilot-Plant Cellulose Acetate Gel Bead Column for the Removal of Limonin from Citrus Juices," J. Sci. Food Agr., 32(1981b)1183-1190.

Levi, A., et al., Lebensm.-Wiss. Technol., 7 (1974)234-235.

Matthiasson, E., "Fouling in Membrane Filtration," in Fouling-and-Cleaning in Food Processing, D. Lund, B. Hallstronm. E. Plett, C. Sander, and C. Tragadh (eds.). University of Wisconsin-Madison Duplicating, Madison, Wisconsin (1985).

Mitchell, D.H., R.M, Pearce, C.B. Smith, and S.T. Brown, "Removal of Bitter Naingin and Limonin from Citrus Juices Containing Them," U.S. Patent 4,514,427, April 30, 1985.

Puri, A., "Citrus Juices, Concentrates, and Dried Powders Which Are Reduced in Bitterness," U.S. Patent 4,439,458, March 27, 1984.

Shaw, P.E., J.H. Tatum, and C.W. Wilson, "Improved Flavor of Novel Orange and Grapefruit Juices by Bitter Components with B-Cyclodextrin Polymer," J. Agr. Food Chem., 32(1984)832-836.

Shaw, P.E., and C.W. Wilson, "A Rapid Method for Determination of Limonin in Citrus Juices by High Performance Liquid Chromatography," J. Food Sci., 49(1984)1216-1218.

Taselskis, B., and J. Nelapaty, "Spectrophotometric Determination of Microamounts of Ascorbic Acid in Citrus Fruits," Anal. Chem., 44(1972)379-381.

The Almana of the Canning, Freezing, Preserving Industries, E. Judge and Sons, Wesminster, Maryland (1979) p. 576.