Bioresource Technology 101 (2010) 3859–3863

Contents lists available at ScienceDirect

Bioresource Technology

journal homepage: www.elsevier .com/locate /bior tech

Nutrient recovery from the dry grind process using sequential microand ultrafiltration of thin stillage

Amit Arora a, Bruce S. Dien b, Ronald L. Belyea c, Vijay Singh a, M.E. Tumbleson a, Kent D. Rausch a,*

a Agricultural and Biological Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USAb National Center for Agricultural Utilization Research, Agricultural Research Service, USDA, 1815 North University Street, Peoria, IL 61604, USAc Animal Sciences, University of Missouri, Columbia, MO 65211, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 24 July 2009Received in revised form 23 December 2009Accepted 31 December 2009Available online 6 February 2010

Keywords:CornEthanolThin stillageMembrane filtration

0960-8524/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.biortech.2009.12.127

* Corresponding author. Address: Agricultural andversity of Illinois at Urbana-Champaign, 1304 West PIL 61801, USA. Tel.: +1 217 265 0697; fax: +1 217 24

E-mail address: krausch@illinois.edu (K.D. Rausch)

The effectiveness of microfiltration (MF) and ultrafiltration (UF) for nutrient recovery from a thin stillagestream was determined. When a stainless steel MF membrane (0.1 lm pore size) was used, the content ofsolids increased from 7.0% to 22.8% with a mean permeate flux rate of 45 L/m2/h (LMH), fat increased andash content decreased. UF experiments were conducted in batch mode under constant temperature andflow rate conditions. Permeate flux profiles were evaluated for regenerated cellulose membranes (YM1,YM10 and YM100) with molecular weight cut offs of 1, 10 and 100 kDa. UF increased total solids, proteinand fat and decreased ash in retentate stream. When permeate streams from MF were subjected to UF,retentate total solids concentrations similar to those of commercial syrup (23–28.8%) were obtained.YM100 had the highest percent permeate flux decline (70% of initial flux) followed by YM10 and YM1membranes. Sequential filtration improved permeate flux rates of the YM100 membrane (32.6–73.4 LMH) but the percent decline was also highest in a sequential MF + YM100 system. Protein recoverywas the highest in YM1 retentate. Removal of solids, protein and fat from thin stillage may generate apermeate stream that may improve water removal efficiency and increase water recycling.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

In the corn dry grind ethanol industry, thin stillage is recycledas backset which is added in the slurry tank. The rest of the thinstillage is sent to evaporators for concentration which involves sig-nificant energy inputs and results in evaporator fouling problems.To decrease the load upon evaporators and reduce demand forfresh water, recycle of water in the thin stillage stream should beincreased. However, there are concerns with recycling thin stillageabove a certain percentage, generally P30–50% depending onplant operating conditions. As percent recycle is increased, the con-centrations of various compounds, especially lactic and aceticacids, inhibit yeast growth and reduce ethanol yields (Chin andIngledew, 1993; Ingledew, 2003).

Ultrafiltration (UF) is an efficient process for selective removalof compounds by convective solvent flow through a membrane.Membrane filtration involves no evaporation of water; hence en-ergy consumption is lower than with thermal methods. Applica-tion of UF membranes to process thin stillage obtained from

ll rights reserved.

Biological Engineering, Uni-ennsylvania Avenue, Urbana,4 0323..

conventional and an enzymatic corn dry grind (E-Mill) processeshave been described (Arora et al., 2009, 2010). Total solidsrecovered through batch UF membrane separation were similarto solids levels obtained from commercial evaporators. Themembrane fouling issue has also been addressed during thin stil-lage concentration and found to be primarily reversible using fil-tration parameters and cleaning methods (Arora et al., 2009). Insmall scale experiments, the flux rates of thin stillage throughMF and UF membranes was high enough to appear economicallyfeasible (Arora et al., 2009). However, application of microfiltra-tion (MF) and sequential MF + UF processes have not been usedto filter and recover nutrients from commercial thin stillage; dis-tribution of yeast inhibitors in membrane retentate and perme-ate streams have not been investigated. Water recycling ratescould be increased if MF and UF methods were effective inremoving compounds that inhibit yeast growth and metabolism.We evaluated nutrient recovery using MF and UF membranesand evaluated permeate streams on the basis of organic com-pound removal. Specific objectives were to: (1) compare filtra-tion characteristics of thin stillage for MF, UF and sequentialMF + UF processes, (2) evaluate solids recovery and nutrientcompositions of permeate and retentate streams and (3) evaluatethe permeate streams for potential water recycling based on or-ganic acid contents.

3860 A. Arora et al. / Bioresource Technology 101 (2010) 3859–3863

2. Methods

2.1. Experimental material

Thin stillage was collected from a commercial dry grind ethanolfacility and stored at 4 �C. To characterize thin stillage and reten-tate streams, one 500 mL sample was analyzed for total solids(TS) using a two stage oven method (Approved Method 44-15A,AACC International, 2000). Composition of thin stillage and reten-tate streams obtained from MF and UF processes were analyzed forprotein (total nitrogen � 6.25), fat and ash content using standardmethods (AOAC, 2003) at the University of Missouri, Columbia.Compositional analyses were performed in duplicate.

2.2. Microfiltration equipment

Microfiltration (MF) was carried out using a bench top mem-brane unit. A tubular stainless steel module with six tubes having0.61 m length, nominal diameter of 0.64 cm, 0.1 lm pore size,0.31 cm wall thickness and 0.07 m2 membrane area (Scepter model2.5-250A-2P6, Graver Technologies, Glasgow, DE) was used in acrossflow filtration arrangement. The unit was equipped with abatch tank of 15 L capacity, a heat exchanger and positive displace-ment pump (model M-03, Hydra-Cell, Minneapolis, MN). The unitwas operated in total recycle mode (permeate returned to tank)or batch concentration mode (permeate collected separately andretentate recycled) with 4.75 m/s crossflow velocity. The permeatethat passed through the membrane was termed MFP and the mate-rial that was retained and returned to tank was termed retentate.Permeate was collected in a graduated cylinder during batch con-centration. Thin stillage (15 L) was used for each batch experiment;permeate flux rates measurements were taken during concentra-tion and presented as LMH (L/m2/h) until batch material was ex-hausted. Five replicates were performed for concentrationprofiles. Permeate flux rate was determined manually, with gradu-ated cylinder and stopwatch. The selected operating conditionswere 690 ± 13 kPa transmembrane pressures (TMP) with a cross-flow velocity and fluid temperature of 4.75 m/s and 75 ± 2 �C,respectively. Thin stillage temperatures in an ethanol plant are60–75 �C. Therefore, the operating temperature chosen for theMF filtration study was 75 ± 2 �C. Average permeate flux rate val-ues were calculated using Eq. (1) as described in the sectiondescribing membrane performance.

2.3. Stirred cell ultrafiltration unit

A stirred ultrafiltration cell (400 mL Amicon, model 8400, Milli-pore Corporation, Bedford, MA) was used for concentrating thinstillage at room temperature. An argon gas cylinder was used toapply pressure to the stirred cell. A magnetic stir bar was used to

R2 P2

YM100

R1 P1

MF

Thin Stillage

Fig. 1. Thin stillage filtration throug

simulate crossflow filtration. Two regenerated cellulose mem-branes, YM10 and YM100 (Millipore Corporation, Bedford, MA)with pore sizes of 10 and 100 kDa, respectively, and effective mem-brane area of 41.8 cm2 were used. Five replicates of thin stillage(300 mL) were used in filtration.

2.4. Membrane selection

Since thin stillage contains a range of molecular weight com-pounds such as amino acids, peptides, proteins, fat and minerals(Jones and Ingledew, 1994; Kim et al., 2008; Arora et al., 2009), itis important to choose a membrane or membranes that recoversolids and produce cleaner permeate stream with higher flux rates.Therefore, four membranes with different pore sizes were chosento evaluate permeate flux rates and solids recovery. In phase I fil-tration experiments (Fig. 1), thin stillage batches were filteredthrough stainless steel MF (0.1 lm pore size) and regenerated cel-lulosic UF membranes YM1, YM10 and YM100 with 1, 10 and100 kDa molecular weight cutoff (MWCO), respectively. Permeateflux rates were measured for all membranes. Thin stillage, reten-tates and permeates were analyzed for compositions. Five replica-tions were used for each treatment.

In phase II, permeates obtained from MF runs were filtered fur-ther using YM100, YM10 and YM1 membranes (Fig. 2) and ana-lyzed for lactic and acetic acid concentrations. UF experimentswere conducted at 25 �C (room temperature) and used pressuresrecommended by the manufacturer for each membrane (380, 207and 70 kPa for YM1, YM10 and YM100 membranes, respectively).Both MF (15 L batch) and UF (350 mL/batch for phase I, 300 mL/batch for phase II) experiments were operated in batch concentra-tion mode and experiments were continued until the material wasexhausted. Permeate flux rates were measured using a graduatedcylinder during MF experiments. During UF, permeate flux rateswere determined gravimetrically by measuring the cumulativeweight permeated, collected from the bottom of the cell as a func-tion of time using an electronic balance. In sequential filtration, MFpermeate was further filtered using a UF membrane. Permeate fluxrates for UF and MF + UF membranes were calculated from start ofthe UF experiments.

2.5. Measurement of membrane separation performance

The average permeate flux rate (Jav) was calculated by

Jav ¼VAt

ð1Þ

where Jav was the average flux rate (LMH), V was the total volume(L) of permeate, A was the effective area of the membrane, and t(h) was the permeate collection time.

R4R3 P3 P4

YM10 YM1

h various membranes (phase I).

Thin Stillage

Permeate (P1)Retentate (R1)

MF

R7R5 R6P5 P6 P7

YM100 YM10 YM1

Fig. 2. Sequential filtration (MF + UF) of thin stillage (phase II).

Table 1Average flux rates of membranes.*

Membrane Type Pore size Jav (LMH)

Microfiltration (MF) Stainless steel 0.1 lm 45.0 ± 9.2bYM100 Regenerated cellulose 100 kDa 32.6 ± 5.8bcMF + YM100 73.4 ± 7.4aYM10 Regenerated cellulose 10 kDa 29.8 ± 5.0cMF + YM10 32.2 ± 2.3bcYM1 Regenerated cellulose 1 kDa 9.4 ± 0.8dMF + YM1 11.0 ± 0.8d

* Mean ± standard deviation. Values in same column with different letters aredifferent (p < 0.05).

A. Arora et al. / Bioresource Technology 101 (2010) 3859–3863 3861

In the case of MF + UF, Jav was calculated using Eq. (1) by consid-ering total permeate volume passed through each UF membrane,the area of UF membrane and total time required for permeate topass through the membrane. The Jav calculation did not includethe MF process as the purpose of MF was to preprocess thin stillage(i.e., to remove suspended solids from thin stillage).

Solids recovery, equivalent to the rejection of solids by eachmembrane type, was defined by

R ð%Þ ¼ 100 1� CP

CR

� �ð2Þ

where CP and CR were permeate and retentate total solids concen-trations, respectively.

2.6. Permeate composition and effect on water recycling

Organic acids such as lactic and acetic acid above a certain con-centration (lactic acid > 0.8% w/v and acetic acid > 0.05% w/v) affectyeast performance and therefore reduce ethanol production(Narendranath et al., 2001). Original thin stillage and permeatestreams obtained from different pore sized membranes were ana-lyzed for lactic and acetic acid concentrations using HPLC methods.Based on these concentrations from previous work, the effective-ness of membrane filtration methods used in reducing lactic andacetic acids was evaluated. If lactic and acetic acid concentrationswere above the thresholds described in Narendranath et al. (2001),it is likely that MF, UF and MF + UF permeate streams would needadditional separation through membranes such as nanofiltration orreverse osmosis in order to affect the amount of recycling withinthe plant.

2.7. HPLC analyses

Thin stillage and permeate samples were collected to determineconcentrations of ethanol, lactic acid, acetic acid, glucose, fructose,maltose, maltotriose, DP4+ and glycerol using HPLC. Each samplewas centrifuged at 1789g for 5 min (Centra CL3, Thermo IEC, Need-ham Heights, MA); supernatant was filtered through a 0.2 lm syr-inge filter into 1 mL vials. Filtrate was injected into an ionexclusion column (Aminex HPX-87H, Bio-Rad, Hercules, CA) main-tained at 50 �C. At a rate of 0.6 mL/min, sugars, alcohols and organ-ic acids were eluted from the column with HPLC grade watercontaining 5 mM sulfuric acid. Components were detected with arefractive index detector (model 2414, Waters Corporation, Mil-ford, MA). Data processing was performed using HPLC software(Waters Corporation). For calibration, we used standards contain-

ing the above components at known concentrations. Two repli-cates of each sample were injected for analysis.

2.8. Statistical analysis

Average flux rates of permeate streams and compositions ofretentate streams for thin stillage and permeate streams weredetermined. Thin stillage and permeate streams obtained from var-ious pore sized membranes were analyzed using HPLC. Five repli-cates for each process (MF, UF, and MF + UF) were used. Acompletely randomized design was used in each case. Results wereexpressed as means and standard deviations. Statistical analyseswere performed by using statistical software (SAS, 1989) using asignificance level of p < 0.05. One-way ANOVA was used to detectdifferences among means. The LSD method was used to comparemean compositions of streams and HPLC results.

3. Results and discussion

3.1. Flux profiles

MF had the highest average flux rates compared to UF mem-branes (Table 1) at 25 �C. Jav for MF + YM100 was higher than forthe YM100 membranes alone because removal of suspended solidsusing MF reduced fouling resistance and increased Jav for theYM100 membrane by more than 200% (32 and 73 LMH forYM100 and MF + YM100 membranes, respectively). In the case ofYM10 and YM1 membranes, sequential filtration (MF + YM10 andMF + YM1) did not increase permeate flux rates. The YM1 mem-brane had higher solids recovery (87%) than MF (81%) as calculatedusing Eq. (2). YM100 had the highest percent flux decline in UF(Figs. 3 and 4).

For sequential processes, MF + YM100 had the highest percentflux decline relative to the amount of permeate collected; only30% of initial flux was recorded after 285 ± 5 mL of permeate wasrecovered from 300 mL of input stream volume (Fig. 5). The YM1membrane had the lowest percent flux decline for UF and MF + UFfiltrations. About 60% of initial flux was observed after collection of275 ± 5 mL permeates (Fig. 5).

Permeate flux rate values declined more slowly over time forYM10 compared to YM100 membrane (Fig. 3). Decline in permeateflux rate was expected due to gel layer formation on the membranesurface as thin stillage concentration increased and deposition oc-curred (Zeman and Zydney, 1996; Cheryan, 1998; Ghosh, 2003;Yorgun et al., 2008). For pressure driven processes, a greater fluxdecrease signified build up of solute at the interface. The YM100membrane initially had higher flux rates than tighter membranes,YM10 and YM1, but eventually had lower flux rates (Figs. 3 and 4).Since original thin stillage had a broader particle size distributionthan permeate, particles might have lodged into the larger poresof the YM100 membrane and blocked them resulting in a flux ratedecrease. Initial filtration of thin stillage using MF did not have any

0

20

40

60

80

100

120

140

160

0 50 100 150 200

Flux

Rat

e (L

MH

)

Time (min)

YM1 YM10 YM100

MFP+YM1 MFP+YM10 MFP+YM100

Fig. 3. Permeate flux rates (LMH: L/m2/h) of UF and sequential MF + UF filtration.

0

20

40

60

80

100

120

140

160

0 50 100 150 200 250 300

Flux

rat

e (L

MH

)

Volume Collected (mL)

MFP+YM1 MFP+YM10MFP+YM100 YM1YM10 YM100

Fig. 4. Permeate flux rates (LMH: L/m2/h) and permeate volume collected forsequential MF + UF filtration.

0

20

40

60

80

100

120

0 50 100 150 200 250 300

Initi

al F

lux

(%)

Volume Collected (mL)

MFP+YM1 MFP+YM10 MFP+YM100

Fig. 5. Normalized percent flux rate with respect to initial permate flux rate forsequential MF + UF filtration.

3862 A. Arora et al. / Bioresource Technology 101 (2010) 3859–3863

observed effect on the YM1 membrane permeate flux profile(Fig. 4). YM1 pores, 100 times smaller than those of the YM100membrane, were too small for particles to cause plugging. Im-proved flux profiles of YM10 following MF (MF + YM10) was no-ticed compared to permeate flux obtained using only YM10membrane and demonstrates how solute size and pore size inter-acted during filtration. In the sequential MFP + YM10 process, theYM10 membrane was plugged by macromolecules that were ableto penetrate the pores. YM10 pore size, intermediate betweenYM100 and YM1, rejected macromolecules larger than pore sizeand therefore had higher flux rate compared to YM10 alone.

The flux values observed during MF and UF experiments(Jav = 9–73 LMH, Table 1) were of similar magnitude observed inother work. Fukumoto et al. (1998) used ceramic MF and UF mem-branes to clarify apple juice in stirred cell systems. Most fluxeswere between 30 and 75 LMH for membranes having 0.02–0.2 lm pore size and operating at 23� and 276 kPa. They observedincreased flux rates when using membranes with smaller poresizes (0.2 lm). Padilla-Zakour and McLellan (1993) filtered applejuice in a crossflow filtration system using a 0.2 lm ceramic mem-brane at 50 �C, 250 kPa, 14.6 m/s (relatively high velocity) and ob-served 285 LMH. Ben Amar et al. (1990) reported 100 LMH with0.2 lm ceramic membrane operated at 50–55 �C, 300 kPa and2.8 m/s. Ghosh and Balakrishnan (2003) conducted a pilot scalestudy using sugarcane juice and using spiral wound UF modules(809 m2 membrane area) operating at 91–97 �C and observed7 LMH. Yorgun et al. (2008) compared UF, NF and RO membranefiltration methods to treat whey from a cheese processing plant.Steady fluxes of 15, 10–15 and 2–3 LMH for UF, NF and RO, respec-tively, were observed.

Mean total solids in retentates were similar among MF and UFmembranes (Table 2). YM1 membrane had higher solids recoverythan the MF membrane. While these data indicate potential to in-crease solids contents of retentate streams, it should be kept inmind that MF and UF apparatus were not designed to determinemaximum solids levels possible for each membrane type. MF andUF experiments were conducted in batch mode; thus, when mate-rial was depleted in the recirculation loop, each experiment wasterminated. Therefore, higher concentrations of retentate solidsmay be possible. There were no differences detected among pro-tein contents of MF, YM100 and YM10 membrane retentatestreams. Retentate streams from the smaller pore size membranes(YM1 and YM10) had lower fat content values than retentate ob-tained from the larger pore size YM100 membrane but similar tothe MF membrane. Ash contents were reduced more than 50% inretentate streams of all membranes. Reduction of ash contents inretentates, which would eventually be combined with streams toproduce DDGS, may be an advantage from an animal nutritionstandpoint. In thin stillage and DDGS, phosphorus contents arequite high relative to animal nutrient needs (Belyea et al., 2004,2006; Rausch and Belyea, 2006). By reducing ash content, phos-phorus content may also be reduced and improve DDGS value.

3.2. Permeate composition and effect on water recycling

There were no differences in lactic acid concentrations of thinstillage and permeate streams (Table 3). Thin stillage samples ob-tained from a dry grind facility at different periods of time hadmean lactic acid concentrations (0.1% w/v) eight times lower thaninhibitory level to the yeast (0.8% w/v; Narendranath et al., 2001).Acetic acid contents of original thin stillage samples were belowthe HPLC detection limit. Glycerol concentrations were not differ-ent between thin stillage and MF permeate. Since stillage is con-centrated using evaporators and eventually added to DDGS,valuable coproducts such as organic acids and glycerol end up asa part of animal diets.

Table 2Retentate compositions from various membranes.*

Treatments Recovery (%) Total solids Protein (% db) Fat (% db) Ash (% db)

Thin stillage 6.5 ± 0.7a 23.5 ± 2.3a 16.7 ± 1.6a 10.5 ± 0.5aMF 81 ± 2a 22.4 ± 1.2b 27.6 ± 1.3ab 31.1 ± 2.4b 3.8 ± 0.5bYM100 82 ± 1a 24.6 ± 3.1b 28.4 ± 0.3b 39.0 ± 4.0c 3.1 ± 0.7bYM10 84 ± 3ab 26.2 ± 2.0b 29.0 ± 0.7b 34.2 ± 1.7b 4.3 ± 0.1bcYM1 87 ± 1b 28.8 ± 1.4bc 32.2 ± 0.1c 31.5 ± 1.7b 5.1 ± 0.7c

* Mean ± standard deviation. Values in same column with different letters are different (p < 0.05).

Table 3HPLC analysis of thin stillage and permeate streams using various pore sized membranes.a

TSb MF YM100 YM10 YM1 MF + YM100 MF + YM10 MF + YM1

Glycerol 1.66 ± 0.06c 1.68 ± 0.08c 1.94 ± 0.03a 1.94 ± 0.02a 1.96 ± 0.04a 1.86 ± 0.05ab 1.80 ± 0.03b 1.79 ± 0.04bcAcetic acid Bdl* Bdl Bdl Bdl Bdl Bdl Bdl BdlLactic acid 0.1 ± 0.02a 0.09 ± 0.007a 0.14 ± 0.09a 0.11 ± 0.06a 0.11 ± 0.04a 0.10 ± 0.008a 0.10 ± 0.06a 0.09 ± 0.02aGlucose 0.04 ± 0.002 0.04 ± 0.002 0.08 ± 0.003 0.04 ± 0.008 0.03 ± 0.007 0.05 ± 0.004 0.04 ± 0.003 0.04 ± 0.005Maltose 0.21 ± 0.006 0.20 ± 0.01 0.23 ± 0.01 0.23 ± 0.01 0.16 ± 0.05 0.19 ± 0.02 0.20 ± 0.04 0.21 ± 0.01Maltotriose 0.05 ± 0.007 0.05 ± 0.007 0.05 ± 0.002 0.06 ± 0.001 0.04 ± 0.001 0.04 ± 0.001 0.05 ± 0.001 0.05 ± 0.002DP4+ 0.95 ± 0.08 0.41 ± 0.02 0.64 ± 0.03 0.45 ± 0.01 0.36 ± 0.02 0.44 ± 0.04 0.40 ± 0.03 0.30 ± 0.02

a Mean ± standard deviation. Values in same row with different letters are different (p < 0.05). Means were compared only for lactic acid and glycerol. N = 5.b TS = thin stillage from commercial dry grind facility.

* Below detection limit.

A. Arora et al. / Bioresource Technology 101 (2010) 3859–3863 3863

4. Conclusions

During MF and UF membrane filtration experiments, MF hadhigher permeate flux rates. Solids concentrations obtained usingmembranes were similar to commercial condensed distillers solu-bles. Ash contents were reduced more than 50% in retentatestreams of membranes tested. Protein recovery was highest inYM1 retentate compared to other membranes. Fat removal fromretentate could increase retentate protein content, thus, a bettervalue as an animal food. Sequential filtration increased permeateflux rates of YM100 membrane (32.6–73.4 LMH). Sequential mem-brane filtration (MF + UF) changed permeate flux rates significantlyand may provide methods to increase water recycle within the drygrind process.

References

AACC International, 2000. Approved Methods of the American Association of CerealChemists, 10th ed. Method 44-15A, 44-19. The Association, St. Paul, MN.

AOAC, 2003. Official Methods of the AOAC, 17th ed. Methods 920.39, 942.05,990.03. The Association of Official Analytical Chemists, Gaithersburg, MD.

Arora, A., Wang, P., Belyea, R.L., Tumbleson, M.E., Singh, V., Rausch, K.D., 2009. Thinstillage fractionation using ultrafiltration: resistance in series model.Bioprocess. Biosyst. 32, 225–233.

Arora, A., Wang, P., Singh, V., Tumbleson, M.E., Belyea, R.L., Rausch, K.D., 2010.Laboratory yields and compositions of conventional and E-mill dry grindprocesses. Cereal Chem. 87, 100–103.

Belyea, R.L., Rausch, K.D., Tumbleson, M.E., 2004. Composition of corn and distillersdried grains with solubles in dry grind ethanol processing. Biores. Technol. 94,293–298.

Belyea, R.L., Clevenger, T.E., Singh, V., Tumbleson, M.E., Rausch, K.D., 2006. Elementconcentrations of dry grind corn processing streams. Appl. Biochem. Biotechnol.134, 113–128.

Ben Amar, R., Gupta, B.B., Jaffrin, M.Y., 1990. Apple juice clarification using mineralmembranes: fouling control by backwashing and pulsating flow. J. Food Sci. 55,1620–1625.

Cheryan, M., 1998. Ultrafiltration and Microfiltration Handbook, second ed.Technomic, Lancaster, PA.

Chin, P.M., Ingledew, W.M., 1993. Effect of recycled laboratory backset onfermentation of wheat mashes. J. Agric. Food Chem. 41, 1158–1163.

Fukumoto, L.R., Delaquis, P., Girard, B., 1998. Microfiltration and ultrafiltrationceramic membranes for apple juice clarification. J. Food Sci. 63, 845–850.

Ghosh, R., 2003. Protein Bioseparation Using Ultrafiltration: Theory Applicationsand New Developments. Imperial College Press, London, United Kingdom.

Ghosh, A.M., Balakrishnan, M., 2003. Pilot demonstration of sugarcane juiceultrafiltration in an Indian sugar factory. J. Food Eng. 58, 143–150.

Ingledew, W.M., 2003. Water reuse in fuel alcohol plants: effect on fermentation. Isa ‘zero discharge’ concept attainable? In: Jacques, K.A., Lyons, T.P., Kelsall, D.R.(Eds.), The Alcohol Textbook, fourth ed. Nottingham University Press,Nottingham, United Kingdom, pp. 343–353.

Jones, A.M., Ingledew, W.M., 1994. Fuel alcohol production: appraisal of nitrogenousyeast foods for very high gravity wheat mash fermentation. Process Biochem.29, 483–488.

Kim, Y., Mosier, N.S., Hendrickson, R., Ezeji, T., Blaschek, H., Dien, B., Cotta, M., Dale,B., Ladisch, M.R., 2008. Composition of corn dry-grind ethanol by-products:DDGS, wet cake, and thin stillage. Biores. Technol. 99, 5165–5176.

Narendranath, N.V., Thomas, K.C., Ingledew, W.M., 2001. Effects of acetic acid andlactic acid on the growth of Saccharomyces cerevisiae in a minimal medium. J.Ind. Microbiol. Biotechnol. 26, 171–177.

Padilla-Zakour, O., McLellan, M.R., 1993. Optimization and modeling of apple juicecross-flow microfiltration with a ceramic membrane. J. Food Sci. 58, 369–374.388.

Rausch, K.D., Belyea, R.L., 2006. The future of coproducts of corn processing. Appl.Biochem. Biotechnol. 128, 47–86.

SAS, 1989. SAS Use’s Guide: Statistics, fifth ed. SAS Institute, Cary, NC.Yorgun, M.S., Balcioglu, I.A., Saygin, O., 2008. Performance comparison of

ultrafiltration, nanofiltration and reverse osmosis on whey treatment.Desalination 229, 204–216.

Zeman, L.J., Zydney, A.L., 1996. Microfiltration and Ultrafiltration: Principles andApplications. Marcel Dekker, New York.