Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=lsst20 Download by: [Georgia Tech Library] Date: 04 September 2017, At: 18:50 Separation Science and Technology ISSN: 0149-6395 (Print) 1520-5754 (Online) Journal homepage: http://www.tandfonline.com/loi/lsst20 Membranes for Kraft black liquor concentration and chemical recovery: Current progress, challenges, and opportunities Nikita S. Kevlich, Meisha L. Shofner & Sankar Nair To cite this article: Nikita S. Kevlich, Meisha L. Shofner & Sankar Nair (2017) Membranes for Kraft black liquor concentration and chemical recovery: Current progress, challenges, and opportunities, Separation Science and Technology, 52:6, 1070-1094, DOI: 10.1080/01496395.2017.1279180 To link to this article: http://dx.doi.org/10.1080/01496395.2017.1279180 View supplementary material Accepted author version posted online: 10 Jan 2017. Published online: 10 Jan 2017. Submit your article to this journal Article views: 133 View related articles View Crossmark data
Transcript
Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=lsst20
Membranes for Kraft black liquor concentrationand chemical recovery: Current progress,challenges, and opportunities
Nikita S. Kevlich, Meisha L. Shofner & Sankar Nair
To cite this article: Nikita S. Kevlich, Meisha L. Shofner & Sankar Nair (2017) Membranes for Kraftblack liquor concentration and chemical recovery: Current progress, challenges, and opportunities,Separation Science and Technology, 52:6, 1070-1094, DOI: 10.1080/01496395.2017.1279180
To link to this article: http://dx.doi.org/10.1080/01496395.2017.1279180
View supplementary material
Accepted author version posted online: 10Jan 2017.Published online: 10 Jan 2017.
Membranes for Kraft black liquor concentration and chemical recovery: Currentprogress, challenges, and opportunitiesNikita S. Kevlicha,b, Meisha L. Shofnerb,c, and Sankar Naira,b
aSchool of Chemical & Biomolecular Engineering, Georgia Institute of Technology, Atlanta, Georgia, USA; bRenewable Bioproducts Institute,Georgia Institute of Technology, Atlanta, Georgia, USA; cSchool of Materials Science & Engineering, Georgia Institute of Technology, Atlanta,Georgia, USA
ABSTRACTMembranes can significantly reduce energy consumption during concentration of black liquor (BL) inthe Kraft papermaking process, but the harsh conditions (pH ~12, 80°C–95°C, ~15 wt% solids) makethis challenging. We elucidate challenges and opportunities for membranes in BL applications. Wecritically reviewmembrane materials, processes, and operational modes investigated in the literature.Future advances will involve fabrication of higher-rejecting (≥95% lignin and inorganics), BL-resistant,NF, and RO membranes. Opportunities exist for molecular sieving and electrically driven membranesto recover other valuable chemicals such as carboxylic acids. We also discuss the economics of BLconcentration with a single-stage membrane process.
ARTICLE HISTORYReceived 23 July 2016Accepted 3 January 2017
The production of forest-based products such as paperinvolves highly energy-intensive pulping processes. TheKraft process (Fig. 1) is the dominant pulping process, inwhich NaOH and Na2S are used as the primary pulpingchemicals (known as cooking liquor or white liquor).[1] Inthis process, wood chips and the pulping chemicals aremixed in a pressurized, heated digester to break up thelignin that holds the wood fibers together and remove thepulp, which is rich in cellulose and used for paper produc-tion. The waste stream from this process is known asblack liquor (BL). The BL stream that exits the digesteris called weak BL (WBL) and is at a typical temperature of80°C–95°C, high pH ~12) and about 15% total solids(~175 kg/m3 solutes).[2, 3] It contains a number of inor-ganic and organic compounds; the most important ofwhich are the pulping chemicals, degraded lignin (usedprimarily to generate energy), and carboxylic acids (sig-nificantly degraded hemicelluloses, which could be usedfor higher-value products). The weak BL stream is thenconcentrated in multiple-effect evaporators to strong BL(SBL) that contains about 75%–80% total solids (~875–933 kg/m3 solutes).[2, 3] This concentrated slurry can becombusted to generate steam, and the resultant inorganicsmelt is dissolved and causticized to recover the pulpingchemicals, which are sent back to the digester.[4–6]
With more than 500 million tons/year of BL generatedworldwide and almost 173 trillion kJ/year (0.2 Quads/year)of energy used for BL concentration in the US alone, theprocessing of BL is an important industrial and environ-mental issue.[7] Membranes are a potential alternative forthe energy-efficient concentration of BL and for recoveringvaluable organic and inorganic components.[8] Because ofthe high fouling potential and high viscosity of concen-trated BL, a complete replacement of the entire evaporationprocess with membranes is unlikely.[8] However, even par-tial concentration (up to about 30%–40% total solids)would result in a major reduction in energy usage.Membranes also have advantages over alternateapproaches for lignin removal that rely upon inducing itsprecipitation via acidification or non-membrane electro-chemical techniques.[9,10] One relevant example of the acidprecipitation technique is the LignoBoost process devel-oped by Innventia and Chalmers University ofTechnology, which has been scaled up in two plants. Thisprocess focuses on lignin recovery post-precipitation withCO2 by utilizing membrane filtration and resuspension ofthe filtrate cake to achieve high purity and high recovery oflignin with a low ash content (~1 wt%).[11, 12] However,techno-economic comparisons of the acidification or non-membrane electrochemical techniques support the generalconclusion that while acidification processes can achieve a
CONTACT Sankar Nair [email protected] School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, Atlanta, GA30332-0100, USA.Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/lsst.
Supplemental data for this article can be accessed on the publisher’s website.
SEPARATION SCIENCE AND TECHNOLOGY2017, VOL. 52, NO. 6, 1070–1094http://dx.doi.org/10.1080/01496395.2017.1279180
high degree of lignin removal, they may also involve highcosts of acidifying chemicals, electrical energy, anti-corro-sion, and process safety measures.[13–19] In light of thesedrawbacks, membrane processes—mainly pressure drivensuch as nanofiltration or reverse osmosis, but also poten-tially electrodialysis and forward osmosis processes—offerthe potential of a lower-cost, easily scalable process inexisting pulp mills to directly concentrate BL as well as toseparate valuable organic and inorganic compounds.
Although BL concentration with membranes is animportant technological challenge, the literature on thistopic has not yet been organized in the context of adetailed review and critique of the state of the art. Thereare two main purposes of the present review article.Firstly, we overview and understand the current state ofthe art in membrane science and engineering for BLconcentration and related applications. Secondly, weidentify the key challenges related tomembranematerials,membrane processing, development of structure–perfor-mance relationships, and operational requirements thatmust be addressed in order to obtain membranes that areviable for BL treatment. In addition to the ubiquitouschallenge of achieving high membrane flux and goodselectivity for desired components, a number of impor-tant issues exist in relation to the long-term stability ofmembranes under harsh operating conditions (especiallypH ~12 and temperature 80°C–95°C), the complex com-position of BL, and its fouling characteristics. To addressthe above two objectives, this article is structured asfollows. We begin with a discussion of the key composi-tional characteristics and physicochemical properties ofBL and outline the general membrane characteristics andmaterials that have been used. We then present a con-sideration of the membrane characteristics as a functionof membrane type and operating configurations. The
effects of fouling are also investigated. We conclude witha discussion of the economics of different BL concentra-tion membrane configurations.
Composition and physicochemical properties ofblack liquor
BL composition depends significantly on the pulpingprocess and on the type of biomass feedstock used.Common wood sources include softwoods (such aspine), hardwoods (such as eucalyptus), and fibrous plants(such as bamboo).[2] The dry solids content in BL can berepresented either as the total dissolved solids (TDS) or asthe total solids (TS). TS include both the dissolved solidsand the suspended solids that exist as particles or colloidsin BL, such as lignin. Tables 1 and 2 summarize keyinformation relevant to the discussion in this section.The main inorganics (Table 1) include NaOH, Na2S,Na2CO3, Na2SO4, Na2S2O3, NaCl, and SiO2 (if non-wood BL), with most of the sodium present in BLbound to the phenolic hydroxyl groups in lignin.[3, 20–23]
Table 2 summarizes the typical relative amounts,concentrations, and weight-average molecular weights(Mw) of the three major organic components in BL.The most abundant organic component in BL is thedegraded lignin, which exists in BL as colloidal macro-molecules that have a high degree of cross-linking.[24,57]
Recovered lignin is valuable because in addition to its
Figure 1. Simplified schematic of the Kraft pulping process andBL recycling (adapted from Refs.[5,6]).
Table 1. Main inorganic solutes in Kraft black liquor.Compound (% TDS Basis) kg/m3 Refs.
Na2CO3 6.6–12.3 2.05–3.5 [3, 20–26]
NaOH 1.3–2.4 0.14–40 [3, 21–23, 25, 27, 28]
Na2S2O3 0.5–4.1 4.7–5.8 [21–23, 29]
Na2S – 0.88–1.11 [3, 25, 26]
Na2SO4 0.9–8.3 – [20–24]
NaCl 0.5 – [21]
SiO2 0.2–0.7 (wood)1–30 (grasses)
– [24, 30]
Table 2. Main types of organic species in Kraft black liquor andsome typical ranges.
ability to produce energy from the combustion, ligninhas many component species, which can be used inplace of petroleum-derived molecules as buildingblocks for higher-value chemicals. The most commoncurrent commercial uses of lignin are as dispersants, asbinders, and in steam/power production.[14,52,58]
The phenolic hydroxyl and carboxyl functional groupsin lignin help to stabilize it in BL (e.g., as in Eq. (1)):[2, 24]
R� OH þ NaOH ! R� ONaþ H2O (1)
Because BL is highly alkaline, the hydroxyl groups dis-sociate at higher pH, making the usually hydrophobiclignin more hydrophilic with a negative surface charge.[34]
If a hydrophilic membrane is used, lignin could interactstrongly with themembrane surface and decrease the waterflux.[34] At pH above 12.5, enough phenolic groups havebecome ionized to enable lignin to take a more compactand spherical particle shape, thusmaking it easier to handlein feed streams.[1, 2] Even though the Mw of lignin in BLcan vary widely (with extremes <1 and >50 kDa), theMw isusually 0.8–9.9 kDa (Table 2), and the polydispersity indexis 3.6–4.5, with lignin in softwood BLs generally having ahigher Mw than lignin in hardwood BLs.[1, 4, 38, 39]
In addition to lignin, the other main organic com-pounds in BL are hemicelluloses (including their broken-down products in the form of carboxylic acids) and woodextractives. The kinds and amounts of hemicellulosesdiffer between different wood species. For example, themain hemicellulose in hardwood is glucuronoxylan, whileit is galactoglucomannan in softwoods.[23] Although theMw of hemicelluloses vary in BL because of variedamounts of degradation based on different cooking con-ditions and biomass species, hemicelluloses also tend toentangle, taking up more volume/molecule than lignin,generallymaking them easier to remove during filtration.-[1, 51] Table 2 shows some typical ranges of hemicelluloseproperties in Kraft BL. Hardwood xylans have a greaterresistance to degradation than their softwood counter-parts; hence, softwood hemicelluloses tend to be moredegraded and thus have lower Mw.
[59] Extractives are themuch lower Mw compounds that come from the wooditself and include resins, fats, waxes, oils, proteins, ter-penes, and other small organic species.[19] Extractivescould be refined to make biodiesel and hemicelluloses(including the degraded products as carboxylic acids)could be used in hydrogels and as a paper additiveamong other applications.[9,52,60,61]
During the concentration (up to 30–40 wt% TS at ~80°C–95°C) and chemical recovery of BL withmembranes, BLbehaves as a Newtonian fluid, although it can be shearthinning at low enough temperatures and high enoughTS concentrations.[2, 20, 62] Figure 2 shows how typical BLviscosity can vary with shear rate, temperature, and total
solids. As the BL TS concentration increases, BL viscosityalso increases significantly. Higher BL concentrations alsoresult in greater osmotic pressure, which lowers the fluxdrastically at a given transmembrane pressure differential(TMP).[3] Nordin et al.[46] showed that beyond a TS con-centration of 31–35 wt%, the pump energy costs substan-tially increase (e.g., by 400% from 35 to 44 wt% at velocityof 2 m/s) and the permeate flux significantly decreases (e.g.,by 86% from 31 to 44 wt%) at 2 m/s) (Fig. 3).[46] Because ofthe above factors, membrane concentration of BL is prac-tically limited to about 30–40 wt% TS, with further con-centration requiring traditional reboilers.
Membranes for BL treatment: Generalcharacteristics
The full range of membrane processes has been investi-gated for BL treatment: microfiltration (MF),[4,33,63–65]
ultrafiltration (UF),[9,31,33,35,38,41,44,45,54,66] nanofiltration(NF),[3,9,31,45,54], and reverse osmosis (RO),[67] with UFstudied the most. Table 3 summarizes the approximaterange of membrane molecular weight cutoff (MWCO)values, pore sizes, and transmembrane pressures (TMPs)that can be used for BL treatment.[58,68–73] MF membranesinvestigated for BL treatment are typically symmetric andhave a single layer.[74] This is in contrast toUF, NF, and ROmembranes, which are asymmetric and consist of a thin,active membrane layer on top of a thicker and highlyporous support, which provides mechanical strength andstability.[74] The different types of membrane pores can befunctionally characterized by their MWCO values, whichcorrespond to the Mw of a test species that has a rejectioncoefficient of about 0.90.[34, 74, 75] RO would reject all
inorganic and organic species, NF is capable of almostcomplete rejection of lignin and hemicellulose (non-car-boxylic acids), while UF (and MF) would allow only thelargest lignin and hemicelluloses to be rejected while per-meating the smaller organics and inorganics. In BL
concentration studies, membranes have been operatedeither in dead-end mode (wherein the direction of fluid-phase flow is perpendicular to the membrane surface) or incross-flow/tangential-flow mode (in which the direction offluid flow is parallel to the membrane surface), as shown inFig. 4.[74] More specifically, BL concentration membraneshave been operated in stirred-cell,[25, 33, 76–83] unstirred-cell,[18] or rotated-cell modules,[81, 82] whereas larger-scaletests have been carried out in flat-sheet,[3, 25, 26, 31, 33, 60, 84]
Figure 4. Membrane permeation setups commonly used in the literature for concentration of black liquor using (a) cross-flow(tubular shown here) or (b) dead-end operational modes (adapted from Refs.[34, 49]).
SEPARATION SCIENCE AND TECHNOLOGY 1073
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
While most of the BL membrane literature focuses onpressure-driven separation of high-MW components, onemay also consider electrodialysis (ED) membranes todirectly recover Na+ ions (in the form of NaOH) from aBL stream while simultaneously removing a portion of thelignin. Generally speaking, ED can remove up to 40%–60%of the dissolved ions.[88] The amount of energy used isdirectly proportional to how many ions pass through themembrane.[73] In the traditional ED process (Fig. 5a), onlyions (and not water) pass through the membrane. Thepolymericmembranes are either anion permeable or cationpermeable. At the cathode, reduction occurs and water isconverted into OH− ions andH2 gas. At the anode, H
+ ionsand O2 gas are produced. ED systems are typically fabri-cated as flat-sheet membranes in plate-and-framemodules.[69] The configuration in Fig. 5b is a recent exam-ple showing the removal of sodium and lignin from BL.Na+ ions from BL pass through the cation-selective mem-brane to react with OH− ions produced from the waterreduction on the cathode to achieve 43%–51% sodiumrecovery.[10] The reduction in pH due to NaOH removalalso results in precipitation of lignin. Salt solutions (NaOHand Na2SO4) were added to the cathode side to reducevoltage drop, thereby increasing efficiency.[89] While thegeneration of large amounts of O2 and H2 gases is likelyundesirable from operational and safety perspectives, theH2 could be recovered and sold to help offset the costs ofthis process.[90]
More advanced ED techniques can be used, such asthose involving bipolar membranes (Fig. 5c). Anion-permeable and cation-permeable membrane pairs are stillused, but the two types of polymeric membranes are nowlaminated together. In between the two membranes, thereis an intermediate/junction layer (J-layer) where water issplit into H+ and OH−without the generation of any gases.Unlike conventional ED, the presence of water in theJ-layer is due to the diffusion of water through the mem-branes in addition to ion permeation. In conventional ED,the generation of H2 and O2 gases uses about half of theelectrical energy used for the process. Thus, ED with bipo-lar membranes can be much less energy intensive. Thetheoretical amount of energy consumed to maintain thevoltage drop near the water splitting voltage of 0.83 Visabout 600–700 kWh per ton of NaOH produced.[91]
BL membrane flux and rejection: Relevanttheory
In this section, we summarize the currently acceptedtheoretical description and experimental understandingof the key performance parameters of pressure-drivenBL concentration membranes: flux, rejection
(selectivity), and fouling behavior (including concentra-tion polarization effects). The TMP is the driving forcethat pushes the permeate through the membrane pores.The osmotic differential pressure, caused by the pre-sence of different solute concentrations on the feed andpermeate sides, opposes permeation. As a result, theflux of the solvent can be written in the well-knownform of Eq. (2), where J is the solvent flux, �P is themembrane permeance (permeability divided by themembrane thickness), Π is the osmotic pressure, (ΔP-σΔΠ) is the net driving force, and σ is the osmotic
reflection coefficient.[69] A σ value of unity impliescomplete solute rejection, while a σ value of zeroimplies that all of the solute passes through themembrane.[92] The osmotic pressure is directly relatedto the solute concentration. Various correlations existin the literature, the simplest of which is the linearproportionality of osmotic pressure and solute concen-tration obtained from the van’t Hoff equation andwhich is valid for dilute solutions.[3] In more concen-trated solutions, the osmotic pressure is often corre-lated to power-law or polynomial functions of thesolute concentrations in which the constant parametersare fitted experimentally.[92,93] The osmotic reflectioncoefficient couples the solvent transport to the rejectionof the solutes.
J ¼ �P ðΔP � σΔ�Þ (2)
Because MF and UF are driven by bulk flow, it isalso common to express Eq. (2) as Eq. (3), wherepermeance is rewritten to express the effects of solutionviscosity (µ) and the membrane’s inherent resistance toflow, Rm. Since σ is ~0 due to the large pore sizes in MFand UF, the osmotic contribution can be neglected.Additional resistances can be added to quantify theeffects of factors that decrease steady-state flux (e.g.,concentration polarization, Rcp).
[68, 92]
J ¼ ΔP � σΔ�
μRm(3)
The BL membrane literature usually reports soluterejection values/percentages or solute retentions as ameasure of the membrane selectivity toward (oragainst) that solute. The “observed (or apparent)rejection” (also called “retention” in some works) isdefined as Robs = 1-Cp/Cb, where Cp and Cb denotethe permeate and bulk feed concentrations,respectively.[3, 34, 76, 77, 94] Some authors also usethe retentate concentration in place of bulk concen-tration, particularly when reporting rejections inexperiments where the retentate concentration signif-icantly changes (e.g., dead-end cells).[33, 74, 94, 95] The“real” or “intrinsic” rejection is Rint = 1-Cp/Cm, whereCm denotes the actual concentration of the solutenear the membrane-feed interface. For solutes with ahigh osmotic reflection coefficient, Rint is greater thanRobs because the solute concentration at the mem-brane surface is greater than in the bulk feed due toconcentration polarization.[77] A useful relationshipbetween the two rejections, the volumetric flux ofthe permeate through the membrane (Jv), and thesolute mass-transfer coefficient, k, can be derivedthat accounts for the backdiffusion of a polarizedsolute into the bulk feed as shown in Eq. (4). A plot
of the right side versus the volumetric flux gives alinear relationship, allowing Rint and k to bedetermined.[75, 76, 83]
lnð1� Robs
RobsÞ ¼ lnð1� Rint
RintÞ þ Jv
k(4)
Bhattacharjee et al.[83] examined models based uponirreversible (non-equilibrium) thermodynamics[96–98] forpredicting solvent and solute fluxes as well as the soluterejection in BL concentration for UF membranes.[83, 99]
The Spiegler-Kedem (SK) model (combined with filmtheory to also solve for t k) shown in Eq. (5) relates theobserved solute rejection to the mass transfer coefficientof TDS or TS, solvent volumetric flux, solute permeance(�Ps), and σ:[83]
Robs
1� Robs¼ σ
1� σ½1� exp � Jvð1� σÞ
Pm
� ��
� expð�JvkÞ (5)
The study examined how well the widely used SKand Kedem–Katchalsky (KK) models matched theexperimental data pertaining to: (1) flux versustime (and steady-state flux) at different TMPs, (2)flux versus time at different stirring speeds (i.e.,different Reynolds numbers), and (3) Cp of TS orTDS as a function of TMP for different stirringspeeds. Figure 6a shows the flux decline over timeusing the same feed concentration and TMPs but attwo stirring speeds. The SK model fitted the experi-mental data reasonably well, while the KK modelsignificantly under predicted the flux. Similarly, inFig. 6b, flux over time is plotted at different TMPs.Again, the SK model outperformed the KK modelwith closer agreement to experimental data. The SKmodel was also better at predicting the permeate TSconcentration based on the TMP. The poorer per-formance of the KK model is likely due to its coef-ficients being dependent on concentration itself,while the SK model’s coefficients are not dependenton concentration.[100] This independence wouldallow better predictions of flux and other properties,which are influenced by concentration.
Membrane materials for BL concentration
Figures 7 and 8 illustrate the important milestones inthe history of development or evaluation of membranematerials for BL concentration. Primarily polymericmembranes and ceramic (metal oxide) membraneshave been investigated. Tables S1 and S2 (SupportingInformation) give an overview of the main types ofpolymeric and ceramic membranes, respectively, that
SEPARATION SCIENCE AND TECHNOLOGY 1075
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
have been investigated for MF, UF, and NF, and ROconcentration of BL. Some of the key developments arediscussed in more detail in this section.
The first important usage of membranes for BLconcentration appears to have been made by DDSCorporation in Norway in the 1970s, although sulfide(not Kraft) BL was used. This early work included the
application of cellulose acetate RO membranes to con-centrate BL from 6 to 12 wt% TS. This trial was fol-lowed by another usage of DDS cellulose acetatemembranes in 1978 in Quebec, Canada, which achievedconcentrations of TS from 12 to 18 wt%. The celluloseacetate RO membranes used in both of the above plantshad lifetimes of about 1 year. Additionally, these
Figure 7. Milestones in the development or use of different membrane materials for black liquor concentration.
1076 N. S. KEVLICH ET AL.
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
membranes required cleaning 2–6 times per weekbecause the TS content of BL clogged the pores andfouled the membranes.[63, 67] About a decade later, Rosset al.[18] studied UF concentration of Kraft BL with 10-and 20-kDa MWCO polysulfone membranes.[18]
However, these membranes displayed very low rejec-tion of lignin at industrially realistic pHs of 11.3 and12.5, likely due to membrane degradation.[18] Also, in1985, Jӧnsson et al.[67] reviewed the applications ofmembranes in the paper industry and found thatwhile UF could concentrate Kraft BL from 17 to 23wt% BL, the heating value of the retenate and perme-ates were lower than those of the feed BL, making theconcentration not worthwhile unless there was a sig-nificant market for lignin. Also, an RO laboratory andpilot-plant study was performed on 1–5 wt% BL, butthe flux fluctuated widely as feed temperature and pHvaried. The RO membrane (DDS HR 95, material notreported) used was stable for at least 2 months.[67]
Cellulose acetate and polysulfone-based membraneswere initially used, likely due to their commercial avail-ability. However, while some of these membranes couldprovide fairly high rejections (e.g., up to 75% rejectionof lignin, though at low pH of 10.2),[18] their majordrawback has been long-term stability, likely fromdegradation in the high pH environment. In recentyears, polyethersulfone (PES) membranes have gained
prominence because of their higher stability, yet thechallenge is still to achieve high rejections of organicsand ideally even inorganics. Arkell et al. showed that at70°C, the latest industrial membranes could achieve90%–97% lignin rejection, with the 90% rejection mem-brane having the best performance due to its higherflux.[9] Others have proposed combining carbonation(to precipitate lignin) with UF and/or NF to achievehigher rejections.[3, 85] Nevertheless, it is expected thatsalt rejections are lower than those of organics becauseof the smaller Mw and size of the inorganic ions. Thereis a need for stable membranes that can perform RO toseparate not only the organics, but also the inorganicspecies from water in BL.
A general problem with many polymeric membranesused to date is their short (<1.5 years) lifetime andrequirement for frequent cleaning, thereby leading topotentially higher capital and maintenance costs.[63] Toaddress these issues, there has been considerable focus inrecent years on inorganic (specifically metal oxide)membranes that are more robust and resistant to thehigh pH and temperature of BL feeds.[63] Ceramic mem-branes, such as the commercially available Kerasepmembranes (made from Al2O3–TiO2) have a generallifetime of about 6 years in caustic solutions.[14, 41]
Ceramic membranes also tend to have higher mechan-ical strength and have been operated successfully at feed
Figure 8. Milestones in the development of membrane processes for black liquor concentration.
SEPARATION SCIENCE AND TECHNOLOGY 1077
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
pressures up to 10,000 kPa.[63] Typically, these commer-cially available membranes are Al2O3, TiO2, or ZrO2
membrane layers on top of ceramic supports. All ofthese membranes, however, have pore sizes that are toolarge for NF or RO applications. For example, aluminumoxide membranes synthesized using anodic oxidationcan achieve a minimum pore size of ~10 nm, which isnot sufficient for good NF or any RO separation.[101]
Although ceramics are generally considered as bettermembrane materials than polymer membranes for BLconcentration because of their thermal and chemicalstability, the membrane cost must also be considered.Ceramic membranes are typically more expensive thanpolymeric membranes. As an example, Arkell et al.[9]
applied their laboratory-scale results of lignin rejectionby both ceramic and polymeric NF membranes toestimate the economics of an industrial-scale process.[9]
The cost of polymeric NF membranes used in theirstudy was only about 10% of the ceramic NF membranecost. On the other hand, the ceramic membrane had arated lifetime of 6 years compared to the 1.5 years lifeof the polymeric membrane. Although the ceramicmembrane was more stable, calculations showed thatthe higher lignin rejection of the polymeric NF and itslow purchase cost resulted in it being the most eco-nomical BL concentration choice. In comparison withusing just the NF polymeric membrane, the ceramic NFmembrane was 48% more expensive, and the combina-tion of a ceramic UF and the polymeric NF was 161%more expensive.[9] These estimates indicate the need fornon-polymeric membranes with lower manufacturingcosts and better lignin rejection if they are to be com-petitive with current polymeric membranes.
As mentioned earlier, ED membranes may havepotential due to their ability to recover the Na+ ionsin the form of NaOH. Both conventional and bipolarED membranes must have both acid and base chemicalstability since they are exposed not only to BL but alsothe concentrated base/acid solutions that result fromion transfer through the membranes as well as anysweep solutions used to minimize the applied voltage.In ED using bipolar polymeric membranes, the acid orbase concentrations can be as high as 4 M. Similar tothe case of NF membranes, many ED membranes arestable in acids but have stability issues at high pH.Membrane swelling and higher temperatures also poseproblems because the ion-selective membrane func-tional groups can be lost and the membrane may ther-mally expand.[102] Examples of “stable” membranematerials that have been used for ED at high pHsinclude polysulfone (PSF) and PES and thus could beuseful in BL applications. Also, the commercially avail-able Nafion (e.g., N117,[10] 324[89]) (from Dupont,
USA) and CMB[103] (Neosepta, Japan) membraneshave been used in ED studies for BL concentrationand are generally stable at high pHs. However, thecapital cost of such membranes could be prohibitivelyexpensive at the large scales required for BL processing.The energy costs of ED-based processes vis-à-vis pres-sure-driven membranes are yet to be compared indetail. Unlike the case of NF membranes, there doesnot currently exist a detailed technoeconomic analysisfor the use of ED membranes, or even other types ofmembranes such as forward osmosis (FO) membranes,for BL processing applications. A technoeconomic ana-lysis of this nature, if found to yield promising results,could increase the available process options andbroaden the challenges to be overcome in membranedevelopment.
Separation of BL components
As mentioned previously, the large range of membraneseparations (MF, UF, NF, and RO) have been investi-gated for concentration of BL with both polymeric andceramic membranes, as detailed in the previous sectionand in Tables S1 and S2 (Supporting Information).Ceramic-based membranes have pore sizes in the MF/UF range and at best, the high molecular weight end ofNF. Consequently, polymeric membranes are still usedfor NF, sometimes in combination with ceramic UF.Future research is likely to move toward introductionof non-polymeric NF and even RO membranes thatcould offer the advantages of stability and efficientoperation at every molecular weight range. RO wouldbe particularly beneficial since it would allow the fullcapabilities of separation of BL components. In thepublished literature, NF is the primary membranetype currently being considered for lignin rejection.The development of better NF and RO membraneswould permit not only the almost complete rejectionof lignin (and all other organics), but would also allowthe separation of free salts, further reducing the energyload and scale formation issues in the downstreamevaporators and boilers. Below, we discuss some ofthe key advances in the literature regarding the separa-tion of each of the major components in BL.
Concentration of lignin
The concentration and removal of lignin from the WBLfeed stream (and thus the TS concentration of WBL toSBL) are the most common objectives in the BL separa-tion literature. This involves the use of UF or NFmembranes to transport water and other low-Mw BLcomponents into the permeate stream, thus producing
1078 N. S. KEVLICH ET AL.
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
a retentate stream concentrated in lignin. Selected lit-erature results representative of the main findings arediscussed below. Liu et al.[63] investigated the effects ofmembrane pore size on lignin rejection. They usedthree different tubular α-alumina (Al2O3) membraneswith pore sizes of 50 nm, 0.2 μm, and 0.8 μm (i.e., UFand MF range).[63] These membranes achieved similarlignin rejections of about 75% (likely due to the feedpH of 11, resulting in larger/aggregated lignin mole-cules that could be rejected by even the largest poresizes studied) while also maintaining high long-termfluxes. Interestingly, the flux of the 0.2-μm membranewas the highest (0.2–0.3 m3 h−1 m−2) for most of theconcentration range), while the 50-nm and 0.8-μmmembranes had similar lower fluxes. The reason forthis is not known. Also, the 0.2-μm membrane had highlong-term fluxes, not experiencing major fouling for upto 35 days.[63] Arkell et al.[9] examined the flux, ligninretention, and overall cost of removing lignin fromsoftwood kraft BL.[9] Specifically, they were interestedto examine the results of a UF and NF combinationcompared to a NF only. The UF membrane was cera-mic (Al2O3 with a surface layer of TiO2), while oneceramic (TiO2) and three polymeric composite mem-branes were tested for NF. At the same MWCO of 1kDa, the polymeric NF membrane outperformed theceramic membrane, achieving lignin retention of 90%compared to 80% with the ceramic membrane (with noUF pretreatment).[9] This was because while the man-ufacturer’s MWCO ratings were the same, a literaturestudy showed that the NF’s pore size was actually some-what larger than the polymeric membrane.[9] Thisambiguity in the MWCO has been observed before.Causserand et al.[75] showed that the observed rejection(used to determine the MWCO) can vary strongly withthe operating conditions (which are rarely specified).
For example, the observed rejection decreased from90% to 40% for a 20-kDa PEG as the flux increased(from increasing the TMP).[75] Interestingly, adding anUF prefiltration step decreased the lignin rejection ofthe ceramic NF membrane from 80% to 56%, whilelittle change was observed with the polymeric mem-brane of similar MWCO. Again, this was likely becauseof the larger pore size (larger molecules were retainedby the UF but passed through the ceramic NF moreeasily than the smaller pore-sized polymeric NF).However, the permeate flux was always higher in theNF membranes after a UF pretreatment, likely from thereduction in fouling species and osmotic pressure.[9]
Several authors have reported the interesting obser-vation of a non-monotonic behavior of lignin rejectionas a function of membrane MWCO in MF and UF. Forexample, the results of Hill et al.[38] on polymericmembranes are shown in Fig. 9a.[38] The rejectionvalue passes through a minimum as a function ofMWCO, with a 500-kDa membrane having almost ashigh rejection coefficient as a 20-kDa membrane. It isalso surprising that membranes with very highMWCOs have any lignin rejection at all. This trend isalso reported in other works and has been attributed togel layer formation.[34, 38, 45] Lignin is hypothesized toform a relatively dense, dynamic gel layer either at themembrane surface or within the large pores of high-MWCO membranes, resulting in a lower effectiveMWCO and an unexpectedly high rejection.[77] Forlower-end UF and NF, the trend is more clear, asshown in Fig. 9b, which plots data from Tables S1and S1 for different types of polymeric and ceramicmembranes under known or likely conditions of high(unchanged) pH (if pH has been artificially lowered,lignin tends to get larger and thus will likely showhigher rejection values than normal). The pore sizes
Figure 9. (a) Non-monotonic dependence of lignin rejection upon UF membrane MWCO (adapted from Ref.[38]) and (b) non-monotonic dependence of lignin rejection of NF membranes (plotted using the data in Tables S1 and S2) upon the estimated poresizes, based on MWCOs and a literature correlation.[80]
SEPARATION SCIENCE AND TECHNOLOGY 1079
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
of the membranes were not reported in these studies,but were estimated using a literature correlation andthe reported MWCOs.[80] Lignin rejection sharplyincreased at a critical pore size of ~3.5 nm (~15-kDaMWCO) to achieve a maximum lignin rejection of~60% and again at ~1.3 nm (~1-kDa MWCO) toachieve >64% rejection. It is also worth mentioninghere that hardwood hemicellulose was easily retainedin most UF and NF membranes due to its generallyhigher Mw than lignin. For example, Jӧnsson et al.[45]
reported greater than 80% retention of hemicelluloseeven with a 100-kDa membrane, whereas a ligninretention of 61% was observed with a 4-kDamembrane.[45] Overall, Arkell et al.[9] reported thehighest lignin concentrations (from 74 to 252 kg/m3
with their 1-kDa ceramic membrane and from 63 to282 kg/m3 with their 1-kDa polymeric membrane).[9]
Unfortunately, no data on the TS of the retentate weregiven. Otherwise, the highest published TS concentra-tions are from Jӧnsson et. al., who concentrated BLfrom 16 to 29 wt% TS.[45]
Separation of valuable components and value-added products
As discussed earlier, BL contains a number of organicspecies. The main organic species are lignin and hemi-cellulose/carboxylic acids. Due to the lack of appropriateseparation technology, most of the valuable BL compo-nents are not industrially usable at the moment. Ligninhas several direct uses and potential uses for higher value-added products.[104] Concentrated lignin can be gasifiedto H2, CH4, and CO for use as energy-efficient and low-
carbon fuels.[10] Lignin is non-toxic and has FDAapproval for usage in food and packaging.[104] The biggestusage of lignin is in commodity markets while specialtymarkets have less market share because they require addi-tional processing or modification of lignin.[104] Actualuses in commodity markets include as cement/concreteadditives (~50% of what little lignin is diverted from therecovery boiler goes to this use alone), binders, animalfeeds, and viscosity reducers in molasses and oil welldrilling muds.[18, 27, 33, 52, 58, 104, 105] Specialty marketuses have included production of vanillin, pesticides, oilwell cement retarders, gypsum board, dispersants, carbonblack, inks, industrial cleaners, micronutrients, and leadacid batteries.[14, 18, 27, 33, 34, 38, 52, 54, 58, 104]
Hemicelluloses and their degraded products as car-boxylic acids (Mw as small as 100 Da) are the secondmost abundant organic component in BL solids. The heat-ing value of carboxylic acids is lower compared to lignin;thus, greater benefit could potentially be derived if thecarboxylic acids were instead used to make higher-valueproducts such as biodegradable polymers.[60, 66] Due to thediversity of carboxylic acids present in black liquor, theirwide range of Mws, and lack of effective separation andpurification processes, little work has been done to recoverthem from BL. Two recent articles have proposed using UFfollowed by other unit operations (Fig. 10) not only toremove and purify lignin but also to recover the valuablecarboxylic acids such as hydroxy acids.[60, 66] To accom-plish the recovery of hydroxy acids from soda BL, Hellstѐnet al.[60] demonstrated a separation process combining a 1-kDa MWCO PES UF membrane, size exclusion chroma-tography, ion exchange, adsorption, and evaporation(Fig. 10a).[60] UF removed up to 75% of the lignin, with
the remainder being removed by an adsorption column.These two unit operations combined resulted in 99%removal of lignin. Alkalis in the UF permeate wereremoved by size exclusion chromatography and ionexchange. In the final step, evaporation was used to sepa-rate volatile acids and water from hydroxy acids.Specifically, hydroxy acid purities of up to 81 wt% andyields of up to 91% could be obtained. The final separationstep could potentially be carried out using a hydrophilic NFmembrane that permeates water to lower energy costs byreducing the amount of water that needs to be evaporated.
Manttari et al.[66] first filtered BL with an UF mem-brane (1.2-kDa MWCO, 75% lignin rejection) and thenused acid precipitation (to remove the rest of the lignin),followed by cooling crystallization, NF filtration (0.15–0.3-kDa MWCO) and adsorption (with vacuum filtrationin between steps) to recover and purify hydroxy acids(Fig. 10b).[66] In this configuration, the UF stage separatedthe much smaller hydroxy acids from lignin, concentrat-ing the hydroxy acids in the permeate (since charge effectsfavored hydroxy acid permeation). However, it affecteddownstream processes by reducing flux, lignin recovery,and hydroxy acid purity. A NF membrane downstreamcan be used to separate the hydroxy acids from theremaining inorganic species and water. In addition tosize of solutes, charge effects are very important in themembrane filtration steps. At the high pH of the UF stage,lignin and hydroxy acids have negative charges. As thenegatively charged lignin was concentrated, the signifi-cantly smaller hydroxy acids permeated faster, resulting inpermeate concentrations 1.4x that of feed black liquor tomaintain feed electroneutrality by the Donnan exclusionprinciple. However, the NF membrane feed was in acidicconditions wherein the hydroxy acids were protonated,and thus, size alone achieved their separation and con-centration from inorganics and water. Without the UF,the hydroxy acid purity increased from 21% in BL to 80%in the NF permeate.[66]
The above discussion also suggests that purificationof BL components will benefit from the development ofmore stable UF, NF, and molecular sieving membraneswith tunable pore sizes. A multistage membrane pro-cess could be envisioned wherein the first one or twostages would separate the lignin and carboxylic acids(without the cost of acidifying chemicals) yet maintain-ing high fluxes. The final NF (or molecular sieving)membrane stage would concentrate the carboxylicacids by permeating water and inorganics. Ideally,these membranes should be negatively charged tofavor lignin retention and the Donnan exclusion prin-ciple would still apply to maintain feed electroneutral-ity, thereby first helping carboxylic acids permeate inthe initial two stages but then favoring inorganics
permeation in the final NF stage to purify and concen-trate carboxylic acids.
Under certain conditions (e.g. in hardwood pulping,as mentioned in section on “Composition andPhysicochemical Properties of Black Liquor”), somehemicelluloses (like xylan) can remain relatively largeand are not significantly degraded into the small car-boxylic acids. Xylan is a dominant hemicellulose inhardwood (Mw 6–19 kDa, ~0.1–1.3 wt% in BL), whichcan be used as a feedstock for production of building-block chemicals like furfural, as a colorant, and as aviscosity reducer in drilling fluids.[52, 106] In 2013, theworld market for furfural was approximately 300,000tons per year and is forecast to reach about 650,000tons by 2020.[107, 108] Lake proposed the use of acidprecipitation in conjunction with membranes and otherunit operations to purify the xylan component in BLfor use in furfural synthesis (Fig. 11).[85] The processbegins by conventional acid precipitation of ligninusing CO2, until the pH is reduced to 8.5–9.5. Thiswould remove most of the lignin, which would beretained along with the xylan hemicellulose in mem-brane separation processes. Furthermore, the reducedpH allows greater flexibility in the choice of membranematerials, since many membrane materials are quitestable at pH 8.5–9.5 but unstable at the high pH(~12) of raw BL. However, as mentioned previously,acidification may have high chemical costs and othersupporting costs (e.g., piping and storing pressurizedCO2).
[19] The next step is a tubular UF membrane(MWCO 1.5–2 kDa) process to concentrate the xylanhemicellulose as retentate while allowing most of thewater, salts, and low-Mw organics (including any exten-sively degraded hemicelluloses as carboxylic acids) topermeate. A spiral-wound NF membrane (MWCO0.15–0.5 kDa) is then used to purify the xylan-richretentate by completely rejecting xylan while permeat-ing water, salts, and the lowest Mw organics. It is notedthat while the NF membrane alone could achieve thedesired xylan purity and concentration without UFpretreatment, this may be difficult in practice due tolow fluxes and severe fouling. UF pretreatment helps toeliminate most of the undesired fouling components aswell as decrease the volume of liquid for processing bythe NF unit operation. The concentrated and high-purity hemicellulose solution is then sent to a catalyticreactor wherein xylan is converted to furfural. A finalstep involves production of pure furfural by its separa-tion from water and xylan. This can be performed bydistillation or solvent extraction, but could also poten-tially be carried out by nanoporous membranes. BL alsocontains a number of other low-Mw (on order of 0.1kDa) dissolved organic compounds that appear in the
SEPARATION SCIENCE AND TECHNOLOGY 1081
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
permeate streams of UF or NF membranes. These low-Mw organics include formic and acetic acids, andextractives such as tall oil and resins.[49, 85, 109] Newmembrane technology for separation of dissolvedlower-Mw BL components from each other may allowincreased usage of these BL components as feedstocksin a biorefinery scenario (e.g., for the production ofbiofuels).
Separation of BL inorganics from organics
Conventionally, the inorganic salts present in BL(Table 1) are recovered as smelt after combustion ofthe SBL stream exiting the evaporators (Fig. 1).[44] Themain monovalent ions are primarily sodium (Na+),with potassium (K+) present in smaller amounts.Current studies on BL concentration membranesfocus on UF or NF to concentrate lignin and hemicel-luloses while permeating water and ions. An energy-efficient membrane process for BL concentration andinorganics recycling would include a downstream ROsystem that permeates water and produces concen-trated alkaline brine containing the inorganic ions anddissolved low-Mw organic solutes. In addition to theusual requirements of high salt rejection, good waterflux, and fouling resistance, these membranes should befabricated from materials capable of withstanding thehigh pH of the UF/NF permeate stream. This require-ment is beyond the capabilities of current polymericRO membranes, and there are significant opportunitiesand challenges in the development of such membranes.The monovalent ions tend to be free in the bulk
solution, while the multivalent ions tend to be boundto lignin or other colloids and hence are found in theretentate of UF membranes.[3, 4, 12, 42] The total con-centration of monovalent ions is in the range of 3.1–42.3 kg/m3. The main multivalent inorganic ions thathave been typically retained are Mg, Mn, Fe, and Ca,with Mg and Mn typically having higher retentions(70%–90%[44]) and Fe and Ca having lower retentions(40%–60%[44]).[27, 36, 41–44] These inorganics make uponly a small portion of the TS in BL. Typical rejectionsfor multivalent ions have been 80%–100% for Mg,45%–85% for Mn, 40%–71% for Fe, and 40%–81% forCa, while the monovalent inorganics had rejections inthe range of 6%–19% since they are permeated by UF/NF membranes.[27, 42, 44]
We highlight here the work of De et al.[3] whoused cellulose acetate membranes (1 kDa for UF and0.5-kDa MWCO for NF) to explore three differentmembrane separation strategies to recover inorganicsfrom BL.[3] Scheme A involved carbonation followedby UF and NF, Scheme B involved carbonation fol-lowed by NF, and Scheme C relied solely on NF. InScheme A, carbonation lowered the pH of BL to 7.5,which precipitated the now protonated lignin,thereby releasing salt ions that were formerly asso-ciated with the negatively charged lignin. The UF andNF stages would then have permeated mostly waterand inorganics. Scheme B used only a NF membraneafter the carbonation step, while in Scheme C, asingle NF stage directly produced a permeate con-taining mostly inorganic ions and water from raw BLwithout carbonation pretreatment.
Figure 11. Process of combining acid precipitation of lignin with membrane (UF and NF) separations to produce a feedstock forsubsequent furfural production. (adapted from Ref.[85]).
1082 N. S. KEVLICH ET AL.
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
Figure 12 shows the TDS rejection, inorganic recov-ery, and permeate flux in each scheme. The reductionof pH by carbonation should have precipitated most ofthe organic species, thus leading to high TDS rejectionsin Schemes A and B. Scheme C (with NF only) has alower TDS rejection, likely due to the permeation ofsmall organic fragments that carbonation successfullyprecipitated. Schemes A and B both showed goodrecovery of inorganics due to the carbonation stepwhich released bound cations from lignin, thus allow-ing for their recovery in the NF permeate stream whileScheme C showed much lower inorganics recoverysince a large fraction of alkali cations remained bound
to lignin in the absence of carbonation. The higherpermeate flux achieved in Scheme A over Scheme Bwas likely because the UF membrane removed thehigher-Mw TDS components and mitigated fouling/pore blockage in the NF membrane. It is interestingto note that Scheme C had the highest permeate flux,with the NF membrane operating at the same pressuredifferential in all three schemes. This is likely due to thehigher concentrations of feed inorganic salts inSchemes A and B, which would increase the osmoticpressure and thus lower the flux. The increased saltconcentrations are likely from both the addition ofCO2 from carbonation and the release of inorganic
ions from the lignin when the pH decreased. It shouldbe noted that De et al.[3] did not discuss the economicsof carbonation versus membrane permeation. The costof supplying enough CO2 to achieve the complete pHreduction on industrial-scale BL feeds is likely highercompared to using membranes (see section on“Economics and Energy Requirements”).
Water separation
The forest products sector is water intensive.[110] Itwould thus be desirable to separate water from BLand recycle it, while recovering the inorganic compo-nents as concentrated brine that would place a lowerload on the inorganics recycling boiler. To ourknowledge, there are no works exploring the produc-tion of “pure” water from BL using membranes. Thedirect production of water from raw BL appears pro-hibitively challenging for two main reasons. First, theosmotic pressure of WBL (~15 wt% TS) is about 7000kPa as estimated from the correlation for BL based onTS.[69] This is a high osmotic pressure compared tothat encountered in the desalination of seawater(about 2500 kPa[111]) and would likely require signif-icant energy costs for pressurization of the BL feed.Second, the presence of large quantities of organicsolids (such as lignin and hemicellulose) and dis-solved organic molecules in raw BL will certainlylead to high concentration polarization and foulingeffects in RO membranes. Thus, it is more reasonableto consider RO membranes for separation of waterfrom the permeate stream of an efficient NF pretreat-ment membrane. Therefore, the industrial use of ROmembranes for producing water for recycle in BLapplications is likely to be contingent on the devel-opment and acceptance of NF membranes in theKraft process. Assuming that the NF membrane canachieve near-complete retention of lignin and most ofthe other higher-Mw organics while allowing permea-tion of nearly all monovalent inorganics and some ofthe other low-Mw organics, the TS content of the NFpermeate would likely be lowered to <8 wt% (madeup mostly of inorganics and lower-Mw organics),which corresponds to an osmotic pressure of ≤3800kPa. The pH of this stream would likely be close tothe feed pH. This represents a feasible opportunity forthe development of inorganic/non-polymeric ROmembranes that can produce water suitable forrecycle while concentrating the inorganics. It is alsonoted that the salt rejection requirements for suchmembranes could be somewhat lower than those ofdesalination membranes, since this application wouldnot require the RO membrane to produce a water
stream of potable quality. Also, there is potential thatforward-osmosis membranes, which use a high-osmo-tic pressure draw solution on the permeate side toremove water from the feed by the osmotic pressuredriving force, could be used to help lower the energycosts of inorganic concentration.[112, 113]
Electrochemical membrane techniques such as EDoffer another interesting method to produce water froma BL stream, by simultaneously removing the inorganicions as well as precipitating the lignin. This could beachieved using bipolar ED membranes as illustrated inFig. 5c. The BL feed solution passes through the centerand the inorganic cations (Na+) and anions (SO4
2-)permeate in opposite directions through the mem-branes under the applied voltage. However, water splitsin the junction layer, and the generated OH− and H+
ions react with the inorganic ions. Thus, the processyields three product streams: the ion-depleted BLstream, concentrated aqueous acid (e.g., H2SO4), andconcentrated aqueous base (e.g., NaOH). The BL pro-duct will also be depleted of lignin due to precipitationinduced by the pH change and hence yield a muchmore dilute aqueous stream suitable for other mem-brane processes such as NF or RO to produce water.For example, Cloutier et al.[89] showed (using bothsoftwood and hardwood BL feeds) that ~75% of lignincould be precipitated and about 80% of the Na+ saltsconverted to NaOH. Two polymeric ionomer mem-branes (Nafion 324 and R-4010) were tested, with theNafion membrane having superior performancebecause of its lower energy consumption. On theother hand, Kumar et al.[48] applied ED after acidifica-tion. The addition of acid allowed f ~59% of initiallignin precipitation, and the ED operation furtherincreased the lignin removal to 90% along with 90%sodium recovery and 80% of the carboxylic acids).[48]
In a third example, Haddad et al.[114] used bipolarmembranes with 20 wt% BL at pH 12.25 to reducethe lignin concentration from 40% to ~30% TS whileoperating at a somewhat higher temperature of 55°C.[114, 115]
ED membrane techniques may become more feasibleif they are used with other unit operations, particularlyfiltration and/or acidification. One such proposedstrategy[116] involves first rejecting most of lignin byUF (thereby concentrating BL from 18 to 33 wt% TS),then performing ED on the 16 wt% TS permeate fromthe UF membrane (which primarily contains inorgan-ics, carboxylic acids, and residual, low-MW lignin). Thelatter step could simultaneously remove more than 80%of the inorganic ions, precipitate low-MW lignin, andleave carboxylic acids in solution. These carboxylicacids could then be recovered by other processes,
1084 N. S. KEVLICH ET AL.
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
yielding a purified water stream as a by-product. Thisprocess was patented in 1986, at which time the avail-able UF and ED membranes were unreliable. To cir-cumvent the membrane stability issues, a 1991patent[117] proposed to first remove most of the ligninby acidification to pH 9–10, followed by ED to removethe inorganics. However, with ongoing and futureadvances in the stability and performance of UF/NFand ED membranes, such a process may become eco-nomically feasible.
Polarization, fouling and flux decline in BLconcentration membranes
One of the disadvantages and challenges of pressure-driven membranes is the significant flux decline overtime because of concentration polarization and/orfouling.[8] There are several works describing the fluxdecline behavior in BL concentration membranes.[27, 33,63–65, 118] Dafinov et al.[118] reported that the fluxdropped ~ 40% from 0.062 to 0.037 m3 h−1 m−2 fortheir 5-kDa ceramic (TiO2) membrane and ~69% from0.058 to 0.018 m3 h−1 m−2 for their 15-kDa ceramic(ZrO2) membrane at 0.5 MPa TMP, within about 100minutes of operation.[118] This is unexpected, sincemembranes with higher MWCOs usually have higherfluxes. To explain these results, the authors suggestedpore blockage of the membranes by lignin macromole-cules (which had sizes similar to the 15-kDa pores).The flux then remained stable, with experimental datameasured until almost 10 and 24 hrs for the 5- and 15-kDa membranes, respectively. Wallberg et al.[27]
showed that with regular cleaning, ceramic (Al2O3–TiO2) UF membranes maintained flux for over twoweeks.[27] As can be expected, it was found that theflux decline is more pronounced as the retentatebecomes more concentrated. This is generally depictedin plots of flux versus volume reduction, VR ([feedvolume-retentate volume]/feed volume).[14] For exam-ple, in one study, the flux declined from 0.12 to 0.02 m3
h−1 m−2 when VR changed from 0 to 0.9.[44] Liu et al.[33,63] also described similar behavior using ceramic (α-Al2O3) MF and polymeric MF and UF membranes(pore sizes/MWCOs: 0.8, 0.2, and 0.22 μm for MFand 3, 6, 10, 30, and 60 kDa for UF) for lignin removalfrom wheat straw BL, achieving a lignin retention of70-80%.[33, 63] The polymeric MF was made of celluloseacetate, and the polymeric UF membranes were madefrom polyacrylonitrile (PAN), polyaryletherketone(PAEK), or polyethersulfone (PES).[33, 63] Periodiccleaning and the use of high cross-flow velocities (toreduce concentration polarization resistance) enabledhigh fluxes of 0.15–0.24 m3 h−1 m−2 to be maintained
over more than 40 days of operation. For membraneswith MWCO < 10 kDa, resistances from pore pluggingand gel layer formation dominated whereas at MWCO>10 kDa, concentration polarization was dominant.The intrinsic membrane resistance was minor. Whilepore plugging and gel layer formation/surface adsorp-tion probably did occur in both types of membranes,the smaller-pore membranes were likely more affectedby these phenomena because even slight pore pluggingwould have significantly reduced the pore size.
Fouling is also an issue for electrically driven EDmembranes. As pH decreases during ED, the lignintends to aggregate and precipitate on the membranesurfaces and on the electrodes.[114, 119] Lignin is nega-tively charged and particularly fouls the anode[10]. Thiseffect is more pronounced at higher temperatures [89],and therefore, ED cannot operate at realistic BL tem-peratures of ~90°C. As in the case of pressure-drivenmembranes, fouling by lignin deposition could beaddressed by periodic backflushing[89, 103] or electricalpulsing.[114] For example, Cloutier et al.[89] achievedmore than 100 hrs of continuous operation using back-flushing and consumed 7000 kWh of energy per ton ofcaustic (NaOH) produced.[89]
Economics and energy requirements
Because of the significant industrial interest in BL con-centration by pressure-driven membranes, several tech-noeconomic studies on this subject have been publishedsince the late 1980s. Table 4 summarizes the mainfindings of these works. In an early work on membranefiltration of BL, Uloth et al.[15, 16] compared economicfeasibility of lignin removal by acid precipitation andan unspecified UF membrane material.[15, 16] Acid pre-cipitation performed better than UF, being about halfas expensive as UF and having better lignin recovery.UF had at best 54% recovery, while acidification couldremove up to 95% of the lignin in BL at pH 4.Additionally, the capital (operating) costs for UF wereabout 1.7 (2.2) times greater. This is initially surprising,since a large amount of acid must be added to precipi-tate lignin. However, the membrane replacement costand steam usage were found to be much higher thanthe chemical cost of precipitation, hence the greateroverall cost. Thus, membrane filtration of BL was con-sidered to be prohibitively expensive in the late 1980s.
Since then, advances in membrane technology haveimproved the economics of membrane-based BL con-centration due to higher water fluxes and solute rejec-tions. Table 4 shows that the economic potential hasimproved significantly, with the total cost/ton of ligninbeing much lower than even for acidification in most
SEPARATION SCIENCE AND TECHNOLOGY 1085
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
Table4.
Summaryof
techno
econ
omicdata
onmem
brane-basedBL
concentration.
Costvalues
calculated
inthecitedworks
areadjusted
inthisworkto
2016
USdo
llars.
Reference#
[15,
16]
[45]
[9]
Year
1989
2008
2014
Inletsolid
sconcen
tration(wt%
TS/TDS)
32.5
17~17
wt%
)BL
feed
rate
0.00925tons
pulp/s
0.05,0
.056
m3 /s
0.028m
3 /s
Feed
lignin(kg/m
3)
X59
(UF),5
4(NF)
74(ceram
icNF),6
3(polym
ericNF)
Retentatelig
nin(kg/m
3)
X90
(UF),1
65(NF)
252(ceram
icNF),2
82(polym
ericNF)
Mem
bran
ede
tails
UF(unspecifiedmaterial,6-60
kDa)
UF(Al 2O3-TiO2,15
kDa);
NF(Polym
er,1
kDa)
UF(Al 2O3-TiO2,2
0kD
a),N
F(TiO
2,1kD
a),N
F(polym
ercompo
site,1
kDa)
Perm
eate
flux
(m3/m
2h)
0.070,
0.090
0.082,
0.110
0.088,
0.17
Max
ligninrejection(%
)X
52–60max
80(ceram
icNF),9
0(polym
ericNF)
Mem
bran
earea
(m2)
X1200,2
300,
4200
XMem
bran
elifetim
e(yrs)
11.5(polym
eric),6(ceram
ic)
1.5(polym
eric),6(ceram
ic)
Ope
rating
time(hrs/yr)
8000
8000
8000
Capitalcost
(k$/yr)
10,215
(6029acidificatio
n)301,
953,
1742
XOpe
rating
cost
(k$/yr)
2192
(977
acidificatio
n)539,
1379,2
757
3158,3
384,
3948,4
061
Totalcost
(k$/yr)
12,407
(7006acidificatio
n)840,
2331,4
499
XLign
inprod
uced
(ton
/yr)
24,000
72,000,7
8,000,
108,000
8300,3
0,000,
41,000,6
8,000
Lign
inprod
uction
cost
($/ton
lignin)
517(292
acidificatio
n)11,3
0,41
52,7
7,135,
485
1086 N. S. KEVLICH ET AL.
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
cases. Recent studies have compared the usage of UFand NF combinations versus NF or UF alone, as well ascomparisons of polymeric and inorganic membranesfor these processes. Jӧnsson et al.[45] analyzed the eco-nomics of removing lignin from various parts of thedigesting process.[45] A ceramic (Al2O3–TiO2) mem-brane (15-kDa MWCO) was used for UF, and a poly-meric membrane (unspecified material) was used forNF (1-kDa MWCO).[45] Three scenarios were consid-ered: (A) treating BL directly from the digester withUF, (B) same as (A) but adding a NF step, and (C)using the traditional evaporators to concentrate BL to31% TDS and then sending that stream to UF.Although the cost ($41/ton of lignin; adjusted to 2016value) of the UF/NF combination (B) was about thesame as that of using evaporators followed by UF (C),the UF/NF combination resulted in both the greatestlignin concentration and the greatest purity. UF alone(A) at $30/ton (adjusted to 2016 value) of lignin wasthe cheapest, but its low lignin rejection (~8.5%) is notsufficient for industrial use. It is noteworthy that evenwith inefficient UF membranes, the UF/NF combina-tion was a comparable alternative to the evaporators,thus demonstrating the value of replacing some of theevaporation steps with membranes.
Of final note in Table 4 is the 2014 study by Arkellet al.[9], which used laboratory-scale NF membranes withbest lignin rejection reported to date (~90% for a polymericmembrane and ~80% for a ceramic membrane, both withMWCOs of ~1 kDa). The laboratory-scale experimentalresults were used to estimate the economics of an indus-trial-scale process with a feed flow rate of 0.028 m3/s and aretentate lignin concentration of 230 kg/m3.[9] Both thepolymeric and the ceramic membranes had a MWCO of1 kDa. Based on their estimates (adjusted to 2016 value),the total cost of removing lignin with polymeric (ceramic)NF membranes only was $52 ($77)/ton of lignin. In com-parison, a combination of UF and NF was much moreexpensive, at $135 ($485)/ton of lignin for the polymeric(ceramic) membranes.[9] It is apparent that regardless ofthe membrane material, NF alone achieved greater recov-ery at a much lower capital cost and that the polymericmembrane was considered more economical due to itslower fabrication cost despite its much shorter lifetime(1.5 years compared to 6 years for the ceramic membrane).
In order to obtain a more generalized picture of theeffects of permeance and selectivity on the economicsof BL concentration membranes, and to relate thisanalysis to currently available membranes, here weconsider a simple mass-balance analysis of a single-stage membrane process assign that both the feed andpermeate sides are well-mixed. Figure 13 shows the
flow diagram of the input and output streams fromthis process, and the Supporting Information lists allthe equations used to describe the mass balances onthis membrane process. The feed is softwood BL whosecomposition based on a BL analyzed in our laboratory,with the mass fractions of lignin (XLf), salts (XSf), andother organic species (XOf) being 0.039, 0.048, and0.063, respectively. The feed flow rate is 0.032 m3/sand is being concentrated from 15 to 30 wt% totalsolids at 85°C. The membrane is operating at a TMPof 3000 kPa, and the osmotic (ΔΠ) term is estimatedfrom a literature correlation using the same reflectioncoefficient extracted from experimental membrane fluxdata.[9, 69] For the curves, we assume an ideal mem-brane with 99% lignin rejection, 15% salt rejection, and57% rejection of other organic species (an arithmeticaverage of the lignin and salt rejections, since the otherorganics are intermediate in size between lignin andinorganic salts). Data from the literature use the givenlignin rejections, but assume a 15% salt rejection and anarithmetic average of the lignin and salt rejections toestimate the other organic species rejection, since thatdata were not provided. It is assumed that all mem-branes can be operated at 85°C and 3000 kPa withoutany fouling and no flux decline. Equations (S1)–(S11)(Supporting Information) are then used to solve for theunknown quantities, which are the retentate andpermeate mass flow rates ( _R and _P, respectively); themass fractions of lignin, salts/inorganics, and otherorganic species in the retentate (XLr, XSr, and XOr,
respectively) and permeate (XLp, XSp, and XOp, respec-tively), as well as the water mass fraction in the perme-ate (XWp).
Figure 14a shows the calculated membrane area versuspermeance for the above membrane process. Figure 14bshows the calculated total levelizedmembrane cost per yearversus permeance for different values of the per-unit-arealevelized membrane cost. Levelized cost estimates werebased on a period of 7 yrs, consistent with that for themanufacture of pulp and paper and of a waste reductionand resource recovery plant.[120] The levelized membranecost is obtained as the unit-area cost of the membrane($/m2) multiplied by the number of membranes (newand replacement) needed for a 7-yr recovery period andthe surface area needed (from Fig. 14a). The number ofmembranes needed was determined by dividing the recov-ery period by the rated lifetime of a membrane (1.5 yr forpolymers and 6 yr for ceramics). Note that operating costs(e.g., electricity for pressurizing and pumping the feed) arenot considered here. The capital cost (including replace-ment costs due to limited stability in BL) of themembranesis the key issue in their industrial adoption, since it is
SEPARATION SCIENCE AND TECHNOLOGY 1087
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
reasonable to assume that all the membranes consideredhere will provide similar energy savings relative to theevaporation processes. The positions of four
experimentally studied membranes (three polymeric andone ceramic) from the literature[9] are also shown as opensymbols in Figs. 14a–14b. Two of the polymeric
Figure 13. Flow diagram of a single-stage membrane process used to estimate the membrane surface area and costs.
Figure 14. (a) Calculated single-stage membrane surface area versus permeance and (b) calculated single-stage membrane levelizedyearly cost versus permeance. Experimental literature data (square open symbols) are based upon Ref.[9]
1088 N. S. KEVLICH ET AL.
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
membranes have higher lignin rejections (97% and 93%)and lower permeances, due to their lower MWCOs (0.2and 0.6 kDa, respectively). The other polymeric membraneand the ceramic membranes have lower lignin rejections(90% and 78%, respectively) and higher permeances due totheir higher MWCO (1 kDa).
Figure 14a shows that the state-of-the-art high-rejec-tion membranes have much higher required areas thanthe desired “ideal case” high-performance single-stagemembrane (with the dotted vertical lines denoting thedifferential between what is currently available andwhat is desired). In other words, new membraneswith higher permeances are needed. The low-rejectingmembranes, on the other hand, are close to the optimalmembrane area requirements. However, as seen inFig. 15, these two membranes are estimated to havesingle-stage lignin recoveries of only 30%–75% andtherefore cannot be used in an economical single-stage membrane process. Therefore, increases in ligninrejection are also needed. Clearly, there is no currentlyavailable commercial membrane with the requiredcombination of performance characteristics. Similarly,Figure 14b shows that high-rejection membranes arevery expensive and require a significant cost reductionto bring them to $50-100/m2 cost levels that are com-mon in the case of polymeric gas separation mem-branes. The currently available low-rejectionmembranes appear to have lower costs, but are notusable for the reason of low single-stage recovery asmentioned earlier.
Conventional ED technology is, for economic rea-sons, practically limited to TDS concentrations of ~0.5–15 g/L[69, 121] in applications such as brackish watertreatment. BL feed streams have much higher inorganic
concentrations (Table 2), and it can therefore beassumed that direct treatment of BL with ED mem-branes is not feasible. For example, Davy et al.[122]
performed an economic estimate on acidification costscompared to ED for lignin precipitation and found thatED was more than 3 times more expensive.Furthermore, ED membranes (both conventional andbipolar) cannot currently operate at typical BL feedtemperatures. However, it is possible that electricallydriven membranes could be used in combination withpressure-driven membrane. Bipolar membranes havelower energy costs than conventional ED[91] andwould probably be used. For example, NF membranescould reject virtually all of the lignin at 90 °C, and thepermeate stream could be treated with ED at a lowertemperature to achieve NaOH recovery without signif-icant lignin fouling. Given the efficiency limitations ofED, complete inorganic recovery may not be feasible.Hence, the use of a RO/FO membrane could be neces-sary to separate the low-concentration ED permeateinto concentrated brine and purified water.
Overall, the following technoeconomic conclusionscan be drawn. If simple concentration of BL to higherTDS content is desired at a lower cost than traditionalevaporators, NF appears to be the most desirable optionbut it requires long-lived membranes that are at least50% cheaper than the best membranes currently eval-uated in the literature. Furthermore, lignin rejectionsapproaching 99% are desirable. There is a significantchallenge in developing membranes that satisfy theserequirements. Although we do not wish to speculatehere regarding details of the composition and fabrica-tion routes of such membranes, a general observationcan be made. To lower membrane fabrication costs,low-cost polymers with good stability in BL can stillbe used as support materials, on which high-perfor-mance functional membranes made from new materialscould be fabricated using economical and scalablemethodologies. If more detailed separation of the TSinto different Mw fractions (lignin, other organics likehemicellulose, and inorganics) is desired, then a com-bination of tunable NF or molecular sieving mem-branes would be necessary. Neither of the abovescenarios addresses the full recovery of inorganics,which would mostly be permeated by the NF mem-branes. Conventionally, these would still be recoveredby combustion of the further concentrated SBL and byevaporation of water from the NF membrane permeate.However, the addition of a RO membrane downstreamof the NF membrane could be beneficial in producingpurified water and further concentrating the inorganicsand low-MW components. In this case, significantchallenges also exist in developing robust RO
Figure 15. Estimated lignin recoveries for four membranesstudied in the literature, based upon data in Ref.[9]
SEPARATION SCIENCE AND TECHNOLOGY 1089
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
membranes for inorganics recovery, since the pH of theNF membrane permeate is expected to be essentiallythe same as that of the BL feed stream. It may also bepossible to use ED (likely with bipolar membranes) inconjunction with RO/FO membranes to remove inor-ganics and carboxylic acids, but the economic viabilityof this scenario is unclear at present. The third andmost conceptually simple scenario is direct RO for rawBL concentration (retaining all suspended and dis-solved components and permeating only water). Thisscenario will likely remain unfeasible in the foreseeablefuture due to the very high TMPs required for directRO of raw BL and the prohibitively large amount offouling that would occur.
Conclusions
Significant reductions in cost and carbon emissionsfrom the Kraft process are expected to result fromsuccessful development of a BL concentration mem-brane technology. Additionally, membranes mayallow efficient separation of valuable BL componentssuch as low-Mw organic molecules and inorganicsalts. Progress has been made in the developmentand evaluation of membranes for BL concentrationand for separation of its valuable components.Polymeric membranes are challenged in BL applica-tions because of their low lifetimes upon exposure tothe harsh BL feed conditions (high pH, relatively hightemperature, and many fouling species). Because oftheir greater stability, non-polymeric (e.g., ceramic)membranes are expected to play a major role in theconcentration of black liquor. However, a major chal-lenge exists in improving their rejection and manu-facturing costs to decisively improve the economics ofmembrane-based BL concentration. Furthermore, toachieve the best possible separation (removal of waterand ions), NF and RO membranes are also required.These membranes are still largely polymer based, andfuture research will need to focus on the developmentof robust, lower-cost NF/RO membranes based uponnew materials that are suited for BL applications.While high fluxes are necessary, lignin rejectionvalues (say, 99%) need not be as stringent as thoserequired in desalination NF/RO membranes since it isnot sought to produce potable water from BL. Thedevelopment of membrane technologies driven byelectrical power (ED) or concentration gradients(FO) is emerging, but they face significant challengesof stability and cost in BL applications. Overall, thedevelopment of viable membrane technologies for BLconcentration—departing from classical polymeric orceramic membranes—is an important need for the
forest products industry, leading to significant energysavings and increased opportunities for water andmaterial recycling.
Acknowledgements
The authors also thank Steven Lien and Scott Sinquefield(Renewable Bioproducts Institute) and Zhongzhen Wang(Georgia Tech) for valuable discussions and assistance.
Funding
The authors acknowledge financial support by the RenewableBioproducts Institute and the Paper Science & EngineeringFellowship, both at the Georgia Institute of Technology.
References
[1] Frederick, J.W. (1997) Black Liquor Properties. In:Adams, T.N. (Ed) Kraft Recovery Boilers; TAPPIPress: Atlanta, GA
[2] Cardoso, M.; de Oliveira, E.D.; Passos, M.L. (2009)Chemical composition and physical properties ofblack liquors and their effects on liquor recoveryoperation in Brazilian pulp mills. Fuel, 88 (4): 756–763.
[3] De, S.; Bhattacharya, P.K. (1996) Recovery of water andinorganic chemicals from kraft black liquor usingmembrane separation processes. Tappi Journal, 79(1): 103–111.
[4] Kirbawy, S.A.; Hill, M.K. (1987) Multivalent ionremoval from kraft black liquor by ultrafiltration.Industrial & Engineering Chemistry Research, 26 (9):1851–1854.
[5] Boniface, A. (1992) Introduction and principals of che-mical recovery in green, In R.P., Hough, G. (Eds)Chemical Recovery in the Alkaline Pulping Process,Third Edition; TAPPI Press: Atlanta, GA
[6] Olsson, M.R. (2009) Simulations of Evaporation Plantsin Kraft Pulp Mills: Including Lignin Extraction andUse of Excess Heat. Ph.D. Thesis, Chalmers Universityof Technology: Göteborg, Sweden.
[7] Forest Products Industry Technology Roadmap: Agenda2020 In 2015; Vol. Oct. 3, 2015.
[8] Kong, L.B.; Hasanbeigi, A.; Price, L. (2016) Assessmentof emerging energy-efficiency technologies for the pulpand paper industry: a technical review. Journal ofCleaner Production, 122: 5–28.
[9] Arkell, A.; Olsson, J.; Wallberg, O. (2014) Process per-formance in lignin separation from softwood blackliquor by membrane filtration. Chemical EngineeringResearch & Design, 92 (9): 1792–1800.
[10] Jin, W.; Tolba, R.; Wen, J.L.; Li, K.C.; Chen, A.C.(2013) Efficient extraction of lignin from black liquorvia a novel membrane-assisted electrochemicalapproach. Electrochimica Acta, 107 (0): 611–618.
[11] Tomani, P. (2010) The Lignoboost process. CelluloseChemistry and Technology, 44 (1–3): 53–58.
[12] Ziesig, R.; Tomani, P.; Theliander, H. (2014)Production of a pure lignin product part 2:Separation of lignin from membrane filtration
1090 N. S. KEVLICH ET AL.
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
permeates of black liquor. Cellulose Chemistry andTechnology, 48 (9–10): 805–811.
[13] Ohman, F.; Theliander, H. (2007) Filtration propertiesof lignin precipitated from black liquor. TAPPI Journal,6 (7): 3–9.
[14] Jӧnsson, A.S.; Wallberg, O. (2009) Cost estimates ofkraft lignin recovery by ultrafiltration. Desalination,237 (1–3): 254–267.
[17] Kirkman, A.G.; Gratzl, J.S.; Edwards, L.L. (1986) Kraftlignin recovery by ultrafiltration - economic-feasibilityand impact on the kraft recovery-system. TAPPIJournal, 69 (5): 110–114.
[18] Ross, D.F.; Cartwright, T.B.; Doshi, M.R.; Dugal, H.S.(1986) Ultrafiltration of kraft black liquor. In IPSTTechnical Paper Series, Appleton, Wisconsin.
[19] Pettersen, R.C. (1984) The chemical composition ofwood in Rowell, In R.M. (Ed) The Chemistry of SolidWood. American Chemical Society: Washington, DC
[20] Venkatesh, V.; Nguyen, X.N. (1992) Evaporation andconcentration of Black Liquor in Green, In R.P., Hough,G. (Eds) Chemical Recovery Process in the AlkalinePulping Process, 3rd Ed.; TAPPI Press: Atlanta, GA
[21] Vakkilainen, E. (1999) Chemical recovery in Gullichsen,In J. F., C.-J. (Ed) Papermaking Science and Technology:Chemical Pulping; TAPPI Press: Finland.
[22] Niemelä, K.; Alén, R. (1998) Characterization ofPulping Liquors in Sjöström, In E.A., R. (Ed)Analytical methods in wood chemistry, pulping andpapermaking; Springer: Berlin Heidelberg.
[23] Sjöström, E.; Alén, R. (1998) Analytical Methods inWood Chemistry, Pulping and Papermaking; Springer:Berlin Heidelberg.
[24] Grace, T.M.; Sachs, D.G.; Grady, H.J. (1977)Determination of the inorganic composition of alkalineblack liquors. TAPPI Journal, 60 (4): 122–125.
[25] Satyanarayana, S.V.; Bhattacharya, P.K.; De, S. (2000) Fluxdecline during ultrafiltration of kraft black liquor usingdifferent flow modules: a comparative study. Separationand Purification Technology, 20 (2–3): 155–167.
[26] De, S.; Bhattacharya, P.K. (1996) Flux prediction ofblack liquor in cross flow ultrafiltration using low andhigh rejecting membranes. Journal of MembraneScience, 109 (1): 109–123.
[27] Wallberg, O.; Jonsson, A.S. (2006) Separation of ligninin kraft cooking liquor from a continuous digester byultrafiltration at temperatures above 100 degrees C.Desalination, 195 (1–3): 187–200.
[28] Xu, K.M.; Liu, H.Y.; Dong, L.; Tong, S.; Liu, Y.J. (2012)Lignin recovery from the black liquor of Reed pulping.In Ren, N., Che, L.K., Jin, B., Dong, R., Su, H. (Eds)Renewable and Sustainable Energy, Pts 1-4; Trans TechPublications Ltd: Stafa-Zurich.
[29] Easty, D.B.; Borchardt, M.L.; Webb, A.A. (1985)Analysis of Pulping Liquors By Ion Chromatography- Evaluation And Validation. Pap. Ja Puu-Pap. Timber,67 (9): 501–&.
[30] Passinen, K. (1968) Chemical composition of spentliquors. In Symposium on Recovery of PulpingChemicals, Kauppakirjapaino Oy: Helsinki, Finland.
[31] Niemi, H.; Lahti, J.; Hatakka, H.; Karki, S.; Rovio, S.;Kallioinen, M.; Manttari, M.; Louhi-Kultanen, M.(2011) Fractionation of organic and inorganic com-pounds from black liquor by combining membraneseparation and crystallization. Chemical Engineering &Technology, 34 (4): 593–598.
[32] Arhippainen, B.; Jungerstam, B. (1968) Kraft LiquorEvaporation. In Symposium on Recovery of PulpingChemicals, Kauppakirjapaino Oy: Helsinki, Finland.
[33] Liu, G.L.; Liu, Y.S.; Ni, J.R.; Shi, H.C.; Qian, Y. (2004)Treatability of kraft spent liquor by microfiltration andultrafiltration. Desalination, 160 (2): 131–141.
[34] Zabkova, M.; da Silva, E.A.B.; Rodrigues, A.E. (2007)Recovery of vanillin from lignin/vanillin mixture byusing tubular ceramic ultrafiltration membranes.Journal of Membrane Science, 301 (1–2): 221–237.
[35] Tanistra, I.; Bodzek, M. (1998) Preparation of high-purity sulphate lignin from spent black liquor usingultrafiltration and diafiltration processes. Desalination,115 (2): 111–120.
[36] Wallberg, O.; Holmqvist, A.; Jonsson, A.S. (2005)Ultrafiltration of kraft cooking liquors from a con-tinuous cooking process. Desalination, 180 (1–3):109–118.
[37] Alekhina, M.; Ershova, O.; Ebert, A.; Heikkinen, S.;Sixta, H. (2015) Softwood kraft lignin for value-addedapplications: Fractionation and structural characteriza-tion. Industrial Crops and Products, 66 (0): 220–228.
[38] Hill, M.K.; Violette, D.A.;Woerner, D.L. (1988) Loweringkraft black liquor viscosity by ultrafiltration. SeparationScience and Technology, 23 (12–13): 1789–1798.
[39] Brodin, I.; Sjoholm, E.; Gellerstedt, G. (2009) Kraftlignin as feedstock for chemical products: The effectsof membrane filtration. Holzforschung, 63 (3): 290–297.
[40] Niemelä, K.; Tamminen, T.; Ohra-aho, T. (2007) Blackliquor components as potential raw materials. In 14thInternational Symposium on Wood, Fibre and PulpingChemistry, Durban, South Africa, pp. 25–28.
[41] Holmqvist, A.; Wallberg, O.; Jonsson, A.S. (2005)Ultrafiltration of kraft black liquor from two Swedishpulp mills. Chemical Engineering Research & Design, 83(A8): 994–999.
[42] Wallberg, O.; Jonsson, A.S. (2003) Influence of themembrane cut-off during ultrafiltration of kraft blackliquor with ceramic membranes. Chemical EngineeringResearch & Design, 81 (A10): 1379–1384.
[43] Wallberg, O.; Jonsson, A.S.; Wimmerstedt, R. (2003)Fractionation and concentration of kraft black liquorlignin with ultrafiltration.Desalination, 154 (2): 187–199.
[44] Wallberg, O.; Jonsson, A.S.; Wimmerstedt, R. (2003)Ultrafiltration of kraft black liquor with a ceramicmembrane. Desalination, 156 (1–3): 145–153.
[45] Jӧnsson, A.S.; Nordin, A.K.; Wallberg, O. (2008)Concentration and purification of lignin in hardwoodkraft pulping liquor by ultrafiltration and nanofiltra-tion. Chemical Engineering Research & Design, 86(11A): 1271–1280.
[46] Nordin, A.K.; Jonsson, A.S. (2009) Flux and energyrequirement during ultrafiltration of a complex
SEPARATION SCIENCE AND TECHNOLOGY 1091
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
industrial process stream. Desalination and WaterTreatment, 2 (1–3): 317–324.
[47] Nie, A.N.S. (2008) Characterization of recovered blackliquor and isolated lignin from oil palm empty fruitbunch soda pulping for semichemical and chemicalpulps. M.S. Thesis, Universiti Sains Malaysia.
[48] Kumar, H.; Alén, R. (2014) Partial recovery of aliphaticcarboxylic acids and sodium hydroxide from hardwoodblack liquor by electrodialysis. Industrial & EngineeringChemistry Research, 53 (22): 9464–9470.
[49] Rojas, O.; Song, J.; Argyropoulos, D.; Bullón, J. (2006)Lignin separation from kraft black liquors by tangentialultrafiltration. La Chimica e L’Industria, 88 (1): 88–95.
[51] Jansson, Z.L.; Brannvall, E. (2011) Characterisation ofdissolved spruce xylan in kraft cooking. Nordic Pulp &Paper Research Journal, 26 (4): 380–385.
[52] Xu, G.; Yan, G.-y.; Yang, J.-h. (2012) An integratedgreen process for beneficial utilization of pulpingblack liquor. Waste Biomass Valor, 4 (3): 497–502.
[53] Aldaeus, F.; Schweinebarth, H.; Torngren, P.; Jacobs, A.(2011) Simplified determination of total lignin contentin kraft lignin samples and black liquors.Holzforschung, 65 (4): 601–604.
[54] Keyoumu, A.; Sjodahl, R.; Henriksson, G.; Ek, M.;Gellerstedt, G.; Lindstrom, M.E. (2004) Continuousnano- and ultra-filtration of kraft pulping black liquorwith ceramic filters - A method for lowering the load onthe recovery boiler while generating valuable side-pro-ducts. Industrial Crops and Products, 20 (2): 143–150.
[55] Pirttinen, E.; Stenius, P.; Kovasin, K. (2007) Surfactantphase equilibria and separation of amphiphilic extrac-tives from black liquor in kraft cooking of wood. Journalof Dispersion Science and Technology, 28 (1): 5–10.
[56] Hazlewood, P.E.; Singh, P.M.; Hsieh, J.S. (2006) Role ofwood extractives in black liquor corrosiveness.Corrosion, 62 (10): 911–917.
[57] Brodin, I. (2009) Chemical properties and thermalbehaviour of kraft lignins. In Fibre and PolymerTechnology, KTH Royal Institute of Technology:Stockholm, Sweden.
[58] Toledano, A.; Garcia, A.; Mondragon, I.; Labidi, J.(2010) Lignin separation and fractionation by ultrafil-tration. Separation and Purification Technology, 71 (1):38–43.
[59] Johakimu, J.; Andrew, J. (2013) Hemicellulose extractionfrom south african eucalyptus grandis using green liquorand its impact on kraft pulping efficiency and paper mak-ing properties. Bioresources, 8 (3): 3490–3504.
[60] Hellstén, S.; Lahti, J.; Heinonen, J.; Kallioinen, M.;Mänttäri, M.; Sainio, T. (2013) Purification processfor recovering hydroxy acids from soda black liquor.Chemical Engineering Research and Design, 91 (12):2765–2774.
[61] Alen, R. (2011) Principles of biorefining. Biorefining ofForest Resources; Paper Engineer’s Association/Paperija Puu Oy Porvoo, Finland.
[62] Zaman, A.A.; Fricke, A.L. (1994) Newtonian viscosityof high-solids kraft black liquors: effects of temperature
and solids concentrations. Industrial & EngineeringChemistry Research, 33 (2): 428–435.
[63] Liu, G.L.; Liu, Y.S.; Shi, H.C.; Qian, Y. (2004)Application of inorganic membranes in the alkalirecovery process. Desalination, 169 (2): 193–205.
[64] Cortinas, S.; Luque, S.; Alvarez, J.R.; Canaval, J.;Romero, J. (2003) Microfiltration of kraft black liquorsfor the removal of colloidal suspended matter (pitch)(vol. 147, pg 49, 2002). Desalination, 151 (3): 309–310.
[65] Cortinas, S.; Luque, S.; Alvarez, J.R.; Canaval, J.;Romero, J. (2002) Microfiltration of kraft black liquorsfor the removal of colloidal suspended matter (pitch).Desalination, 147 (1–3): 49–54.
[66] Manttari, M.; Lahti, J.; Hatakka, H.; Louhi-Kultanen,M.; Kallioinen, M. (2015) Separation phenomena in UFand NF in the recovery of organic acids from kraftblack liquor. Journal of Membrane Science, 490: 84–91.
[67] Jonsson, A.S.; Wimmerstedt, R. (1985) The applicationof membrane technology in the pulp and paper-indus-try. Desalination, 53 (1–3): 181–196.
[68] Schäfer, A.I.; Fane, A.G.; Waite, T.D. (2005)Nanofiltration: Principles and Applications; Elsevier:Amsterdam.
[73] Baker, R.W. (2002) Membrane Technology.Encyclopedia of Polymer Science and Technology; JohnWiley & Sons, Inc: New York.
[74] Foley, G. (2013) Membrane Filtration: A Problem SolvingApproach with MATLAB®; Cambridge University Press:Cambridge, UK.
[75] Causserand, C.; Rouaix, S.; Akbari, A.; Aimar, P. (2004)Improvement of a method for the characterization ofultrafiltration membranes by measurements of tracersretention. Journal of Membrane Science, 238 (1–2):177–190.
[76] Rajnish; Bhattacharya, P.K. (1985) Ultrafiltrative soluterejection behavior of black liquor. ACS SymposiumSeries, 281: 313–323.
[77] Poddar, T.K.; Singh, R.P.; Bhattacharya, P.K. (1989)Ultrafiltration flux and rejection characteristics of blackliquor and polyethylene-glycol. Chemical EngineeringCommunications, 75: 39–56.
[78] Bhattacharjee, C.; Bhattacharya, P.K. (1992) Predictionof limiting flux in ultrafiltration of kraft black liquor.Journal of Membrane Science, 72 (2): 137–147.
[79] Bhattacharjee, C.; Bhattacharya, P.K. (1993) Fluxdecline analysis in ultrafiltration of kraft black liquor.Journal of Membrane Science, 82 (1–2): 1–14.
[80] Li, J.; Ohagan, T.; MacLeod, J.M. (1996) Using ultra-filtration to analyze the molar mass distribution ofkraft lignin at pH 13. Canadian Journal of ChemicalEngineering, 74 (1): 110–117.
1092 N. S. KEVLICH ET AL.
Dow
nloa
ded
by [
Geo
rgia
Tec
h L
ibra
ry]
at 1
8:50
04
Sept
embe
r 20
17
[81] Bhattacharjee, C.; Bhattacharya, P. (2006)Ultrafiltration of black liquor using rotating disk mem-brane module. Separation and Purification Technology,49 (3): 281–290.
[82] Bhattacharjee, S.; Datta, S.; Bhattacharjee, C. (2006)Performance study during ultrafiltration of Kraftblack liquor using rotating disk membrane module.Journal of Cleaner Production, 14 (5): 497–504.
[83] Bhattacharjee, C.; Sarkar, P.; Datta, S.; Gupta, B.B.;Bhattacharya, P.K. (2006) Parameter estimation andperformance study during ultrafiltration of Kraftblack liquor. Separation and Purification Technology,51 (3): 247–257.
[84] Adikane, H.V.; Thakar, D.M.; Nene, S.N. (2004)Optimisation of colour and sugar rejection of blackliquor using membranes. Separation and PurificationTechnology, 36 (3): 229–234.
[85] Lake, M.A.; Blackburn, J.C. Process for ProducingFurfural from Black Liquor. US Patent Application#14/119,517, 2014.
[86] Olsen, O. (1980) Membrane technology in the pulpand paper-industry. Desalination, 35 (1–3): 291–302.
[87] Ren, X. (2016) Evaluation of membrane filtration fortreatment of black liquor in small-scale pulp and papermills in India. M. S. Thesis, Massachusetts Institute ofTechnology.
[88] Adnan, S.; Hoang, M.; Wang, H.T.; Bolto, B.; Xie, Z.L.(2010) Recent trends in research, development andapplication of membrane technology in the pulp andpaper industry. Appita Journal, 63 (3): 235–241.
[89] Cloutier, J.-N.; Azarniouch, M.; Callender, D. (1995)Electrolysis of weak black liquor Part II: Effect ofprocess parameters on the energy efficiency of theelectrolytic cell. Journal of Applied Electrochemistry,25 (5): 472–478.
[90] Ghatak, H.R. (2009) Economic potential of black liquorelectrolysis as a treatment option for small agro-basedmills. Tappi Journal, 8 (11): 4–11.
[91] Franken, T. (2000) Bipolar membrane technology and itsapplications. Membrane Technology, 2000 (125): 8–11.
[92] Jonsson, A.S.; Tragardh, G. (1990) Fundamental prin-ciples of ultrafiltration. Chemical Engineering andProcessing, 27 (2): 67–81.
[93] Wijmans, J.; Nakao, S.; Smolders, C. (1984) Flux lim-itation in ultrafiltration: osmotic pressure model andgel layer model. Journal of Membrane Science, 20 (2):115–124.
[94] Cheryan, M. (1998) Ultrafiltration and MicrofiltrationHandbook; CRC Press: Lancaster, PA.
[95] Cheryan, M. (1986) Ultrafiltration Handbook;Technomic Pub. Co.: Lancaster, PA.
[96] Spiegler, K.S.; Kedem, O. (1966) Thermodynamics ofhyperfiltration (reverse osmosis): criteria for efficientmembranes. Desalination, 1 (4): 311–326.
[97] Kedem, O.T.; Katchalsky, A. (1958) Thermodynamicanalysis of the permeability of biological membranes tonon-electrolytes. Biochimica et Biophysica Acta, 27:229–246.
[98] Thiel, S.W.; Lloyd, D.R. (1988) Application of thestefan-maxwell equations to the pressure-driven mem-brane separation of dilute multicomponent solutions ofnonelectrolytes. Journal of Membrane Science, 37 (3):233–249.
[99] Sarkar, P.; Datta, S.; Bhattacharjee, C.; Bhattacharya, P.K.; Gupta, B.B. (2006) Performance study on ultrafil-tration of Kraft black liquor and membrane character-ization using Spiegler-Kedem model. Korean Journal ofChemical Engineering, 23 (4): 617–624.
[100] Williams, M.E. (2003) A Review of Reverse OsmosisTheory. EET Corporation and Williams EngineeringServices Company, Inc.
[101] Thormann, A.; Teuscher, N.; Pfannmöller, M.; Rothe, U.;Heilmann, A. (2007) Nanoporous Aluminum OxideMembranes for Filtration and Biofunctionalization.Small, 3 (6): 1032–1040.
[102] Wilhelm, F.G. (2001) Bipolar membrane electrodialysis-membrane development and transport characteristics. Ph.D. Thesis, University of Twente: Enschede, TheNetherlands.
[103] Haddad, M.; Mikhaylin, S.; Bazinet, L.; Savadogo, O.;Paris, J.R. (2016) Electrochemical acidification of kraftblack liquor: effect of fouling and chemical cleaning onion exchange membrane integrity. ACS SustainableChemistry & Engineering. doi: 10.1021/acssuschemeng.6b01179
[104] Gargulak, J.D.; Lebo, S.E. (2000) Commercial use oflignin-based materials in Northey, In R.A., Glasser, W.G., Schultz, T.P. (Eds) Lignin: Historical, Biological, andMaterials Perspectives; American Chemical Society:Washington, D.C.
[105] Wang, Z.; Xue, J.W.; Zhang, J.S.; Liu, W.X. (2012)Ultrafiltration treatment and physical properties ofammonia sulfite pulping liquor of wheat straw. In Li,Y. (Ed) Material Sciences and Technology, Pts 1 & 2;Trans Tech Publications Ltd: Stafa-Zurich.
[106] Ban, W.; Van Heiningen, A. (2011) Adsorption ofhemicellulose extracts from hardwood onto cellulosicfibers: I. Effects of adsorption and optimization factors.Cellulose Chemistry and Technology, 45 (1–2): 57–65.
[108] Furfural Market Analysis By Application (FurfurylAlcohol, Solvent) And Segment Forecasts To 2020.Grand View Research: 2015.
[109] Holmbom, B. (1999) Extractives. In Sjöström, E., Alén,R. (Eds) Analytical Methods in Wood Chemistry,Pulping and Papermaking; Springer: Berlin Heidelberg.
[110] Bhattacharjee, C.; Sen, D. (2012) Treatment of kraft blackliquor using membrane-based separation process. InMohanty, K., Purkait, M.K. (Eds) MembraneTechnologies and Applications; CRC Press: Boca Raton, FL.
[112] Cath, T.Y.; Childress, A.E.; Elimelech, M. (2006) Forwardosmosis: Principles, applications, and recent develop-ments. Journal of Membrane Science, 281 (1–2): 70–87.
[113] Zhang, J. (2013) Forward Osmosis Black LiquorConcentration Device and Operating MethodThereof. Chinese Patent CN103194925.
[114] Haddad, M.; Bazinet, L.; Savadogo, O.; Paris, J. (2016)Electrochemical acidification of Kraft black liquor:Impacts of pulsed electric field application on bipolarmembrane colloidal fouling and process intensification.Journal of Membrane Science. DOI: 10.1016/j.memsci.2016.10.043.
[115] Haddad, M.; Labrecque, R.; Bazinet, L.; Savadogo, O.;Paris, J. (2016) Effect of process variables on the per-formance of electrochemical acidification of Kraft blackliquor by electrodialysis with bipolar membrane.Chemical Engineering Journal, 304: 977–985.
[116] Gregor, H.P.; Rowe, J.W. (1986) Membrane Processes forSeparation of Organic Acids from Kraft Black Liquors. USPatent 4,584,057.
[117] Azarniouch, M.K.; Prahacs, S. (1991) Recovery ofNaOH and Other Values from Spent Liquors andBleach Plant Effluents. US Patent 5,061,343.
[118] Dafinov, A.; Font, J.; Garcia-Valls, R. (2005) Processing ofblack liquors by UF/NF ceramicmembranes.Desalination,173 (1): 83–90.
[120] Peters, M.S.; Timmerhaus, K.D.; West, R.E. (2003) PlantDesign and Economics for Chemical Engineers, 5th Ed.;McGraw-Hill: New York.
[121] Water Treatment Technology Fact Sheet: Electrodialysis.ALL Consulting (2010).
[122] Davy, M.F.; Uloth, V.C.; Cloutier, J.N. (1998)Economic evaluation of black liquor treatment pro-cesses for incremental kraft pulp production - A com-parison of two processes at Paprican. Pulp & Paper-Canada, 99 (2): 35–39.
![Identification of lignin oligomers in Kraft lignin using ... · dependent neutral loss-MS3 approach for carbonyl com- pounds in salvia [24]. This approach can be adapted for LOs and](https://static.documents.pub/doc/80x56/604dacf3fdfc2a0b862add9d/identification-of-lignin-oligomers-in-kraft-lignin-using-dependent-neutral-loss-ms3.jpg)
