The potential of pervaporation for biofuel recovery from fermentation:An energy consumption point of view☆

Peiyao Zheng,Chong Li,Naixin Wang,Jie Li,Quanfu An,,*

1 MOE Key Laboratory of Macromolecular Synthesis and Functionalization,Department of Polymer Science&Engineering,Zhejiang University,Hangzhou 310027,China

2 Beijing Key Laboratory for Green Catalysis and Separation,College of Environmental and Energy Engineering,Beijing University of Technology,Beijing 100124,China

Keywords:Bio-alcohol Pervaporation Energy Selectivity Vapor-liquid equilibrium

A B S T R A C T Recovering alcohols from dilute fermentation broth is an emergent task in bio-fuel production process.Since they are primary planned for fuels,energy required to separate these alcohols should be considered in evaluating the potential of a separation technology.A membrane-based process,pervaporation,is of special interest because of its environmental friendliness and easy integrating character.This review probes into the fundamentals of pervaporation especially in terms of the heat required for evaporation.Meanwhile,the separation data of the most representative alcohol-selective pervaporation membranes reported in the literatures are collected and compared with the vapor-liquid equilibrium curve,which represents the distillation selectivity.They include:inorganic membranes, silicon rubber based membranes,Mixed Matrix Membranes and some other special materials.By doing so,the status of alcohol recovery via pervaporation would be more clear for researchers.For ethanol recovery,it is selectivity rather than flux that is in urgent need of solution.While for butanol recovery,membranes with satisfactory selectivity have been developed, increasing the separation capacity would be more pressing.

1.Introduction

The production of green, renewable energy is greatly motivated by the concerns of energy security and environmental problem.Under this circumstance,bio-fuels,as a form of solar energy utilization,promises to become the future energy resources.Ethanol and n-butanol are of special interest because of their potential as an alternative to common gasoline [1]. The USA and Brazil had developed bioethanol as modern energy for long[2,3].In 2017,the National Energy Administration of China issued the implementation plan for the application of vehicle ethanol-gasoline,identifying the significance of bio-alcohol production in the next few years.

The microbial synthetic pathway now allows to produce a variety of advanced biofuel[4].However,the fermentation broths only provide a low concentration alcohol mixture. Separation technology must be applied to extract the target products from the dilute solution before it is ready to serve as fuels.On the other hand,the bioreactor productivity would also be greatly enhanced if an in-situ separation technology is integrated to remove the sufficient alcohols,which may cause inhibition to the micro-organisms.Distillation is currently the dominate method to recovery alcohol from dilute fermentation broths,taking advantage of its matured technique,high efficiency and easy to scale-up property[5]. The demand for lower energy consumption and more mild operation condition encourages the exploration of other alternative separation technologies,including adsorption,liquid-liquid extraction,gas stripping,membrane distillation and pervaporation[5-7].Pervaporation is probably the most potential process owing to the following advantages:1)High separation factor potential-if a highly selective membrane is adopted.2)Moderate operating conditions and no harm to the microorganisms. 3) No need for additives which may cause contamination.4)Continuous process.A pervaporation involved bio-alcohol production flow is diagrammed in Fig.1.

A general comparison between different separation technologies is highlighted in Table 1.The Vapor-Liquid Equilibrium(VLE),on which many of these technologies rely,shows a love-hate relationship with alcohol/water separation.Owing to the VLE behavior,the composition difference between liquid and vapor phase allows the spontaneous separation of components depending on their relative volatility.But in the same time,leaving an upper bound of separation selectivity for those without a third party(entrainer)involved[8].Membranes can be applied in organic recovery in the form of membrane distillation or pervaporation.In membrane distillation,a porous membrane(microfiltration or ultrafiltration membrane)is adopted.Conventional wisdom suggests that the separation depends on the VLE behavior,which means the separation efficiency cannot outperform the relative volatility.Pervaporation,however,is able to make use of the relative volatility of components as well as the selective transport of membranes simultaneously,which endows it with great research value.

Fig.1.Pervaporation involved bio-alcohol production flow.

Table 1Advantages and disadvantages of several separation technologies in alcohol recovery

Since these bio-alcohols are primary planned for fuels,the energy required to produce and separate them should, apparently, become one of the most important factors to be considered[9,10].First of all,the total energy requirement should not higher than the heat of combustion,e.g.,the combustion heat of ethanol and n-butanol is 29.7 and 36.1 MJ·kg-1,respectively,otherwise the game will not worth the candle.Fortunately,this is not a big problem for the modern technique[9].Secondly,to replace a matured and widely used separation technology,such as distillation,the alternative would better show higher energy efficiency at similar operation conditions. Logically, a pervaporation membrane with a separation factor higher than the relative volatility(calculated from VLE curve)is expected to substitute at least one ideal plate of distillation column in terms of latent heat of vaporization required[11].

Most of previous reviews about organophilic pervaporation focus on the membrane materials[6,12-14],few had discussed energy aspects systematically. Here we aim at the energy consumption during the pervaporation process for biofuel recovery from dilute fermentation,mainly the heat required to evaporate the alcohols and water.Hereby,the status and future demands of pervaporation in comparison with those VLE based separation methods can be clearly seen.

2.Fundamentals of Pervaporation

The history of pervaporation can be traced to 1917, when Kober et al. first proposed this concept [15]. The features of pervaporation can be concluded as: 1) liquid phase in feed side;2)vapor phase in permeate side;and 3)separation achieved through a selective membrane.A schematic illustration is presented in Fig.2.Permeation is driven by the vapor pressure difference between the two sides of the membranes,since the downstream is always applied with a vacuum or sweep gas.The most widely used model to describe the molecular transportation during pervaporation is solutiondiffusion model[16],which assumes that the feed liquid first dissolve in the membrane,then diffuse across the membrane at concentration gradient and finally desorb into vapor phase at the permeate side.Owing to the different affinities and diffusion rates of species through the membrane,one or several substances can be enriched in the permeate side,thereby separation occurs.

In most publications,the separation performance of pervaporation was reported as permeate flux(J)and separation factor(β),which is defined as:

Fig.2.Schematic illustration of pervaporation process.

and

where Q is the amount of permeate,A is the effective membrane area,t is the operation time,x and y are concentrations of components i and j in feed and permeate,respectively.

Baker et al.emphasized that pervaporation performance data should be presented as permeability(Pi),permeance(Pi/l)and selectivity(α)[17].The equations are as follows:

where xiand γiare the mole fraction and activity coefficient of component i in the feed,pisatis the saturated vapor pressure of component i at feed temperature, pilis the partial pressure of component i in the permeate side and l is the effective membrane thickness.

Accordingly,the selectivity is defined as the ratio of the permeabilities or permeances of components i and j through the membrane:

The advantage of using permeability and selectivity is that it reflects the true property of a membrane,instead of that of the entire separation process.In other words,the driving force,which varies with operation conditions such as temperature, pressure and feed concentration, is normalized or eliminated.

Also, one must notice that even without the membrane, a free evaporation process is able to separate the species in most cases(except for azeotropes) according to the VLE curve. This is the principle of distillation.However,a highly selective membrane is expected to obtain much higher enrichment factor than free evaporation,i.e.not limited by the thermodynamic VLE behavior.This is particularly true for organic dehydration process [6], for example, the permeate water content could be higher than 99 wt%even when the feed is a near azeotropic water/ethanol mixture[18].

We can find that the definition of separation factor is in the same form as relative volatility(βv):

where z represents the vapor-phase concentration at vapor-liquid equilibrium state and

In many cases, especially in laboratory studies, the downstream pressure of pervaporation is usually very low,hence pilin Eq.(3)can be neglected and Eq.(3)therefore be simplified to:

Under these circumstances, the relationship between separation factor(β),relative volatility(βv)and selectivity(α)can be expressed as(here selectivity should be given in mole form):

This result indicates that both VLE behavior and membrane property contribute to the separation efficiency of pervaporation.A non-selective membrane(α=1)is supposed to share the same separation efficiency with distillation or other VLE-based methods. It also provides a benchmark to evaluate a pervaporation membrane by comparing the feed and permeate composition with VLE curve.That is,a real selective membrane should possess a permeate concentration higher than that of VLE data.Otherwise,the membrane is not necessary,if not playing a negative role,in the separation process[11,17].

3.Energy Consumption

Pervaporation,unlike most other membrane technologies,involves a phase change (evaporation) during the separation process, which costs energy.As mentioned before,the energy consumption is a crucial factor in bio-alcohol recovery via pervaporation.Generally,the energy required in pervaporation includes latent heat of vaporization,sensible heat to a certain feed temperature,coolant(condenser),vacuum,liquid pump and so on. It is difficult or maybe unrealistic to take all these aspects into account as few literature or industrial case and be referred.On the other hand, the latent heat of vaporization contributes the majority of total energy consumption according to Vane[9]and many of the rest energy expenditure like pumps also exist in other separation techniques.Therefore,we only look into the latent heat of vaporization as a representative energy requirement in this review.

An ideal membrane with infinite selectivity or separation factor only allows the desired component to permeate(for bio-alcohol recovery,which is ethanol or butanol).In such a case,we only need to provide the heat to evaporate the target product. However, the selectivity is never infinite, in fact, far from it for current membranes in alcohol recovery.The undesired molecules(water)will also permeate through the membrane unavoidably and,at the same time,cost energy.

The normalized energy consumption (Q), the total heat input to permeate per unit of alcohol (marked as component i), can then be calculated as:

For alcohol(i)-water(j)binary system,Eq.(9)is rewritten as:where ΔHivapis the heat of vaporization of bio-alcohol and Ciis alcohol concentration in permeate.

Although ΔHivapchanges with temperature and pressure,the variation at common pervaporation conditions is usually negligibly small that it can be regarded as constant. Specifically, ΔHivapof ethanol,n-butanol and water are 838, 593, and 2260 kJ·kg-1, respectively.Then the normalized energy consumption Q is only the function of permeate alcohol concentration Ci. In other words, higher alcohol concentration in permeate is always preferred,no matter it is achieved by improving the membrane selectivity or simply increasing the feed concentration. It is seen that ΔHivapof water is much higher than that of bio-alcohols, which means a large amount of energy will be applied on the phase transition of water.For example,if the permeate mixture contains 33 wt% ethanol and 67 wt% water (a common permeate composition in separating 5 wt% ethanol/water mixture with PDMS membrane), it will cost 5.36 MJ energy to evaporate each kg of ethanol, and 84% of that is wasted in the evaporation of undesired water molecules.One should be aware that multistage separation is needed to achieve the desired concentration(～95 wt%),which can be further dehydrated to meet fuel specifications,usually by adsorption or hydrophilic pervaporation.Moreover,in real applications,the feed concentration is not kept constant as we usually do in the lab(the so-called fractional evaporation),but decreases with the separation process.So the final feed concentration(also known as residual concentration) depends on the required alcohol recovery. And logarithmic average alcohol concentration is usually adopted to represent the average feed concentration (Eq. (11)) [5]. As such, the average feed concentration decreases with increasing alcohol recovery and is more rapid at high recovery.That is,the less allowable product loss,the higher energy will need.Since Ciis significantly affected by the feed concentration, it is of great value to investigate the pervaporation separation performance in a wide range of feed concentration (not just 5 wt%ethanol/water or 1 wt%n-butanol/water),especially at low concentration.

To evaluate the potential of pervaporation in bio-alcohol recovery,the energy consumption should be compared with other separation technologies, especially the current benchmark: distillation, which completely relies on the relative volatility according to VLE behavior.Eqs. (9) and (10) hold correct for other separation technologies as well,so we can compare the energy consumption by directly comparing the permeate alcohol concentration Ciwith VLE data.In the following sections, the permselectivity of various pervaporation membranes towards ethanol an n-butanol will be discussed and compared with the VLE curve.

Realizing that he had let away someone so important in his life, he decided to call her immediately. His whole mind was flooded with fear. He was afraid that she might have found someone new or no longer had the same feelings anymore... For once, he felt the fear of losing someone.

4.The VLE Behavior

Before we start to make any comparison,the vapor-liquid equilibrium of ethanol/water mixture should be understood. The VLE data can be obtained by collecting experimental data and/or predicted with proper physical property method, such as Non-Random Two Liquid Model(NRTL),Universal Quasi-Chemical Model (UNIQUAC),Wilson Model,Margules model and so on.All these physical property methods rely on the regression of phase equilibrium data.For common liquid mixtures,the raw VLE data are adequate enough so that most methods are able to give accurate prediction. Fig. 3 presents the predicted VLE curve for ethanol/water at atmospheric pressure. It is seen that all these methods coincide well with the experimental result and the liquid temperature shows no significant inference on the vapor phase composition.When the ethanol mole fraction in liquid phase is 0.02 (ca. 5 wt%), the corresponding composition in vapor phase is 0.18(ca.35 wt%).This result brings out the reality that when fed with 5 wt% ethanol/water mixture, only those with a permeate ethanol concentration above 35 wt% (or separation factor ＞10.2) can be regarded as ethanol perm-selective membranes.

When it comes to n-butanol, things get even more complicated(Fig. 4). First, n-butanol is only partial soluble in water, the mixture turns to two phase in the region of 0.2 mol%-49 mol%(7.7 wt%-80 wt%)[5].Thereby a platform exhibited in the VLE curve of n-butanol/water system.Second,and more importantly,the experiment phase data at low butanol concentration is so rare that it created troubles for the prediction of phase behavior.The curves predicted by different methods match the experimental data to a certain extent when referring to the entire range of composition.Yet at low butanol concentration,the predictions obtained from different methods are no longer consistent with each other because few experimental phase equilibrium data can be referred within this range. For instance, at 60 °C and atmospheric pressure,the vapor phase composition which is in equilibrium with a 0.25 mol%(1 wt%) n-butanol/water liquid mixture is expected to be 6.1 mol%(21.1 wt%)according to the NRTL fitting from DECHEMA Vapor-Liquid Equilibrium Data Collection [19].While NRTL fitting form Aspen Plus gives a lower value of 2.6 mol%(9.9 wt%),and 3.1 mol%,4.1 mol%and 18.9 mol% (11.6 wt%, 14.9 wt% and 48.9 wt%) for UNIQUAC, UNIFAC and Wilson fitting,respectively.Consequently,researchers may draw confusing conclusions when different physical property methods are adopted. As an example, if a membrane shows a permeate butanol content of 25 wt%when fed with 1 wt%dilute butanol aqueous solution at 60°C,the selectivity α would be calculated to be 1.25 or 3.04 or 2.54 or 1.91 according to the above mentioned physical property methods.The difference is too vast to ignore and there is a very real possibility to result in a confusing or even wrong conclusion in comparing the separation performances(both permeability and selectivity)from different literatures. We suppose that this is because the raw VLE data of nbutanol/water are not adequate enough to give accurate predictions,since all these physical property methods rely on the regression of phase equilibrium data. To date it seems that NRTL model given by DECHEMA meets the experiment data best.Unfortunately,DECHEMA only provides the data at 60°C for constant temperature,which makes it difficult to discuss the permeability and selectivity at other temperatures. Unlike ethanol/water system, the VLE behavior of n-butanol/water seems to be obviously affected by the liquid temperature(Fig.4 c,d).The relative volatility of n-butanol is higher at higher temperature,which favors the evaporation of butanol.Based on the above analysis,standard and authoritative VLE data sets of n-butanol/water are in urgent need.

Fig.3.The vapor-liquid equilibrium(VLE)curve of ethanol/water binary system.(a)Experimental data and curves predicted by different methods and(b)the enlarged figure of(a)at low alcohol content.(c)The effect of liquid temperature predicted by NRTL method with data form Aspen Plus 8.8 and(d)the enlarged figure of(c)at low alcohol content.

Fig.4.The vapor-liquid equilibrium(VLE)curve of n-butanol/water binary system.(a)Experimental data and curves predicted by different methods and(b)the enlarged figure of(a)at low alcohol content.(c)The effect of liquid temperature predicted by NRTL method with data form Aspen Plus 8.8 and(d)the enlarged figure of(c)at low alcohol content.

5.Ethanol Recovery

In the past few decades, bio-ethanol is mainly produced by corn glucose (in the USA) and sugarcane sucrose (in Brazil). Recently,technology from nonfood-plant cellulosic sources has been developed for bio-ethanol production,which diminishes the predicament of food crisis[3].During the enzyme hydrolysis process,cellulose is hydrolyzed to glucose,and the resultant concentration of ethanol produced in the fermentation broth mainly ranges from 5 to 90 g·L-1and typically located at 45-50 g·L-1[2].That's why most researchers reported the separation data of ethanol recovery at a constant feed concentration of 5 wt%.

The permeate ethanol concentration versus feed ethanol concentration collected from previous literatures is presented in Fig.5[19-90].The dash line represents the VLE behavior of ethanol/water at 60 °C,since temperature does not have significant inference on this phase equilibrium,this curve is universally applicable.

Inorganic membranes based on hydrophobic zeolites have shown outstanding separation performance compared with other membranes and VLE curve. MFI type zeolites with high Si/Al ratio (ZSM-5) or pure-silica(silicalite-1)are structured with uniform,molecular-sized and hydrophobic pores and are featured with extraordinary sorption selectivity towards alcohols. Meanwhile, molecules have difficulty passing each other in the membrane pores so that the diffusion of ethanol is sped up while the diffusion of water is slowed down[91].Permeate ethanol concentrations between 70 wt%and 80 wt%can be easily achieved by zeolite membranes when fed with 5 wt%mixture[60,64,66]. Which means that the permeate mixture can be directly delivered to a hydrophilic pervaporation device to produce anhydrous ethanol with no intermediate process needed.However,the complex and expensive synthesis routes as well as the relatively low flux(compared with hydrophilic pervaporation)in removal of organic compounds from water make zeolite membranes less industrially attractive[92].Dispersing zeolite particles in polymer matrix could be regarded as a compromise between profile and membrane costs and the performance mainly depends on the type and loading of particles[12].The reported selectivity of mixed matrix membranes (MMMs) differs a lot with each other because the key factor is to fabricate membranes without agglomerates or defects. For polymeric membranes,poly(1-trimethylsilyl-1-propyne) (PTMSP) seems to be the best performing material apparently[9,12].This material is famous for its large free volume, which endows PTMSP with high permeability and hence can be applied in both gas separation and pervaporation.The permeate ethanol content of PTMSP membranes is located slightly above the VLE curve, revealing the inherent selectivity of PTMSP towards ethanol.Nevertheless,PTMSP membranes have been proved to be unstable with flux and selectivity declining against continuous operation, probably due to the fouling and collapse of free volume.For this reason,PTMSP as pervaporation membrane material gradually fade from people's view in the past decades.An interesting and subversive result can be observed in Fig.5a.The current benchmark polymeric membrane material, poly(dimethyl siloxane) (PDMS), can hardly outperform the simple evaporation process. More specifically, most PDMS membranes display separation selectivity within 0.6-1,the separation factor is usually lower than the relative volatility,only the best of which is able to reach the VLE line.It is seen that the optimal performance of PDMS and its derivatives rightly overlaps with the dash line.This result revealed that PDMS membranes are actually water selective for ethanol/water system,despite being investigated as ethanol permselective membranes for long.Similar conclusions are drawn by Baker et al.[17]in 2010 that pure silicone rubber membranes deliver selectivity lower than 1.In other words,PDMS membranes are playing negative roles in the separation of ethanol/water mixture while other polymeric membranes, such as polyether block amide (PEBA) and other block copolymers, have seldom reach the level of PDMS yet. In such cases(selectivity ＜1),pervaporation is not preferred.Other methods,such as membrane distillation is more suitable due to its higher flux and separation efficiency.A special remark should be given to the F-Ag-PD@Al2O3membrane prepared by Huang et al. [91]. The membrane was prepared by the polymerization of dopamine on a Al2O3disk at the presence of Ag+,followed by the modification with thiolfuoroalkanes.The as-prepared membrane was superhydrophobic with a thickness of only 0.2 μm. In separating 5 wt% ethanol/water mixture at 50 °C,the flux and separation factor were ca.2.5 kg m-2·h-1and 102,respectively.The selectivity was nowhere behind that of zeolite membranes and the flux was somehow higher. If such membrane is low cost in fabrication and is able to scale up,then it would be of great potential in ethanol recovery from fermentation broth.

Fig.5.The permeate ethanol concentration(a)and selectivity(b)versus feed ethanol concentration collected from previous literatures.Energy consumption for concentrating each kg ethanol in the first stage separation (c). Relative energy consumption for the reported pervaporation membranes in comparison with simple evaporation process according to vapor-liquid equilibrium(d).Dash lines represent the level of VLE based simple evaporation process and are shown for reference purposes.The abscissa values in c and d should be interpreted as the average feed concentration.Legends are common to all the sub-figures and are given only once in(a).

Fig.5c presents the energy consumption(Q)to concentrate each kg of ethanol in the first separation stage calculated according to Eq.(10).And the relative value of pervaporation membrane/simple evaporation is given in Fig. 5d. For all separation technologies involving phase change, the normalized energy consumption increases dramatically with the decreasing of feed ethanol concentration, especially when the average feed concentration is lower than 3 wt%. Note that the abscissa values in Fig. 5c,d should be interpreted as the average feed concentration(Ci,averagein Eq.(11))rather than the initial feed concentration.Given that the combustion heat of ethanol is 29.7 MJ/kg,distillation at Ci,average＜0.6 wt% cannot even afford the latent heat of evaporation in the first stage(Fig.5c),hence alternative separation technologies are in demand. For example, feed solution with an initial ethanol concentration of 5 wt%and a residual concentration of 0.5 wt%(ethanol recovery ～90%)equals the 2 wt%average concentration.Under this circumstance,distillation will have an average permeate ethanol concentration of 18 wt%and the energy consumption will be 10.5 MJ·kg-1ethanol.Zeolite membranes with ultra-high selectivity are expected to lower the energy consumption by ～70%compared with the first stripper of distillation. Membranes with selectivity ＜1 will require more energy input than simple evaporation.The overall level of PDMS membranes costs 100%-150%heat that of simple evaporation and PEBA membranes generally cost twice of that and most other polymeric membranes may cost even more than the energy content in ethanol. So far, one has to admit that most polymeric membranes,including the benchmark material PDMS, are not suitable for the recovery of ethanol from fermentation broth in energy perspectives.Zeolite membranes can meet the demand of energy efficiency,but at the same time,are known to be expensive and more likely to be affected by the fermentation byproducts[9,92],which makes zeolite membranes less industrially practical.The challenge and future research direction in bio-ethanol upgrading should be focused on developing polymeric membranes with real ethanol selectivity (at least α = 1, and α ＞5 would be better) or solving the difficulty and cost of manufacturing defect-free large-scale zeolite membranes.Mixed matrix membrane is probably a feasible pathway since the preparation and application of these membranes are somehow similar to standard polymeric membranes and the separation profile could be significantly enhanced compared with pure polymers[9].

In addition,there is an annoying trade-off between fermentation productivity/recovery and energy consumption in separation.Specifically,in order to keep a high fermenting rate,the alcohol concentration in the broth(Ci,initial)should be maintained at a low level.And to minimize the waste of alcohol products in the residual,the lower the final feed concentration Ci,initialthe better.However,low Ci,initialand Ci,initiallead to low Ci,averageaccording to Eq.(11).As is discussed before,the normalized energy consumption Q is only the function of permeate alcohol concentration.For membranes with certain separation property, the permeate concentration decreases with decreasing feed concentration,thereby leading to the increasing energy input [5]. Thus, striking a balance between productivity and cost is of significant value in practical separation process.

6.Butanol Recovery

Butanol,in a sense,is a better candidate as fuel alternative even than ethanol[93,94].The advantages of butanol include:1)Higher heat of combustion than ethanol and is closer to that of traditional gasoline.2)Lower vapor pressure at operation condition,which fascinates its application in existing gasoline supply channels without major modification.3)Not sensitive to water so that it is compatible with existing fuel storage infrastructure and is able to mix with gasoline at any proportion or even completely replace it with no need to worry about the water absorption issue.4)Less flammable and less hazardous to handle.On the other hand, the recovery of butanol is also more favorable compared with ethanol. Because butanol is only soluble in water at concentrations ＜7.7 wt%or ＞80 wt%,after one step recovery from dilute aqueous solution,the permeate would separate into two immiscible layers so that liquid-liquid separation units can be utilized [5].Currently, the main bio-butanol production pathway is acetonebutanol-ethanol (ABE) fermentation, where a mixture of acetone,butanol and ethanol is produced.The concentration of butanol in the product generally varies between 1 wt%and 2 wt%,and the most popular fermentation broth using Clostridium acetobutylicum as ferment bacteria gives a butanol concentration of around 1 wt%together with a acetone:butanol:ethanol mass ratio of approximately 3:6:1[7].

Fig.6.The permeate ethanol concentration(a)and selectivity(b)versus feed ethanol concentration collected from previous literatures.Energy consumption for concentrating each kg butanol in the first stage separation (c). Relative energy consumption for the reported pervaporation membranes in comparison with simple evaporation process according to vapor-liquid equilibrium(d).Dash lines represent the level of VLE based simple evaporation process and are shown for reference purposes.The abscissa values in c and d should be interpreted as the average feed concentration.Legends are common to all the sub-figures and are given only once in(c).

It is seen that the membrane materials applied in butanol recovery is generally the same as those in ethanol recovery.For the same material applied in butanol recovery and ethanol recovery,the butanol-water selectivity is normally higher than ethanol-water selectivity(Fig.6b).Fortunately,most researchers had achieved satisfactory performances with permeate butanol concentrations outperform the VLE curve.The membrane materials above the VLE curve can be roughly classified into four types,namely,hydrophobic zeolites,PDMS and its derivatives,PDMS based Mixed Matrix Membranes and PTMSP.Zeolites,especially silicalite-1,had shown great potential in butanol recovery.A permeate butanol concentration of ～80 wt%can be achieved at low feed butanol content of ～1 wt%.Remarkably,tubular silicalite-1 membrane prepared by Kanemoto et al. [109] exhibited a permeate butanol content of～31 wt%when the feed was only 0.1 wt%and ～75 wt%when fed with 1 wt%solution,the separation factors were up to 300-450.Distillation at 0.1 wt% average feed concentration is unacceptable because it requires more than 50 MJ·kg-1butanol right in the first separation stage,which exceeded the combustion heat of butanol(36.1 MJ·kg-1)(Fig. 6c) while this silicalite-1 membrane costs less than 5 MJ·kg-1heat in the first step,only ～10%that of distillation.If this level of selectivity could persist in large scale and real-time fermentation broth,then there would likely be a great possibility for pervaporation to replace distillation as the standard butanol recovery technology.

Moreover,PDMS membranes here are proven to be energy efficient compared with simple evaporation.The inherent selectivity of pure or modified PDMS in butanol recovery are in the range of 1-2 and the optimal data of PDMS membranes reportedly reduce roughly half the energy needed.Zeolite/PDMS mixed matrix membranes demonstrate slightly higher selectivity(2-4)than PDMS and further reduce the energy consumption by ca.25%.This is valuable for practical engineering design. For example, commercial pervaporation membranes Pervap 1070(zeolite/PDMS membrane,40 wt%zeolite loading)produced by Sulzer Chemtech displayed a flux of 137 g·m-2·h-1and a separation factor of 50 in separating 1 wt%butanol/water solution at 50°C[108].Although Pervap 1070 is no longer available now,it still acts as a pioneer in pervaporative recovery of alcohols.

There are also some other materials investigated in butanol recovery.PTMSP shares similar selectivity with mixed matrix membranes,but still limited by its instability.Recently,mental-organic frameworks(MOFs) have been extensively investigated as advanced materials in gas separation and other membrane processes. Although pure MOF membranes do not show favorable separation performance in organophilic pervaporation [155,156],dispersing their nanoparticles in PDMS matrix has gain gratifying achievements in butanol recovery.Examples that can be referred include ZIF-8/PDMS prepared by Fan et al.[126]and ZIF-7/PDMS prepared by Li et al.[105],which possess butanol-water separation factor of 81 and 66, respectively. Supported ionic liquid membranes(SILMs)are receiving increasing attention in recent years due to their organophilicity.A wide variety of ionic liquids with different properties can be obtained by tailoring the ions and therefore the separation performances various in a wide range.PEBA is still not an ideal membrane material in consideration of low alcohol selectivity. Again, F-Ag-PD@Al2O3membrane shows satisfactory separation performance, a flux of ca. 5 kg·m-2·h-1and permeate alcohol concentration of 82 wt% is achieved in separation of 3 wt%iso-butanol/water solution [91]. Currently, materials suitable for butanol recovery via pervaporation are mainly zeolite and PDMS,but other materials are very likely to be competitive with the VLE and are of significant potential.

In this review,membrane flux or permeability are not discussed,because flux does not affect the energy input during the phase change process while flux does,however,play an important role in the membrane cost and capital cost[11].For PDMS or zeolite/PDMS membranes with selectivity ＞1,the energy consumption has already been reduced compared with distillation, thus increasing the flux would be more efficient for economical.

At last,one must realize that Figs.5c,d and 6c,d only calculate the latent heat of vaporization required in the first stage of evaporation,the following separation stages and other aspects are not taken into account.Things would be more complicated if those issues were considered. Vane [9] had clarified that pervaporation membranes should have ethanol-water separation around 20 and/or 50 with and without heat recovery from condensation process to meet the energy efficiency of distillation.But this simplified case is still instructive to membranologists because the first step takes a major and relatively fixed proportion of the total energy consumption under similar situations.Hecke et al.[139]simulated the energy consumption in a scenario that consists of a simple stripper with decanter and subsequent distillation columns with 2 wt% n-butanol solution as the initial feed. The energy needed to obtain fuel grade n-butanol (99.76 wt%)with 99%butanol recovery in such a scenario was 21.4 MJ/kg and the first stripper required two-thirds of that(14.18 MJ·kg-1).

7.Conclusions and Perspectives

Pervaporation as an emerging membrane technology is potential in recovering bio-alcohol from fermentation broth. Seeing that these bio-alcohols are primarily produced as petroleum fuel substitutes,the energy required to manufacture and purify them should be minimized.Energy issues in pervaporation separation ethanol and butanol from dilute solution were discussed in detail with special attentions paid on the first separation stage.Specifically,the alcohol-water selectivity of different membranes reported in previous literatures were collected with energy consumptions calculated and compared with the vaporliquid equilibrium.In doing this,the profile ad potential of different membranes can be clearly seen and conclusions can be generalized as follows:

1) For ethanol recovery, the energy efficient membranes include zeolites, PTMSP and mixed matrix membranes. Most polymeric membranes, including the benchmark material PDMS, generally cost more energy compared with simple evaporation process.

2) Butanol is easier to separate compared with ethanol.Thereby the same membrane material shows higher selectivity in butanol recovery than in ethanol recovery.The optimal silicalite-1 membrane is able to reduce ca.90%energy consumption in the first separation stage,and the value is approximately 50%for PDMS membranes.

3) A few special materials mentioned above show satisfactory separation performance in the lab,but the prospects of industrialization are not investigated in depth.

Propelling pervaporation to industrial application faces several inevitable issues namely,selectivity,scale-up and long-term stability.Selectivity,or the separation performance in comparison with the VLE data,determines the energy consumption in pervaporation.This review has mainly examined the aspect of selectivity and energy requirement.While scale-up issues and long-term stability limit the application potential of two best performed membranes, pure zeolite and PTMSP.The former is expensive and difficult to manufacturing defect-free large-scale membranes, and the latter is criticized for the aging and performance degradation.Altogether,some perspectives and proposals can be provided,they are:

1) For ethanol recovery,efforts should be made in searching membranes with real ethanol perm-selectivity(selectivity ＞1).Membranes with high permeate flux but selectivity ＜1 make no contribution to the separation in energy perspective.

2) For butanol recovery,current polymeric membranes like PDMS have already proven to be energy effective,increasing the permeability may have better profits to reduce the membrane and capital cost.

3) If the scale-up problem can be solved, then zeolite membranes would be most potential for bio-alcohol recovery.

With the ever-increasing study of organophilic pervaporation,we believe that promising membrane materials can be discovered and further considered for application in alcohol perm-selective pervaporation in the near future.