Ethanol-Water Separation by Countercurrent Reverse Osmosis

modern RO systems operate at below 1000 psi, concentrating ethanol beyond about 15 vol% ..... Fluid Systems Division of UOP Inc. (San Diego,. Californ...
2 downloads 0 Views 1MB Size
Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on September 19, 2015 | http://pubs.acs.org Publication Date: December 12, 1985 | doi: 10.1021/bk-1985-0269.ch019

19 Ethanol-Water Separation by Countercurrent Reverse Osmosis ERIC K. L. LEE, W. C. BABCOCK, and P. A. BRESNAHAN Bend Research, Inc., Bend, OR 97701-8599

Countercurrent reverse osmosis (CCRO) is a process design that helps solve a major problem in enriching ethanol by reverse osmosis: the high osmotic pressure of concentrated ethanol solutions. The effective osmotic pressure gradient across a membrane is reduced by supplying the permeate side of the membrane with a solution more concentrated in ethanol than the permeate but less concentrated than the feed. This causes ethanol to back-diffuse from the recirculation solution into the membrane. The concentration increase inside the membrane lowers the concentration difference (and thus the osmotic pressure difference) between the feed-solution and membrane phases. Membranes with open porous sublayers are preferred for use in CCRO because they allow ethanol to diffuse relatively unhindered through the sublayer and accumulate inside the membrane. With a new thin-film-composite membrane, designated 3N8, it has been shown that CCRO is about seven-fold more energy-efficient than distillation for enriching 10 vol% ethanol to 50 vol%. However, this and other reverse-osmosismembranes developed for desalination cannot be used at higher ethanol concentrations because of their low ethanol-water selectivity and their tendency to degrade.

Ethanol produced by fermentation i s conventionally dehydrated by d i s t i l l a t i o n , an i n e f f i c i e n t process that consumes energy equivalent to a large f r a c t i o n of the energy content of the product ethanol.Q.) Reverse osmosis (RO) has been considered before f o r ethanol-water separation because of i t s inherent energy e f f i c i e n c y . However, a d i f f i c u l t y encountered i n using RO i s the high osmotic pressures associated with concentrated ethanol solutions. For example, the osmotic pressure of a 15-volZ ethanol solution i s about 960 p s i , and that of a 50-vol% solution i s about 3700 psi.(2) Because most 0097-6156/85/0269-0409$06.00/0 © 1985 American Chemical Society In Materials Science of Synthetic Membranes; Lloyd, D.; ACS Symposium Series; American Chemical Society: Washington, DC, 1985.

Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on September 19, 2015 | http://pubs.acs.org Publication Date: December 12, 1985 | doi: 10.1021/bk-1985-0269.ch019

410

M A T E R I A L S SCIENCE O F SYNTHETIC M E M B R A N E S

modern RO systems operate at below 1000 p s i , concentrating ethanol beyond about 15 v o l % has not been considered possible, according to previous studies.(2,3) Countercurrent reverse osmosis (CCRO) i s a process design o r i g i n a l l y conceived by Loeb and Bloch(4) for overcoming the problem of high osmotic pressure gradients i n the production of concentrated s a l t solutions. E s s e n t i a l l y , the osmotic pressure across a membrane i s reduced by r e c i r c u l a t i n g , on the permeate side of the membrane, a solution whose osmotic pressure i s s l i g h t l y below that of the feed. Only a modest operating pressure would i n p r i n c i p l e sustain the flow across the membrane and effect separation. CCRO may be applied to ethanol enrichment as i l l u s t r a t e d i n Figure 1. A fermentation beer containing about 8 to 10 v o l % ethanol i s pressurized and fed into a membrane u n i t . Assuming that the membrane i s p r e f e r e n t i a l l y permeable to water, ethanol would be enriched i n the d i r e c t i o n of feed solution flow. Part of the concentrated product solution leaving the feed side of the membrane unit i s depressurized and recirculated to the permeate side, flowing countercurrent to the feed. Mixing between the r e c i r c u l a t i o n solution and the permeate brings the permeate-side concentration (and thus i t s osmotic pressure) closer to that of the feed. In t h i s way, the osmotic pressure difference can be kept below the operating pressure even as the feed solution becomes highly concentrated. The CCRO concept has not been proven i n practice; thus, an objective of the present work was to demonstrate the process concept experimentally. Various RO membranes were characterized to determine i f their use for ethanol enrichment by CCRO would be more energy-efficient than by d i s t i l l a t i o n , and to i d e n t i f y membrane c h a r a c t e r i s t i c s that affect the performance of the process. Principles Comparison of RO and CCRO. In v i r t u a l l y a l l RO membranes, a t h i n , selective skin layer i s supported by a much thicker microporous sublayer. During RO operation, the composition of the permeate i s determined by the s e l e c t i v i t y of the skin layer, the feed solution composition, and the operating pressure. The concentration of the permeate i s established as the feed solution flows through the skin layer, and i t remains constant inside the sublayer. This concentration p r o f i l e i s shown i n Figure 2a. The osmotic pressure across the membrane, Δ π - π^, increases with increasing concentration difference, 2"" 3* the skin layer. I f Δπ exceeds the operating pressure, no permeate w i l l be obtained. Since the osmotic pressure of a 14-vol% ethanol solution already exceeds 800 p s i , f o r example, an operating pressure of 800 p s i cannot be used to enrich ethanol solutions past 14 v o l % i f the membrane exhibits perfect r e j e c t i o n . β

c

c

a

c

r

o

s

s

Real membranes are not perfectly permselective; the rate of ethanol permeation across the skin layer of the membrane determines the actual osmotic pressure gradient. The lower the membrane s e l e c t i v i t y , the less l i k e l y i t i s that the flow of permeate would be stopped due to high osmotic pressures. However, using a membrane with poor ethanol r e j e c t i o n necessarily compromises the e f f i c i e n c y of separation.

In Materials Science of Synthetic Membranes; Lloyd, D.; ACS Symposium Series; American Chemical Society: Washington, DC, 1985.

19.

LEE ET AL.

411

Countercurrent Reverse Osmosis

CCRO represents a more a t t r a c t i v e method of reducing the osmotic pressure gradient. As shown i n Figure 2b, an ethanol concentration gradient ^ " C ^ i s established across the porous sublayer of the membrane by supplying a r e l a t i v e l y concentrated solution to the low-pressure side of the membrane. This concentration gradient causes ethanol to diffuse toward the skin layer against the convective flow of the permeate. An accumulation of ethanol inside the sublayer increases the ethanol concentration Controlling '3 the permeate-sii e concentration i n this way decreases the e f f e c t i v e osmotic pressur^ difference and increases the permeate f l u x . We refer to the fljix increase as the "CCRO e f f e c t . "

Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on September 19, 2015 | http://pubs.acs.org Publication Date: December 12, 1985 | doi: 10.1021/bk-1985-0269.ch019

C

CCRO Process Mojiel. The CCRO process i s i l l u s t r a t e d by the schematic diagram of the membrane unit shown i n Figure 3. A model of the process i s derived based on the o v e r a l l mass balance around the entire membrane u n i t , the l o c a l mass balance around a d i f f e r e n t i a l section ( x — y ) of the u n i t , and the membrane model which correlates the fluxes of water and ethanol within that section to the operating conditions. The o v e r a l l mass balance can be written as: V

= V, - V ρ f r V c - V.c. - V c , ρ ρ f f r r'

and

(1) (2)

where c and V represent the ethanol concentration (as volume fraction) and volumetric flow rate, respectively. The subscripts p, f, and r denote product, feed, and r e c i r c u l a t i o n solutions, respectively. Subscripts 2 and 4 denote the feed and permeate sides of the membrane, respectively. External concentration p o l a r i z a t i o n i s assumed to be n e g l i g i b l e , so that 2 i * 4 5 ' ** ip c

— j-

e c

an
Membrane ! Solution

Feed ; ; Permeate Solution ι Membrane ι C1

Co Solute Diffusion

Water Flow c

Water Flow C Solute Flow[

Solute Flow C C3

Ca

SoluteRejecting Skin Layer

Porous Sublayer (b) CCRO

(a) RO

Figure 2. Comparison of Concentration P r o f i l e s Inside a Skinned Reverse-Osmosis Membrane Under Reverse-Osmosis and Countercurrent Reverse-Osmosis Conditions

Gross Output of Product Solution

Membrane Vf

Dilute Ethanol Feed Solution"

Diluted Recirculation Solution to be Reprocessed at the Fermentation Stage C Vr r

Figure 3.

c"

> /Vp/(1-r)

//

Net Output of Concentrated Ethanol Product

2

!Cp*rm! !Vp»rm!

^(Pressure Reduction and Recycle Ratio Control)

C4" V » r

Partial Recirculation of Concentrated Ethanol Product Cp

Schematic Diagram of the Countercurrent Reverse-Osmosis Process

In Materials Science of Synthetic Membranes; Lloyd, D.; ACS Symposium Series; American Chemical Society: Washington, DC, 1985.

19.

413

Countercurrent Reverse Osmosis

L E E ET A L .

Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on September 19, 2015 | http://pubs.acs.org Publication Date: December 12, 1985 | doi: 10.1021/bk-1985-0269.ch019

gross product solution recirculated to the permeate side. Referring to Figure 3, i f the net product output flow rate i s V , then the gross product flow rate at the exit of the membrane unit i s V /(1-r) and the recycled stream flow rate i s rV / ( 1 - r ) . ^ The recycle r a t i o affects the performance of the CCRO process i n two ways. With a high recycle r a t i o , the osmotic pressure gradient i s more e f f e c t i v e l y reduced, and the consequent permeate flux increase means that the desired degree of ethanol enrichment can be accomplished i n a smaller membrane area. However, a higher recycle r a t i o also decreases the net product flow rate such that more feed solution i s required to reach a given production goal. At some optimum recycle r a t i o , the desired product purity and productivity i s obtained with the minimum membrane area and feed rate. Membrane Model.

The function of the membrane model i s to allow J ν

and J to be calculated under a given set of operating conditions. The properties of both the skin layer and the porous sublayer are involved i n deriving a suitable CCRO membrane model. The material of the skin layer governs i t s permselectivity, and the porous sublayer characteristics determine how e f f e c t i v e l y the osmotic pressure gradient across the skin layer could be reduced by permeate-side r e c i r c u l a t i o n . A v a r i e t y of RO membrane models exist that describe the transport properties of the skin layer. The solution-diffusion model(5) i s widely accepted i n desalination where the feed solution i s r e l a t i v e l y d i l u t e on a mole-fraction b a s i s . However, models based on i r r e v e r s i b l e thermodynamics usually describe membrane behavior more accurately where concentrated solutions are involved.(6) Since high concentrations w i l l be encountered i n ethanol enrichment, our present treatment adopts the i r r e v e r s i b l e thermodynamics model introduced by Kedem and Katchalsky. (7.) In this model, the volumetric flux Jv and solute flux J 8 g

.

through the skin layer are described by two equations: J and

J

g

ν

= L

(ΔΡ-σ[π -π ])

ρ

Z

(9)

J

= J (l-0)(c +c )/2 + ω ' ^ - ^ ) , v

2

(10)

3

where ΔΡ i s the operating pressure. There are three membrane parameters: L^, the hydraulic permeability; σ, the r e f l e c t i o n c o e f f i c i e n t ; and ω ' , the solute permeability. The values of these parameters are obtained from RO experiments i n which and c^ (equal to c^) are d i r e c t l y measurable. The s e l e c t i v i t y of the membrane i s expressed i n terms of the r e j e c t i o n : Rejection (%) = ( l - c / c ) χ 100%. 3

2

(11)

Equations 9 and 10 have the same form as the model equations derived by Kedem and Katchalsky, but they d i f f e r i n two ways: 1) the use of (c +c^)/2 instead of the logarithmic mean of c and c^; and 2) the 9

2

In Materials Science of Synthetic Membranes; Lloyd, D.; ACS Symposium Series; American Chemical Society: Washington, DC, 1985.

414

M A T E R I A L S SCIENCE OF SYNTHETIC M E M B R A N E S

use of c j ' ( c - c ) instead of ω ί ^ - π ^ ) , as given by those authors. These modifications were made to simplify the subsequent d e r i v a t i o n of the CCRO membrane model; some loss of accuracy i s expected at high ethanol concentrations, 'at which the osmotic pressure does not vary l i n e a r l y with concentration. In the porous sublayer, the net ethanol flux i s the difference between convective permeation and back-diffusion:

Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on September 19, 2015 | http://pubs.acs.org Publication Date: December 12, 1985 | doi: 10.1021/bk-1985-0269.ch019

2

3

J

v

dc(x)

c(x)

~

(12)

S

dx where D i s the e f f e c t i v e d i f f u s i v i t y of ethanol i n the porous sublayer, and χ i s the distance measured from the skin-sublayer interface toward the low-pressure side of the membrane. D e incorporates the effects of a l l sublayer c h a r a c t e r i s t i c s that affect mass transfer of ethanol, such as porosity and t o r t u o s i t y . At steady state, the ethanol flow across the skin layer equals that across the porous sublayer. Thus, combining Equations 10 and 12 yields the o v e r a l l ethanol balance across the entire membrane: e

J ( l - a ) ( c + c J / 2 + u>'(c -c ) = J c(x) - D ν L 5 2 3 ν e 9

0

dc(x)

Q

(13)

dx

Equation 13 can be integrated with the following boundary conditions : c(x) = and

at χ = 0

c(x) • c^ at χ = t ,

where t i s the thickness of the sublayer.

Thus:

(1-σ)/2 + ' / J + ( c / c ) ( Q / l - Q ) W

c /c 3

v

4

2

(14)

2

l/U-Q) - (1-σ)/2 + ω'/J

where Q exp[-J K]. Κ i s the fourth parameter i n the membrane model, defined as K; :t/D , and quantifies the d i f f u s i o n a l resistance to ethanol i n the sublayer of the membrane. With Equation 14, "3 can be e x p l i c i t l y calculated for known values of and c ^ / c the concentration r a t i o of the r e c i r c u l a t i o n and feed solutions. Assuming osmotic pressure to be proportional to ethanol concentration, Equations 9 and 14 are combined to give v

e

2>

( l - Q c / c ) / ( l - Q ) - (1-σ) 4

J

ν

=

L ρ

ΔΡ - σ π

2

(15)

0

l/U-Q) - (1-σ)/2 + ω'/J

In Materials Science of Synthetic Membranes; Lloyd, D.; ACS Symposium Series; American Chemical Society: Washington, DC, 1985.

19.

Equation 15 i s an i m p l i c i t function of i n terms of measurable quantities and the four membrane parameters L^, o*» ω > and K, and i t can be solved by numerical methods. Substituting the values of J into Equation 14 yields c^, which i n turn allows J to be calculated from Equation 10, Then the quantities V and c can be calculated using Equations 7 and 8, perm perm The value of Κ i s determined from an RO experiment i n which the permeate side of the membrane i s exposed to a r e c i r c u l a t i o n solution to produce a known ^/ 2 i° simulate CCRO conditions. The measured permeate flux i s substituted into Equation 15 to compute the value of Κ that would match the experimental value of J . Combining the process and membrane models, the o v e r a l l CCRO system i s described by Equations 1 through 8, 10, 11, 14, and 15, and the value of the recycle r a t i o , r . The performance of the CCRO process i s expressed in terms of the t o t a l membrane area (which determines the equipment costs), and the input flow rate of the pressurized d i l u t e feed solution (which determines the energy c o s t s ) . For a given membrane, these two quantities are fixed at a given recycle r a t i o and operating pressure. v

Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on September 19, 2015 | http://pubs.acs.org Publication Date: December 12, 1985 | doi: 10.1021/bk-1985-0269.ch019

415

Countercurrent Reverse Osmosis

LEE ET AL.

g

c

c

r a t

t 0

Experimental Membranes. Seven membranes were evaluated i n this work. These were flat-sheet and hollow-fiber membranes o r i g i n a l l y developed for RO applications. A l l have the anisotropic, skinned structure depicted in Figure 2. 1. NS-100.

2.

3.

4.

5.

This i s a polyurea thin-film-composite membrane formed by i n t e r f a c i a l l y crosslinking polyethyleneimine (PEI) with tolylene diisocyanate (TDI) on a microporous polysulfone support membrane. Flat-sheet and hollow-fiber membranes were prepared in our laboratory. The soluterejecting skin layer was deposited on the lumen surface of the hollow f i b e r s . The f i b e r s were pressurized i n t e r n a l l y during use. NS-101. This i s a polyamide TFC membrane prepared by crosslinking PEI with isophthaloyl chloride. Flat-sheet NS-101 membranes were prepared by a procedure similar to that used for producing NS100 membranes. TFC-801. Flat-sheet samples of this commercial polyetherurea TFC membrane were supplied by the Fluid Systems D i v i s i o n of UOP Inc. (San Diego, California). Polybenzimidazolone (PBIL). This i s an asymmetric membrane(8) that exhibits good r e j e c t i o n of many organic compounds. Flat-sheet samples were provided by T e i j i n Ltd. (Tokyo, Japan). Cellulose t r i a c e t a t e (CTA). CTA hollow-fiber membranes were obtained from Dow Chemical Company (Midland, Michigan). The solute-rejecting layer i s on the

In Materials Science of Synthetic Membranes; Lloyd, D.; ACS Symposium Series; American Chemical Society: Washington, DC, 1985.

Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on September 19, 2015 | http://pubs.acs.org Publication Date: December 12, 1985 | doi: 10.1021/bk-1985-0269.ch019

416

M A T E R I A L S SCIENCE OF SYNTHETIC

MEMBRANES

outside surface of these f i b e r s , which are pressurized externally. 6. Poly(vinyl alcohol) (PVA). A commercial flat-sheet PVA f i l m (Mono-Sol Type 1-0015-3, Chris Craft Industries, Inc., Gary, Indiana) was surfacecross linked with TDI to render i t water-insoluble. 7. 3N8. This i s a TFC polyamide membrane developed at Bend Research. It i s formed by i n t e r f a c i a l l y crosslinking a monomeric amine on a microporous polysulfone support membrane. Membrane Testing. The membranes were characterized on an RO test loop. Hollow-fiber modules were equipped with f i t t i n g s as shown i n Figure 4 to allow c i r c u l a t i o n on both sides of the membrane. F l a t sheet membranes were tested i n special c e l l s of the type shown i n Figure 5; r e c i r c u l a t i o n solution could be pumped into a port i n the center of the c e l l and forced to flow through a sintered stainless steel support plate to reach the permeate side of the membrane. RO tests were conducted at various pressures to measure ethanol r e j e c t i o n and permeate flux as functions of feed concentration. To mimic CCRO conditions, a solution equal i n concentration to the feed was used for permeate-side r e c i r c u l a t i o n , and the changes i n flux was monitored as r e c i r c u l a t i o n was switched on and o f f . Results and Discussion Reverse-Osmosis Tests. The performance of the membranes at low ethanol feed concentrations i s summarized i n Figure 6. The 3N8, TFC-801 and NS-100 flat-sheet membranes exhibited the best s e l e c t i v i t y . Notably, the highest ethanol r e j e c t i o n observed was about 60%—much lower than the rejection of these membranes for s a l t s . Further tests were conducted at higher feed concentrations; the results are shown i n Figures 7 to 11. The membranes generally exhibited decreasing r e j e c t i o n and permeate flux with increasing ethanol feed concentration. Obviously, the increase i n osmotic pressure with increasing concentration offset the effect of lower r e j e c t i o n , which tends to reduce the osmotic pressure gradient across the membrane. The net result was a decrease i n driving force for permeation, and thus a decrease i n the flux observed. For the CTA and PBIL membranes, the permeate f l u x increased somewhat as the membranes became e s s e n t i a l l y non-selective at feed concentrations approaching about 50 v o l % ethanol. A l l of the membranes tested degraded to some extent upon exposure to concentrated ethanol solutions. Degradation was manifested by lower rejections and higher fluxes when membranes exposed to high ethanol concentrations were retested at low feed concentration. The hollow-fiber geometry i s preferred for CCRO because i t i s more convenient to provide c i r c u l a t i o n on both sides of fibers than i t i s for flat-sheet membranes. A number of hollow-fiber NS-100 membrane modules were tested at 250 p s i . The f i b e r s were stable f o r several days at t h i s operating pressure at feed concentrations up to 25 v o l % ethanol. However, higher operating pressure and/or feed concentration caused these membranes to f a i l rapidly due to p l a s t i c i z a t i o n weakening of the polysulfone support. The CTA

In Materials Science of Synthetic Membranes; Lloyd, D.; ACS Symposium Series; American Chemical Society: Washington, DC, 1985.

Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on September 19, 2015 | http://pubs.acs.org Publication Date: December 12, 1985 | doi: 10.1021/bk-1985-0269.ch019

LEE ET AL.

Countercurrent Reverse Osmosis

Feed Solution Entrance (NS-100 only)

Ports for Permeate-Side Circulation (for NS-100) or Pressurized Feed (for CTA)

Unch

+Feed Solution Exit (for NS-100) or Permeate Outlet (for CTA)

Figure 4. Construction of a Hollow-Fiber Membrane Module

In Materials Science of Synthetic Membranes; Lloyd, D.; ACS Symposium Series; American Chemical Society: Washington, DC, 1985.

418

M A T E R I A L S SCIENCE OF SYNTHETIC

MEMBRANES

Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on September 19, 2015 | http://pubs.acs.org Publication Date: December 12, 1985 | doi: 10.1021/bk-1985-0269.ch019

Recirculation Solution Entrance (also used as permeate outlet during RO testing) Bolt (1 of 4) Sintered Stainless Steel Plate

Recirculation Solution Exit

Ο Rings

Feed Solution Entrance

Feed Solution Exit

Recirculation Solution Entrance

Bolt (1 of 4)

Recirculation Solution Exit Sintered Stainless Steel Plate

SECTION A-A'

Figure 5. Flat-Sheet Membrane Test C e l l

In Materials Science of Synthetic Membranes; Lloyd, D.; ACS Symposium Series; American Chemical Society: Washington, DC, 1985.

19.

Ethanol Rejection

Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on September 19, 2015 | http://pubs.acs.org Publication Date: December 12, 1985 | doi: 10.1021/bk-1985-0269.ch019

419

Countercurrent Reverse Osmosis

LEE ET AL.

0

20

40

0

5

10

(%)

• 60

100

Flat Sheet Membranes

Hollow-Fiber Membranes

Permeate Flux

(gfd)

Figure 6. Average Reverse-Osmosis Performance of Membranes Evaluated

Figure 7.

Effect of Ethanol Feed Concentration on the ReverseOsmosis Performance of T r i p l i c a t e 3N8 Membrane Samples (400 p s i , ambient temperature)

In Materials Science of Synthetic Membranes; Lloyd, D.; ACS Symposium Series; American Chemical Society: Washington, DC, 1985.

420

M A T E R I A L S SCIENCE O F S Y N T H E T I C M E M B R A N E S

Downloaded by UNIV OF CALIFORNIA SANTA BARBARA on September 19, 2015 | http://pubs.acs.org Publication Date: December 12, 1985 | doi: 10.1021/bk-1985-0269.ch019

Ethanol Rejection (%) 0 20 40 60 80 Ethanol Concentration in Feed (vol %) 10Γ Permeate Flux (gfd)

0 20 40 60 80 Ethanol Concentration in Feed (vol%) Figure 8. Effect of Ethanol Feed Concentration on the ReverseOsmosis Performance of T r i p l i c a t e TFC-801 Membranes (800 p s i , ambient temperature)

Ethanol 20 Rejection (%)

10 20 30 40 50 60 Ethanol Concentration in Feed (vol%)

Permeate Flux 5 (gfd)

10 20 30 40 50 60 Ethanol Concentration in Feed (vol%) ( · Performance after