EPA-R2-72-103
DECEMBER 1972           Environmental Protection Technology Series
Hollow Fiber Technology
for Advanced Waste Treatment
                                 Office of Research and Monitoring
                                 U.S. Environmental Protection Agency
                                 Washington, O.C. 20460

-------

-------

-------
              EPA Review Notice
This report has been reviewed by the Environmental
Protection Agency, and approved for publication.
Approval does not signify that the contents necessarily
reflect the views and policies of the Environmental
Protection Agency, nor does mention of trade names
or commercial products constitute endorsement or
recommendation for use.
                      11

-------
                         ABSTRACT
The utility of hollow fiber reverse osmosis membranes in reno-
vation of secondary municipal effluent was investigated through
construction, laboratory evaluation,  and monitoring in field service
of various  hollow fiber modules.  All units incorporated cellulose
acetate hollow fibers,  annealed for sodium chloride rejections of
80-95% at 250 psi external operating pressure.  Product water
capacities  ranged from 50-300 gallons  per day.  Module designs
considered included the single seal end,  looped fiber bundle;
double seal end,  parallel bundle; radial flow parallel bundle; and
a rolled, woven hollow fiber fabric.  The typical flux-rejection
characteristics  of the  basic fiber system (4 gfd-95%) were
observed in waste water service, but steady-state flux, maintained
only with regular detergent flushes,  was usually less than 1 gfd,
with an accompanying decline in selectivity.  A notable exception
was the woven hollow  fiber fabric design, which showed improved
retention of start-up characteristics and minimum effects of shell-
side fouling during short-term field tests.

This report was submitted in fulfillment of Contract 14-12-926,
Project #17040 FEE,  under the sponsorship of the Environmental
Projection Agency by  the Monsanto Research Corporation, Chemstrand
Research Center, Inc., Box  731, Durham, North Carolina 27702.  The
Project Engineer was J. D.  Bashaw.
                               111

-------
                   TABLE OF CONTENTS

                                                          PAGE

I    CONCLUSIONS                                          1

II    RECOMMENDATIONS                                   3

III   INTRODUCTION                                        5

IV   LABORATORY CHARACTERIZATION OF HOLLOW       7
     FIBER PERFORMANCE

         Preparation of Hollow Fibers                        7
         Construction of Test Facilities                      9
         Performance with Key Solutes                       9

V    DESIGN AND CONSTRUCTION OF  HOLLOW FIBER     23
     MODULES

         Basic Assembly Techniques                       23
         Series D:  single seal looped fiber                 25
         Series D-2:  double seal parallel fiber             25
         Series R:  radial flow                             29
         Series W: woven fabric                           30

VI   INVESTIGATIONS OF MODULE FOULING              41

         Laboratory Performance with  Surface-             41
            Active Agents
         Laboratory Simulation of Fouling                  43
         Analysis of Fouling in Field Service               46,

VII  FIELD RESULTS                                     51:

         Field Test Facilities                              51
         Series D Modules                                 56
         Series. D-2  Modu-les                               59
         Series W Modules                                 60

VIII ACKNOWLEDGMENTS                                87

EX   REFERENCES                                        89

X    PATENTS AND PUBLICATIONS                       91

-------
                         FIGURES


                                                          PAGE

 1  Schematic of laboratory reverse osmosis test            10
       loop for hollow fiber units

 2  Schematic of basic hollow fiber module designs          24

 3  Engineering design of WA Series hollow fiber            33
       module

 4  Construction of woven fabric hollow fiber modules       35

 5  Photograph of woven fabric module WA001               37

 6  Le Clerc,  Medico Model 22-4 loom used in              39
       assembly of module WA002

 7  Schematic of field reverse osmosis test loop for         53
    hollow fiber units

 8  Field unit for testing of hollow fiber modules             54
    with secondary municipal effluent

 9A-9E    Field results for module DQ09                   62-66

10A-10E   Field results for module D010                   67-71.

11A-11E   Field results for module D008                   72-76.

12A-12E   Field results for module D0013-2                77-81

13A-13E   Field results for module WA001                  82-86
                             VI

-------
                       TABLES
                                                        PAGE

1      Summary of Reverse Osmosis Performance          11
       for 40:60 Cellulose Acetate Hollow Fibers
       at 250 psi

2      Dependence of Product Water Flux on Time for       13
       Cellulose Acetate Hollow Fibers Operating with
       1000 ppm NaCl at 250 psi and Zero Recovery

3      Rejection of Orthophosphate by Cellulose             14
       Acetate Hollow Fibers

4      Effect of pH on Orthophosphate Rejection             14

5      Rejection of Nitrate by Cellulose Acetate             15
       Hollow Fibers

6      Acetate Rejection as a Function of pH for            1.6
       Cellulose Acetate Hollow Fibers

7      Ammonia Rejection by Cellulose Acetate             16
       Hollow Fibers

8      Reverse Osmosis Performance of Cellulose       __   17
       Acetate Hollow Fibers with Urea Feed Solutions

9      Reverse Osmosis Performance of Cellulose          18
       Acetate Hollow Fibers with Glycine Feed  Solutions

10     Flux-Time Response of Cellulose Acetate Hollow     19-20
       Fibers with Selected Solutes

11     Flux-Pressure Results for Cellulose Acetate         21
       Hollow Fibers with Triton X-100 Feed

12     Specifications for Hollow Fiber Module  D009         26

13     Specifications for Hollow Fiber Module  D0013-2      27
                             vn

-------
                                                       PAGE

14     Design Parameters for WA Series Hollow           31-32
       Fiber Modules

15     List of Materials for WA Series Hollow Fiber       34
       Module

16     Laboratory Reverse Osmosis Performance for      36
       Hollow Fiber Fabric Module  WA001

17     Laboratory Reverse Osmosis Performance for      38
       Hollow Fiber Fabric Module  WA002

18     Performance of Hollow Fiber Module D008 with     42
       Benzalkonium Chloride Solutions

19     Flux-Time Relationship with Benzalkonium         42
       Chloride Feed

20     Flux-Time Relationship with Sodium Alkyl-         43''
       Arylsulfonate Feed

21     Analysis of Flow Data for Module D008             44-45

22     Mean Content of Certain Solutes in Filtered         52
       Secondary Effluent
                            vi u

-------
                     CONCLUSIONS
In general,  two facts seem evident from operations with the  series
of hollow fiber modules tested in wastewater service.  First,
efficiency of module operation is definitely a function of  the effi-
ciency of module cleaning and the degree of module cleanliness
maintained.  Best results were almost always obtained with  modules
recently subjected to a detergent treatment in the field,  or with
those which had been renovated in the laboratory and returned to
service.  Our results also indicate that a  systematic cleaning and
conditioning regime in the field promotes  improved module per-
formance.   Possibly the best evidence for this conclusion is to be
found in the performances of modules D010 and D008.

A second,  significant observation is  that module designs which
promote a smooth, regular flow of water through the unit,  with
minimum opportunity for packing or  channeling, are likely to give
improved performance,  both with respect to flux stability and
efficiency of rejection over an extended time.  The primary  evi-
dence for this generalization is the field performance of module
WA001.  The woven fabric concept, which this unit represents,
appears to be an attractive way of achieving uniform, low resistance
shell-side  flow, while still retaining the desirable aspects of hollow
fiber geometry.  With fibers of dimensions used in this program,
the woven fabric design permits a ratio of effective membrane area
to pressure vessel volume of approximately 1000 ft2/ft3 -  still an
order of magnitude larger than for flat membrane  designs.

The frequently noted failure of BOD rejection results to  turn out
like those from sodium ion or chloride ion analysis was  a source
of some concern in our work.  In a number of specific cases, we
seemed to be dealing with indications that cellulose acetate fiber
is more permeable to such gases as  chlorine or hydrogen sulfide
than it is to water.  Accumulation of hydrogen sulfide or other
oxidizable  gases in the product water would have the net effect of
indicating poor performance with respect  to BOD rejection.

-------
                 II  RECOMMENDATIONS
The principal objective of this program was evaluation of various
hollow fiber configurations in renovation of waste water by reverse
osmosis.   On the basis of the exceptionally good performance shown
by units incorporating the woven fabric configuration,  it is recom-
mended that this module concept be further investigated.

The woven fabric design specifically investigated in this program
could be immediately improved in subsequent versions.  First,  it
is likely that smaller, and more durable warp fibers could be
employed.  Second,  overall module efficiency could be improved by
reduction in the diameter  of the center mandrel.

The results obtained here, more  broadly interpreted,  indicate that
hollow fiber bundles providing uniform, low-resistance shell-side
flow are much less prone  to fouling in sustained service.   This
advantage is gained  at the expense of  a substantial reduction in
fiber  packing density.   Accordingly, it seems worthwhile to  investi-
gate the utility of related, structured fiber assemblies in order to
retain the desired hydraulic characteristics, with potential for
improvement in packing density.

-------
                  Ill  INTRODUCTION
This report provides  a detailed account of all  work carried
out from 15 October 1970 to 30 November 1971 under Contract
14-12-926,  granted to Monsanto Research Corporation by the Water
Quality Office, U.  S. Environmental Protection Agency.   The
objective of the program was to evaluate the utility of high-flux
hollow fibers  in treatment of secondary municipal effluent.  Essen-
tially no development work on preparation of hollow fibers was
conducted under this contract, although current results and
improvements from other   related programs receiving U.  S.
Government support were  utilized whenever applicable.   The
organization of the report  follows the technical approach origi-
nally proposed  and carried out: laboratory evaluation of the basic
cellulose acetate hollow fiber system in rejection of waste water
constituents; design and construction of modules to meet the
requirements of waste water service; laboratory research and
engineering studies of fouling phenomena to aid in module design
and operation; evaluation of module designs in field service and
selection of appropriate operation and maintenance procedures.

-------
  IV  LABORATORY CHARACTERIZATION OF HOLLOW
                 FIBER PERFORMANCE
The  hollow fiber  system  available at the  beginning of this  pro-
gram was a dry-jet wet spun cellulose acetate composition,  typically
300/100 micron OD/ID dimensions, capable of being annealed  to
provide a range of flux-rejection properties at 250 psi.  Later in the
program, a modified cellulose acetate hollow fiber was developed to
provide especially attractive flux-divalent ion rejections at 100-150
psi.   Laboratory test facilities were constructed to evaluate fiber
performance in rejection of sodium chloride; phosphate, nitrate,  and
acetate ion; ammonia, urea, glycine, glycols and detergents.
Preparation of Hollow Fibers

The basic cellulose  acetate fibers used in  this program  are
essentially a hollow fiber form of the density-gradient or skin-core
Loeb   membrane structure.  By virtue of the very thin, dense skin
formed on the outer surface of the hollow fiber, rejection of dissolved
solutes may be raised to a high level  (98%, for example, with sodium
chloride).  The bulk of the hollow fiber wall is a relatively  porous
structure which easily  transports product water.  Thus, the only
appreciable resistance to passage of water through the fiber wall is
that created by the salt-rejecting skin.  Since the transport rates of
both salt and water are inversely proportional to skin thickness,
decrease in the latter does not alter the rejection capability; product
water flux,  on the other hand,  is  correspondingly increased.   In the
preparation   of hollow fibers incorporating this desirable structure,
considerable latitude is afforded by the unique  spinning process in
control of skin thickness and degree of perfection.  As a result, the
methods developed permit trade-off in flux and rejection over a wide
range.  Cellulose acetate hollow fibers under sustained test at 250 psi
external pressure with 3000 mg/1  sodium chloride feed typically pro-
vide from about 2 gal/ft/day product water flux (gfd) with  98% salt
rejection to 10 gfd with 50% salt rejection.    A particularly
useful intermediate selection is 4 gfd with 95% rejection.   Both flux
and rejection increase  with pressure over the permissible  range
(maximum 350 psi).

-------
 The basic steps in the continuous solution spinning process are analo-
 gous to those customarily employed in Loeb membrane preparation.
 The spinning dope  consists of about 35% cellulose acetate dissolved in
 a mixture of acetone and formamide.   Various proportions are suit-
 able for spinning - including cellulose acetate in pure acetone -
 depending upon the selection of other  process parameters.  A typical
 composition of polymer,  acetone and  formamide is 35:20:30 by weight.

 The viscous spinning dope is  metered through a thermostated jacket
 and extruded  through a noil of il spinnerette.   The spinnerette  may be
 of several basic designs, including a  segmented arc and a tube-in-
 orifice arrangement.  The latter provides greater flexibility  in con-
 trol of fiber diameters; the chief virtue of the former is its easier
 adaptability to multihole  spinnerette construction.  With both types,
 the outside diameter of the hollow  fibers extruded may be  made sub-
 stantially less than that.of the orifice  opening by increasing the take-
 up rate on the thread line.

 The first important step  in the formation of the actual membrane fila-
 ment occurs after emergence of the hollow fiber from the  orifice and
 during passage through an air gap.  The evaporation of the volatile
 component  (acetone) which occurs  establishes the concentration gra-
 dient essential for  skin formation.   The continuous thread line, par-
 tially stabilized by passage through the evaporative  air gap, next enters
 the coagulation bath - usually  water at low temperature.  Most of the sol-
 vent near the  fiber outer  surface is replaced by water in this  process,
 precipitating the polymer and stabilizing the fiber.  Virtually complete
 removal of solvent and final stabilization of the hollow fiber occurs in
 subsequent passage through water  wash baths on godet rolls.

 The  final tailoring  of membrane properties is obtained through con-
 solidation of the fiber wall structure in post or in-line immersion in
 hot water.  The hollow fibers, which  must be maintained wet in storage
 and subsequent operations,  are now wound on bobbins and  are ready for
use in construction of reverse osmosis modules.

 Cellulose acetate hollow fibers  spun by minor variants of the  above
procedures have been tested under a wide variety of end-use  conditions.
Fibers ranging in outside diameter from about 100 to about 300 microns
have been prepared.  Typical rejections of sodium chloride have been
 cited earlier;  in general,  a given fiber shows much improved rejection
of phosphate,  sulfate and other multivalent ions.   Rejection of all
solutes at 250 psi is essentially  constant for feeds ranging from 1000

-------
to 10, 000 mg/1. Flux stability is generally quite good and has been
established in continuous reverse osmosis testing for periods
exceeding two years.
Construction of Test Facilities
An eight-cell reverse osmosis  test loop was assembled early in
the program, to permit characterization of small fiber bundles and,
in later work, to accommodate 100-200 gpd modules for process
investigation.  Key components of the test loop included:  two inde-
pendent feed systems  to permit changing feed compositions with no
down-time, a Milton-Roy duplex diaphragm positive displacement
pump rated at 12. 6 gph, and a dome-loaded diaphragm back pressure
regulator of our own design.  Extensive use was made of PVC conduits,
to minimize accumulation of iron deposits in the recirculation system.
A schematic of  the laboratory test loop is given in Figure 1.

The standard fiber test bundle  consists of  a supported loop of
30 wraps of hollow fibers (60 fiber ends), potted into a 1/8" pipe
nipple with an epoxy resin (Shell Epon  815® and curing agent T-:
The pipe nipple is installed in the test  loop, through which feed solution
can be circulated at 200-400 psi,  in contact with the exterior  surfaces
of the hollow fibers.   Flux (in gfd) is obtained by measuring volume of
product water from the fiber bundle of known surface area per unit
time; rejection  r by conductivity or chemical analysis  of feed stream
composition (C^) and product composition (CD), in accordance with the
relationship

              r  =  100 (1 - Cp/Cf).
Performance with Key Solutes

Laboratory reverse  osmosis  test facilities were utilized in
evaluation of flux and rejection of hollow fiber samples with aqueous
solutions of solutes representative of those in municipal waste water.
All experiments were conducted with the small fiber bundles described
in the preceding section, at ambient temperatures, and at essentially
zero recovery.   The hollow fibers tested were cellulose acetate,  spun
from 40:60 acetone:formamide,  and  annealed at various temperatures
to provide a range of flux-selectivity relationships.  A summary of
results is given in Table I.

-------
ru
    1  2
   Feed
  Supply
   Tank
       (Product Water)
34       ^5
                      6
       ^L~lJU
                   Pressure
                   Gauge
                 S-,
                 o
                 -i-j
                 a
3
o
o
                                Back Pressure
                                 Regulator
                                Feed
                                Supply
                                Tank
                                                I
          Figure 1.  Schematic of laboratory reverse osmosis test
                  loop for hollow fiber units

-------
                                      TABLE 1

              Summary of Reverse Osmosis Performance for 40:60 Cellulose
                             Acetate Hollow Fibers at 250 psi
 Annealing	Rejection,  %  	
Temperature   Flux   NaCl   orthophosphate  nitrate  acetate   ammonia    urea   glycine
   0 C         gfd   (pH 6. 1)    (pH 7)	(pH 6. 1) (pH 6. 6)  (pH 6. 6)   (pH 5. 9)  (pH 6)
75
78
82
85
87
7.
6.
4.
3.
3.
0
3
4
7
0
•4 	 -
86
92
95
97
98
98
99
(100)
(100)
(100)
68
81
92
93
93
82
91
93
97
95
80
90
90
94
97
15
16
19
25
28
89
93
97
99
(100)

-------
 Continuous testing for extended periods was confined to experiments
 with sodium chloride feed at 1000 mg/1.   Stability of both product
 water flux and selectivity was good over the total test period of
 approximately three months.  By about the sixtieth day of operation,
 substantial flux decline had occurred,  especially with fiber samples
 annealed at lower temperatures (higher flux samples).   However,
 a thirty-minute low-pressure flush with 2% citric acid solution
 removed ferric hydroxide, accumulated from the test loop,  and
 restored performance.  Details of the sodium chloride test
 sequence are given in Table 2.

 The first solute of specific interest in advanced waste treatment to
 be investigated was the orthophosphate ion.  The feed solution was
 prepared from NaH^PC^ with phosphoric  acid added to adjust the feed
 pH in the range of 4. 0 to 7. 0.  Phosphate rejection varied from 96. 7%
 for fiber annealed at 75° C at a pH of 4. 0  to greater than 99. 9% for
 fiber annealed at 87° C over the whole pH range investigated.  The
 results obtained at a feed pH of 7. 0 are given in  Table  3.  Phosphate
 was determined colorimetrically by  spectrophotometric absorption
 at 705 mM- of the color developed with  ammonium molybdate.

 At lower pH,  there was a slight loss in selectivity to orthophos-
 phate  (especially for fiber annealed at 75  and 78° C) due to the
 existence of more undissociated phosphoric acid at lower pH.   Since
 the rejection is quite high for all samples, the effect of this loss of
 selectivity on rejection is quite  small.  However, the "salt reduction
 factor, " Cf/C  (Cf = feed concentration,  C  = product concentration),
 is sensitive to small changes in selectivity.  This is  illustrated in
 Table  4.  Similar trends were observed for all fiber  samples which
 were sufficiently permeable to orthophosphate to give a detectable
 product concentration.  Although the effect of pH on selectivity to ortho-
 phosphate was slight,  the fact that there was any effect  at all suggested
 that pH would be a very important variable in the case  of weak  acids
 and bases, such as, acetic acid and  ammonia.

 Another solute of specific interest in advanced waste treatment is
 the nitrate ion.  The standard 40:60  hollow fibers examined in this
 phase of the program  showed good selectivity toward sodium nitrate,
but less  so than toward sodium chloride.   Rejections for hollow fibers
 annealed at different temperatures are given in Table 5.

 The rejection of sodium acetate (as a model organic solute) was
determined in the  pH range 4. 0 to 6. 6. As expected, rejection was
a strong function of pH, ranging from  33% at pH 4. 0 to  95% at pH 6. 6


                               12

-------
                            TABLE  2

Dependence of Product Water Flux on Time for Cellulose Acetate Hollow
  Fibers Operating with 1000 mg/1 NaCl at 250 psi and Zero Recovery
                               Flux, gfd
3 ays
1
4
H
32
52
65
70
72
74
81
7
6
5
6
5
4
3-3-75
. 56 (86.
. 94
. 70
. 71
. 55
. 41 (74.
Samples
6
6
.81 (83.
.28 (83.

l)a




2)
cleaned
2)
8)
813-3-78
6. 31 (92. 5)
6. 15
5. 70
6. 28
6. 08
4. 82 (87.2)
with 2% citric
6. 37 (91. 8)
5. 94 (92. 8)
813-3-82
4. 37
4. 05
4. 03
4. 43
4. 59
3. 99
acid
4. 54
4.25
(96.4)




(92. 9)

(93. 6)
(94.2)
813
• 3.45
3. 51
3. 67
3. 82
4. 11
3. 67

3. 77
3. 57
-3-85
(97. 7)




C96. 9)

(97. 2)
(97. 8)
813-
2.
2.
2.
3.
T
(54f 2.
(62) 2.
(67)
(69) 3.
(75) 2.
74
75
46
00
89
71

00
75
3-87
(98.8)




(95.4)

(98.2)
(97.8)
  a   Rejection of 1000 mg/lNaCl.

  b   Operating days for Sample 813-3-87.

-------
                  TABLE  3

    Rejection of Orthophosphate by Cellulose
             Acetate Hollow Fibers
                                         _3
    250 psi, feed pH = 7. 0, 1600 mg/1 PO4
                        Product
Sample             Concentration, mg/1          Rejection, %
81




3-3-75
-78
-82
-85
-87
38
13
2
1
<1
97. 6
99. 1
99. 8
>99. 9
>99. 9
                 TABLE  4

    Effect of pH on Orthophosphate Rejection

           Sample 813-3-75, 250 psi


      pH           Cf/C            Rejection,  %


     7.0           42.0                97.6
     6.0           38.7                97.4
     5.0           36.8                97.3
     4. 0           30. 6                96. 7
                       14

-------
                           TABLE 5

                 Rejection of Nitrate by Cellulose
                     Acetate Hollow Fibers

                  1000 mg/1 NaNO3 at 250 psi

                   Sample        Rejection,  %

                 813-3-75            68
                       -78            81
                       -82            92
                       -85            93
                       -87            93
for the fiber sample annealed at 87° C.  The data are summarized in
Table 6.  To the extent that organic carbon in secondary effluents is
present in the form of short-chain carboxylic acids, the  results indi-
cate an operation pH near neutrality would be desirable.

The acetate ion feed solution described above also contained 50 mg/1
of ammonia.  Rejections  for this constituent are given in Table 7.
Selectivity in this case is also pH dependent,  but much less so, owing
to the predominance of ammonium ion in the pH  range examined.

The laboratory hollow fiber test series was also extended to exami-
nation of urea rejection.  Analysis was by the Kjeldahl method, data
supplied by the University of North Carolina Water Research
Laboratory.  The results are summarized in Table 8. Selectivity of
all fiber samples toward  urea was found to be poor,  a result which
appears to be characteristic of cellulose acetate membrane systems
reported in the literature also.

A similar test series on glycine solutions, representative  of some
amino acid constituents in waste water,  was also completed.  Analysis
was done by spectrophotometry of the glycine complex with cupric ion in
the UV spectrum (work done  at Chemstrand Research Center).  Results
                                15

-------
                            TABLE 6
            Acetate Rejection as a Function of pH for Cellulose
                    Acetate Hollow Fibers,  250 psi
                                            n
                               Rejection,  %
Sample

813-3-75
     -78
     -82
     -85
     -87
pH = 4. 0
25
28
28
t. 0 ?
33
5.0
69
74
75
76
77
6. 1
75
86
87
90
91
6. 6
82
91
93
97
95
        Analyzed as organic carbon by the University of North
        Carolina Water Research Laboratory, Chapel Hill,
        North Carolina.
                           TABLE  7

               Ammonia Rejection by Cellulose Acetate
                     Hollow Fibers, 250 psi
                                        Minimum-Maximum
                Sample                    % Rejection	

               813-3-75                      72-84
                    -78                      83-92
                    -82                      81-90
                    -85                      86-94
                    -87                      93-97
                            16

-------
                         TABLE  8

        Reverse Osmosis Performance of Cellulose Acetate
           Hollow Fibers with Urea Feed Solutions

         (Approximately 500 mg/1 feed at 250 psi and 26° C)
                         Run #1
Sample            Flux,  gfd        r,  %       Flux, gfd       r, %
Run #2
813-3-75              6.2            16           5.5            9
      -75              5.6            16           5.6           14
      -78              5.9            17           5.6           14
      -78              5. 6            22           5. 3           12
      -82              4.4            22           4.3           16
      -85              3.8            29           3.7           20
      -87              3.1            33           2.9           23
      -92              2.1            33           2.0           28

      Note:   The apparent variation in urea selectivity between
              the two  runs is unexplained.  Possibly, an error
              in feed concentration is responsible.
are given in Table 9.  Rejection is seen to be quite good, as might be
expected for a highly polar zwitterion compound.

Subsequent laboratory characterization of the hollow fiber test series
included runs with aqueous solutions of poly(ethylene glycol)-600 and
Triton X-100.  Flux-time data are given in Table 10.  The object of this
work  was to identify differences in flux decline which should be associated
with surface activity of the solute species.  In the PEG-600 runs, flux
decline was negligible over the  80-hour test period.  This solute pre-
sumably establishes an equilibrium concentration in the membrane (fiber
wall)  rather quickly,  which dictates subsequent transport of solute and
water in sustained operation. With Triton X-100, significant flux decline
was noted,  particularly with the higher-flux fibers.  As one might postu-
late,  the polar/non-polar asymmetry of the molecule in this case appears
to promote partial solubility of  the solute at the membrane surface, with
attendant accumulation and surface fouling.  Recent results 3 suggest
that such effects are  an important aspect of fouling in practical appli-
cations of reverse osmosis to waste water renovation.
                                17

-------
                      TABLE  9

     Reverse Osmosis Performance of Cellulose Acetate
         Hollow Fibers with Glycine Feed Solutions

     (Approximately 500 mgA feed at 250 psi and 26°  C;
                     pH 5. 8-6. 3)
     Sample              Flux, gfd            r,  %
813-3-75
813-3-78
813-3-82
813-3-85
813-3-87
1302-82
1302-85
1302-88
5. 8
6.2
4. 5
3. 5
2. 9
2. 2
1. 8
1. 0
89
93
97
99
99.
98
99
99.




6


6
The flux-pressure response in such systems is a more direct
measure of surface fouling and these effects were also examined.
Results are given in Table 11. Although there is some departure of
the flux-pressure response from linearity (a linear relation would be
expected, because of the negligible osmotic pressure of the feed),  the
effects are small and could easily be  obscured by normal compaction.

It appears from work completed to date that surface active  agents  con-
tribute to flux decline in our hollow fiber systems, but much less so
than anticipated from results on flat membranes.  We have  established
from detailed theoretical analyses of flow in hollow fiber systems, and
confirmed in practice,  that salt polarization is also very much less of
a problem with hollow fiber units.  Thus,  it is reasonable to conclude
that the secondary membrane built up by convection of surface active
agents to the primary membrane surface is less easily formed in hol-
low fiber units.  In subsequent laboratory work (cf. Section VI), however,
the effects  of surface active solutes on prototype modules,  operating
at more realistic end-use  conditions,  were  also investigated.
                                18

-------
                                       TABLE 10

               Flux-Time Response of Cellulose Acetate Hollow Fibers with
                                     Selected Solutes
Operating Time
    (hours)
   Control*
      1
      4
     80
     83
   Control*
      1
      2
      3
      4
     21
     23
Flux (gfd) for Various Fiber Samples
813-3-75 813-3-78
000 mg/1
6. 7
6. 3
6. 4
6. 2
6.2
500 mg/1
7.2
7. 0
6. 9
6'. 5
5. 8
6. 3
5. 6
813-3-82 813-3-85
1302-82
813-3-87
1302-88
PEG-600 at 250 psi
6. 3
6. 0
6. 1
5. 9
6. 0
Triton X-100
6. 7
6. 4
5. 4
5. 8
5. 3
5. 6
5. 0
5. 3
5. 0
5. 0
4. 8
4. 8
at 250 psi
4. 8
4.8
4. 7
4.8
4. 0
4. 3
3. 8
4.2
4. 0
4. 0
4. 0
4. 1

_
5. 6
5.3
3. 7
3. 1
3. 6
3.2
3. 4
3. 3
3. 4
3.3
3. 3

3. 5
3. 5
3. 0
3. 0
2. 9
3. 0
2. 7
3. 3
3.1
3.2
3. 1
3.2

3. 0
3. 1
3. 1
2. 9
2. 7
2. 8
2. 4
1. 18
1. 18
1. 18
1. 15
1. 13

1. 05
1. 11
1. 07
1. 18
0. 96
1. 03
0. 89

-------
       TABLE 10 (continued)
       Flux-Time Response of Cellulose Acetate Hollow Fibers with Selected Solutes
N5
o
       Operating Time
          (hours)
           25
           29
           46
           52
           69
           73
Flux (gfd) for Various Fiber Samples
813-3-75
813-
ise to 1000 mg/1
6. 0
5. 9
6. 1
5. 9
6. 0
6. 1
5.
5.
5.
5.
5.
5.
3-78
813-3-82
Triton X-100 at
4
5
6
4
4
3
4.
4.
4.
4.
4.
4.
0
1
2
2
2
1
813-3-85
1302-82
813-3-87
1302-88
250 psi
3.
3.
-
3.
3.
3.
3
4

4
9
4
3.
2.
3.
2.
3.
3.
0
8
0
9
0
0
2.
2.
2.
2.
2.
2.
7
6
8
8
8
8
1. 06
0. 96
I. 04
1. 07
1. 04
1. 07
              *  Run with deionized water at 250 psi.

-------
                    TABLE 11

Normalized Flux-Pressure Results for Cellulose Acetate
        Hollow Fibers with Triton X-100 Feed

                (500 mg/1 at 21° C)

Sample
813-3-75
813-3-78
813-3-82
1302-82
813-3-87
1302-88
Flux, gfd
at 250 psi
4. 97
4.88
4. 15
2.80
2. 69
0. 99
10 gfd/psi at various pressures
50
2. 12
2. 14
1. 66
1.18
1. 10
0. 37
100
2.22
2.22
1.85
1. 36
1.25
0. 49
150
2. 23
2. 11
1. 76
1.31
1.14
0. 41
200
2.02
1. 97
1. 64
1.34
1.09
0.39
250
1. 99
1. 95
1. 66
1. 12
1.08
0.40

-------
  V   DESIGN AND CONSTRUCTION OF HOLLOW FIBER  MODULES
A major objective of this program was to evaluate the effectiveness
of various hollow fiber module designs in wastewater service.   Four
basic module design concepts, illustrated schematically in Figure 2,
were evaluated in the laboratory; three of these were checked out in
field service.  In addition,  variations in  fiber packing density and in
arrangement of feed entrance ports were included in the basic designs
examined.  All designs involved feed stream pressurization on the
outside of the hollow fibers, with product water flowing through the
fiber bores.
Basic Assembly Techniques

Each of the four module designs represented in Figure 2 involves
a parallel bundle of hollow fibers, one or both ends of which are
embedded in an epoxy tubesheet.  The hollow fibers are wet spun and
must be maintained in the wet state throughout module assembly.
Accordingly, a number of procedures for construction of the units are
identical for all designs.

The  first step involves the assembly of as-spun,  five-filament,
hollow fiber yarn taken from bobbins  into hollow fiber bundles.  For
Designs 1, 2,  and 3 of Figure 2, this  involves winding hollow fiber
yarn onto a suitable mandrel, meanwhile  keeping the yarn  supply and
the bundle moist with water  sprays.   Design 4 entails a fabric  weaving
operation,  also done in such a way that the emerging fabric can be
kept moist.

With all four designs, in  the  course of bundle assembly, a secondary
or dam seal of water compatible, Dow Corning RTV ® bonds the fibers
together at  a point  just below the (subsequent) location of the epoxy
main seal(s).   This operation divides  the  fiber bundle into  two  (or
three)  parts:  the main,  active fiber portion and the end portion(s)
above the RTV seals.

The  next step involves treatment of the  fiber bundle end portion(s) with
100°  C water, to stabilize the outer fiber  diameter against later shrink-
age during end-drying.   Then,  the active fiber portion of the bundle
is immersed in gelatin solution and cooled to  solidify the gelatin.
                                23

-------
  1
       A
        m
 	"	I
Jfc
III III


rr
•
•
•
R
•
|
H
«
^
^
*
M
rr

                           ,	yzŁ\_ y^_	

        V
                                             1
I
Figure 2.  Schematic of basic hollow fiber module designs:  1-single seal
          looped fiber (Series D); 2-double seal parallel fiber (Series D-2);
          3-radial flow (Series R);  4-woven rolled fabric (Series W).
          Feed, product, and reject flows are indicated by F,P, and R,
          respectively, with optional combinations indicated by primes.
                               24

-------
This operation holds sufficient moisture in the active fiber bundle to
retain the reverse osmosis properties, yet immobilizes this water
from wicking along the bundle length.  Next, the water from the end
portion(s) is removed  by air drying and the necessary high pressure
tubesheets are cast, sealing the fiber bundle to the pressure vessel.
The gelatin may now be removed by warm water washing and the
bundle is ready for insertion into the rest of its pressure vessel.
Series D: Single Seal Looped Fiber

The basic design of Series D modules is that depicted in Figure 2,
Design 1. This is, perhaps, the simplest arrangement of hollow fibers
and served as a point of reference in our work.  In this concept, hollow
fiber  is laid down precisely onto a mandrel, to form a closely packed,
parallel  array of fibers looped at each end.  Packing factors (fractional
cross-sectional area of pressure vessel filled with fibers) can  range up
to 0. 70.  One looped  end remains in the final module.  The other is
severed  after the  tubesheet is cast to  expose the fiber bores for prod-
uct water removal.  Feed enters the bundle under pressure just below
the seal  area,  most conveniently by means of a Victaulic   side-entry
coupling. The feed stream flows axially through the bundle,  emerging
as concentrated brine from the end of the pressure vessel.  Product
water, passing through the wall of each hollow fiber, travels upward
along the fiber bore and exits at atmospheric pressure.  A typical
Series D design is detailed in Table 12, in this  case for Module  Number
D009, one of those utilized in field service (cf. Section VII).
Series D-2:  Double Seal Parallel Fiber

The construction of Series D-2 hollow fiber modules is indicated in
Design 2 of Figure 2.  A potential advantage of this design is the more
streamlined shell-side flow afforded by the elimination of fiber loops,
and the ready provision for reversal of feed flow.  Assembly is
similar to that employed for Series D units, except that tubesheet
seals are formed at both ends of the bundle (thus,  no looped fibers
remain) and feed/brine ports are cast into the tubesheets.  As indi-
cated in Figure 2,  inclusion of a side entry port in the middle of the
pressure  vessel permits operation in three different modes.   In all
modes, product water is removed simultaneously from both ends of
the unit.  Design details for Module Number D0013-2 are given in
Table 13.
                                 25

-------
                     TABLE  12

      Specifications for Hollow Fiber Module D009

      (Test conditions:  3000 mg/1 NaCl feed at 250 psi,
                       50% recovery, 70° F)
Hollow Fibers:

  Dimensions, OD/TD, microns                    300/100
  Number of fiber ends                             1 6, 000
  Fiber type                                40:60 CA, 88° C
  Packing density                                   0. 65
  Active fiber area,  ft2                               70

Pressure  Vessel:

  Construction                            316 stainless  steel
  Length,  in.                                         18
  Inside diameter, in.*                              1.85

Performance Under  Test Conditions:
  Product water flux, gfd                            3. 0
  Rejection, %                                      90
  Productivity, gpd                                  200
  Shell-side pressure drop, psi                       5
  Bore-side pressure drop, psi                      15
    *  Pressure vessel is nominal 2" pipe.  Dimension
       given is that for an insertable PVC liner encasing
       the fiber bundle.
                        26

-------
                            TABLE 13

            Specifications for Hollow Fiber Module D0013-2
Type:  Parallel fiber, with an RTV secondary seal and an epoxy primary
seal at each end; hollow  fiber bundle fits into metal pressure vessel and
is held in position by snap rings; brine-side seal is provided by an O-ring
at each end, fitted to a groove in the epoxy tubesheet; multiple brine
discharge ports penetrate the tubesheet at each end.
Hollow Fibers:   Approximately 16, 000; 300 micron OD, 100 micron ID;
40:60 cellulose acetate; packing factor 50%.
Module Dimensions:  2. 072" ID; 56" in length;  total membrane area
227 ft2.
Operation:  Feed enters pressure vessel through an inlet port situated
at midpoint; brine and product water discharge at both ends.  Alternate
operation is feed through one of the multiple tubesheet ports, brine
discharge at the other, and collection of product water at both ends.
A third operational mode involves feed through both ends with discharge
through the center port.


Performance:  Approximately 500 gpd product water capacity with 90%
rejection of 1500 mg/1 NaCl-feed at 250 psi and 70° F.   Initial laboratory
data (3000 mg/1 NaCl,  250 psi, 45%  recovery,  fed at one  end, discharged
at other):  1. 84 gfd overall flux; 92% rejection at feed end,  88% at reject
end; 2.21 gfd and 90%/90% at 300  psi.  Approximate shell-side and
bore-side pressure losses under cited laboratory test conditions  at
250 psi are 15 and 22 psi, respectively.
                               27

-------
An extensive laboratory investigation of Module D0013-2 was also
completed.  Measurements of flux and rejection were made as func-
tions of pressure, feed salinity,  and recovery.  The flux provided by
the module, about 2  gfd at 250 psi, or about 500 gpd, was less than
expected.  Otherwise, however,  performance under a variety of test
conditions was quite satisfactory. The three operational modes were
examined in the laboratory: co-current,  with shell-side feed intro-
duced at the center of the cylindrical pressure vessel,  brine discharge
and product water discharge from each end;  counter-current, with
feed introduced at each end and brine discharged from  the center;
unidirectional, with  feed introduced at one end and brine discharged
at the other (center port inoperative).  The latter mode is co-current
with respect to approximately half the  fiber bundle length and counter-
current with respect to the remaining length. Because of shell-side
pressure drops and progressive increase in steady-state salt concen-
tration along the length of the bundle during normal operation at
recovery, this mode necessarily provides a different flux and rejec-
tion (water quality) at one end of  the module  than at the other.
Different product water concentration levels at opposite ends of the
unit were,  in fact, observed in the lab checks with unidirectional flow.

Laboratory  results for unidirectional flow, with combination of
product and brine streams, may  be summarized as follows:

   (a)  Flux was linear with pressure up to  300 psi, using
        water, 3000 mg/1, and 5000 mg/1 NaCl  feeds.  Inter-
        cepts  on the pressure axis approximated the osmotic
        pressures of the feeds.

   (b)  Rejection of  NaCl  increased  with increasing pressure,
       appearing to level off at about 90%.   An  asymptote less
       than 100%, according to the solution-diffusion model,
       would  result from leaks (through which contributions to
       both salt and water flux are pressure dependent).  In our
       case,  it is quite likely that slight leakage past the O-ring
       pressure seals contributed.

   (c)  Flux as a function of recovery, from 10% to 80%, passed
       through a shallow  maximum.  This result is quite reason-
       able and is attributable to decreased average shell-side
       pressure (at low recovery, where pressure loss is  signifi-
       cant) and to decreased driving pressure (large osmotic
       pressure contributions at high recovery).
                               28

-------
    (d)  Process rejection as a function of recovery,  determined from
        product and feed concentration, decreased with increasing
        recovery.  Correction of process rejection for average,  local
        salt concentration and for normal decrease accompanying
        decrease in flux, yielded a  fairly constant "intrinsic"
        rejection of about 90% for the entire recovery range 20% to
        80%.

Following the lab checks, Module D0013-2 was readied for field ser-
vice at the Chapel Hill field installation (cf. Section VII).
Series R: Radial Flow

As illustrated by Design 3 of Figure 2, the radial flow arrangement
involves feeding a pressurized stream through the center of a hollow
fiber bundle by means of a porous feed tube.  Potential advantages in
waste water treatment appeared to be superior flux and rejection
available through minimization  of flow channeling.  Feed flows  radially
outward, through the bundle,  collecting at the outer surface of the
bundle and discharging through  a port located midway along the
pressure vessel.

As with Designs 1 and 2,  assembly of the radial-flow hollow fiber
bundle entails winding the hollow fiber yarn onto a mandrel to form a
parallel array of hollow fibers.  In this design,  however, the winding
is done about the porous feed tube, which therefore becomes a perma-
nent part of the assembly.  RTV and epoxy seals are formed at both
ends (a single seal,  looped fiber arrangement is an alternate design).
The fiber bundle may be consolidated at intermediate stages of  the
winding by wrapping it with a porous fabric.  The fabric also becomes
a permanent part of the unit.

Several radial-flow modoules of 200-500 gpd capacity were constructed^
during the course of this program and one was specifically evaluated
for use in waste water service.  Laboratory performance was quite
satisfactory.  On the recommendation of EPA personnel, however,
who had experienced severe fouling problems with this design,  no
radial-flow units were actually  operated with secondary municipal
effluent. Instead, efforts were  directed to more promising designs.
                                29

-------
Series W:  Woven Fabric

The final hollow fiber concept evaluated in this program was the
woven fabric,  Design 4 in Figure 2.  Our objective in considering
this arrangement was to construct  a module more resistant to
fouling by decreasing the fiber density and enlarging the water
passages throughout the fiber bundle.  The method involved weaving
a fabric on a hand loom, using cotton twine as warp filaments,  and
hollow fiber yarn as fill material.  When rolled and assembled into
a working module, the hollow fibers end up running axially and
parallel, as in the D-Series design (Design 1, Figure 2), but spaced
and positioned by the (inert) warp filaments running circumferentially
through the bundle.

Initial work involved evaluation of weaving parameters and materials
and testing of  small  fabric samples under reverse osmosis conditions.
On the basis of the preliminary work,  detailed design parameters
for a  demonstration  module (WA  series) were specified.  These are
summarized in Table 14 and Figure 3.  Module  components required
are given in Table 15.   The cotton  warp element was selected
primarily for  convenience; nylon or other synthetic  filaments of
appropriate characteristics would be more suitable  for long-term
waste water service.

Two 4" x 6" test samples of hollow fiber fabric  were loosely rolled,
potted in epoxy,  and assembled for standard evaluation of reverse
osmosis properties  on our test board.  Figure 4A shows one such
test sample.  Initial performance of both was very good: 4. 8 gfd
with 93. 2% rejection and 5. 2 gfd  with 95. 4% rejection, under test
with 3000 mg/1 NaCl feed at 250 psi and zero recovery.   Performance
on both samples subsequently declined, but this was believed due to
kinking of the  (unsupported) fabric.  As the latter problem  would not
arise  in WA series modules, we  interpreted the test board results
as evidence that fabric  preparation does not harm the hollow fibers
and that the design proposed should therefore be practicable.

Work  next proceeded on preparation of the five-foot length of fabric
for demonstration module WA001.  Weaving was done on a hand loom,
using  a manually inserted shuttle.  The 50:50 acetone:formamide fiber
was employed.  To keep the fibers  wet,  it proved simplest to submerge
the entire loom in a  shallow pan and weave the fabric under water.
Weaving time per  one-foot length of full-width fabric was approximately
21/2  hours.   (The process could,  of course, be automated and vastly
                               30

-------
                           TABLE  14

      Design Parameters for WA Series Hollow Fiber Modules
General Description:  A reverse osmosis unit with a woven fabric fiber
    bundle incorporating cellulose acetate hollow fibers aligned for
    axial shell-side flow;  looped fiber configuration with single end
    seal; stainless steel hardware.

Hollow Fibers:  Standard  40:60 cellulose acetate,  300 micron OD,
    100 micron ID,  zero-length flux of 4 gfd with 97% rejection
    of 3000 mg/1 NaCl at 250 psi and zero recovery; pressure
    capability to 300 psi.  Alternate fiber is 50:50 cellulose
    acetate, 300 micron OD, 100 micron ID,  zero length flux
    of 3 gfd with 97% rejection.

Fiber Bundle:  Incorporates a 5-foot length of annealed, 15-inch width
    fabric involving closely packed, five-filament hollow fiber fill
    yarn woven with 30 mil  diameter soft cotton twine warp spaced
    at 0.2-inch intervals.  Fabric thickness in  single layers is
    approximately 70 mils;  in wound layers, approximately 45 mils.
    The (experimentally observed) fiber count is such that 85 feet
    of single  filament hollow fiber is consumed per inch fabric width
    per foot fabric length.  The fabric is wound on a solid nylon
    mandrel such that the hollow fibers are co-axial with the mandrel,
    overlaid with a protective wrap, and sealed to the mandrel with
    an RTV dam at one end.  Protruding, looped fibers at this end
    are  sealed in an epoxy tubesheet and cut to  provide the product
    water outlet.  Of the 15-inch fabric width (bundle length),  3 inches
    is allotted for both seals and end trimming, leaving an active
    length of 12 inches.  Fabric as woven is prepared on a 16-inch
    loom (as-woven width is 15 1/2 inches), allowing for shrinkage
    of 1/2-inch on annealing.   Total active  fiber area is 16. 6 ft  .

Pressure Vessel: Nominal 2" stainless steel pipe fitted with a tapped,
    threaded pipe cap at the b-rine discharge end,  a pipe nipple en-
     closing the bundle tubesheet,  a PVC product water cap, and a
     Victaulic side entry coupling for the feed stream.

-------
TABLE 14 (continued)
Design Parameters for WA Series Hollow Fiber Modules
Module:  Operation is with counter-current,  axial shell-side flow.
    Design productivity is 50 gpd (3. 0 gfd) with 95% rejection of
    3000 mgA NaCl at 250 psi and 50% recovery.  Bore-side
    pressure drop under  test conditions is approximately 30
    psi; shell-side pressure drop is small.  Local fiber packing
    density (within fabric bundle) is approximately 0. 26;
    module area utilization is about 1000 ft  of membran
    per ft  of active pressure vessel.
                             32

-------
Ł*•''
       (A)   FABRIC  CROSS -
              SECTION  DETAIL:
MANDREL DETAIL:
                                                                         *-o.oao nope
                                                                           WART* SCOTS
                    Figure 3.  Engineering design of WA Series hollow fiber module

-------
                         TABLE  15
     List of Materials for WA Series Hollow Fiber Module
    Item
Hollow fiber


Warp fiber

RTV

Epoxy

Pressure vessel



Tubesheet nipple


Brine discharge seal


Coupling


Product water cap


Bundle mandrel

Cartridge liner


Discharge port fittings
Quantity
Description
1320 ft    As-spun,  5-filament hollow CA fiber
            yarn

 410 ft    30 mil diameter soft cotton twine

  2 oz    Secondary seal

  5 oz    Tubesheet seal

   1      Nominal 2"  (2. 067" ID) 316 stainless
           pipe, 14" length, one Victaulic
           grooved, one-threaded end

   1      Nominal 2"  stainless pipe, 3" length,
           one  Victaulic grooved, one-threaded end

   1      2" 316 stainless pipe cap,  drilled and
           tapped for  brine discharge tube

   1      2" side-entry Victaulic coupling,
           Style 72

   1      2" PVC threaded cap,  drilled for
           product discharge tube

   1      1" diameter nylon rod, 14" length

          4'  length of  2" wide Cerex '  spun-
           bonded fabric,  6 mils thickness

   2      Brine discharge pipe and product
           water pipe, sized to  suit test
           facility
                               34

-------

   (C)   fabric for demonstration
       Tnodule, rolled and ready
        for annealing
(B) 15" x 60" fabric for
    demonstration module
(A) 4"x6"
     rolled
     fabric
     test
    sample
 Figure 4.  Construction of woven fabric hollow fiber modules '.

                              35

-------
accelerated. )  A section of  the first five-foot fabric sample is shown
in Figure 4B.   The loosely  rolled bundle,  ready for hot water
annealing,  is shown in Figure 4C.

Assembly of the module involved attaching and rolling the annealed
fabric onto a mandrel and,  at the  same time, forming a 3/4" seal of
RTV cementing the fabric roll to the mandrel.  A final wrap of Cerex
consolidates the bundle. End-treatment,  gelatinization,  end-drying,
and potting of the tubesheet proceeded as with conventional looped
fiber  designs.   (The warp strings were removed from the tubesheet
end prior to potting. )

Module WA001  was assembled with its pressure vessel and laboratory
tested.  Results of the laboratory check are compiled in Table 16.
They  were very satisfactory,  and within performance specifications
for this hollow  fiber. A photograph of the assembled module, ready
for field testing, is shown in Figure 5.
                            TABLE 16

      Laboratory Reverse Osmosis Performance for Hollow Fiber
                        Fabric Module WA001

                     (3000 mg/1 NaCl feed at 70°F)
                         50:50 A:F CA fiber

Test Day
1
2
2
2
Pressure
psi
135
125
205
250
Recovery
%
20
20
20
20
Flux
gfd
0. 81
1. 0
2. 0
2. 6
Productivity
gpd
14
18
33
43
Rejection
%
93. 6
93. 3
95. 1
96. 3
Following the successful construction of woven fabric Module WA001,
and its subsequent excellent performance in field service (see
Section VII),  a second unit, WA002,  was constructed.  Design
specifications were identical with those for the first unit, except that
                                36

-------
Figure 5.  Photograph of woven fabric module WA001
                       37

-------
WA002 was assembled with the 40:60 cellulose acetate fibers used in
all other modules examined in this program and, as a result of using
different equipment, fiber count increased from 85 to 105 feet of
single filament per inch-foot. Assembly of WA002 was done on  a
slightly more sophisticated loom,  with foot operated heddle  and
manual  shuttle,  shown in Figure 6.  Total active fiber area  was
17. 7 ft  .  Laboratory performance is recorded in Table 17.

Evaluation of the woven fabric concept completed our investigation
of hollow fiber module designs.  It was apparent from all data collected
that this design was the most successful in  waste water service.
Minimization of dead volume occupied by the center mandrel,  sub-
stantial improvements in efficiency of fabric weaving, and module
scale-up to large diameters were  all deemed feasible.
                        TABLE 17

     Laboratory Reverse Osmosis Performance for Hollow Fiber
                      Fabric Module WA002

                    (4400 mg/1 NaCl feed at 70° F)
           Pressure   Recovery  Flux   Productivity   Rejection
Test Day      psi          %        gfd         gpd            %

   1          250         40       4.9         87            87
   2          250         13       4. 3         76            93
   2          250         35       4. 3         76            90
   2          300         23       5.2         92            92
   3          250         20       4.0         71            92
                             38

-------
Figure 6. Le Clerc, Medica Model 22-4 loom
          used in assembly of Module WA002
                      39

-------
           VI  INVESTIGATIONS OF MODULE FOULING
 Laboratory Performance with Surface-Active Agents

 To examine more closely the results which had been obtained on the
 small test loop,  some of the work conducted before was repeated with
 a full-size, D-series module, D008.  This work was considered the
 more important because test loop results had not indicated an effect
 as  pronounced as it was with flat membranes.  The discrepancy had
 been particularly noticeable with the  cationic surf ace-active agent,
 benzalkonium chloride. Therefore, new experiments were performed
 with a solution which contained approximately 500 mg/1 of dissolved
 Hyamine 3500 ®  '(Rohm and Haas).  This product is an alkylbenzyl-
 dimethylammonium chloride,  in which the alkyl group is in the  range
 10-16 carbon atoms long.

 The apparatus used for this test was  similar in principle to the small
 test loop, but considerably larger; liquid pumping was performed by
 a Goulds multi-stage centrifugal pump which operated at 3430 rpm.
 Its output was sufficiently large that a substantial part of the circulated
 liquid stream had to be by-passed, rather than sent through the module
 if measurements were to be made at  a reasonably high recovery.  The
 system  operated as a closed loop; except for samples removed, all
 circulated liquid was returned to the  starting bath.

 Results obtained with the benzalkonium chloride are indicated in
 Table 18. Basically, they show that  there is a slight,  but definite
 fall-off  in flux with time  which takes place mainly in the early
 stages.   The effect,  however, is considerably less pronounced  than
 it is with flat membranes.

 A second test was performed with this solute under modified conditions
 the whole test being  conducted at constant pressure (250 psi) and flux
 being measured at fixed time intervals.  Since previous work had
 indicated that the greatest effect took place in the early stages of
 operation, this test was run only for  a few hours.  In  this case,
•temperature-corrected fluxes showed only a slight change with  time,
 as  shown in Table 19.   This was not only slight,  but indecisive  as
 well.  It was concluded that this particular solute, per se,  caused
 no  specific fouling effect of appreciable magnitude.
                                 41
                                        AWBERC UBRARV „

-------
                 TABLE 18

  Performance of Hollow Fiber Module D008  with
          Benzalkonium Chloride Solutions
Pressure,
psi
100
150
200
250
300
300 (1 hr
later)
Flux, Rejec
.g*
1.
2.
2.
3.
3.
3.

•d 	
49
08
79
32
92
92

tion,
Flux/ pressure %
0.
0.
0.
0.
0.
0.

01
01
01
01
01
01

49
36
395
34
306
306

98.
95.
95.
97.
97.
97.

6
8
0
0
2
0

                TABLE 19

  Flux-Time Relationship with Benzalkonium
         Chloride Feed at 250 psi
Time (hr)

  Start
  0. 67
  1. 67
  2. 67
Flux, gfd,  corrected
 for temperature

        3. 91
        4. 08
        4. 08
        3. 77
                     42

-------
An effort was then made to determine whether such an effect was
demonstrable for an anionic surface-active agent (sodium alkylaryl-
sulfonate).  In this case, severe foaming was observed in experi-
mental work using a 500 mg/1 solution, but,  once more,  the results
were of such a nature as to suggest that no significant effect was
operative.   Data are given in Table 20.
                        TABLE 20

           Flux-Time Relationship with Sodium Alkyl-
                Arylsulfonate Feed  at 250 psi

                                  Flux,  gfd,  corrected
         Time (hr)                   for temperature

           Start                          3. 83
           0.5                            3.10
           1.5                            4.28
           2.5                            4.18
These results prompted us to make a theoretical study of the probable
effects to be expected in hollow fiber systems.  From analysis of flow
behavior in such systems,  it became clear that polarization^ phenomena
are usually much less important in hollow filament systems than they
are with flat membranes.   Thus, the build-up of a "secondary mem-
brane, " consisting of a  layer  of the fouling solute, is far less likely to
occur in hollow filament systems, and our failure to observe a very
pronounced effect was a predictable result.

Laboratory Simulation of Fouling

Gelatin was adopted as a typical fouling water component because it is
representative of a class of materials which may well be found in
waste waters--partially degraded proteins.  Tests with this solute were
performed with a D-series module  (D008), using the same  test loop
mentioned above.  Fouling was caused by adding a gelatin solution to  a
sodium chloride solution which was already being circulated.  Results
obtained are indicated in Table 21;  it is interesting to note  that no
great effect was observed at the 1000 mg/1 level, but that a very
                                  43

-------
                                  TABLE 21

                       Analysis of Flow Data for Module D008
    Investigation:   Intentional fouling of a 2", standard looped fiber unit, 65%
    packing density.  Experiment involved:  (1) test with water/Nad; (2) foul
    with gelatin/Nad; (3) flush with water/Nad; (4) flush with Biz®; (5) check
    with water/Nad.  All tests were with 1000 mg/1 NaCl  at 250 psi and 21°  C.
    Gelatin feed concentration was 1000-2000 mg/1.
               2          3          4          5          6                7
                                            mg/1 in   mg/1, corrected  mg/1, corrected
Condition % Recovery  gfd Flux  % Rejection  product  for recovery        for flux _

  (1)           8         3.70       88         119          114             114
              48         3. 70       79         210          141             141

  (2)          47         3.31       85         146          101              94
              64         2.34       63         370          198             125
              53         1.88       72         280          178              90
              53         1.75       70         300          191              90
              42         1.13       34         650          478             146

  (3)          37         0.69       11         860          666             124
               7         0.71       37         610          590             113
  (4)          Biz flush
  (5)          45         3.48       82         182          130             122

-------
          TABLE 21  (continued)
          Analysis  of Flow Data for Module D0008
-p-
Ol
Column 6   Corrected rrigA is observed mg/1 times 2/(I + 1/(1-R) ),
            where R is decimal recovery.   (Correction assumes
            a linear increase in axial salt concentration from feed
            end to reject end. )

Column 7   Corrected mg/1 is Column 6 value times observed flux,
            divided by reference flux (3. 70).

-------
pronounced fouling effect was observed at the 2000 m.g/1 level.  More-
over, this effect increased with time.  Also, this fouling was evidently
not readily removed by simple replacement of the gelatin solution
with tap water.  On the other hand, the fouling was removed very
efficiently by treating the module with a Biz ^ solution which had  been
adjusted to approximately pH 7.   (Biz is an enzyme-deter gent product
manufactured by the Procter & Gamble Company.  Use of this prod-
uct does not constitute an endorsement or recommendation.)

These results confirm that fouling with purely proteinaceous matter
can be readily removed by treatment with suitable cleaning agents,
particularly those which contain enzymes,  and should not represent
a serious long-term problem.
Analysis of Fouling in Field Service

In contrast to laboratory results, fouling of experimental modules
in field service showed quite another pattern in operation on
secondary effluent from the University of North Carolina Waste-
water Research Center, a pilot plant installation.  It should be
emphasized that all secondary effluent processed had been filtered
before circulation through the modules; this filtration was made
increasingly thorough with  the passage of time, as it became
apparent that a severe fouling problem was present.  Initially,  an
opportunity was provided for some  settling to take  place in the
supply water; and, after settling, the water was passed through  a
50-micron Cuno filter.  It soon became apparent that little real
settling was taking place, and the settling tanks were replaced with
a sand filter (Swimquip Model FRP-20,  20" diameter,  high-rate
sand filter,  designed primarily for swimming pools),  while still
retaining the 50-micron cartridge filter.  Subsequently, a 5-micron
filter was placed in line with the 50-micron filter,  as well.

In spite of the efforts  made  to ensure thorough removal of suspended
material, operation of all modules  except the W series was charac-
terized by a steady drop in product water flux which could be only
partially restored by systematic Biz treatments.  In all cases, we
ultimately found it necessary to remove  the module from service
and subject it to a radical cleaning treatment,  which consisted of
completely  removing the hollow filaments from the housing and
physically removing the fouling material.  A portion of this
material was greasy in nature.  It may have resulted either from
                                46

-------
oil leakage in the pumping system of the Wastewater Research
Center (a known occurrence), from incomplete degradation of
animal or vegetable fat in the primary and secondary treatment
stages,  from other hydrocarbon oil  present in the wastewater,
or from a combination of some  of these.  Since repeated Biz
treatments seemed to have been relatively ineffective in
removing it,  we believe that hydrocarbon oils probably  account
for much of the trouble.

The greasy component of the fouling material, however, was only
a relatively minor part of the whole mass.  The greater part of
it consisted of a solid substance, part of which was amorphous
and evidently organic, and the remainder of which was a roughly
crystalline solid, light brown in color, which appeared  to be in-
organic.  Several sets of analyses were performed on this solid,
both in our own laboratories and by  personnel of the Robert  A.
Taft Water Research Center, Cincinnati, Ohio.  These  invariably
showed a high content of inorganic substances  (residue on ashing)--
usually in the range 67-75%.  However,  the analyses by the
Robert A.  Taft Water Research Center personnel also showed a
fairly high chemical oxygen demand, which is presumably
attributed to the organic components of the solid.  The nature of
the inorganic component has not been ascertained.   Quantitative
analysis has shown that the major cation is calcium (16. 6%).
Emission spectroscopy showed  the presence of a number of  other
metals, but not in large amount (Fe, 0. 7%; Mn,  0.2%; Na, 0. 2%).
Analysis for CC^ showed only 0. 5%  in the original sample, or
1. 8% in the ash.  (Calcium carbonate requires 45. 5%.)  Similarly,
only 1. 18% sulfur was found present (much too low for calcium
sulfate).  However,  a phosphorus content of approximately 15. 4%
was determined.  This is  too low for tricalcivtm phosphate (20. 0%),
which might not be stable under the  prevailing pH conditions in  any
case (pH 6. 8).  It is also too low for CaHPCs, which is a con-
ceivable constituent (P,  22. 8%). The calcium analysis  is about
right for Ca(H2PC>4)2 (Ca, 17. 1%), but this substance is moderately
water-soluble, and would not be expected to accumulate in the
module,  even under conditions  of extreme polarization.  More-
over, the phosphorus analysis is too low (theory for calcium
superphosphate,  26. 5%).   The atomic ratio of phosphorus to cal-
cium is apparently close to 1:1  (actually, 5:4).   This rules out
the assignment of this material as hydroxyapatite,  Ca5(OH)(PC>4)3,
which had been suggested by some of the personnel at the University
of North Carolina Wastewater Research Center. Possibly analysis
                              47

-------
of carefully extracted solid, containing only the inorganic portion,
will be more fruitful.

Attempts to minimize or eliminate this organic or inorganic
fouling by regular and systematic cleanings with Biz were largely
unsuccessful.  Controlled chlorination of the feed and adjustment
of its pH to 5. 5 also proved ineffective.  For some reason,
modules of the W design appeared to be much less subject to this
type of fouling; the reason for this difference is not  known with
certainty, but it appears that it must be related to the uniform,
consistent flow profile which  is maintained in such modules,  with
very limited opportunity for stagnant regions of high local super-
saturation to develop.

There is no reason to assume that the  particular design used in
this demonstration module is optimum.  Further work along the
lines  suggested by it is strongly indicated.   Work should include
designs providing for higher overall productivity, as well as  even
more uniform flow characteristics.

When we first became aware  of the fouling problem  in field work,
our initial assumption was that fouling was the result of bacterial
multiplication in the module and connecting lines. To some
extent,  this was probably true.  In efforts to eliminate or control
fouling from this cause, we adopted measures  which should have
eliminated bacterial growth.  First,  a simple chlorinator of the
type used to treat home swimming pools was installed.  This was
a cartridge which contained large lumps of trichlorotriazinetrione
(trichloroisocyanuric acid).   It was installed in such a way that it
constituted a by-pass loop in  parallel with the sand filter, so that
the pressure drop across the filter created a driving force to move
liquid through the chlorinator.  Flow was then  throttled  with a
simple needle valve,  in such  a way that the chlorine content of the
feed water  was adjusted to 0. 5-2. 0 mg/1, as determined with a
chlorine test kit.

While this treatment undoubtedly did some  good in reducing bac-
terial multiplication, it did not solve the basic difficulty.  Module
fouling continued at about the same rate, as judged by the rate of
flux decline with time.
                              48

-------
Attention was then turned to pH adjustment, to see if a different
pH would help prevent the deposition of inorganic  solids.   Cellulose
acetate hollow filaments  can be used down to a pH of about 4. 0, but
this pH is at the threshold of polymer degradation, so a pH of 5-5. 5
was sought.  An experimental titration  of the feed water showed
that it had an appreciable buffering capacity, so that a definite feed
of acid was required to reduce the original pH  (6.  8) to the desired
level.   In practice, we settled upon a system in which 0. 5 molar
sulfuric acid was fed into the feed water line from a large reser-
voir by using a  perisaltic pump which operated continuously, at a
delivery rate of 3. 5-4. 0  ml/min.  The  rate of  water flow  through
the system at the time was approximately 2100-2150 ml/min.
This system proved somewhat less effective than  we had hoped,
since the Tygon capillary bore tubing used in the pump hardened
with use and sometimes failed to deliver acid.  At about this time
we discovered that the chlormation system itself was reducing the
pH of the feed water to about 5. 5, and therefore there was no
real need to continue with the acid feed.

The use of the acid pretreatment did not seem  to affect the rate
of fouling greatly, either.  While no measure adopted ever fully
solved the fouling problem,  the best set of conditions which  we
were able to work out included the following procedures:
(1) thorough filtration of the feed water (sand filter, followed
successively by a 50-micron Cuno filter and a  5-micron Cuno fil-
ter; (2) relatively high liquid flow rate through the system, to
minimize opportunity for purely physical fouling;  (3) chlorination;
and (4) acid treatment.   These measures were supplemented,  of
course, by regular Biz cleanings (at least once a  week, some-
times  twice), regular change  of filter cartridges,  and systematic
backwashing of  the sand filter (daily).   Only the woven fabric
module did not  require Biz treatment.

Typical performance characteristics of the various modules with
respect to fouling, as indicated by flux-time variation, are  shown
in Figures  9A-13A of the following Section VII.   These show quite
graphically, the type of behavior which was characteristic of both
single-ended and double-ended modules; Figure 13A shows the
performance of the woven fabric module tested; the improved flux
stability with time is apparent.
                                 49

-------
Fouling by inorganic solids appeared to be the greatest single  fac-
tor in causing decline in module efficiency.  Such accepted measures
as careful filtration, Biz treatment, chlorination,  and pH adjust-
ment had only limited effect.   Thus, one may conclude that such
factors  as bacterial multiplication, slime deposition,  etc. , were
not decisive in this particular operation.  On the other hand,
deposition of solids  whose formation was not affected by these
procedures appeared to take place steadily, ultimately requiring
removal of the module for thorough clean-up.  For the most part,
these solids appeared to be inorganic.
                              50

-------
              VII  FIELD RESULTS
Field Test Facilities

The wastewater used in our experimental work was taken from the
pond into which secondary effluent from the University of North
Carolina Research Center trickling filter ran, by simply tapping
off from a line which supplied such water for test purposes.  Over
the period during which we determined content of various sub-
stances  in the feed stream (filtered) to our module, the mean con-
tents  of  certain key solutes were as  shown in Table 22.

As noted previously, the pH of the feed water used in this work
was about 6. 8.  The water itself did not normally contain visible
suspended particles; nevertheless,  it had a distinct grayish
opalescence which was due to very finely  divided suspended
particulate matter characteristic of trickling filter effluent.  In
early work with the sand filter,  turbidity  measurements were
used to establish  the frequency requirements for backwashing;
turbidity of the sand filtered effluent equal to about 15 Jackson
Turbidity Units (JtU) was considered marginally satisfactory.
This value was reached in 1-2 days after  each backwashing
operation.

In spite  of the fact that visible suspended  solids were normally not
present  in the secondary effluent, deposition of black solid  matter
began to take place  in the connecting lines very soon after start-
up, and  this was particularly noted near the 50-micron filter.   It
was normally necessary to replace  the cartridge in this filter at
least  once per week, sometimes more of ten--otherwise pump
starvation began to  take place.   Moreover,  during  the period
when  settling tanks  were being used, the formation of floating
solid material in  these was quite evident.   Initially this was
assumed to be due to bacterial growth; however,  it seems very
likely that a part  of it was also due  to  simple  agglomeration of
finely divided solids present in the water.  This phenomenon was
one of the reasons for eliminating the  settling tanks; the princi-
pal other reason being that no great  amount of settling ever
appeared to take place in any case.
                                 51

-------
                          TABLE 22

      Content of Certain Solutes in Filtered Secondary Effluent
     University of North Carolina Wastewater Research Center


Solute                   No.  of Observations          Content, mg/1

BOD                          47                      35-36
Kjeldahl nitrogen              42                      20-21
Na+                          54                      35-36
Cl-                           56                      36-37
Organic  carbon                22                      48-49
Inorganic carbon              22                      33-34
Phosphate (total)              17                       6-7
NO2 '                         *                        ca. 0. 01
NO ~                         **                      ca- °- l
   O

        *  Usually not even reported.
       **  Data extremely limited.  Occasionally ran
           as high as 1. 1 mg/1.
The test assembly which was used in evaluating module performance
is shown in Figures 7 and 8.  It was originally designed to provide
for operation of two modules simultaneously.  Except for a brief
period when both module D009 and module D0013-2 were operating,
the entire pump flow was routed into a single module.

Secondary effluent was  taken from one of the final clarifiers  of
the Wastewater Research Center by tapping on to a supply line
already in existence.

The test assembly did not operate as a closed loop in the  ordinary
sense.  Water  was  taken from the clarifier, processed through the
test unit, and continuously returned to the reservoir (both product
water and brine), samples being withdrawn only as needed.  The
solute content of the clarifier reservoir changed continuously,
                              52

-------
                                                                                       pressure
                                                                                        gauge
  TO
WASTE
                  back
                pressure
                regulator
         flow meter
               reject stream
               TO
            WASTE
r
'
back
f pressure
\ regulator
"4, /
wi y
ter
mple for ,
stream #2 /


pressure
gauge
9 :
1
I/ *
i ^
d
-d


pressure
gauge C
>
•S

> !
\ '
/ ^
S^^»
, 1 "
accumulator
9 _
               TO
            WASTE
              TO
            WASTE
                     flow meter
                                                                                                                          SECONDARY  EFFLUENT FEED
                                             T
                                                                                                               chlorinator
                                                     pressure
                                                      gauge
                                                      o
                                                                                               accumulator
   \ '
   / v
   ^
product
                               ressure
                                pump
                                                                                                   :artri
                     flow meter
                                                                                                   cartridge
                                                                                                    filter
                                                                                                                   feed       feed
                                                                                                                 pressure  temperature
                                                                                                                 recorder  recorder
                                                                                                                                                        110 AC
                                          product #1
                                                          Figure 7.   Schematic of field reverse osmosis  test  loop for hollow fiber units


                                                                                       53

-------
                                    >wr-o
Figure  8.  Field unit for testing of hollow fiber modules with
            secondary municipal effluent
                                54

-------
owing to influx of water from the trickling filter and discharge of
overflow into.nearby Morgan Creek.

Reference has already been made to settling tanks which had been
originally installed in the feed water line; these are not shown in
the schematic sketch of Figure 7,  since they were removed at an
early date.  The idea behind them  was that, since particulate
matter  occurs in the feed water (although almost colloidally
divided),  it should be possible to remove at least part of it by
settling.  When they were in use, the procedure was as follows:
supply water entered a 200-gallon  polyethylene tank via a simple
valve which was opened and closed by a float;  this provided a fairly
constant water level near the top of the tank.  A connecting line was
provided from this tank to a second one,  passing through a wide-
bore solenoid valve which was controlled by a simple clock motor
and adjustable cam operating against a microswitch. The "on"
period of the microswitch was so adjusted that the second tank never
became completely empty.   Feed for the test system was taken from
a point  somewhat above the bottom of the second tank.  In operation,
this system proved unwieldy and ineffective; no great amount of
settling appeared to take place, but much floating insoluble  matter
was soon seen.  This may have been either a product of bacterial
multiplication (which was known to occur to some extent within the
lines) or a product of  simple agglomeration of particulate matter
already present.  In any case, the size and complexity of this
portion of the assembly was steadily reduced, and it was finally
eliminated altogether.

The bacterial multiplication alluded to above was evidenced by the
development of a foul  odor in the water from the module test
system at points where there was any opportunity at all for stagna-
tion.  There was also evidence of gas pressure development; in
early operation this pressure had to be bled off regularly.   Later  on,
means were found to combat it; the most effective ones seemed to
be relatively high liquid velocity through the system (operation at
low recovery), and chlorination of the feed.  The effect of a pH
change  was less conclusive, but may have contributed to reducing
the severity of this problem.  It is to be noted that these measures
were  relatively effective  against anaerobic bacterial multiplication,
but were far less effective against the fouling problem which has
been mentioned above.
                                 55

-------
A summary of field experience with hollow fiber  modules of three
designs is given below.  Design specifications and laboratory per-
formance for  each were described in Section V.   Unless otherwise
noted, field operating pressure was 250 psig and all flux data are
corrected to 70° F.  Graphical records of field results are contained
in the set of Figures  9A through 13E.
Series D Modules
Module D009:   The first module tested was of the single-end type,
previously described,  incorporating parallel,  looped fibers within
a housing.   This module was installed on 2 March 1971,  and was
operated over a period of 108 days, including  a total of some 3. 5
weeks down-time for cleaning.  In its initial running period of 29
days, its performance was  characterized by a steady decline in
flux,  with brief improvements following each treatment with Biz.
Initially these treatments at times  gave a higher product flux
than was obtained at start-up; however, the response gradually
became less significant.  Finally,  the last two Biz treatments
failed to create a very great response at all.   During the same
period, rejections of key solutes  were moderate and tended to
fluctuate; it was noticeable  that the best results were normally
obtained in the period following a Biz  flush, but this effect did not
persist until the end.  It is  also interesting to note that the best
BODg rejection did not necessarily parallel the best rejection of
sodium or chloride ion.  The  reason for this discrepancy is  not
known;  a possible reason may be  that some of the  substances
responsible for BOD5 in this system were liquids  or gases rather
than dissolved solids.   Some of these  may well display anomalous
transmission characteristics  through fiber walls.   The first
cleaning of the module was  carried out when flow of brine  and
product through it had completely stopped.  Accumulation  of the
large amount of solid in fouled modules was first noted at  this
time.

Upon re-installation, this module performed in a way which  fairly
closely simulated its earlier performance, even though a sand
filter was installed at this time.   Again,  BOD5 rejection did not
necessarily parallel rejection of sodium or chloride ion.   The
time  required for the flux to drop to a completely  unacceptable
level was approximately the same as before.
                               56

-------
Maintenance of flux level was about the same in the third period of
use after cleaning as it had been in the first two periods.  Under
the conditions of use which were in existence at the Wastewater
Research Center,  the useful service period before radical cleaning
appeared to be somewhere in the range 20 to 30 days.  In general,
this module appeared to be somewhat less effective in rejecting
sodium or chloride ions than was predicted; and the rejection of
substances responsible for BODs was erratic,  although generally
high when the module was in good condition--either freshly cleaned
or freshly washed with Biz.

When module D009 was taken apart for the  third cleaning, large
numbers of opaque white patches were noted among the translucent,
shiny hollow filaments.  These were quite brittle,  and resulted in
many broken filaments during handling.   Attempts to patch them all
were  extremely tedious,  and an attempt to  prepare an intact module
of slightly lower throughput by excising them and  sealing the  ends
failed altogether.  Otherwise, work would have been continued with
this module.

Examination of material from the opaque sections provided little
useful information.  Microscopic study showed that the hollow fila-
ments in these areas were  still intact--the walls were not broken,
nor were they collapsed.  The material was still soluble in acetone,
so it may be concluded that it was not an extensive hydrolysis
product of cellulose acetate.  Only traces of inorganic material
were  found present by emission spectroscopy.  The opaque areas
presented an appearance like that of cellulose acetate whictf has
been precipitated from a solvent into water.  A possible explanation
is that a physical phenomenon similar to devitrification of  glass had
occurred.  Another is that some product of bacterial origin may
have actually caused a partial solution of fibers in these regions,
and this cellulose  acetate was subsequently "reprecipitated" in a
very amorphous form during the removal and cleaning process.

Module D01Q:   This module, also of the single seal,  looped  fiber
design, was installed in service on  10 August 1971 and it was set
up from the start with the intention  of minimizing the  fouling which
had been such a problem with module D009.  To this end, the water
flow rate through the module was greatly increased, and a program
                                57

-------
of chlorination was begun.  The original objective was to maintain a
chlorine level of approximately 2  mg/1 in the feed water.  In early
attempts to adjust this level,  a very highly chlorinated stock was
obtained; this made BOD5 determinations meaningless in samples
taken at these times.  These measures  were only partially success-
ful; flux decline with time was almost as severe as it had been
before, even though the frequency of Biz treatments was also
increased.  The principal comment that can be made about results
with this module is that BOD5  rejection appeared to be better than
it had been in the first case; this may possibly be due to the chlori-
nation technique, although the  analysts adopted a procedure of
treating the samples  with sodium  sulfite to destroy residual chlorine.
On the other hand,  sodium ion and chloride  ion were rejected rela-
tively poorly with this module, and  this accounted for the decision
to replace  it after a considerably  shorter period  of operation.

Module D008:   Installation of this module  (on 4 October 1971)
was  coupled with a revamping of the chlorine feed system, to ensure
more reliable and effective operation  (a chlorinator of our own
design was fabricated, and it was connected in parallel with the sand
filter, to utilize the pressure drop across the sand bed to ensure a
steady flow of water through the chlorinator cartridge).  Moreover,
an acid feed was provided, to make sure that the pH of the feed water
was  maintained at 5 to 5.5.  This was accomplished by pumping 0. 5
molar sulfuric acid from a large reservoir  into the feed line with a
small,  adjustable peristaltic pump, at a rate of 3. 5-4 ml/min.
(The feed water proved to have a considerable buffer capacity; other-
wise this amount of acid would have been far too  great. )

The  effect of the added acid was apparently  small as compared with
the effect of acid liberated from the trichlorotriazinetrione; on
several occasions when the pump  line was found obstructed,  the  pH
of the treated feed water  was not noticeably affected.  The manu-
facturers of this product  note that water treated with it normally
develops an acid reaction.

On the whole, these measures appear to have helped in maintaining
a somewhat higher flux level,  or, at least,  in slowing down the rate
of flux decline.   On the whole, also, the overall balance between
BOD5 rejection and rejection of sodium and chloride ions was
improved somewhat.  However, it should be noted that in each case
there was  at least one day where the rejection was very  poor.
                                58

-------
Opportunity was taken with this module to evaluate the possibility
of improving module performance by decreasing the filament packing
density of fibers in the module.  When the module had been operated
for two weeks, it was  removed,  thoroughly cleaned, and reassem-
bled without the liner.  This change caused a decrease in  filament
packing density from 57% to 49%.  Results of this test were made
difficult to evaluate  by the fact that the module sat completely idle
for a period of ten days  (23 October - 31  October 1971) when the
Wastewater Research Center was completely flooded out during a
period of exceptionally heavy and prolonged  rainfall.  However, as
nearly as one can estimate,  the  rate of flux  decline and the rejection
behavior of the module were not improved; and they may even have
been affected somewhat adversely.
Series D-2 Modules
Module D0013-2, representing the double seal, parallel fiber design,
was installed on 1 June 1971 and was operated for 44 days.  For
seventeen days of this period it was operated jointly with module
D009; during this period the output capacity of the pump in service
was not sufficient to supply each module  with water at 250 psig,
so module D0013-2 was operated at 150 psig for the  first 17 days.
When module D009 was shut down, operating pressure was raised
to 250 psig.  On the whole, D0013-2 appeared to resist decrease
in flux due to fouling somewhat more effectively than D009 (looped
fiber design), with which it was operating simultaneously.  Product
flux through it started at a lower gfd level, but never dropped off
quite so far as it did with the single-ended module.  A portion of
this beneficial effect may be the result of the fact  that several
modes of operation are possible with this module, and each of them
was used from time to time.  This included the ability to pass feed
from one end of the bundle to the other, and to reverse periodically
the direction of flow.  Such changes may well have had the effect of
breaking up the pattern of solids deposition in the  module (fouling).

It is difficult to compare accurately the performance of modules
D009 and D0013-2 during the period when they were  operating simul-
taneously,  because they were operating at different pressures.  How-
ever, it appears  that module D0013-2 was somewhat less effective  in
rejecting solutes during the  time when it operated at 150 psig, and
that its performance improved predictably when the  pressure was
increased.  This, of course, is normal for reverse  osmosis units.
                               59

-------
It was noteworthy that BODs rejection paralleled rejection of
sodium and chloride ions more closely in this case.

Module D0013-2 was removed for cleaning on 15 July 1971. In
the ordinary course of events it would have been re-installed.
Examination of the fiber bundle,  however, disclosed the same
opaque formations which were present in module D009.

Attempts were made to put this module (and module D009) back into
serviceable condition by techniques which have usually been quite
successful with cellulose acetate hollow filament modules.  First,
a systematic effort was made to locate and seal off each broken
filament.  When feasible,  this is easily done by first tying a simple
overhand knot in the loose end, then dipping the end into a small
amount of  "Duco" cement.  In actual operation,  new broken ends
seemed to  be generated by handling operations about as fast as the
first  ones were repaired.

For this reason, a more drastic approach was tried; the opaque
areas of embrittled fiber were simply cut out entirely.   Then the
problem reduced to sealing off the ends of a fairly substantial
fiber  bundle, all at once.  This proved to be very difficult; in
spite  of repeated efforts, either by dipping into Duco cement or
by treating with solvents,  some of the ends remained unsealed,
as demonstrated by testing with low-pressure nitrogen while the
fiber  ends  were under water.
Series W Modules

The description of this woven fabric module has been given pre-
viously.   Unit WA001 was installed in service on 17 November
1971, and was operated for 15 days.  No change was made in the
general parameters of the test board:  chlorination, acid feed, and
rate of water pumping through this module were maintained just
as they had been with the last module  used (D008).   Since the
effective  area of the fiber surface  in this module was considerably
less than in the  other modules,  its maximum throughput of product
was also  less,  and therefore the percentage of product recovery
was quite low.   This circumstance may have affected the results
obtained in the early phases of work with this module, although
later results make this unlikely.
                                60

-------
Basically,  results with this type of module were extremely satis-
fying. ' After an initial period of a few days in which flux declined
approximately 25%,  flow of product water stabilized at an approxi-
mately constant level which did not change significantly over the
next week or more.  Moreover,  rejection of sodium and chloride
ions was outstandingly good, never dropping below 85%,  even when
the percentage of product recovery was increased tenfold.  BOD,-
rejection was not so satisfactory; although individual rejections
were occasionally quite good, overall performance  was only
average.  During the last three days of use, module WA001 was
operated at approximately 65% recovery.  Possibly the  most
interesting fact found here is that rejections did not drop greatly
at this increased recovery.  In  fact,  the available data indicate
that the BODg rejection improved.
                                61

-------
 o
 U5
 Cvl
  O
  o
 esi
  to
>Co
  o
  to
  o
  o
                        _L
_L
J_
     Feed:   filtered secondary effluent

     Pressure:  250 psi

     Maintenance:   Biz flushing of fiber unit indicated

                   by solid points,  lab cleaning by

                   asterisk.
               0.00      40.00    80.00

                            TEST DflY
        120.00
        160.00
 FIGURE   9fl.VflRIRTION  OF PRODUCT  WflTER  FLUX

              WITH  TIME-NODULE  D009
                              62

-------
  o
  o
  o_
  o.
  00
LJO
CJ
as:
 o.
 CSJ
  o
  o
     Feed:  filtered secondary effluent

     Pressure:   250 psi
               0.00     40.00    80.00    120.00   160.00

                           TEST  DRY
 FIGURE   9B.VflRIRTION  OF SODIUM ION REJECTION

             WITH TIME-MODULE  D009
                           63

-------
 o
 o

 o
 o.
 o.
 CO
Zo
Ujo
UJ
0.
UJ
 o.
 CM
 o
 o
    	1	I	I


     Feed:   filtered secondary effluent

     Pressure:   250 psi
              0.00    40.00    80.00     120.00    160-00
                          TEST DRY
 FIGURE   9C.VflRIflTION OF  CHLORIDE  ION REJECTION

             WITH  TIME-MODULE DOQ9
                            64

-------
  o
  o
  o
  o.
  o
  o
  o.
  CD
UJO
LU

Q_
O

LU
QC
  o.
  CO
  o
  o
                                I
     Feed:  filtered secondary effluent

     Pressure:  250 psi
               0.00     40.00    80.00    120.00    160.00

                           TEST DRY
 FIGURE   9D.VflRIflTI0N  OF BOD  REJECTION

             WITH HUE-MODULE  0009
                            65

-------
 o
 o
 o
 o.
 o
 o
 o.
 CD
LUO
o
LU
a:
  o.
  (M
  O
  O
                       _L
_L
     Feed:   filtered secondary effluent

     Pressure:   250 psi
              0.00     40.00    80.00
                          TEST DRY
        120.00
160.00
 FIGURE   9E.VflRIflTION OF  KJELDRHL  NITROGEN

             REJECTION WITH  TINE-MODULE D009
                           66

-------
  o
  00
  (NJ
  O
  CO
  
-------
  o
  o
  o
  o.
  o
  o
  o.
  CD
UJO
UJ

0_
So.
CJ

UJ
  o.
  CM
  O

  O
               _L
J.
J.
_L
                                                J_
     Feed:  secondary effluent, filtered and chlorinated

     Pressure:  250 psi
               0.00     20.00    40.00

                           TEST DRY
                60.00
                80.00
 FIGURE  lOB.VflRIflTION  OF SODIUM  ION REJECTION

             WITH TIME-MODULE  D010
                           68

-------
  o
  o
  0_
  o
  o
 o.
 GO
Ujo

°
UJ
a.
UJ
"=
 0.

 CM
     Feed:   secondary effluent, filtered and chlorinated

     Pressure:   250 psi
               0.00     20.00    40.00    60.00    80.00

                           TEST  DflY
 FIGURE  lOC.VflRIRTION  OF CHLORIDE  ION  REJECTION

             WITH TIME-MODULE  D010
                           69

-------
 o
 o
 o
 (NJ.
  O
  O
 O
 O.
LUO
o
LU
  o
  O
  O.
  CVI
               _L
J_
_L
J_
                                                _L
     Feed:  secondary effluent, filtered and chlorinated
     Pressure:   250 psi
               0.00     20.00    40.00
                           TEST DRY
                60.00
                80.00
 FIGURE  10D.VPRIRTI0N  OF BOD  REJECTION
             WITH TIME-MODULE  0010
                            70

-------
  o
  o
  o
  o.
  o
  o
 o.
 CO
Ujo
2s-

o
UJ
  o.
  CM
  o
  o
     Feed:  secondary effluent, filtered and chlorinated

     Pressure:   250 psi
               0.00     20.00    40.00    60.00    80.00

                           TEST  DRY
 FIG-URE  lOE.VflRIFITIGN  OF KJELDflHL NITROGEN

             REJECTION  WITH TIME-MODULE  0010
                          71

-------
  
-------
  o
  o

  o
  o.
  o
  o
  o.
  00
Ujo
LU

Q_
__
Zo

S
UJ
  o.
  CM
  o
  O
  . ^
     Feed:  secondary effluent, filtered,  chlorinated,

            pH adjusted to 5. 5

     Pressure:  250 psi
               0.00     20.00    40.00    60.00    80.00

                           TEST DRY
 FIGURE  llB.VflRIRTION OF  SODIUM  ION REJECTION

              WITH TIME-MODULE D008
                            73

-------
 o
 o
 o
 o_
 o
 o
 o.
 CO
LiJO
o
UJ
Od
  o.
  (M
  o
  o
               _L
_L
_L
                                        _L
    Feed:  secondary effluent,  filtered, chlorinated,

           pH adjusted to 5. 5

     Pressure:   250 psi
               0.00     20.00    40.00

                           TEST DRY
                60.00
                80.00
 FIGURE  HC.VflRIflTIQN  OF CHLORIDE  ION  REJECTION

             WITH TIME-MODULE D008
                            74

-------
  o
  o.
  o.
  00
Ujo
UJ

0_
 "o
Zo
O
  CM
  O

  O
     Feed:  secondary effluent,  filtered, chlorinated,

            pH adjusted to 5. 5

     Pressure:  250 psi
               0.00     20.00    40.00    60.00    80.00

                            TEST OflY
 FIGURE  llD.VRRIflTION OF  BOD REJECTION

              WITH TIME-MODULE 0008
                          75

-------
                       _1_
_L
     Feed: secondary effluent, filtered, chlorinated,

           pH adjusted to 5. 5

     3r ;-ssare:    250 psi
  o
  o
  o
  o.
  o
  o
  o.
  00
Lao
LU

Q_
O •
.	,0.
UJ


Lug
<Ł. •
 o.
 OJ
  o
  o
               0.00     20.00    40.00

                           TEST DRY
        60.00
80.00
 FIGURE  llE.VflRIRTI0N  OF KJELDRHL  NITROGEN

             REJECTION  WITH  TIME-MODULE  D008
                            76

-------
  o
  CM
  O
  o
o
u.
Xo

 . •.
 Jo





 o


 o'
  p
  IN
    Feed:   filtered secondary effluent

    Pressure:   150 psi to day 17; 250 psi afterwards

    Maintenance:  Biz flushing of fiber unit indicated

                  by solid points, lab cleaning by
                  asterisk.
                0.Q6     20.00
                              TEST  fl*Y
 FIGURE
               WITH
                             77

-------
 o
 o
 o
 o.
 o
 o
 o.
 00
UJO
CJ

UJ
  o.
  CM
  o
  o
               _L
J_
_L
     Feed:  filtered secondary effluent

     Pressure:  150 psi to day 17; 250 psi afterwards
               0.00     20.00    40.00

                           TEST  DRY
                60-00
                80.00
 FIGURE  12B.VRRIRTI0N  OF  SODIUM ION  REJECTION

             WITH  TIME-MODULE  D0013-2
                            78

-------
  o
  o
  o
  o.
  o
  o
 o.
 oo
Zo
LUO
So-
UJ


2
a: •
 o.
 CO
  o
  o
     Feed:  filtered secondary effluent

     Pressure:   150 psi to day 17; 250 psi afterwards
               0.00     20.00    40.00    60.00
                           TEST  DRY
80.00
 FIGURE  12C.VflRIflTION  OF CHLORIDE ION  REJECTION

             WITH  TIME-MODULE  D0013-2
                            79

-------
 o
 o
 o
 o.
 o
 o
 o.
 00
Zo
UJO
O
UJ
  O.

  CJ
  o
  o
     Feed:  filtered secondary effluent

     Pressure:  150 psi to day 17; 250 psi afterwards
               0.00     20.00    40.00    60.00    80.00

                           TEST  DRY
 FIGURE  12D.VRRIRTION  OF  BOD REJECTION

             WITH  TldE-nODULE D0013-2
                          80

-------
  o
  o
  o
  o.
  o
  o
  o,
  co
Zo
LLJO

~S-
o
UJ
 Ł
  o.
  CM
  O
  o
     Feed: filtered secondary effluent

     Pressure:  150 psi to day 17; 250 psi afterwards
              0.00     20.00    40.00    60.00    80.00
                           TEST  DRY
 FIGURE  12E.VflRIflTION OF  KJELDflHL NITROGEN

             REJECTION WITH TIME-MODULE  00013-2
                            81

-------
 o
 00
 CM
 O
 CO
Xo
 !~
 'CM
  o
  o
  o
  ao
Feed:  secondary effluent, filtered,  chlorinated,
       pH adjusted to 5. 5
Pressure: 200 psi
Maintenance:  Biz flushing of fiber unit indicated
             by solid points, lab cleaning by
             asterisk.
               0.00     4.00     8.00     12.00    16.00
                            TEST DRY
 FIGURE 13fl.VRRIflTION  OF PRODUCT  WflTER FLUX
              WITH  TIME-MODULE WflOOl
                             82

-------
  o
  o
  o
  o
Ujo
 "o
  o
O

UJ
oc
  00.

  CO
  o
  o
  00
     Feed:  secondary effluent, filtered, chlorinated,

            pH adjusted to 5. 5

     Pressure:   200 psi
               0.00     4.00     8.00     12.00     16.00

                           TEST DRY
 FIGURE  13B.VflRIRTION  OF SODIUM  ION REJECTION

             WITH TIME-MODULE  WflOOl
                           83

-------
 o
 o
 o
 o.
 o
 o
 (O.
 CO
Zo
UJO
LU

Q_
2s

O
UJ
  o
  o
  o
  CD
     Feed:  secondary effluent, filtered, chlorinated,

           pH adjusted to 5. 5

     Pressure:  200 psi
               0.00     4.00     8.00     12.00     16.00

                           TEST  DRY
 FIGURE  13C.VRRIRTION  OF CHLORIDE  ION REJECTION

             WITH TIME-MODULE  WR001
                           84

-------
  o
  o
  o
  CM_
  o
  o
  o
  o.
LUO
°
LU
  o
  o
  o.
  CSJ
     Feed:  secondary effluent, filtered, chlorinated,

           pH adjusted to 5. 5

     Pressure:  200 psi
               0.00     4.00     8.00     12.00    16.00

                            TEST DflY
 FIGURE  13D.VflRIflTION OF BOD REJECTION

             WITH TIHE-HODULE WflOOl
                            85

-------
  o
  o
  M3-
  O)
 O
 O
  CD.
  CD
Zo
Ujo
LU
Q_
ED •
O
UJ
  (O
  o
  o
  IO.
  LO
     Feed: secondary effluent,  filtered,  chlorinated,
           pH adjusted to 5. 5
     Pressure:  200 psi
               0.00     4.00     8.00      12.00    16.00
                           TEST DRY
 FIGURE  13E.VflRIflTION  OF KJELOflHL  NITROGEN
             REJECTION  WITH  TIME-ttOOULE WflOOl
                            86

-------
             VIII  ACKNOWLEDGMENTS
Personnel at Chemstrand Research Center,  Inc. who participated
directly in this program were J. D. Bashaw (Principal Investigator
from 10-70 to 2-71),  J.  K.  Lawson (Principal Investigator from
2-71 to 12-71), J. C. Berry,  E. S. Gothard, R. L.  Leonard,
R. F.  Cole,  L.  C. Locust, and M.  C.  Readling.  Program
Manager was T. A. Orofino.  EPA Project Officer was
John M.  Smith.

Experimental field work was carried out in laboratories of the
University of North Carolina Wastewater Research Center, at
Chapel Hill,  North Carolina.  We wish to thank the personnel of
this  facility for the use of space and for their cooperation on many
occasions.  Particular thanks are due to Dr.  James  C. Brown,
manager of the Wastewater Research Center,  Dr. Linda  Little,
Dr.  Don Francisco,  Mr.  William Walker, and Mr. John  Street.
Some of the analyses which we required were performed  either at
the Wastewater Research Center, or else were conducted by
University of North Carolina laboratories for us at the request of
Dr.  Brown and Mr. Walker.  These included BOD, Kjeldahl
nitrogen,  ammonia nitrogen, phosphate, organic and inorganic
carbon, nitrate,  and  nitrite.  When the sand filter was installed,
we were able to arrange for Mr. Street to backwash  it for us on
a regular basis.   On  the few occasions when some of our equipment
misfunctioned, various members of their staff were  prompt to
notify  us.
                                 87

-------
                   IX  REFERENCES
1.   S. Loeb and S. Sourirajan,  University of California (Los
    Angeles) Eng. Dept. , Report No.  60-60,  July 1960.
2.   T. A.  Orofino et al. , Office of Saline Water Research and
    Development Progress  Report No.  549,  May 1970.
3.   H. B.  Hopfenberg, Department of Chemical Engineering,
    North  Carolina State University, Raleigh,  N. C. , private
    communication.  See also A.  S. Michaels,  Chem. Eng.
    Progs. 64,  No.  12,  31  (1968)
4.  Ulrich Merten,  "Desalination by Reverse Osmosis, "
    U.  Merten, Ed., The M. I. T. Press,  1966 (Chapter 2).
5.  Work conducted under Contract 14-30-2773, Office of
    Saline Water, U. S. Department of the Interior to Monsanto
    Research Corporation.
                               89

-------
          X  PATENTS AND PUBLICATIONS
No patent applications nor publications by Monsanto Research
Corporation have resulted from work carried out in the course of
this contract.
                                 91

-------
  SELECTED WATER
  RESOURCES ABSTRACTS

  INPUT TRANSACTION FORM
                                           J.KeportNo.
                           3. Accession No.
                          w
  4. Title
          Hollow Fiber Technology For Advanced Waste
          Treatment
                           e.  November 1971
                           8,
  7. Authot(s)
    J. D. Bashaw, J. K. Lawson, and T. A.  Orofino
                                                                     t No,
  9, Organization                -
     Monsanto Research Corporation
     Chemstrand Research Center, Inc.
     Durham, North Carolina
                          10. Project No.
                                17040 FEE
                              Contract/Grant No.

                                14-12-926
                          13.  Type vfRepQit and
                              Period Coveted
  12.  Sffiasoriof Orfaai-atioo
  IS. Supplementary Notes
                     Environmental Protection Agency report
                    '" 1972.
          °  The utility of hollow fiber reverse osmosis membranes in renovation of
    secondary municipal effluent was investigated through construction, laboratory
    evaluation,  and monitoring in field service of various hollow fiber modules.  All
    units incorporated cellulose acetate hollow fibers,  annealed for sodium chloride
    rejections of 80-95^6 at 250 psi external operating pressure.  Product water
    capacities ranged from 50-300  gallons per day.  Module designs considered
    included the single seal end, looped fiber bundle; double seal end, parallel bundle;
    radial flow parallel bundle; and a rolled,  woven hollow fiber fabric.  The typical
    flux-rejection characteristics of the basic fiber system (4 gfd-95%) were observed
    in waste water service, but steady-state flux,  maintained only with regular
    detergent flushes, was usually  less than 1 gfd,  with an  accompanying decline in
    selectivity.   A notable exception was the woven hollow  fiber fabric design,  which
    showed improved retention of start-up characteristics  and minimum effects of
    shell-side fouling during short-term field tests.

    This report was  submitted in fulfillment of Contract 14-12-926,  under the
    sponsorship of the U. S.  Environmental Protection Agency.
  i7a .Descriptors *Reverse osmosis,  *pesalination processes,  *Membranes, Wastewater
    treatment, Semi-permeable membranes
  i7b. identifiers  *Hpllow fiber technology,  *Cellulose acetate hollow fine fibers, fiber
    spinning,  hollow fiber modules
  17c. COWRR Field & Group  05D
18. Availability 19. Security Class,
fReport}
20. Security Class.
(Pmee)
21. Wo,«f
"I8
?2. J»r«>®
Send To:
WATER RESOURCES SCIENTIFIC INFORMATION CENTER
U.S. DEPARTMENT OF THE INTERIOR
WASHINGTON. D. C. 2O24O
  Abstractor J. D. Bashaw
I institution  Monsanto Research Corporation
WRSIC 102 (REV. jyNE 1971)
                                                 U. S. GOVERNMENT PRINTING OFFICE : 1972 724-770/169

-------