United States
Environmental Protection
Agency
Risk Reduction
Engineering Laboratory
Cincinnati OH 45268
Research and Development
EPA/600/S2-91/011 June 1991
Project Summary
Arsenic(lll) and Arsenic(V)
Removal from Drinking Water in
San Ysidro, New Mexico
Dennis Clifford and Chieh-Chien Lin
The removal of a natural mixture of
As(lll) (31 ^ig/L) and As(V) (57 ^g/L) from
a groundwater high In total dissolved
solids (TDS), and also containing fluo-
ride (2.0 mg/L), was studied in San
Ysidro, NM, using the University of
Houston (UH)/U.S. Environmental Pro-
tection Agency (EPA) Mobile Drinking
Water Treatment Research Facility. The
raw water in this study was deliberately
unchlorinated so as not to oxidize the
As(lll) present. The objective of the study
was to establish a cost-effective means
of removing As(lll), As(V), and fluoride
from this and similar waters.
Arsenic adsorption onto fine-mesh
activated alumina gave better-than-ex-
pected results in view of the knowledge
that As(lll) is known to be poorly re-
tained on alumina. Approximately 9000
bed volumes (BV) could be treated at
pH 6 before the arsenic maximum con-
taminant level (MCL) (0.05 mg/L) was
reached. At the natural pH of 7.2, how-
ever, only 1900 BV could be treated
before exceeding the MCL. Approxi-
mately 70% of the adsorbed arsenic
was recoverable by cocurrent regen-
eration with 6.5 BV of 4% NaOH, but
after two regenerations, the column ca-
pacity was reduced to 72% of its virgin
performance. Coarser, 12 x 28 mesh,
alumina did not perform as well in ad-
sorption or regeneration. The spent alu-
mina regenerant was treated by lowering
its pH to 8.5 and quantitatively
coprecipitating the arsenic with the bulk
AI(OH), precipitate. The sludge pro-
duced was not hazardous as determined
by the EPA's extraction procedure (EP)
toxicity test. Analyses of the spent
regenerant solution showed that un-
avoidable oxidation of the As(lll) to As(V)
occurred on the alumina, which helps
to explain its better-than-expected col-
umn performance.
Reverse osmosis (RO) treatment with
either a cellulose triacetate (CTA) or
polyamide (PA) hollow fiber membrane
resulted in > 97% arsenic removal and
> 94% TDS removal. Electro dialysis (ED)
removed 73% of the arsenic and was
able to meet the arsenic MCL on the
City Water containing 89 (ig/L total ar-
senic; however, ED removed only 28%
of the As(lll) from a new well containing
100% As(lll) at a level of 230 ng/L.
Chloride-form anion exchange also
performed better-than-expected but not
well enough for It to be considered seri-
ously for treatment. About 200 BV could
be treated before the arsenic MCL was
reached. Point-of-use (POU) RO treat-
ment with a thin film composite (TFC)
membrane was effective in removing
> 91% of the arsenic and > 94% of the
TDS at low (< 15%) water recovery.
Because of the small (70-dwelllng)
community, the difficulty of central treat-
ment, and the poor water quality, EPA
chose San Ysidro as a test community
for POU RO treatment. That study (EPA/
600/2-89-050) showed POU RO treatment
to be a viable alternative to central treat-
ment.
This Project Summary was devel-
oped by EPA's Risk Reduction Engi-
neering Laboratory, Cincinnati, OH, to
announce key findings of the research
project that Is fully documented In a
&A> Printed on Recycled Paper
-------
separate report of the same title (see'
Project Report ordering Information at
back).
Introduction
San Ysidro, NM, is a small community
45 miles northwest of Albuquerque. The
current water supply, a 12-ft-deep infiltra-
tion gallery, is inadequate in quantity dur-
ing the summer months and exceeds,
year-round, the EPA MCL of 0.05 mg/L for
arsenic. It also contains 2.0 mg/L of fluo-
ride; at the time of this study, that also
exceeded the old 1.4 mg/L MCL Addition-
ally, this "City Water" has high IDS (810
mg/L), hardness (282 mg/L), and alkalinity
(468 mg/L). From 1975 through 1981, the
mean arsenic concentration was 0.074 mg/
L and fluoride ranged from 1.6 to 3.0 mg/L.
In an effort to obtain an adequate quan-
tity of arsenic-free water, three test wells,
ranging in depth from 44 to 128 ft, were
drilled in 1982. Unfortunately, these new
wells contained much higher levels of ar-
senic, fluoride, dissolved solids, iron, sul-
fate, chloride, and manganese than did the
existing City Water from the shallow infil-
tration gallery.
Field research was performed in San
Ysidro over a period of 9 mo using the UH/
EPA Mobile Drinking Water Treatment Re-
search Facility (Mobile Inorganics Pilot
Plant). This 10-ft wide by 40-ft long trailer
had been used at three previous locations
to study the removals of fluoride, nitrate,
and chromate from groundwater supplies.
One objective of the move to San Ysidro
was to attempt treatment of both the City
Water and the best of the new test wells,
Well No. 4, because of their differing levels
of arsenic, fluoride, and iron. After having
decided that the new wells were too diffi-
cult to treat, however, the basic objective
of the San Ysidro arsenic removal experi-
ments was changed to development of a
simple, cost-effective way to remove the
arsenic from the San Ysidro City Water
and from similar waters.
Experimental Details
Alumina adsorption, electrodialysis, re-
verse osmosis, and ion exchange pro-
cesses were studied in San Ysidro without
oxidative pretreatment of the raw water,
i.e., the treatment processes were fed the
natural mixture of As(lll) and As(V). This
was done because the next planned study
in Hanford, CA, would involve oxidation of
As(lll) to As(V) before its removal by the
same processes. The alumina adsorption
and ion exchange tests were carried out in
lab scale (1-in. diameter) columns rather
than in the 8- or 10-in.-diameter pilot-scale
columns to minimize the production of ar-
senic-contaminated sludges from alumina
and ion-exchange regeneration.
Analytical Methods
With the exception of metals analyses
performed using a Perkin Elmer Model
5500 Inductively Coupled Plasma (ICP)
Spectrometer* at UH, and a check sample
of the City Water run by an independent
laboratory, all analyses were performed
with mobile lab equipment. Analysis proce-
dures from "Standard Methods for the Ex-
amination of Water and Wastewater" were
used for hardness, alkalinity, silica, and
sulfide. Fluoride and pH were analyzed
using electrode methods from an Orion
manual. All methods and instruments were
standardized and calibrated daily. Gener-
ally two or more standards from different
sources were used during the study.
We used a method which we had previ-
ously developed for EPA to rapidly speci-
ate arsenic. It takes advantage of the fact
that, in the pH range of 3.0 to 8.4, As(V) is
ionic existing as monovalent HaAsOi, or
divalent HAsOf", whereas As(lli) is un-
charged arsenbus acid, HaAsOa. When
chloride-form strong-base anion resins are
used for the separation, As(lll) passes
through the resin column unhindered
whereas As(V) is completely retained by
the resin. Following speciation, graphite
furnace atomic absorption spectroscopy
(GFAAS) is used to determine total arsenic
(As(lll) + As(V)) on the untreated sample
and As(lll) on the column effluent.
Arsenic(V) is determined by difference.
Alumina Experiments
Because the removal of natural mix-
tures of As(lll) and As(V) by activated alu-
mina had not been studied previously, it
was the focus of this study. And, because
we had experienced unforeseen oxidation
of As(lll) in previous lab studies, such
unplanned oxidation of As(lll) was also of
interest in San Ysidro. The optimum pH for
the adsorption of arsenic and fluoride is
known to occur in the 5.5 to 6.0 range;
therefore, a pH of 6.0 was fixed for most of
the alumina runs. Because adsorption onto
alumina is known to be a kinetically con-
trolled process, the two common mesh
sizes, 14 x 28 (1.2 x 0.6 mm) and 28 x 48
(0.6 x 0.3 mm), of Alcoa F1 alumina were
compared. Finally, two concentrations of
the NaOH regenerant were used to deter-
mine which was more economical in terms
of the mass of arsenic removed/mass of
NaOH applied.
Mention of trade names or commercial products
does not constitute endorsement or recommenda-
tion for use.
A 100-L pilot-scale feed tank was used
to adjust the pH of the incoming raw water;
however, most of the tank effluent was
bypassed to waste to minimize mixing,
aeration, and detention time. This was done
to prevent oxidation of the As(lll) and to
represent more closely, the actual
feedwater that would exist in a full-scale
treatment process. For all the alumina ex-
periments, the feedwater pH was adjusted
to 6.0 with 2% H2SO4. The 32-in.-deep
alumina beds containing 0.4 L of media
operated at a flow rate of 80 ± 3 mL/min
for an empty bed contact time (EBCT) of 5
minutes. Generally, the alumina runs, which
lasted 20 to 25 days, were continued until
the effluent MCL for arsenic (0.05 mg/L)
was reached, which meant that the runs
lasted far beyond fluoride exhaustion. Only
total arsenic was determined on the col-
umn effluent samples, i.e., there was no
speciation of As(lll) or As(V).
Desalting Tests
Arsenate, As(V), which is a large
mono-, di-, or trivalent anion at pHs above
3, is known to be effectively (> 97%) re-
jected by RO membranes. The percent
removal of arsenite, a nonionic species at
neutral and acidic pH, varies widely (43%
to 81%), however. Thus, it was of interest
to study the removal of a mixture of As(lll)
and As(V) using PA, CTA, and TFC mem-
branes. Additionally, TFC membrane per-
formance was studied using a Culligan
POU RO system. Two different RO mod-
ules were used in the study: a Dow, hol-
low-fiber, CTA type of membrane and a
DuPont hollow-fiber, PA type. Each was
operated separately as a single module at
approximately 50% recovery. The Dow
module was larger, producing 11,700 L/
day compared with 4500 L/day for the
DuPont module. Based on the recommen-
dations from each manufacturer, the feed
to the Dow unit was acidified to pH - 6.3
whereas no acid was added to the DuPont
module feed. For both the Dow and DuPont
systems, a deep-bed filter was used ahead
of the 10-u.m cartridge filter. Also, an
antiscalant, 10 mg/L sodium hexameta-
phosphate (SHMP), was added continu-
ously during each run. Although no
reference to arsenic rejection by ED could
be found, ED was expected to effectively
remove charged arsenate species but be
ineffective for the removal of uncharged
arsenite. There was some speculation that
the potential and current generated in the
ED stack might oxidize and thereby re-
move arsenite.
An Ionics Aquamhe I ED unit with auto-
matic current reversal to prevent fouling
was used in San Ysidro. It operated with-
-------
out pretreatment to produce 1790 L/day
(475 gpd) of product water at 81% recov-
ery utilizing internal brine recycle.
Ion-Exchange Experiments
Attempts were made to verify the ex-
pectation of very short ion-exchange runs
because of (a) the presence of nonionic
arsenite, (b) high IDS, and (c) high sulfate
in the feedwater. Related research before
this study showed As(V) is removed well
by ion exchange in tow sulfate/low IDS
water, but the inherent danger of an elu-
tion peak of high arsenic exists if the col-
umn runs beyond breakthrough. These
considerations, in addition to the interest-
ing possibility of a sulfate/arsenate selec-
tivity reversal due to the high TDS of the
San Ysidro City Water, prompted a limited
study of strong-base anion resins in the
chloride form. In these experiments, the
pH of the feedwater was not adjusted. The
resin bed consisted of 400 ml_ of chloride-
form, strong-base anion resin—either lonac
ASB-1, a type 1 gel resin with microporosity,
or Dowex-11, an isoporous "improved po-
rosity" type 1 resin. Following each ex-
haustion run, the resin was regenerated
with 5 BV of 6% NaCI solution (18.3 Ib/ft3),
i.e., approximately 4 to 5 times the stoi-
chiometric requirement based on total resin
capacity.
Arsenic Sludge Disposal Tests
Wastewater disposal studies were per-
formed on the spent regenerant solutions
from the alumina column regenerations in
an attempt to verify previous research on
arsenic removal in Fallen, NV. In Fallen it
was found that by simply neutralizing the
alumina regenerant solution the resulting
AI(OH)3 precipitate would adsorb the ar-
senate to yield a supernatant water with
less than 0.10 mg/L total arsenic. The
difference between the present study and
prior studies is the presence of a signifi-
cant amount of As(lll) which, because it is
poorly adsorbed on the alumina, could
cause the AI(OH)3 sludge to fail the Extrac-
tion Procedure (EP) toxicity test.
In a typical AI(OH)3 precipitation test,
500 ml_ of spent alkaline regenerant solu-
tion (pH 13) was placed into a 1-L beaker
and acidified to pH 6.5 with HCI. A small
amount of acid was added at 1/2- to 4-hr
intervals during the next 14-hr period to
maintain the pH at 6.5 to prevent
redissolution of AI(OH)3. The solution was
allowed to stand overnight (10 hr) before
being filtered through a quantitative, paper
filter. This paper filter was later dried for 12
hr at ambient temperature (22° C) and
stored for future studies including the EP
toxicity test. Total arsenic, As(lll), and As(V)
were determined on the spent regenerant
before precipitation and on the filtrate after
filtration.
Results and Discussion
Water Quality
Most of the values for the City Water
analyses were single-point determinations,
others, however, notably the As(lll) and
As(V) and total arsenic values, were the
averages of many determinations. During
the study, the mean As(lll) value was 31 ±
8.6 ng/L, As(V) was 57 ± 8.2 u.g/L, and the
total As concentration was 88 ± 8.3 |ig/L.
Occasionally, oxidation of As(lll) to As(V)
was observed when a raw water sample
was allowed to sit for several hours or
more. This oxidation was not uniformly
repeatable, however.
In addition to arsenic, major constitu-
ents of the City Water were TDS (810 mg/
L), alkalinity (468 mg/L), hardness (282
mg/L), sodium (190 mg/L), chloride (123
mg/L), silica (60 mg/L), sulfate (37 mg/L),
and fluoride (2 mg/L). Desalting with RO or
ED would be required to meet the EPA
secondary MCL for TDS (500 mg/L) and to
tower the hardness and sodium levels.
Serious precipitation and fouling problems
would, however, be expected with desalt-
ing because both the City Water and Well
No. 4 were found to be supersaturated
with BaSO4 and CaF?. Acid addition to
prevent CaCO3 precipitation and addition
of SHMP, a precipitation inhibitor, were
recommended for scale control. Further-
more the recovery would be limited to ap-
proximately 50% to prevent silica fouling.
Activated Alumina Results
Typical Alumina Breakthrough Curves
at pH 6.0
Typical breakthrough curves (Figure 1)
for fluoride and arsenic in the effluent from
the activated alumina column showed fluo-
ride breaking through first and reaching a
maximum level of 1.4 mg/L long before
arsenic reached its 0.05 mg/L MCL If acti-
vated alumina is used in this fashion, i.e.,
without oxidative pretreatment, the time to
reach the arsenic MCL will be typically two
to three times as long as the time to reach
a fluoride level of 1.4 mg/L for this particu-
lar water. This may be seen in Table 1,
which contains the summary of the alu-
mina results and a comparison of the fluo-
ride and arsenic run lengths.
Arsenic and Fluoride Breakthrough
Curves
In a related laboratory study at the UH,
alumina-run simulations were made with
synthetic waters similar to the San Ysidro
City Water except that the synthetic waters
contained either 100% As(lll) or 100%
As(V). In this way, rt was possible to com-
pare the arsenic and fluoride removal per-
formances among three runs to quantify
the effect of oxidizing the San Ysidro As(lll)
to As(V). The run length (8760 BV) to
arsenic breakthrough for the San Ysidro
City Water, a mixture of As(lll) and As(V),
fell between that of pure As(lll) (300 BV)
and pure As(V) (23,400 BV). Therefore,
oxidizing the San Ysidro City Water to
100% As(V) should more than double the
alumina run length to about 23,000 BV at
pH 6.0.
The shape of the San Ysidro arsenic
breakthrough curve (Figure 1) was de-
layed, i.e., no leakage before 3,600 BV,
and was surprisingly sharp. A much earlier
As(lll) breakthrough was expected based
on the lab simulation data. By way of ex-
planation, some oxidation of As(lll) to As(V)
occurred in the field column, as proven
later by regeneration studies of eluted
As(lll) and As(V). Also, the trivalent ar-
senic concentration of the field study was
only 32 u.g/L, i.e., one-third the concentra-
tion in the lab study.
The fluoride capacities of the various
columns were remarkably similar at 4160
to 4280 g/cm3. This occurred despite the
fact that the San Ysidro water contained
only 2 mg FVL whereas the laboratory
waters contained 3.0 mg FYL. Finally, in
these column tests, the presence of As(lll)
and As(V), at a level of 100 mg As(total)/L,
did not seem to influence the fluoride ca-
pacity of the alumina.
Effect of Mesh Size
Figure 2, representing the arsenic
breakthrough curves for Runs 1 and 2,
illustrates that the mesh size of the alu-
mina has a dramatic effect on its perfor-
mance for arsenic removal. Referring again
to Table 1, it can be seen that the coarse
mesh grade treats 6840 BV to the arsenic
MCL whereas the fine mesh can treat 8760
BV with corresponding arsenic capacities
of 390 and 575 g/m3, respectively. Such
large differences between coarse and fine
were not noted, however, during fluoride
removal.
Effect of pH
Not reducing the pH of the feedwater to
the optimum range between 5 and 6 re-
sulted in a drastic loss in both the arsenic
and fluoride removal capacity of alumina.
The effects of varying both arsenic con-
centration (from 90 to 230 u.g/L) and pH
(from 6.0 to 7.1) are apparent when Runs
1 and 8 using coarse alumina are com-
pared: the high, trivalent arsenic concen-
tration and unadjusted feedwater pH during
-------
90
80
70
60
50
40
30
20
10
3.2
2.8
2.4
1.6
1.2
0.8
0.4
200
Time, Hours
400 600
800
10
Time, Days
15 20
25
30
35
San Ysidro Alumina Run No. 2
Fine 28x48 mesh, F-1 Alumina
Bed Volume = 400 mL
How Rate = 80 mUmin, City Water
EBCT=5min
Feed pH= 6.0 ±0.1
-a—
r
+' 8760 BV
/lo50u.gAs/L
F,
2544 BV
To 1.4mg F /L
2.0 mg/L
Effluent pH
•, C0 = 80u,g/L [40% As(lll)]
2000 4000 6000
Bed Volumes, BV
8000
Figure 1. Breakthrough curves for fluoride and arsenic from 28x48 mesh activated alumina
column, Run No. 2.
Run 8 resulted in a run length (252 BV)
which was only 4% of that using City Water
at pH 6.0 (8760 BV). These unadjusted-pH
runs were made to illustrate the short run
lengths that would occur in the simplest
POU treatment systems compared with a
pH-optimized system. Allowing the natural
pH (7.3) feedwater to contact fine alumina
produced a similar reduction in both ar-
senic and fluoride removal capacity. For
example, at the optimum pH of 6.0, the
arsenic and fluoride run lengths were 8760
and 2540 BV, respectively; whereas at pH
7.3, the run lengths were reduced to only
22% of the optimum values.
Regeneration of Alumina
Fluoride is more easily and completely
eluted from the exhausted alumina during
NaOH regeneration than is arsenic. This is
evident in Figure 3 containing typical re-
generation elution curves. The fluoride elu-
tion curve always begins slightly ahead of
the arsenic curve, and the arsenic curve
has a much longer tail. Only 60% to 70%
of the arsenic was recovered from fine- or
coarse-mesh alumina even when using
excessive (10 - 17 Ib NaOH/ft3 alumina)
regenerations. For both fluoride and ar-
senic removal, 3.0 gram-equivalents of the
dilute (1%) NaOH/L alumina (7.5 Ib NaOH/
ft3) eluted more arsenic than did the con-
centrated (4%) NaOH, and the fine-mesh
alumina permitted slightly higher arsenic
recoveries with both the 1% and 4% NaOH
solutions.
Neutralization of the NaOH-laden col-
umn with a relatively concentrated 2%
H2SO4 (0.4 N) solution applied at the same
flow rate (3.9 BV/hr) as the regenerant
Table 1.
Summary of Activated Alumina Result*
Run Number
Parameter
Mesh
Mesh Size
Condition*
Feedwaterff
FeedpH§
BVto1.4mgF/L
Days to 1.4 mg F/L
g F/m3 Adsorbed
to 1.4 mg FA.
BV to 50 mg AsA.
Days to 50 \ig As/L
g As/m3 Adsorbed to
50u.g/LAs
1
Coarse
14x28
New
CW
6.0
3080
10.8
3870
6840
23.8
390
2
Fine
28x48
New
CW
6.0
2540
9.0
4160
8760
30.4
575
3
Coarse
14x28
1 xReg.
CW
6.0
2380
8.1
3063
5880
20.4
380
4
Fine
28x48
2xReg.
CW
6.0
2380
8.1
3870
8040
27.9
575
5
Coarse
14x28
2xReg.
CW
6.0
1740
6.0
2260
4500
15.6
305
6
Fine
28x48
2 x Reg.
CW
6.0
2040
7.1
3130
6300
21.9
493
7
Fine
28x48
New
CW
7.3
547
1.9
925
1944
6.8
175
8
Coarse
14x28
New
No. 4
7.1
UL_L
w_^_
252
0.9
53
' once regenerated; 2 x Reg. * twice regenerated.
f For runs 1 through 7, San Ysidro City Water (CW) with 92± 10 \ig AsA. was pH adjusted to 6.0 before using
t For Run 8, Well No. 4 water with 230 \ig AsA. was ted.
§ No pH adjustment was made for runs 7 and 8.
-------
60
50
|> 40
I
I »
I
.o
1 20
10
Arsenic MCL 0.05 mg/L
Coarse
14 x 28 Mesh Alumina
(Run 1)
Fine
28x48 Mesh Alumina
(Run 2)
2000 4000 6000
Bed Volumes, BV
8000
Figure 2.
Comparison of arsenic breakthrough curves for coarse (14 x 28) and fine (28 x 48)
mesh aluminas.
immediately following regeneration works
well. This procedure is simpler than gradu-
ally decreasing the concentration of acid
following NaOH regeneration, a procedure
used in other fluoride removal applications.
The breakthrough curves for fluoride
adsorption on fine alumina are only slightly
affected by one or two regenerations. Re-
generations had a clearly negative effect,
however, on the run length to arsenic break-
through. After two regenerations, the BVto
50 jig As/L for the coarse-mesh alumina
dropped to 4500 from 6840, i.e., a 34%
reduction. The reduction in arsenic capac-
ity was smaller (28%) for the fine- as com-
pared with the coarse-mesh alumina.
Treatment of Spent Regenerant
The spent-regenerant solution was
acidified to pH 6.5 with HCI, settled for 24
hr, and filtered before being analyzed for
arsenic. The copreciprtation/filtration pro-
cedure removed essentially all of the As(V)
but only 36% of the As(lll), and 97% of the
arsenic remaining after precipitation was
As(lll). Therefore, if this procedure is to be
used in a full-scale application, any As(lll)
in the regenerant should be oxidized to
As(V). Based on results of our earlier As(lll)
oxidation studies, chlorine should be added
after the pH has been reduced to 6.5 to
take advantage of the much faster As(lll)
oxidation in the 6 to 10 pH range.
The settled, arsenic-contaminated alum
sludge produced in this manner amounted
to approximately 12% of the total initial
solution volume. Other investigators, using
a similar precipitation procedure on a 4%
NaOH spent regenerant (more concen-
trated), found the settled sludge to be 25%
and the filtered sludge solids to be less
than 1% of the original wastewater vol-
ume. Following the 24-hr extraction proce-
dure, the arsenic (total) concentration in
the EP test filtrate was 0.6 mg/L, i.e., far
below the 5.0 mg/L limit for classification
as a hazardous waste.
Arsenic Copreclpltatlon from Raw
Water
Iron hydroxide floe and hydrous iron
oxide solids can be used to remove ar-
senic from water by a mechanism of
coprecipitation or adsorption. Moreover,
As(V) is much more effectively removed by
ferric hydroxide than is As(lll). Based on
this knowledge, an attempt was made to
remove at least part of the arsenic from
San Ysidro City Water and Well No. 4
water by oxidation and precipitation of the
naturaMron present. The City Water (con-
taining 0.06 mg/L Fe(ll)) did not contain
enough iron to give a visible precipitate,
and no arsenic was removed by chbrina-
tion and filtration at either pH. Well No. 4
water containing 2.0 mg/L Fe(ll) is partially
treatable by this arsenic removal method.
At pH 7.1, 60% of the arsenic was re-
moved; at pH 8.5, somewhat less, 52%,
was removed.
Desalting Results
Based on percent removal of contami-
nants, the polyamide RO membrane clearly
gave the best TDS removals (97%) and
arsenic removals greater than 99% at 50%
recovery with a single pass. As predicted,
ED gave the poorest removal of arsenic—
presumably because molecular As(lll) could
not be transported out of the feedwater
using electrical current. ED cannot be rec-
ommended if As(lll) removal is a major
criterion. For example, with Well No. 4
water containing 188 mg As(lll)/L, only
28% of the arsenic was removed. This
conclusion contrasts with those made when
ED was used to remove fluoride, nitrate, or
chromate; in previous studies, ED per-
formed as well as or better than RO for
contaminant rejection. If ED is to be used
for As(lll) removal, preoxidation with chlo-
rine, for example, is required to convert
molecular As(lll) to ionic As(V).
Ion-Exchange Results
Although immediate breakthrough of
essentially all the As(lll) was expected, it
did not occur. Significant, immediate ar-
senic leakage was apparent, but it did not
reach the 25 to 36 mg As(lll)/L present in
the feedwater. Rather, about 200 BV was
treated before a level of 30 ng/L total ar-
senic was reached in the column effluent.
The 93 u,g/L total arsenic level in the feed
was not reached in the effluent until 570
BV was treated. Most importantly, the ar-
senic concentration in the effluent never
exceeded that of the influent. Thus, chro-
matographic peaking did not occur be-
cause of sulfate driving arsenic off the
column as has been regularly observed in
our prior laboratory studies. (An arsenic
peak, however, cannot be ruled out later in
the run.) In spite of the better-than-ex-
pected performance of these resins, they
did not perform well enough to be consid-
ered seriously as a viable treatment alter-
native (at least 400 BV).
Fluoride was not removed to any sig-
nificant extent by the chloride-form anion
resins. During treatment by Dowex-11 or
ASB-1 resins, fluoride reached 1.4 mg/L at
approximately 4 and 18 BV, respectively.
-------
Time, Minutes
2O 4O 6O 8O 1OO
8OOO .
6OOO .
4OOO -
20OO -
San Ysidro Alumina Regeneration Run No 2R
Regenerant« 4% (1 N) NaOH
Regeneration Rate - 26.7 mUmin, I.e., 4 BV/hr
E8Cr- 15min, BV- 4OO mL
01 234567
360
320
28O
24O
*
200 e
i
160 U
120
80
40
Figure 3. Arsenic and fluoride elution during a regeneration of fine-mesh alumina using 4%
NaOH. Run 2R—regeneration following exhaustion Run No. 2.
Such early breakthroughs resulted from
the very low affinity of fluoride for these
typical strong-base an ion resins which
clearly cannot be recommended for fluo-
ride removal.
The exhausted ion exchange columns
were completely and easily regenerated
with the use of approximately 3 BV of 1.0
N (6%) NaCI (11 Ib NaCI/ft3) in a cocurrent
(downflow) mode. The adsorbed arsenic
(presumably As(V)) was easily elutecl from
the exhausted resin.
The ease with which anion exchange
columns were regenerated in this study
and the fact that arsenic did not peak after
breakthrough suggested that ion exchange
should be further studied. These and pre-
vious results with strong-base anion ex-
change resins indicate a real potential for
chloride-form anion exchange for As(V)
removal. For example, our previous expe-
rience suggests that approximately 400 to
500 BV should be attainable before the
arsenic MCL is reached if the As(lll) is
oxidized to As(V) prior to ion exchange.
Point-of-Use (POU) Treatment
Six months after the San Ysidro project
began, the test results suggested no easy
solution to the combined arsenic/fluoride
contamination problem. Even if the water
was chlorinated to produce As(V), which
would yield alumina runs exceeding 20,000
BV, the fluoride present would force the
alumina runs to end at 2,000 BV so as not
to exceed 1.4 mg/L
The complexity of the alumina adsorp-
tion/regeneration cycle for a small commu-
nity, the anticipated short alumina runs
due to fluoride, the ineffectiveness of ion
exchange, and the anticipated sludge dis-
posal problem led to the consideration of
POU treatment employing RO. The Culligan
H-82 POU RO system that was tested had
a nominal capacity of 8 gal/day product
water. The system comprised a 10-u.m
cartridge filter; a granular activated carbon
(GAG) filter; a TFC RO membrane; a sec-
ond, smaller GAG filter; and, finally, a pres-
surized storage tank. (Other manufacturers
supply similar equipment.)
A salient feature of POU-RO units is
their low percent water recovery—typically
10% to 15%. This is both an advantage
and a disadvantage. With such low recov-
ery, there is no significant concentration of
the brine; therefore, membrane scaling and
fouling problems are minimal compared
with central treatment utilizing the typical
70% to 80% recovery. The disadvantage
is that only 10% to 15% of the feedwater is
recovered for drinking.
The initial results of the POU-RO pilot
test indicated 8 jig/L arsenic in the product
water when the feed water contained 90
jj.g/L. Subsequent arsenic analyses on the
product water from this unit yielded unde-
tectable arsenic levels, i.e., < 2 ng/L.
Conclusions
The existing San Ysidro City Water that
contains 810 mg/L TDS, 282 mg/L CaCO3
hardness, and 190 mg/L sodium and is
contaminated with 57 u,g As(V)/L, 31 p.g
As(lll)/L, and 2.0 mg FVL can employ acti-
vated alumina adsorption, RO, or possibly
ED to remove arsenic. The first two treat-
ment methods can be applied either in
central treatment or at POU. Preoxidation
using chlorine to convert As(lll) to As(V)
will definitely aid in removing arsenic but is
not essential. Significant oxidation of As(lll)
to As(V) appears to have occurred in all
the processes tested and consequently
better-than-expected removal of arsenic
occurred in all cases.
About 8800 bed volumes of
unchlorinated San Ysidro City Water ad-
justed to pH 6 could be continuously passed
through a virgin, fine mesh (28 x 48) acti-
vated alumina column before the arsenic
MCL was reached. Under similar condi-
tions, a run length of 6800 BV was ob-
tained for the coarse (14 x 28) mesh
alumina.
Feedwater pH was the most significant
variable in activated alumina treatment for
arsenic removal. At the natural pH of 7.3,
only 1900 BV could be treated before the
arsenic MCL was reached; this compares
to 8800 BV at pH 6.
In all the activated alumina tests, fluo-
ride broke through long before arsenic. For
example, using the fine-mesh alumina at
pH 6, fluoride reached 1.4 mg/L at 2500
BV whereas arsenic did not reach its 0.05
mg/L MCL until 8800 BV.
Even with excessive cocurrent regen-
erations employing 1% or 4% NaOH, a
maximum of 70% of the adsorbed arsenic
was recovered, and subsequent runs to
arsenic breakthrough were shorter than
with virgin alumina. During the third ex-
haustion cycle, the run lengths were re-
duced to 72% and 66% of the virgin
capacity for the fine- and coarse-mesh
-------
aluminas, respectively. Countercurrent
upflow regenerations were not attempted
in San Ysidro but would probably have
been more effective assuming that chan-
neling was avoided and adequate flow
distribution was achieved.
In the spent NaOH regenerant solution,
99.8% of the As(V) and 36% of the As(lll)
were removed by copreciprtation with the
AI(OH)3, which was produced when the
spent regenerant solution was acidified to
pH 6.5 using HCI. The total arsenic re-
maining in solution after precipitation was
0.92 mg/L, of which 97% was As(lll). The
arsenic-contaminated AI(OH)3 sludge re-
sulting from the pH 6.5 precipitation proce-
dure on the dilute (1% NaOH) regenerant
was 12% of the solution volume after 24-hr
settling. The dried sludge (7.8 g/L of spent
regenerant) was subjected to the EPA EP
toxicity test and easily passed.
ED with no pretreatment except car-
tridge filtration reduced the City Water ar-
senic by 73%, from 85 down to 23 jig/L,
while reducing the TDS by 72%. ED was
not effective, however, in removing As(lll)
from the anaerobic Well No. 4 water. There
arsenic was only reduced by 28%, from
188 ng/L down to 136 u.g/L
Both the CTA and the IPA hollow fiber
RO membranes did an excellent job (>
97% and > 99% removal, respectively) in
removing arsenic from the City Water with-
out prechlorination to convert As(lll) to
As(V). Greater than 94% removals for both
TDS and fluoride were also obtained For
all contaminants, the PA membrane per-
formance was superior. Thus, RO with
pretreatment consisting of SHMP addition,
cartridge filtration, and possible pH adjust-
ment to 6.0 is a technically effective, but
costly, means of treating waters like San
Ysidro City Water.
Even though the City Water contained
40% As(lll) which is nonionic at the natural
pH of 7.2, ion-exchange with chloride-form
strong-base resins worked reasonably well
in reducing the total arsenic concentration.
Before the arsenic MCL was reached, 160
to 220 BV could be treated. Arsenic leak-
age, primarily As(lll), was substantial, how-
ever, and the runs were too short to
seriously consider ion exchange as a treat-
ment method. (Chlorine oxidation of the
As(lll) would probably increase the run
lengths to 500 BV.)
A POU RO system containing a TFC
membrane achieved 95% overall reduc-
tion in TDS and a 91% removal of arsenic,
which appeared to improve with time. POU
RO treatment is attractive for this applica-
tion because of the small size of the com-
munity, the multiple contaminants in the
water, and the fact that no pretreatment of
the raw water would be necessary.
Recommendations
A POU RO treatment system study
was recommended in San Ysidro as a
result of the findings of this research. Such
a study was undertaken in San Ysidro with
the result that POU RO treatment was
found to be an "effective, economical, reli-
able and viable alternative to central treat-
ment" for removing arsenic and other
contaminants (K. R. Rogers, EPA/600/2-
89-050, March 1990).
The full report was submitted in fulfill-
ment of Cooperative Agreement No.
807939 by the University of Houston under
the partial sponsorship of the U.S. Envi-
ronmental Protection Agency.
&U.S. GOVERNMENT PRINTING OFFICE: 1991 - 54*-OUt/400ll
-------
•—• —
^ss^^^sf^^^^
I he complete report entitled *A.C6r (S0G >1«*~—'
Springfield, VA 22161
Th* CDA T°l0Phone: 703-487-4650
United States
Environmental Protection
Agency
BULK RATE
usness
Penalty for Private Use $300
EPA/600/S2-91/011
------- |