Environ. Sci. Technol.

1989, 23,566-572

Silica Polymerization and Other Factors in Iron Control by Sodium Silicate

and Sodium Hypochlorite Additions
Michael G. Browman,+ R. Bruce Robinson,* and Gregory D. Reed
Department of Civil Engineering, 73E Perkins Hall, University of Tennessee, Knoxville, Tennessee 37996

4 The effect of silicate type and dilution on iron control Table I. Properties of Undiluted Sodium Silicate Liquids
Used in Investigations ( 6 )
by silicate/hypochlorite additions was examined. The
effectiveness of four silicate types (Si02/Na20ratios of product wt ratio SiOz, viscos, d,"
3.75, 3.22, 2.50, and 1.80), each undiluted and at four di- name SiOz/NazO % CP g/cm3 PH
lutions ( l / l O , 1/50,1/100, and 1/500), was evaluated by
s35 3.75 25.3 220 1.31 11.1
an optimized method of chemical addition (silicate 15 s N 3.22 28.7 180 1.38 11.3
before hypochlorite) and with the pH controlled at 7. With Star 2.50 26.5 60 1.40 11.9
filterable iron (<0.1 pm) and turbidity as measures of Starso 1.80 24.1 60 1.43 12.9
treatment success, it was observed that more concentrated
silicate feed solutions and silicates with higher SiOz/NazO OAt 20 "C.
ratios were more effective, i.e., better sequestrants. The
four undiluted silicate liquids and dilutions of the 3.22 ratio experiments ( 4 ) suggests the involvement of colloidal in-
silicate were characterized by reaction coefficients related teractions in the mechanism of sequestration.
to silica polymerization (K,and Po). Significantly, Despite the time elapsed since the technique was first
treatment effectiveness was favored by the presence of utilized for iron control and despite potential economic
complex higher molecular weight silica polymers, indicated benefits of tens of millions of dollars, there has been little
by low K1 values, and/or the existence of a high proportion research on the sequestration of iron by silicates. There
of the silica as complex polymers, indicated by Po. U1- is a need to establish (1) the major factors controlling
trafiltration studies of treatment systems revealed a pat- treatment success, (2) the limits of the treatment and the
tern of progressive growth with time of iron-containing optimal conditions for treatment success, and (3) the
colloids that is retarded more effectively by higher ratio mechanism of the sequestration.
silicates and more concentrated silicate solutions. To this end, the specific objectives of this investigation
were (1) to examine the role of sodium silicate type, dilu-
Introduction tion, and addition method, as well as pH in achieving
Dissolved iron is a common problem in US.groundwater successful iron treatment via sequestration and (2) by
supplies. Red water complaints are likely when the means of ultrafiltration techniques, to examine the role
groundwater is pumped to the surface and the ferrous iron of colloids/microparticulates in treatment success.
is oxidized, unless the iron is controlled either by removing Notably, in this paper, the term silica is used in its most
it or by "sequestering" it, i.e., producing a clear water general sense, i.e., to comprise all forms of silicon dioxide
without removing the iron. represented by the formula (SiOz.xHzO),, where x is from
The low cost and operational simplicity of the seques- 0 to 2, and n is from 1to infinity (5). Also, although clear
tration technique make it potentially more attractive to silicate liquids may contain colloids, these are described
small resource-limited communities than the more complex as solutions.
and expensive conventional practice of iron removal. Experimental Section
Nearly simultaneous additions of sodium silicate and Efficiency of Treatment. Model iron-containing
chlorine have been used successfully to sequester iron. groundwaters were treated with sodium silicate and sodium
Notably, sequestration does not remove the iron but re- hypochlorite to compare treatment efficiency with the
duces the turbidity and color problems commonly asso- characteristics of the silicate feed solutions. The silicate
ciated with iron. Iron removal is often unnecessary, be- liquids tested had SiOz/NazOratios by weight of 1.fQ2.50,
cause it is an aesthetic and not a health problem. 3.22 (N-silicate), and 3.75. General physical and chemical
In 1950, Henry (1) conducted field tests with sodium properties of these silicates are presented in Table I (6).
silicate (SiOz/Na20= 3.22, hereinafter 3.22 ratio silicate), These silicates were fed undiluted and at dilutions of 1/10,
in association with chlorination, on well water and obtained 1/50,1/100,1/500. All dilutions were aged 24 h to allow
good iron treatment. Concentrated silicate was diluted the silicate species to approach a new equilibrium and thus
9-fold to provide a 3.2% Si02 feed solution and dosed to increase experimentalreproducibility. Also, a 24-h aging
continuously at 15 mg of SiOz/L. The iron (0.2-0.9 mg/L) time would approximate actual practice. The model waters
was sequestered for 3 days. Subsequent tests were con- were prepared, dosed, and sampled as follows:
ducted by Dart and Foley ( 2 , 3 )on municipal systems with 1. Deionized water, 22 L, was sparged with nitrogen gas
near-neutral pH waters of soft to medium hardness and until the oxygen concentration measured approximately
containing 0.2-1.45 mg of Fe/L and 15-20 mg/L native zero with a D.O. meter.
SiOz. Satisfactory treatment was achieved with the ap- 2. Alkalinity was added at 252 mg/L NaHCO,.
plication of chlorine or hypochlorite and 3.22 ratio silicate 3. Carbon dioxide and nitrogen were blended in to
fed at or near full strength but dosed at 2.7-6 mg of achieve pH 7.
SiOz/L. Regarding the mechanism of iron sequestration 4. Ferrous chloride, FeC12, was added to give 2 mg of
by silicate, Dart and Foley ( 2 , 3 )suggested one based on Fe/L.
an iron-silicate complex. In contrast, other evidence from 5. A 2-L portion of the ferrous iron solution was
actual treatment systems (1)and from allied iron-silica transferred to a Nalgene bottle under a nitrogen/carbon
dioxide atmosphere and a ferrous iron sample collected.
Present address: Tennessee Valley Authority, 270 Haney 6. The solution was mixed (velocity gradient, 5900 s-l)
Building, Chattanooga, TN 37402-2801. under the same atmosphere and dosed appropriately with
586 Environ. Sci. Technol., Vol. 23, No. 5, 1989 0013-936X/89/0923-0566$01.50/0 0 1989 American Chemical Society
sodium silicate (9-15 mg of Si02/L) and sodium hypo- was measured at 400 nm. This latter step is necessary to
chlorite (-2.4 mg of OCl-/L). convert the initial 8-silicomolybdicacid to a-silicomolybdic
7. Dissolved oxygen and pH were measured in situ. acid within a convenient time. The postheating absorbance
8. Samples were taken for subsequent measurement of value and the ratio of extinction coefficients for a- and
turbidity (immediately), free chlorine, alkalinity, silica, and P-silicomolybdate are used to calculate the theoretical
various iron fractions. absorbance of the 8-silicomolybdatecorresponding to the
9. Procedures in steps 5-8 were repeated for each test total amount of SiOz present. Analysis of the data used
treatment. two equations developed by O’Connor to determine
coefficients indicative of the fraction of the totalsilica that
10. Bottles were shaken for 10 days at 20 “C and sam- is polymerized and the degree of silica polymerization in
ples collected on days 0, 1,3,5,7, and 10 for filterable (C0.1
this fraction.
pm) and total iron, turbidity, and pH.
In the first equation, the percent Si02 remaining un-
Total and filterable iron were measured on acidified reacted with molybdic acid is given by
samples (pH < 2) by atomic absorption spectroscopy.
Ferrous iron was measured by the bathophenanthroline 100(ODT- ODJ/ODT = 100 - % color developed (1)
method of Stumm and Lee (7); initial ferrous iron con- where OD, is the optical density of the solution at time t ,
centrations were always 294% of the total iron. Silica was and ODT is the total optical density of the 8-silico-
measured at first by the molybdate method (8) and later molybdate complex corresponding to the total amount of
by atomic absorption. Measurements of alkalinity, pH, SiOz present. The second equation is a combination of eq
and turbidity were according to standard methods (8). 1 and another equation relating the % color developed”
Free chlorine was measured by the NJV-diethyl-p-(DPD) to reaction coefficients characterizing the silica present,
method (8)as modified by the Hach Co. (9) for rapid field assuming sequential first-order reactions for silica depo-
determinations. Dissolved‘oxygen was monitored with a lymerization and conversion of 8-silicomolybdate complex
Yellow Springs Instrument Co. Model 54 meter using a to the a-complex. The resultant equation is given by
submersible polarographic probe.
Ultrafiltration Tests. Model groundwaters were
treated with undiluted and 1/100 dilutions of 3.75 and 3.22
Si02/Na20ratio silicates as above, but extra samples were
% unreacted Si02 = 43.5 - [ K1- 0.565Ko
KO-K1 -

taken for ultrafiltration.

Samples for ultrafiltration were prefiltered (0.1-pm 0.435[ 100 - s ] e - K d (2)
membrane filter) to reduce potential clogging of the ul-
trafiltration membranes. The resultant filtrate was then where K1 is the rate constant for the formation of 8-sili-
passed through either a 1000 or 200 OOO molecular weight comolybdate complex from the depolymerization of the
(MW) cutoff Micro Filtration Systkm (MFS) ultrafiltration silicate polymers, Po is the initial fraction of polymerized
membrane. The 1000 MW cutoff membrane system was silicate, KOis the rate constant for conversion of the /3-
run at a pressure of 55 psig and the 200000 MW cutoff complex to the a-complex, and t is the time elapsed.
system at a pressure only slightly greater than atmospheric. Values of K1 and Po were calculated from graphical
Ultrafiltration membranes were soaked for at least 25 h analysis of data in accordance with eq 2.
in approximately 0.01 N H2S04and then for 10-15 min
in deionized water and finally flushed with deionized water Results and Discussion
(approximately 10 and 90 mL for the 100 MW cutoff and Efficiency of Treatment. Preliminary experiments
200 000 MW cutoff membranes, respectively) prior to were necessary to evaluate the effects of pH and of the lag
sample processing. Immediately before ultrafiltration, the time between the addition of the silicate and the hypo-
prefiltered solution to be processed was made 0.1 M in chlorite. A standard experimental methodology was then
NaCl by the addition of dry salt to reduce polarization selected based on the results. Filterable iron (CO.l pm)
effects. (In a preliminary experiment, 2 mg/L ferric iron is a key variable in the discussions of all of the experiments
at a low pH and 0.1 M NaCl showed 100% filterability on iron sequestration by sodium silicate and hypochlorite.
throughout the 1000 MW cutoff membrane; hence, po- Filterabilities over 95% corresponded to waters that were
larization was not believed to be a problem.) The volume clear and of low turbidity.
of solution for processing was initially approximately 60 The effects of varying lag times from adding the sodium
mL (approximately 5 mL was used to rinse the insides of hypochlorite (the oxidant) 2 min ahead of the sodium
the ultrafiltration unit, the first 5 mL of ultrafiltrate was silicate (the sequestrant) to adding it 1 min after are shown
discarded, and the next 25 mL was saved for analysis). in Figure 1. In these tests the silicate was added undiluted
Total iron and silica analyses were made of the filtrates and the pH was uncontrolled after initial adjustment to
from the 0.1-pm membrane filter and the 1000 or 200 000 pH 7. When the hypochlorite was added before the silicate
MW cutoff membrane. in these experiments, the iron was oxidized, hydrolyzed,
Silicate Characterization. The sodium silicate feed and precipitated before contacting the silicate, producing
solutions used to investigate the efficiency of treatment significant turbidity and resulting in zero filterable iron
were characterized for the percentage of polymer present for the duration of the test. The optimal lag time was the
and the degree of polymerization. The method of addition of silicate at least 15 s ahead of the hypochlorite,
O’Connor was followed (IO): 200 mL of molybdic acid as evidenced by the associated filterable iron of 95100%.
reagent (11) was combined with an aliquot of the silicate Adding the silicate and hypochlorite simultaneously did
feed solution, mixed for 20 min, transferred to a l-cm not provide sufficient time for mixing of the silicate which,
quartz cell, and the rate of color development measured in addition to being quite viscous (Table I), was added in
at 400 nm on a double-beam spectrophotometer. Absor- a much smaller volume (-0.05 mL) than the sodium hy-
bance was followed until the full color developed. Ex- pochlorite (-7 mL). The implications of the results from
periments were performed at 21 and 25 OC. Duplicate the filterable iron were corroborated by the corresponding
samples of the silicate/molybdic acid solutions were heated turbidity results (12) for which higher turbidities were
in capped bottles at 290 OC for 2 3 h, and then absorbance associated, broadly, with lower levels of filterable iron. The
Environ. Sci. Technol., Vol. 23, No. 5, 1989 567
 -
*--e1/50
undiiuted
1/10

*--*i / I O O
0--4 1/500

m
* -
c---.silicate and hypochlorite
simultaneously
siiicate 1 5 sec before hypochlorite

-
~

o----o silicate 30 sec before hypochlorite

c--+ silicate 60 sec before hypochlorite
hypochlorite z min before silicate

0 0 0
0 1 2 3 4 5 6 7 8 9 1 0
TIME ELAPSED, DAYS 0 1 2 3 4 5 6 7 8 9 1 0

Figure 1. Lag-time effects on iron filterability for undiluted 3.22 ratio

sodium silicate.

M 1/10
0.9- +--o 1/50
---*1/100
0.8- +--c 1/500

PH controlled 0.73
at PH = 7.1
PH uncontrolled
PH of subsample
readjusted to 7.1

pH=7.1, I
, I
0 1 i 3 b 5 6 7 ~ 9 l 0
O 01
- b3 5 7 io
TIME ELAPSED, DAYS
TIME ELAPSED, DAYS
Flgure 2. pH effects on iron filterability for 1/50 dilutions of 3.22 ratio Flgure 4. Dilution effects on turbidity for 3.22 ratio sodium silicate at
sodium silicate at 9 mg of SiO,/L. 13 mg of SiO,/L with pH controlled at 7.

success of the silicate-before-hypochlorite sequence is The effects of dilution of the silicate feed solution and
consistent with the findings of Hale [ascited by Henry ( I ) ] . of silicate type on treatment effectiveness were examined
A 15-s lag time gave good results and was chosen for the in detail. Tests were conducted a t pH 7 and a t a 13-15
remainder of the experiments. mg of Si02/L dosage using four sodium silicates with
The effect of pH on treatment effectiveness for a 1/50 Si02/Na20ratios of 3.75, 3.22,2.50, and 1.80. The chosen
dilution of 3.22 silicate is shown in Figure 2. The pH was silica dosage gave good differentiation among the treat-
controlled at 7.1 in one experiment and left uncontrolled ment performances of the various ratio silicates and dilu-
in the other. With the pH controlled at 7.1, filterable iron tions. Lower dosages gave nearly 0% filterable iron from
decreased to zero within 3 days. In the uncontrolled early in the test period for the lower ratios and higher
system, the pH drifted from an initial value of 7.05 to 8.3 dilutions. Conversely, higher dosages gave nearly 100%
within 10 days and produced better results than the other filterable iron for the duration of the test period for the
system. Filterable iron remained a t 100% for the first 3 higher ratios and lower dilutions. The treatment effec-
days and a t >90% for up to 164 days. Adjusting the pH tiveness of 3.22 ratio silicate utilizing undiluted and 1/10,
of a subsample of this water to pH 7.1 on day 161 resulted 1/50,1/100, and 1/500 dilutions dosed at 13 mg of Si02/L
in obvious precipitation and settling out of the iron within is presented in Figure 3. A 3.22 ratio is commonly used
3 days. The observed role of pH agrees with the sugges- in practice. Treatment success was largely inversely related
tions of Dart and Foley ( 3 ) ,who reported that reduced to the dilution of the feed solution. The inverse rela-
silica dosages were required for satisfactory treatment at tionship between treatment effectiveness and silicate di-
pH levels greater than 7.5. Clearly, pH affects treatment lution is also demonstrated in Figure 4, the lower turbid-
success significantly. However, a detailed examination of ities indicating better performance. The pH of these test
the role of pH was not one of the major objectives of this systems increased gradually (-0.1 unit/2-3 days) between
investigation. For this reason and because of the need to samplings, at which time the pH was readjusted to 7.0 with
control this important variable a t a realistic level, the C 0 2 ( g )(12). The total iron concentrations in the system
remaining experiments were conducted with the pH con- remained constant at -2 mg of Fe/L (12)for the duration
trolled at -7. Also, higher silica doses favor better iron of the experiment.
treatment (13). Thus, in experiments investigatingthe role Partial results for the comparative effectiveness of the
of other parameters, silica doses within a narrow range various silicate types tested are shown in Figure 5. Here,
were used. changes in filterable iron with time are presented for the
568 Envlron. Sci. Technol., Vol. 23, No. 5, 1989
 ,,3.75 SiO,/No,O

\ P \

TIME ELAPSED, DAYS

Flgure 5. Effects of different Si02/Na10ratio sodium silicates on iron

fllterabllity at 13-15 mg of S102/L wlth pH controlled at 7.

Table 11. Comparative Effectiveness of Various Dilutions

of Selected Sodium Silicates at Maintaining Iron 10 20 30 4 0 50 60 70 BO 90 100 110 120 130 140
Filterability" TIME ELAPSED, MINUTES

time elapsed to 50% of filterability Figure 6. Percent unreacted silica vs time for the reactbn of molybdic
for silicate with SiOz/Na20ratio, days acid with Undiluted 1.8, 2.5, 3.22, and 3.75 ratio sodlum silicates at
25-26 O C .
dilution 3.75 3.22 2.5 1.8
undiluted 9.8 6.7 5.8 2.4
1/10 8.7 7.6 4.1 2.4
1/50 6.2 5.9 4.0 2.0
1/100 5.5 4.2 3.8 1.9
11500 2.1 1.0 1.1 0
"At 13-15 mg of SiOz/L.

undiluted silicates only. Results for the dilutions are

omitted for clarity. However, Table I1 summarizes the s
treatment data for all of the dilutions for each of the four f e / 5 0 0
silicate types in terms of t m values, i.e., the time for the n 10-
w 9-
filterable iron to reach 50%. These data show the same L4 6-
7-
pattern as for the 3.22 ratio silicate, Le., superior per- 6-
formance by the more concentrated feed solutions for a f 5-

given silica dosage. In addition both Figure 5 and Table

I1 indicate clearly that better treatment success is obtained
with the higher ratio silicates.
Silicate Characterization. Characterization curves
for the reaction with molybdic acid of undiluted solutions
of four silicate types are presented in Figure 6. Because
of the density of the observed data points and the absence
of outliers, for the first 15 min of each curve only every
't
I L , I

IO 20 x)
I I I

40 so 60
I I 1 '

m so so loo
T I M E ELAPSED, MINUTES
I "
110
' 1
120 1x1140

other point is plotted. These curves show that the lower

ratio (SiOz/NazO)silicates react with molybdic acid more Figure 7. Percent unreacted silica vs time for the reaction of molybdic
acid with undiluted, and 1/10, 1/50, 1/100, and 1/500 dilutions of 3.22
rapidly than the higher ratio ones. ratio sodium silicate at 20-21 OC.
Mathematical analysis of each curve in accordance with
eq 2 quantifies the reaction rate of the major silica polymer show that while there is a high proportion of polymerized
fraction, K,, and the fraction of silicate present in polymer silica in each of these materials there is no definitive trend
form, Po.Within at least the first 15 min, the contribution in this characteristic associated with silicate ratio. K1
of KOin eq 2 is insignificant and can be neglected, i.e., KO values, which are related to the depolymerization rates of
= 0. Equation 2 then simplifies to complex silica, reflect the degree of silica polymerization
% unreacted SiOz = Poe-Klt (3) (10). Ideally, K1 values and polymer complexity are in-
versely related. Thus, the silica polymers in the 3.75 ratio
A plot of log (% unreacted SiOJ versus time as in Figures silicate (K,= 0.020 min-') are more complex and/or of
6 and 7 should be linear for the first several minutes until higher molecular weight than those for the 1.80 ratio sil-
the influence of KObecomes significant. The initial slope icate ( K , = 0.18 min-l). These results are consistent with
of the line is -K1, and the y intercept corresponds to Po. findings of Iler (5) and O'Connor (10) for a variety of other
Accordingly, eq 2 is used to characterize the silicates using ratio silicates.
the first linear portion of the data. Reactivity data were also obtained for the 3.22 ratio
The Kl values (Figure 6) show quantitatively that among silicate, undiluted and as 1/10, 1/50, 1/100, and 1/500
the undiluted silicates the higher ratio silicates react more dilutions (Figure 7 ) ,at 21 "C, the temperature at which
slowly than the lower ratio material. Po values (Figure 6) all of the iron treatment experiments were executed. As
Environ. Sci. Technol., Vol. 23, No. 5, 1989 569
 \

"T 11:
\
10.0 - \ e - 100

8.0D -
9 - \o

\
\ - 70
60- \
\
- 60
I
8
'
\e
8 50- I C---+l SILICA,
- . a: 1,000 MW C u t o f f
',
\ c
50
a0 40- --o.ium
\
\
~
40 m
2 30-
\
I
\
'
&---c)

o--+ -
-200,000 MW C u t o f f
1,000 MW cutoff

t1
\ W
30
20-
3.0
2.0
\ 20 4

0.01 , , , , I , ,
i4 lo
0
--Fe %----c---- -c- -- ----a
,
0 i 2 3 4 5 6 7 8 9 IO
-1.S -1.0 -0.5 0.0 0.5 1.0 1.5
T I M E ELAPSED, DAYS
LOG (%SiO$
Flgure 9. Iron filterability for an undiluted 3.22 ratio sodium silicate
Flgure 8. Silicate reactivity coefficients derived from the reaction of system at 13.9 mg of SiO,/L under controlled pH using standard fil-
molybdic acid with dilutions of 3.22 ratio sodium silicate at 21 O C . tration and ultrafiltration.

in Figure 6, in Figure 7 within the first 15 min only the 100

alternate data points are plotted for each curve. Reaction UNDILUTED, 3.75 SiO, /NapO
coefficients for the undiluted and the dilutions of the 3.22
- - -4 1/100, 3.75Si0,/Nap0
ratio silicate are given in Figure 8. The SiOz concentra-
tions on the x axis are given in logarithmic form solely for UNDILUTED, 3 . 2 2 SiO,/Na,O
convenience of presentation. For the linear phase of each i
--
, 1/100,3.22 SiO,/Na,O
of the plots from which the reaction coefficients were
derived, the magnitude of the correlation coefficient for
the fit of the best-fit plot to the data indicated significance
at the 99% level of probability.
The Po and K1 values increase and decrease respectively,
with increasing concentration. The trends in Po indicate
a higher proportion of the total silica to be in polymeric
form in the more concentrated solutions. In addition, the
trends in K1 show the silica polymers to be more complex
in the more concentrated solutions. O'Connor (IO)found TIME ELAPSED, DAYS
similar trends in Po and K1 for dilutions of 3.3 ratio silicate Figure 10. Effect of dilution and SiO,/Na,O ratio on iron filterability
within the tested range of 2-30% SiOz. By applying the at 12.5-14 mg of SiO,/L under controlled pH using a 200000 MW
results from the silicate characterization (Figures 6-8) to cutoff membrane.
those from the efficiency of treatment (Figures 3 and 5;
Table 11),the correlation is unmistakable. Higher Po and consistent with progressive growth in particle size of col-
lower K , values are associated with better treatment loidal iron. Notably, over 90% of the iron did not pass
success. Accordingly, the combination coefficient, Po/Kl, through the 1000 MW cutoff membrane filter on day zero.
can be used as a broad indicator of the sequestration po- Therefore, it is further concluded that the iron-containing
tential of silicate solution, higher values indicating greater colloid is created immediately after treatment. Also, the
sequestration capability. Thus, undiluted 3.22 ratio silicate role of simple complexation appears to be minimal at this
with a Po/K, value of 3930%.min is a much better iron point or a t least transitory. The average pore diameters
sequestrant than the 11500 dilution for which Po/Kl is can be estimated as 0.0013 and 0.0075 pm for the 1000 and
180%.min. The logical conclusion is that more concen- 200 000 MW cutoff of filters, respectively.
trated silicate solutions with higher proportions of silica Figure 10 compares the filterable iron through the
as complex polymer are likely to be more effective se- 200 000 MW cutoff filter for each of the four test systems.
questrants of iron. The order of filterable iron for the various ratio silicates
Ultrafiltration. Further insights into the possible was undiluted 3.75 ratio 1 undiluted 3.22 ratio 1 1/100
mechanisms by which silicates effect iron sequestration dilution 3.75 ratio I 1/100 dilution 3.22 ratio. This order
are provided by the ultrafiltration experiments. Figure is consistent with that observed earlier for the filterable
9 shows filterable iron versus time using a series of pro- iron with a 0.1-pm filter. In addition, with the 200000 MW
gressively finer membrane filters for a system treated with cutoff filter differences among the treatments are observed
undiluted 3.22 silicate. Filterable silica for the 1000 MW much earlier than with the coarser 0.1-pm filter. The
cutoff membrane filter is also included. The results ultrafiltration results reaffirm that the difference among
presented in Figure 9 are typical of the other similarly treatment systems is reflected in the distribution of iron
tested systems-1/100 dilution 3.22 ratio silicate, and particle sizes at any given time and the rate of particle
undiluted and 1/100 dilution 3.75 ratio silicate. growth with time. The more successful treatments main-
Significantly, for each test system the pattern of fil- tain finer particle size distributions for more extended
terable iron through the various membranes filters was 0.1 periods.
pm > 200000 MW cutoff 2 1000 MW cutoff. The filter- The pattern of filterable silica (<lo00 MW cutoff) shown
able iron through the 1000 MW cutoff decreased to nearly in Figure 9 is typical of the four systems tested by ul-
zero first, then for the 200 000 MW cutoff, and last, if at trafiltration. The filterable silica in all instances was less
all, for the 0.1-pm membrane filter. These results are than 82% on day zero and increased progressively with
570 Environ. Sci. Technol., Vol. 23, No. 5, 1989
time over the 3 days of data. In contrast, other tests with Table 111. Theoretical Speciations
a freshly prepared iron-free 10 mg of Si02/L solution of
3.22 ratio sodium silicate at pH 7.2 showed the filterable sDecies concn. mmol/L
~ ~~~

silica to be 97.6% using a lo00 MW cutoff ultrafilter. This no precip amorphous iron
agrees with reports of rapid silica depolymerization when species allowed oxhydroxide precip
silicate solutions are diluted to low concentrations (5). On init NaHC03 1.5000 1.5000
the other hand, the filterability of undiluted 3.22 ratio init FeCl, 0.0358 0.0358
sodium silicate would be approximately 60% (5), because init silica as H4SiO4 0.2660 0.2660
that solution contains a large percentage of high molecular init NaOCl 0.0360 0.1540
weight silica species. These filterable silica and iron results Na+ 1.535 1.535
taken together suggest a close relationship between the NaCOF 1.080 X 1.081 X
NaHC02 3.685 X lo4 8.688 X lo4
release of filterable silica and the loss of iron filterability Fez+ 1.212 x 10-19 <1o-m
or growth of iron colloids. This is consistent with earlier FeOH+ 40-20 a0-m
findings that iron and other metal ions retard silica de- Fe(OH)c <1o-m a0-m
polymerization (5). Fe(0H); <10-20 <lo-%
Fe3+ 2.046 X 1.113 X
Discussion of Treatment Mechanism FeOH2+ 8.080 X lo4 4.398 X lo4
FeC12+ 3.722 X 2.026 X
The suggested mechanisms for the role of silicate in iron FeC12+ 1.478 X 8.052 X
sequestration are (1)formation of an iron-silica complex FeC12 a0-m a0-m
and (2) stabilization of an iron-containing colloid. The Fe(OH)2t 0.03183 1.734 X
results of the current investigation and of allied iron- Fe(OH)t 0.003586 1.954 X lo4
silicate experiments published elsewhere support a col- Fe(OH), 3.739 x 10-4 2.037 X
loidal mechanism and refute an iron-silica complex. The Fez(OH)2+ 3.051 X 9.025 X
Fe3(0H)45t 2.783 X lo-" a0-m
theoretical treatment mechanism must account for at least FeOSi(OH):+ 7.492 X 4.082 X
several phenomena: (1)the results from the ultrafiltration H+ 10"' 10"'
experiments of this investigation, (2) the strong adverse OH- 7.138 X 7.137 X
effects on treatment of calcium as compared with sodium, c1- 0.0895 0.0895
chloride, sulfate, and bicarbonate, (3) the absence of high c0:- 5.798 X 5.794 x 104
turbidity but the observation of Tyndall scattering, (4)the HC0,- 1.218 1.128
importance of pH, (5) the importance of chlorine, and (6) H2C03 0.2803 0.2804
the metastable nature of the treatment.
oc1- 0.0181 0.0181
HOC1 4.685 X 4.686 X
It has been proposed that the formation of an iron- H2Si02- 3.185 X 10" 3.182 X lo4
silicate complex, FeOSi(OH)32+,prevented iron precipi- H3Si04- 2.518 X lo4 2.518 X lo4
tation (2). This was based on the strength of the inter- H4Si04 0.2657 0.2657
action between Fe3+and (OH),SiO- ions as reflected in the amorphous iron oxyhydroxide NA 0.03579
equilibrium formation constant, K = 5.69 X lo9 (14). precip (ferrihydrite)
Notably, competing complexation and precipitation reac-
tions were not considered in this proposed mechanism. for by equilibrium complexes. In contrast, the observed
In other experiments (3)the stabilized iron was removed metastability is consistent with the thermodynamic in-
quantitatively by several anionic resins but poorly and stability of colloidal metal oxides (16) such as hydrous
inconsistently by cationic resins. As a result, an anionic ferric oxide. Also, the presence of a colloidal sol accounts
complexation mechanism was proposed. However, major for the Tyndall scattering reported by Henry (1). Because
inconsistencies in the retention of the iron and the silicon the particle sizes involved are much less than the wave-
from the sequestered iron-silicate combination by these length of light, no significant turbidity is measured (17).
resins make the interpretation of these results ambiguous. In related studies, Hazel (4) demonstrated that sodium
A more thorough examination of the reaction system silicate could coagulate and disperse preformed ferric ox-
allowing the inclusion of other non-silicate complexes and yhydroxide sols (80 mg of Fe/L) in accordance with other
species diminishes the likely role of the FeOSi(OH)32+ observations of the dispersing power of sodium silicates
complex. Table I11 lists the species predicted by the (18). The addition of sodium silicate first coagulated
MINTEQ computer model (15) for a water containing 2 positively charged iron sols by reducing the positive charge.
mg/L ferrous iron, 91 mg/L bicarbonate, 16 mg/L silicic Further silicate addition dispersed the particles by re-
acid as Si02,and sufficient hypochlorite to oxidize the iron. versing the net particle charge to negative. Moreover, it
If no iron precipitate is allowed, then nearly all of the iron was observed that higher ratio silicates were more effective
is predicted to be complexed in the species Fe(OH)2+and at this charge reversal. These results are consistent with
Fe(OH),O;very little is complexed with orthosilicate. If colloidal stability theory and are directly relevant and
an amorphous iron oxyhydroxide solid phase is allowed, applicable to the current investigation. Hazel's findings
then essentially all of the iron is precipitated. Hence, parallel exactly the results of the current investigation of
equilibrium calculations do not support the iron-silica the treatment effectiveness of higher ratio silicates and
complex, FeOSi(OH)32+,but do support the existence of those of others on the effects of calcium (1, 13, 19).
a solid phase. In the silicate/hypochlorite treatment system, as for the
Furthermore, the ultrafiltration results (Figure 9) show preformed, iron sol system, the more complex silica poly-
the major fraction of the sequestered iron to be in the mers, which are provided by the higher ratio silicate, would
particulate form within minutes of treatment, whereas have lower points of zero charge (PZC) (20). For this
simple complexes such as FeOSi(OH)32+should remain reason, the higher ratio silicates at a given pH > PZC are
quantitatively filterable (<lo00 MW cutoff). Other results more effective at producing the negative charges necessary
such as the ineffectiveness of orthosilicates and naturally for colloidal dispersion. Also, higher pH levels promote
occurring silicic acid as sequestrants as compared with the dissociation of particle surface hydrogens. The resultant
synthetic polymeric liquid silicates, as well as the me- residual negative charges favor dispersion and thus a more
tastability of the silicate treatment, are difficult to account stable colloidal system.
Environ. Sci. Technol., Vol. 23, No. 5, 1989 571
 The responses of iron-silicate systems to the presence acknowledge the assistance of Joseph M. Holden, Dianne
of hardness cations such as Ca2+also can be interpreted Delzer, and Daniel David. The sodium silicates were
readily in terms of standard colloidal theory. The in- provided gratis by the PQ Corp.
creased silicate needed to achieve iron dispersion/se-
questration in the presence of Ca2+has been observed in Registry No. Fe, 7439-89-6;SOz, 7631-86-9;NaOC1,7681-52-9;
several instances (3, 4, 19, 21). Robinson et al. (19) sodium silicate. 1344-09-8.
quantified the effect of calcium, sodium, sulfate, chloride,
and bicarbonate on treatment. Statistical regression es- Literature Cited
timated the adverse effect of calcium to be approximately (1) Henry, C. R. J. Am. Water Works Assoc. 1950,42,886896.
100 times that of sodium; anions seemed to have little (2) Dart, F. J.; Foley, P. D. J. Am. Water Works Assoc. 1970,
effect. This agrees roughly with the Schulze-Hardy rule 62, 663-668.
for the relative coagulation power of divalent and mono- (3) Dart, F. J.; Foley, P. D. J. Am. Water Works Assoc. 1972,
valent ions. Additionally, computer equilibrium modeling 64, 244-249.
of a water containing up to 200 mg/L Ca and 16 mg/L (4) Hazel, F. J.Phys. Chem. 1945,49, 520-537.
silicic acid as SiOz predicted essentially no complexes of (5) Iler, R. K. The Chemistry of Silica; Wiley-Interscience:
calcium and silica even though Ca(OSi(OH)s)2,CaOSi- New York, 1979.
(OH),+, and Ca02Si(OH)2were part of the data base. (6) The PQ Corp. Sodium Silicates: Liquids/Solids. Bulletin
17-103. The PQ Corp., Valley Forge, PA, 1979.
Moreover, experimental post addition of calcium to already (7) Stumm, W., Lee, G. F. Ind. Eng. Chem. 1961,53,143-146.
sequestered iron caused the iron to precipitate rapidly. (8) American Public Health Association: American Water
These results are consistent with the formation of a neg- Works Association; Water Pollution Control Federation
atively charged iron-containing colloid by the near si- Standard Methods for the Analysis of Water and
multaneous addition of sodium silicate and hypochlorite. Wastewater, 15th ed.; American Public Health Association:
Finally, the role of hypochlorite is to oxidize the ferrous Washington, DC, 1981.
iron. Holden (22) demonstrated this by sequestering ferric (9) Hach Co. Test Instruction. Free and Total Chlorine Test
iron with sodium silicate in the absence of hypochlorite Kit. Hach Co., Loveland, CO, 1983.
or chlorine by adding to an acidic ferrous chloride solution (10) O’Connor,T. L. J. Phys. Chem. 1961, 65, 1-5.
(11) Alexander, G. B. J. Am. Chem. SOC. 1963, 75,5655-5657.
sufficient dilute 3.22 ratio silicate to yield a pH of 8.1. In (12) Browman, M. G. Role of Silica Polymerization in the
contrast, the addition of dilute sodium hydroxide to similar Treatment of Iron Bearing Groundwaters by the Addition
acidic ferric chloride solutions produced readily settleable of Sodium Silicate and Sodium Hypochlorite. M.S. Thesis,
iron precipitates at similar pH levels. Of course, hypo- The University of Tennessee, Knoxville, unpublished, 1986.
chlorite will oxidize ferrous iron very rapidly at pH 7 (23). (13) Robinson, R. B.; Browman, M. G.; Holden, J. M. Pro-
Post-hypochlorite addition analyses of sequestered iron ceedings, American Water Works Association Annual
show the complete absence of ferrous iron. Conference, Denver, CO, 1985.
(14) Weber, W. J.; Stumm, W. J. Znorg. Nucl. Chem. 1965,27,
Conclusions 237-239.
Our experiments and theoretical calculations and the (15) Felmy, A. R.; Cirvin, D. C.; Jenne, E. A. MINTEQ: A
work of other researchers support the following conclu- Computer Program for Calculating Geochemical Equilibria;
sions: Batelle Pacific Northwest Laboratories: Richland, WA,
1. The nearly simultaneous addition of sodium silicate 1983.
(16) Weber, W. J. Physicochemical Processes for Water Quality
and hypochlorite sequesters iron by dispersing an iron- Control; Wiley-Interscience: New York, 1972.
containing colloid. (17) Vanous, R. D.; Larson, P. E.; Hach, C. In Water Analysis;
2. The iron-containing particles grow over time. Minear, R. A., Keith, L. H., Eds.; Academic: New York,
3. Higher Si02/Na20ratio silicates, which have a more 1982; Vol. 1, pp 164-234.
complex silica, are more effective at iron sequestration than (18) Vail, J. G. Soluble Silicates: Their Properties and Uses;
lower Si02/Na20ratio silicates. Reinhold Publishing Corp.: New York, 1952.
4. For a given ratio silicate, the more concentrated (19) Robinson,R. B.; Minear, R. A.; Holden, J. M. J. Am. Water
solutions, which characteristically have higher proportions Works Assoc. 1987, 79, 116-125.
of complex silica, are more effective a t iron sequestration (20) Falcone, J. S., Jr. In Soluble Silicates; Falcone, J. S., Jr.,
than less concentrated solutions. Ed.; ACS Symposium Series 194; American Chemical So-
ciety: Washington, DC, 1982; pp 134-148.
5. The response of treatment system to pH changes and (21) Robinson, R. B.; Baumann, E. R.; Demirel, T. J. Environ.
additions of divalent calcium ions are consistent with Eng. Diu. (Am. SOC. Ciu. Eng.) 1981, 107, 1211-1227.
colloidal stability theory. (22) Holden, J. M. Treatment of Iron Bearing Groundwaters
6. In the absence of divalent cations, higher pH levels by the Simultaneous Addition of Sodium Silicate and So-
favor more effective iron treatment. dium Hypochlorite. M.S. Thesis, The University of Ten-
7. The primary role of hypochlorite is to oxidize ferrous nessee, Knoxville, unpublished, 1985.
iron. (23) Snoeyink,V. L.; Jenkins, D. Water Chemistry; John Wiley
8. Timing and mixing efficiency of silicate and chlorine and Sons: New York, 1980.
additions are important for iron sequestering because of
the rapid oxidation/hydrolysis of iron and the rapid de-
polymerization of sodium silicate. Received for review February 11, 1987. Revised manuscript
received June 21, 1988. Accepted November 4, 1988. This
Acknowledgments research was supported with funds from the Environmental
Protection Agency, American Water Works Association Research
We acknowledge the advice and counsel of F. James Foundation, and the University of Tennessee Waste Manage-
Dart of the Ontario Ministry of Environment. We also ment Center.

572 Environ. Sci. Technol., Vol. 23, No. 5 , 1989