Dissolved air flotation of bioreactor

effluent using low dosages of

polymer and iron
Esa Melin, Herman Helness and Hallvard
Ødegaard

INTRODUCTION
Many cities still discharge large amounts of inadequately treated wastewater to
marine waters. In many of these cities, the effluent standard that is to be met is
secondary treatment. Secondary treatment is, however, not implemented
because of excessive costs, lack of available land, and costly sludge disposal.
These cities need a treatment method that is compact, reaches secondary
treatment effluent standard, and has a minimal sludge production. A competitive
solution would be a treatment plant consisting of fine screening/sieving for
primary treatment, a highly loaded biofilm reactor, and a highly loaded
separation reactor (Ødegaard et al. 2000).
The major part of the organic loading from municipal wastewater is in
particulate matter (Levine et al. 1985; Ødegaard 1998; 1999). Therefore, with a
good separation process, the major part of the organic loading can be removed.
However, concentration of soluble organic matter is often too high for treatment
plants to meet secondary standards with particle removal only. In the high-rate
treatment concept, the intention is to operate the bioreactor at such high loading
rates that it removes soluble matter but hydrolysis of particulate organic matter

[1]
 2 Book title

does not take place. The removal of particulate organic matter is left for
separation reactor after the bioreactor.
Compact biological treatment systems require biofilm reactors. Biofilters
clog easily at high loads of particulate matter, which results in too frequent filter
washing. For high-rate treatment concepts, a moving bed biofilm reactor
(MBBR) has been shown to be a good alternative since the process can accept
high particulate and soluble organic loading rates (Ødegaard et al. 2000).
One alternative for a high-rate separation process is flotation. The flotation
system can be used with higher loading rates than sedimentation and has been
shown to be effective in secondary wastewater treatment (Ødegaard 2000; Filho
and Brandao 2000). With flotation, sludge settleability is not an issue. This can
be a problem in a MBBR at high soluble COD loading rates (Ødegaard et al.
2000). Flotation has become more attractive after recent developments of a very
highly loaded flotation process (so called turbulent flotation) where the surface
loading can be as high as 25-40 m/h (Kiuru 2000).
Optimal separation processes require addition of a coagulant. Inorganic metal
salts are often used. However, with high metal dosages, the sludge production
becomes high because of chemical precipitation. With use of cationic polymer
the sludge production can be reduced but the dosages required can be relatively
high (Fettig et al. 1990). When an inorganic metal salt is combined with low
dosages of polymer, the metal dosage and sludge production can be significantly
reduced without compromising treatment efficiency (Ødegaard 1998).
This paper reports the results from preliminary screening test of different
cationic polyacrylamide (PAM) and poly-diallyl-N,N-dimethylammonium
chloride (polyDADMAC) polymers. The purpose was to investigate what kind
of polymer is best for flotation of the MBBR effluent and to find optimal dosage
for coagulation with combination of metal salt and polymer.

EXPERIMENTAL SET-UP

Flotation experiments
The principle of the treatment process is presented in Figure 1. A laboratory-
scale MBBR was used to treat domestic wastewater, which was pumped into a
buffer tank from a nearby residential area. The MBBR loading was very high
with a detention time of only 15 min. An Aztec flotation jar tester (Severn Trent
Services, Capital Controls Ltd, England) was used for flotation tests. One-litre
samples collected from the MBBR outlet were used in each jar.
The iron and polymer were dosed with syringes under rapid mixing (400
rpm) which was continued for 0.5-1 min. The water was then flocculated for 20
 Short chapter title 3

min while mixing with 80 rpm. In the flotation step, 150 ml of dispersion water
(15% recycle rate) was used, saturated with air under 5 bar pressure. Distilled
water was used as dispersion water. The dilution effect of dispersion water is
taken into account when calculating in the results. The samples from clarified
water were taken 10 min after dispersion water was applied.
Iron was dosed as JKL (FeCl2SO4), which is a product of Kemira Chemicals.
PAMs were manufactured by Kemira and polyDADMACs were from Cytec.

Coagulant

Pre treatment MBBR Flocculation Flotation

fine sieve HRT: ~30 min HRT: 10 - 15 min HRT: 15 - 30 min
.
Figure 1. Schematic of the treatment process.

Experimental design and data analysis

Designed experiments were used to evaluate the effect of polymer properties
and polymer and iron dosages on the treatment results. Figure 2 shows the
design regions for molecular weight and charge density. Two different designs
were used for PAMs. In addition, three different low molecular weight, high
charge density polyDADMACs were tested.
The low iron dosages varied from 0 to 0.2 mmol Fe/l and polymer dosages
from 0.5 to 3 mg/l. Some polyDADMAC tests were done with polymer dosages
up to 3.4 mg/l and 0.3 mmol Fe/l. Since a real wastewater was used in the
experiments, the wastewater quality could not be used as a design variable.
The results were evaluated using Partial Least Squares Regression (PLSR), a
multivariate analysis method based on analysis of the variation in the data
(Martens and Næs 1989). In the PLSR, a new set of x-variables called PLS
components (PC) are computed in such a way that the first PC lies in the
direction of the largest variation of the data. The second PC lies in the direction
of second largest variation and so on. The logic is that the largest variation in the
data is likely to be caused by important or real effects while small variations in
 4 Book title

the data can be caused by less important effects or noise. One advantage of
PLSR is that one avoids focusing on large variations in the x-data that have little
importance for the variation in the response variables. Another advantage is that
the PCs are orthogonal, i.e. linearly independent. Using PCs in regression can
therefore overcome problems caused by collinear x-variables. However, one
must be aware of the danger of over-fitting and meaningless results and put
heavy emphasis on validation of the models.

16

14

12
MW, (10 6 g/mol)

10

0
0 2 4 6 8
Charge density, (meq/g)

Figure 2. Design regions for charge density and molecular weight

(■ PAMs, ∆ polyDADMACs).

The multivariate regression model has so far been developed for suspended
solids (SS) removal. The conclusions from the model are presented and the
average results from experiments are used to illustrate the observed effects.

Wastewater
The effluent from the MBBR was used in the flotation tests. The water quality is
presented in Table 1 for the tests with PAMs and polyDADMACs. The
temperature of the water was 9-11°C.
Table 1. Raw water quality during the experminents.
PAMs PolyDADMACs
Average Min Max Average Min Max
SS (mg/l) 143 98 187 111 53 163
COD (mg/l) 249 161 316 207 115 306
SCOD (mg/l) 67 51 92 61 39 125
pH 7.71 7.54 7.92 7.70 7.43 7.89
 Short chapter title 5

RESULTS AND DISCUSSION

Effect of polymer and iron dosage

The multivariate regression model shows that the treatment results are primarily
governed by the polymer and iron dosages. The effect of polymer dosage was
linear for both types of polymer. The response for iron dosage showed
curvature, i.e. the effect of iron on the treatment results becomes smaller with
increasing iron dosage. Figure 3 shows the average SS removals in all the tests
with different polymers. Without metal coagulant, only moderate removal of SS
is achieved. At PAM dosages of 1.75 and 3 mg/l, increasing iron dosage from
0.1 to 0.2 mmol Fe/l did not increase removal efficiency as much as
demonstrated between no iron and 0.1 mmol Fe/l, explaining the curved
response in the model. The curved response is difficult to see in the
polyDADMAC results, probably due to variation in the wastewater quality as
discussed below.
100 100
a b
80 80
SS removal (%)

SS removal (%)

60 60

40 40

20 20

0 0
0 1 2 3 4 0 1 2 3 4
Polymer dosage (mg/l) Polymer dosage (mg/l)

removal (♦ 0 mmol Fe/l, □ 0.1 mmol Fe/l, ▲ 0.2 mmol Fe/l, ◊ 0.3 mmol Fe/l). Lines
Figure 3. Effect of iron and PAM (a) or polyDADMAC (b) dosages on suspended solids

show model predictions.

The lines in Figure 3 show the model predictions for SS removal. The model
fits the experimental results well and shows that the variations in the removal
efficiency are a result of varying raw water quality rather than experimental
error. Generally, the removal efficiency is similar between PAMs and
polyDADMACs but polyDADMACs give slightly higher SS removal with
0.2 mmol Fe/l. With 3 mg/l of polymer and 0.2 mmol Fe/l the SS removal
efficiency varied with PAMs from 72 to 89% resulting in 16-46 mgSS/l
 6 Book title

(average 29 mgSS/l) in the treated water. With polyDADMACs, the removal

efficiency varied from 79 to 91% with residual SS of 6-21 mg/l (average 14
mg/l). However, because of the variation in the wastewater quality (Table 1), it
cannot be concluded that polyDADMACs in general give better results than
PAMs.

100 100
a b
80 80
COD removal (%)

COD removal (%)

60 60

40 40

20 20

0 0
0 1 2 3 4 0 1 2 3 4
Polymer dosage (mg/l) Polymer dosage (mg/l)

(♦ 0 mmol Fe/l, □ 0.1 mmol Fe/l, ▲ 0.2 mmol Fe/l, ◊ 0.3 mmol Fe/l).
Figure 4. Effect of iron and PAM (a) or polyDADMAC (b) dosages on COD removal

100 100
a b
80 80
SCOD removal (%)

SCOD removal (%)

60 60

40 40

20 20

0 0
0 1 2 3 4 0 1 2 3 4
Polymer dosage (mg/l) Polymer dosage (mg/l)

(♦ 0 mmol Fe/l, □ 0.1 mmol Fe/l, ▲ 0.2 mmol Fe/l, ◊ 0.3 mmol Fe/l).
Figure 5. Effect of iron and PAM (a) or polyDADMAC (b) dosages on SCOD removal

Figure 4 shows the removal of chemical oxygen demand (COD) in the

flotation tests. The removal patterns are the same as with SS although overall
removal efficiencies are lower. Figure 5 shows the removal of soluble COD
(SCOD). Only minor removal is observed in tests without metal coagulant and
iron improves SCOD removal. Soluble COD is normally defined as the COD
 Short chapter title 7

measured after filtering sample through GF/C filter, which has a nominal pore
size of about 1 µm. This fraction still includes some colloidal material and the
truly soluble fraction is below 0.1 µm. Therefore, the removal of SCOD can
actually be coagulation of the colloidal fraction.

Effect of raw water quality

The multivariate analysis indicates that the raw water SCOD and pH influenced
SS concentration in clarified water with PAMs. In the case of polyDADMACs,
the raw water COD and SS affected the treatment efficiency while pH and
SCOD did not have a significant effect. Figure 6 shows the effect of raw water
SCOD on SS removal for all the tests. With PAMs, removal efficiency
decreases when SCOD is over 65-70 mg/l while the polyDADMAC results are
unaffected by raw water SCOD. However, the results need to be verified by
further tests because the PAM results with high SCOD concentration are from
experiments without iron dosage while the polyDADMAC results with high
SCOD concentration are from experiments with 0.2 mmol Fe/l.
100 100
b
80 80
SS removal (%)

SS removal (%)

60 60

40 40

20 20

0 0
40 50 60 70 80 90 100 20 40 60 80 100 120 140
Raw water SCOD (mg/l) Raw water SCOD (mg/l)

Fig. 6. Effect of raw water SCOD on SS removal in flotation test using PAMs (a) and
polyDADMACs (b) (♦ 0 mmol Fe/l, □ 0.1 mmol Fe/l, ▲ 0.2 mmol Fe/l). Lines show
model predictions.

Figure 7 shows the results for both polymers. While they seem to confirm the
model, the raw water pH varied only in a very narrow range (from 7.4 to 7.9)
and therefore the results are not very conclusive. However, the results indicate
that PAMs are more sensitive to raw water quality like SCOD and pH than
polyDADMACs.
 8 Book title

100 100
a b
80 80
SS removal (%)

SS removal (%)
60 60

40 40

20 20

0 0
7.4 7.6 7.8 8.0 7.4 7.6 7.8 8.0
Raw water pH Raw water pH

Fig. 7. Effect of raw water pH on SS removal in flotation test using PAMs (a) and
polyDADMACs (b) (♦ 0 mmol Fe/l, □ 0.1 mmol Fe/l, ▲ 0.2 mmol Fe/l). Lines show
model predictions.

Effect of polymer properties

The model indicates that with PAMs there is an effect of the molecular weight
that depends on the iron dose. The model predicts that when iron is not used or
the dosage is low, it is a benefit to have a high molecular weight polymer. When
the iron dosage is increased, slightly better results are predicted with a low
molecular weight polymer. Figure 8a shows the average results from all the tests
done with different polymers. It should be noted that the wastewater quality and
average polymer dosages are not the same for the different data points. The
average trends in the data for the different iron dosages support the results from
the model. Without iron, the removal efficiency is lowest with low molecular
weight PAMs. If the polymer is going to be used alone without metal coagulant,
low molecular weight PAMs do not seem to be the best alternative as also
observed by Pilipenko and Ødegaard (2002). In Figure 8b, the model prediction
of the different data points is included. The good agreement between the model
and the experimental results shows that the scatter in the results can largely be
explained by the variation in the wastewater quality and polymer dosage.
The model for PAM indicates that a high charge density is a benefit. Figure
9a shows the removal efficiency versus charge density of the polymer with all
the tested polymers. In these figures, also polyDADMAC is included for
comparison (charge density 6.2 meq/g). The average trend in the PAM-data with
iron doses of 0.1 and 0.2 mmol Fe/l supports the model, while the average trend
for the PAM-data with no iron shows the opposite. However, this is due to two
data points with high removal at charge densities of 0.3 and 1.8 meq/g. In
 Short chapter title 9

Figure 9b, the model prediction of the different data points is included. The
good agreement between the model and the experimental results indicate that
the scatter in the results can largely be explained by the variation in the
wastewater quality and polymer dosage. The results for both molecular weight
and charge density demonstrate that although these polymer properties probably
have an effect on the removal efficiency, they are small compared to the effect
of dosage and wastewater quality.
100 100
a b
80 80
SS removal (%)

SS removal (%)
60 60

40 40

20 20

0 0
0 5 10 15 0 5 10 15
6 6
Molecular weight (10 g/mol) Molecular weight (10 g/mol)

Figure 8. The effect of molecular weight on SS removal efficiency with different iron
dosages for all the tested PAMs (a) and the same results together with the model
predictions (b) (♦ 0 mmol Fe/l, □ 0.1 mmol Fe/l, ▲ 0.2 mmol Fe/l).

100 100
a b
80 80
SS removal (%)

SS removal (%)

60 60

40 40

20 20

0 0
0 2 4 6 8 0 2 4 6 8
Charge density of the polymer (meq/g) Charge density of the polymer (meq/g)

Fig. 9. Effect of charge density of the polymer on the SS removal with all the tested
polymers (a) and the same results together with the model predictions (b) (♦ 0 mmol
Fe/l, □ 0.1 mmol Fe/l, ▲ 0.2 mmol Fe/l).
 10 Book title

There were no significant differences between the tested polyDADMACs.

Also, the best PAMs performed equally with polyDADMACs. When a PAM is
used together with iron, the best choice seems to be a medium molecular weight,
high charge density polymer.

COD fractions in the flotated water

In some tests, COD was analysed from water samples that were filtered through
filters having different pore size. Figure 10 shows the COD fractions in raw
water and flotated water. In the experiment, polyDADMAC was used at variable
dosages (0.6-3.4 mg/l) and the iron dosage was 0.2 mmol Fe/l. The results show
that particles above 11 µm are effectively removed by flotation, which is
consistent with general understanding that flotation is effective in removing
particles down to 10 µm in size (Kiuru 1990). The truly soluble COD fraction
(<0.1 µm) is the largest fraction in the clarified water and is not removed very
well. Since the aim is removal of particulate organic matter, this size fraction is
of no interest. The preceding biological process should be operated so that the
truly soluble fraction is removed to desired levels. The results show that the
observed SCOD removal is mostly the removal of colloidal material (size
fraction 0.1-1 µm). The 1-11 µm size fraction is, however, critical for successful
particle removal and further process optimisation should concentrate on good
flocculation of this particle size range.

120.0

100.0 Raw water

P = 0.58 mg/l
COD (mg/l)

80.0
P = 1 mg/l
60.0
P = 2 mg/l
40.0 P = 3 mg/l

20.0 P = 3.4 mg/l

0.0
>11 µm 1-11 µm 0.1-1 µm <0.1 µm

Figure 10. Particulate COD fractions in raw water and flotated water with different
polyDADMAC dosages and 0.2 mmol Fe/l iron.
 Short chapter title 11

CONCLUSIONS
1. The results show that good SS removal can be obtained by flotation when
low dosages of iron and polymer were combined.
2. The dosages of iron and polymer together with raw water properties like
SCOD are more important than molecular weight or charge density of the
polymers.
3. There were no significant differences between polyDADMACs and the best
PAMs. The results indicate that with metal coagulant, the best PAM is
medium weight, high charge density polymer. Without metal, high
molecular weight PAMs give the best result.
4. Multivariate analysis is a good tool when analysing results with variable
water quality. Although the regression model developed for SS removal
were able to predict the results well, it needs additional tests to fill the gaps
in the data set and an independent verification test.

REFERENCES
Fettig, J., Ratnaweera, H. and Ødegaard , H. (1990) Synthetic organic polymers as
primary coagulants in wastewater treatment, Water Supply 8, 19-26.
Filho, A. C. T. P. and Brandao, C. C. S. (2000) Evaluation of flocculation and dissolved
air flotation as an advanced wastewater treatment. Wat. Sci. Tech. 43(8), 83-90.
Kiuru, H. (1990) Unit operation for the removal of solids and their combinations in water
treatment. In Chemical water and wastewater Treatment (Hahn, H.H. and Klute, R.
(eds). Springer-Verlag, Berlin, pp. 169-186.
Kiuru, H. J. (2000) Development of dissolved air flotation technology from the 1st
generation to the newest or 3rd one (very thick micro-bubble bed) with high flow-
rates (DAF in turbulent flow conditions). Wat. Sci. Tech. 43(8), 1-7.
Levine, A.D., Tchobanoglous, G. and Asano, T. (1985) Characterization of the size
distribution of contaminants in wastewater; Treatment and reuse implications. J.
WPCF, 57(2), 805.
Martens, H. and Næs, T. (1989) Multivariate Calibration, John Wiley & Sons Ltd.
Ødegaard, H. (1998) Optimized particle separation in the primary step of wastewater
treatment. Wat. Sci. Tech. 37(10), 43-53.
Ødegaard, H. (1999) The influence of wastewater characteristics on choice of wastewater
treatment method. In Proc. Nordic Conf. on Nitrogen removal and biological
phosphate removal, Oslo, 2-4 February.
Ødegaard, H. (2000) The use of dissolved air flotation in municipal wastewater
treatment. Wat. Sci. Tech. 43(8), 75-81
Ødegaard, H., Gisvold, B., Helness, H., Sjøvold, F. and Liao, Z. (2000). High rate
biological/chemical treatment based on the moving bed biofilm process combined
with coagulation. In Chemical water and wastewater Treatment VI (Hahn, H.H.,
Hoffmann, E. and Ødegaard, H. (eds). Springer-Verlag Berlin Heidelberg New
York, pp. 245-255.
 12 Book title

Pilipenko, P. and Ødegaard, H. (2002) Removal of suspended solids and sludge

production in coagulation of municipal wastewater with cationic polyelectrolytes. In
Proc. of the International Congress and Trade Fair Water: Ecology and
Technology, ECWATECH, June 4-7, Moscow.

ACKNOWLEDGEMENT
This study was financed by Kemira OYj, Finland, National Technology Agency,
Finland and Kaldnes Miljøteknologi AS, Norway

CONTACT
Hallvard Ødegaard, Department of Hydraulic and Environmental Engineering,
Norwegian University of Science and Technology, N-7491 Trondheim, Norway.
E-mail: Hallvard.Odegaard@bygg.ntnu.no