ARTICLE Critical Evaluation and Modeling of Algal Harvesting Using Dissolved Air Flotation Xuezhi Zhang,1 John C. Hewson,2 Pasquale Amendola,1 Monica Reynoso,1 Milton Sommerfeld,1 Yongsheng Chen,3 Qiang Hu4 1
Department of Applied Sciences and Mathematics, Arizona State University, Polytechnic Campus, Mesa, Arizona 85212; telephone: þ1 480 727 5294; fax: þ1 480 727 1475; e-mail: [email protected] 2 Sandia National Laboratories, Albuquerque, New Mexico 3 School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, Georgia 4 Institute of Hydrobiology, Chinese Academy of Sciences, Wuhan, Hubei, China
Introduction ABSTRACT: In this study, Chlorella zofingiensis harvesting by dissolved air flotation (DAF) was critically evaluated with regard to algal concentration, culture conditions, type and dosage of coagulants, and recycle ratio. Harvesting efficiency increased with coagulant dosage and leveled off at 81%, 86%, 91%, and 87% when chitosan, Al3þ, Fe3þ, and cetyl trimethylammonium bromide (CTAB) were used at dosages of 70, 180, 250, and 500 mg g1, respectively. The DAF efficiency-coagulant dosage relationship changed with algal culture conditions. Evaluation of the influence of the initial algal concentration and recycle ratio revealed that, under conditions typical for algal harvesting, it is possible that the number of bubbles is insufficient. A DAF algal harvesting model was developed to explain this observation by introducing mass-based floc size distributions and a bubble limitation into the white water blanket model. The model revealed the importance of coagulation to increase flocbubble collision and attachment, and the preferential interaction of bubbles with larger flocs, which limited the availability of bubbles to the smaller sized flocs. The harvesting efficiencies predicted by the model agree reasonably with experimental data obtained at different Al3þ dosages, algal concentrations, and recycle ratios. Based on this modeling, critical parameters for efficient algal harvesting were identified. Biotechnol. Bioeng. 2014;111: 2477–2485. ß 2014 Wiley Periodicals, Inc. KEYWORDS: flotation; model; coagulant; flocs; bubble; Chlorella
Correspondence to: X. Zhang Contract grant sponsor: The Laboratory Directed Research and Development program, Sandia National Laboratories and by the Arizona Center for Algae Technology and Innovation at Arizona State University, U.S. Department of Energys National Nuclear Security Administration Received 28 February 2014; Revision received 2 May 2014; Accepted 23 May 2014 Accepted manuscript online 31 May 2014; Article first published online 14 July 2014 in Wiley Online Library (http://onlinelibrary.wiley.com/doi/10.1002/bit.25300/abstract). DOI 10.1002/bit.25300
ß 2014 Wiley Periodicals, Inc.
Dissolved air flotation (DAF) has been widely used to separate microalgae from natural water and wastewater (Edzwald, 2010; Ma et al., 2007), and has often proven more efficient than sedimentation for algae-water separation (Teixeira and Rosa, 2006). DAF has increasingly been used to harvest algae from raceway ponds and photobioreactors for biofuel applications (Christenson and Sims, 2011; Zhang et al., 2012). An important factor in the economic application of DAF to algal harvesting is the energy costs associated with bubble formation, and recent work in efficient bubble generation has been reported by Hanotu et al. (2012, 2014). The application of DAF to algal harvesting differs from algal separation in water and wastewater treatment in that the cell concentration in algal mass culture is typically thousands of times greater than in water and wastewater (Conti et al., 2005; James et al., 1994). Further, the biochemical and chemical properties of algal cells and culture media change considerably depending on algal strains and culture conditions, which may in turn affect coagulant demand and flotation efficiency. Operational protocols for DAF algal harvesting are usually developed on the basis of empirical considerations combined with jar testing. There is a pressing need to critically evaluate and understand the influence of algal properties, bubble properties, and coagulation conditions on DAF algal harvesting. Modeling of DAF algal harvesting would help to identify key parameters involved in the process and thus be valuable in efforts to achieve high harvest efficiencies. The white water blanket model, initially developed by Edzwald and colleagues (Edzwald, 1995; Edzwald and Wingler, 1990; Malley and Edzwald, 1991) has been widely used for modeling the separation of water and particles, including algal cells, in water and wastewater treatment. The term “white water blanket” describes a rising air bubble suspension, a “blanket” in a DAF tank that captures Biotechnology and Bioengineering, Vol. 111, No. 12, December, 2014
2477
particles above as it moves toward the water surface. In this model, the bubble number had not been considered to be a limiting factor, as the ratio of particles to bubbles in water or wastewater typically is not large. The particle number-based efficiency has also generally been calculated using the average particle size instead of a size distribution. However, with higher initial cell concentration in algal culture, the bubble number may be limiting. After coagulation or flocculation, the flocs number distribution is dominated by the smallest particle sizes, while most of the biomass is associated with a few large flocs. It will therefore be necessary to consider the floc size distribution when modeling DAF algal harvesting, as the collection of larger flocs contributes more to the overall harvesting efficiency. In this study, the effects of culture conditions, cell concentration, coagulant type and dosage, and recycle ratio on DAF harvesting efficiency were investigated using the oleaginous green microalga C. zofingiensis as a model organism. A DAF algal harvesting model was developed to predict the mass-based harvesting efficiency by introducing a bubble limitation and mass-based floc size distribution into the white water blanket model. The critical parameters for effective DAF algal harvesting were identified. The objectives of this study are to (1) evaluate the influence of algal cell properties, coagulants and DAF operation conditions on algal harvesting efficiency, and (2) extend the white water blanket model to understand and predict the outcome of the DAF algal harvesting process by incorporating the effects of floc size distribution and limits to the number of bubbles.
Materials and Methods Algal Culture and Characterization The freshwater microalga C. zofingiensis (LRB-AZ-1201) isolated and maintained in the Laboratory for Algae Research and Biotechnology at Arizona State University (Arizona, USA) was grown in both nitrogen-depleted and nitrogen-replete conditions, as nitrogen starvation favors lipid accumulation (Hu et al., 2008). C. zofingiensis was selected as a model organism because it is widely used as a production strain for biofuels and high-value bioproducts (Del Campo et al., 2004; Liu et al., 2011) and has a spherical shape. The nitrogen-depleted cells were cultured indoors in 15 L panel reactors with a 5 cm light path and continuous fluorescent lighting. The nitrogen-replete culture was grown in 50 L panel reactors with the same light path and natural dark-light cycles. The light intensity, culture pH, initial and final NO3-N concentration, and initial and final dry weight are summarized in Table I. The culture was harvested after six days of cultivation, and the dry weights of the nitrogen-depleted and nitrogenreplete culture reached approximately 1.5 and 2.0 g L1, respectively.
2478
Biotechnology and Bioengineering, Vol. 111, No. 12, December, 2014
Table I. Summary of culture conditions for both nitrogen-depleted and nitrogen-replete C. zofingiensis. Parameters
Nitrogen-depleted
Nitrogen-replete
Indoor or outdoor Culture volume (L) Light path (cm) Initial algal concentration (g L1) Initial NO3-N (mg L1) Final NO3-N (mg L1) Algal concentration before harvesting (g L1) Culture period (d) Growth phasea
Indoor 15 5 0.2 33.6 NDb 1.5
Outdoor 50 5 0.5 128 23.5 2.0
6 Stationary
Light intensity (mmol m2 s1) pH range Culture temperature ( C)
220 7.0–8.2 26
6 Exponential growth 1900c 7.0–8.5 15–30
a
Growth phases were identified by changes in cell number (data not shown). b Not detectible. c Daily maximum.
DAF Algal Harvesting C. zofingiensis cultures were harvested using a DAF jar tester (DBT6, EC Engineering) using the procedure previously described (Zhang et al., 2012). In this study, chitosan (Sigma– Aldrich 417963-100G, practical grade), aluminum sulfate (Al2(SO4)3 18H2O), ferric sulfate (Fe2(SO4)3), and cetyl trimethyl ammonium bromide (CTAB) were tested for their efficacy for DAF pretreatment, and aluminum sulfate was then selected for further study. After rapid injection of the coagulant, the pH was adjusted immediately to optimal coagulation conditions (pH of 7.0 0.2, 6.2 0.2, 7.5 0.5, and 8.0 0.2 for chitosan, aluminum, CTAB and iron, respectively) by adding pre-determined amounts of 1 M NaOH. After coagulation, the algal suspension was flocculated for 10 min at 30 rpm (velocity gradient G ¼ 30 s1) before bubbles were injected to float the algal cells. The harvesting efficiency was calculated on the basis of the mass of the algae in the subnatant and in the original media. Dissolved organic matter (DOM)-free cultures were used to exclude the influence of DOM on Al3þ consumption and floc size distribution when harvesting algal cells with different biomass concentrations at a 20% recycle ratio and when harvesting algal cells at the same biomass concentration (7.4 g L1) with different recycle ratios. DOM in the nitrogen-depleted culture was removed by centrifugation (6,000 rpm for 10 min), and cells were re-suspended into fresh BG11 culture medium prior to harvesting. The Al3þ dosage of 48.5 mg g1 was used in these parameter studies based on preliminary experiments.
Floc Size Distribution Measurement Before bubble injection, a sample of the algal suspension was transferred to a microscope slide for floc size measurement.
Samples were carefully taken to avoid disturbing the flocs using a 1 mL transfer pipet cut at the tip to provide a larger opening. A minimum of three samples were taken and at least 2,000 algal cells and flocs for each sample were captured via bright field microscopy. The area of each floc in the images was determined using a pixel-counting algorithm (ImageJ software, version 1.43u), and the equivalent diameter of each floc was calculated from the square root of the fine-grained area. The number of flocs in each size range (1.0–2.0, 2.0– 3.0… 30–40… 390–400 mm) was then counted to obtain a number-based size distribution as a function of the floc diameter, df. The size measurements were verified using 3.0 mm-, 30 mm-, and 100 mm-SiO2 monodisperse particle standards (Microspheres-Nanospheres, New York). As algal biomass harvesting efficiencies are typically expressed in terms of dry weight rather than cell number, the number-based size distribution (dFn) was then converted to a mass-based size distribution (dFm) for model input using Equation (1): idF n dF m ¼ R1 idF n
ð1Þ
dp
where dp is the diameter of individual cells and i is the number of algal cells in a floc. A detailed discussion of Equation (1) can be found in Section S1 (Supplementary Information).
Modeling of DAF Algal Harvesting White Water Blanket Model for DAF Particle Separation To better understand the DAF algal harvesting processes, the white water blanket model was employed to describe floc-bubble interactions and subsequent separation. This model predicts the number of flocs removed relative to the initial number (Edzwald, 2010; Edzwald and Wingler, 1990): nf ðtÞ ¼ expðltnb Þ nf ;0
ð2Þ
where nf is the floc concentration, nb is the bubble concentration, t is the floc-bubble contact time determined by the DAF unit flow characteristics and l ¼ 0:25ahT vb pd 2b is a rate expression in inverse concentration units. In the rate expression, hT is the sum of various dimensionless flocbubble collision efficiencies (Haarhoff and Edzwald, 2004), a is the floc-bubble attachment efficiency, vb is the bubble rise velocity and db is the bubble diameter. Further information on the floc-bubble collision efficiencies employed is provided in Section S2 and Figure S1 of the Supplementary Information. The total number of bubbles is a function of the bubble diameter, recycle ratio R, air mass concentration in the water at saturator pressure (ys), and at ambient pressure (y0), and the combined saturator and nozzle efficiency, k, as described in Equation (3). A detailed
derivation of Equation (3) is provided in Section S3 (Supplementary Information). nb ¼
6kðys y0 ÞR pd b 3 ð1 þ RÞrair
ð3Þ
DAF Algal Harvesting Model With Floc Size Distribution and Bubble Limitation In the DAF algal harvesting process, the floc mass is distributed over a range of floc sizes (Supplementary Fig. S2). In addition, the bubble number may be limiting as the floc number is much higher than in typical DAF water and wastewater treatments. Thus the above model was extended to account for limited bubble availability; these extensions to the model are explained in somewhat greater detail in the Section S4 of the Supplementary Information. As no simple analytical solution jointly solves the evolution of bubbles and flocs over a range of floc sizes, a simpler approach was used based on the greater collision coefficients of bubbles with large flocs (Supplementary Fig. S1A). Bubbles are allowed to interact with flocs of a given size class at the probability given by the original model. Because floc size is often comparable to or exceeds bubble size, once bubble-floc collisions occurred the bubble-floc agglomerates were assumed to not sweep up further algal flocs, although further bubble collisions were allowed up to a maximum number of bubbles determined by the floc size. Bubbles are assumed to interact with the largest flocs first and the prediction of bubble–floc interaction was carried out by integrating across the size distribution from the largest to the smallest flocs; that is, the remaining bubbles were allocated to possible interactions with the next smaller floc size until all the bubbles were assigned or the entire floc size distribution was covered. This assumption that the interaction occurs sequentially from the largest to smaller flocs is certainly imperfect; simulations with greater resolution of floc–bubble interaction dynamics could help resolve the limitations of the model but are beyond the present work scope. The maximum number of bubbles that can attach to a floc of a given size (nmax) is calculated using Equation (4), assuming each bubble occupies a square (db db) on the surface of a floc (Tambo et al., 1986): nmax ¼ p
2 df db
ð4Þ
The minimium number of bubbles needed to float a floc of a given size (nmin) is calculated using Equation (5), assuming the floc–bubbble can be floated if its density (rfb) is lower than the density of water (rw) rfb ¼
rf d f 3 þ nmin rb d b 3 < rw d f 3 þ nmin d b 3
ð5Þ
The total number of bubbles available for interaction with flocs smaller than df depends on the number of bubbles that
Zhang et al.: DAF Algal Harvesting Modeling Biotechnology and Bioengineering
2479
have attached to larger flocs. Similar to Equation (2), an expression for the bubble-floc collision rate gives the number of bubbles that stick to flocs of size df: dBn ¼ minfBn ½1 expðlt nf dFn Þ;
DðdFn Þnmax g
ð6Þ
where Bn is the bubbles available to interact with flocs of size df, and dBn is the number of bubbles that stick to flocs of size df. dBn is limited to the maximum number of bubbles that attach to a given floc, D(dFn)nmax. Note that we have left Bn in concentration dimensions to simplify the notation; it has not been normalized to unity. The number of bubbles available for interaction with smaller flocs is found by integrating dBn over the floc size distribution: 2 3 Zdf ð7Þ dBn5 Bn ¼ max40; nb;0 1
Based on Equation (2), the number fraction of flocs of a given size collected by bubbles is 1exp(ltBn) where Bn varies according to Equation (7). The fraction of these that actually float will be determined by the floc-bubble rise velocity, vfb, depending on rfb in Equation (5). The number fraction of flocs that are actually harvested is: DðdFn Þ ¼ r fb ½1 expðltBn Þ
ð8Þ
where r fb ¼ min 1; max 0; vfbht r , which denotes the fraction of flocs floated, tr is the allowed rise time and h is the water depth. For the experiments in this study, the allowed rise time was effectively infinite so that rfb is approximately unity as long as the floc-bubble velocity is positive. The total mass harvested by bubbles (mH) is the integral over floc distribution and the fraction harvested times the initial dry mass, m0,
Equation (8) and Equation (9) according to floc-bubble collision efficiency and attachment efficiency, using the available bubble number at the given floc size range. Small flocs are collected at reduced (or zero) efficiency as bubbles are depleted. The final mass-based harvesting efficiency was then calculated using Equation (10), the integral of the mass of the flocs collected in all floc size ranges. A schematic diagram of the modeling algorithm is shown in Supplementary Figure S3.
Results Influence of Coagulants on DAF Harvesting Efficiencies Coagulants differ in their capacity to generate flocs and in their ability to enhance floc-bubble attachment efficiency. The influence of different coagulants and their dosages on measured DAF harvesting efficiencies is shown in Figure 1. Harvesting efficiency increased with coagulant dosage and leveled off at 81%, 86%, 91%, and 87% when chitosan, Al3þ, Fe3þ, and cetyl trimethylammonium bromide (CTAB) were used at dosages of 70, 180, 250, and 500 mg g1, respectively. Different efficiency-dosage relationships were obtained for different coagulants with the lowest dosage requirements observed for chitosan, which agrees with the literature (Ahmad et al., 2011; Divakaran and Pillai, 2002; Vandamme et al., 2010; Wyatt et al., 2012). The different efficiencydosage relationships obtained for different coagulants may be due to their different floc formation properties and the different influences on floc-bubble interactions, possibly as a result of the characteristics of charge density of different coagulants. Larger flocs were generated using chitosan, whereas smaller flocs were generated using CTAB
Z1 mH ¼ m0
DðdFn ÞdFm
ð9Þ
dp
The total harvesting efficiency can be obtained by comparing the total mass harvested, mH, with the initial dry mass, m0: hmDAF ¼
mH ¼ m0
Z1 DðdFn ÞdFm
ð10Þ
dp
In summary, the mass-based fraction (dFm) of flocs in each size range was first calculated using Equation (1). Bubbles were allocated to the flocs using collision efficiencies based on the original models, from larger to smaller sizes, until all bubbles were assigned. The number and mass of flocs in a given size range collected by bubbles was then calculated with
2480
Biotechnology and Bioengineering, Vol. 111, No. 12, December, 2014
Figure 1.
DAF harvesting efficiencies for nitrogen-replete C. zofingiensis as a function of chitosan, Al3þ, Fe3þ, and CTAB dosages.
(Supplementary Fig. S4). Since similar trends as a function of relative dose are obtained for all coagulants, subsequent experiments were conducted using Al3þ. Influence of Culture Conditions on Coagulant Dosage Requirements Harvesting efficiency depends on the surface properties of the algae and the ability of the coagulants to alter algal cell surface properties to favor flocculation and floc–bubble adhesion. Physiological stress, as occurs under nitrogen-depleted culture conditions, accelerates algae aging and changes their surface properties. C. zofingiensis cultures were grown in nitrogen-depleted and nitrogen-replete media as indicated in Table I, and used to evaluate the influence of surface properties on DAF harvesting. At the time of harvesting, the nitrogen-depleted culture had reached the stationary growth phase with a dry weight of 1.5 g L1 and the nitrogen-replete culture was still in the exponential growth phase with a dry weight of 2.0 g L1. Figure 2 shows the DAF efficiency-Al3þ dosage relationships for the cultures under a fixed recycle ratio of 20%. Harvesting efficiencies increased with Al3þ dosages, peaking near 90%. However, to achieve the same harvesting efficiencies, the nitrogen-replete culture required more Al3þ than the nitrogen-depleted culture. This result agrees with previous work that showed algae surfaces differ depending on culture conditions, and that aged C. zofingiensis required less Al3þ for harvesting (Zhang et al., 2012).
fixed 20% recycle ratio (Fig. 3A). The harvesting efficiency was greater than 92% when the initial algal cell concentration was 1.2 g L1. The harvesting efficiency dropped to 81% as the initial cell concentration increased to 3.1 g L1, and it further decreased to 41% at a 7.4 g L1 initial cell concentration. The typical range of recycle ratios used for algal cell separation in water and wastewater treatments is between 6% and 12% (Gregory and Edzwald, 2010), but the cell concentrations in these cases are typically less than 107 cells L1 (Conti et al., 2005; James et al., 1994). With a concentration of 1.5 g L1, C. zofingiensis cell number was approximately 8.6 1010 cells L1, with 2.2 108 flocs L1 of
Influence of Initial Algal Biomass Concentration and Recycle Ratio on Harvesting Efficiency The influence of the initial algal biomass concentration on harvesting efficiency was tested with DOM-free cultures at a
Figure 2. DAF harvesting efficiency-Al3þ dosage relationships for both nitrogenreplete (2.0 g L1) and nitrogen-deplete (1.5 g L1) C. zofingiensis at the recycle ratio of 20%.
Figure 3. Influence of initial algal dry weight (A) and recycle ratio (B) on DAF harvesting efficiency of C. zofingiensis. The recycle ratio was 20% in (A), and the initial cell concentration was 7.4 g L1 in (B). DOM in the nitrogen-depleted culture was removed, and algal pellets were suspended in fresh culture medium to make different concentrations.
Zhang et al.: DAF Algal Harvesting Modeling Biotechnology and Bioengineering
2481
different sizes after coagulation. As the initial algal dry weight increased, the number of flocs also increased, thus requiring more air bubbles for their removal. To confirm the impact of the bubble limitation on DAF algal harvesting, the influence of recycle ratio on the harvesting at a high biomass concentration, 7.4 g L1, was tested (Fig. 3B). Bubble concentration increased with recycle ratio as shown in Equation (3). The harvesting efficiency increased with recycle ratio, reaching 72% when the recycle ratio increased to 50%. The results in Figure 3A and B both suggest that the available number of bubbles is a limiting factor at high biomass concentrations: more bubbles are required when harvesting algal biomass with a higher concentration. This behavior has not traditionally been predicted with the white-water blanket model for DAF separations and motivated the model developed in Equations (6–10). In the following section we evaluated the model in the context of the data presented here. DAF Harvesting Model Evaluation—Influence of Floc Size Distribution and Attachment Efficiency Figure 4 shows images of algal cells and flocs after Al3þ flocculation. Without Al3þ, C. zofingiensis cells were uniform in size, averaging 3 mm. When Al3þ was added, flocs were generated. Increased dosages of Al3þ resulted in larger floc sizes and fewer single cells surrounding the flocs. A reduction in the number of single cells was also observed with increased dosages of chitosan, Fe3þ, and CTAB. Both number-based and mass-based size distributions at different Al3þ dosages are given in Supplementary Figure S2, while distributions for different coagulants are compared in Supplementary Figure S4 for relevant respective high-end dosages (Supplementary Information). Although the number distribution was dominated by the smallest particle sizes, most of the
Figure 4. Micrographs of flocs of nitrogen-depleted C. zofingiensis at different Al3þ dosages. The length of the scale bar is 100 mm.
2482
Biotechnology and Bioengineering, Vol. 111, No. 12, December, 2014
biomass was associated with a few large flocs after coagulation, so that a high harvesting efficiency can be achieved if relatively few large flocs are generated through coagulation and then collected by bubbles. The addition of coagulant alters the algal surface allowing floc formation and easier attachment between flocs and bubbles, represented by the floc–bubble attachment efficiency, a. The influences of both Al3þ dosage (or floc size distribution) and a on the predicted DAF harvesting efficiencies of nitrogendepleted C. zofingiensis are shown in Figure 5 along with measurements from Figure 2. Dashed lines connect symbols representing predicted harvesting efficiencies for varying a for a given size distribution (a function of the Al3þ dosage from Supplementary Fig. S2). Values of a at different Al3þ dosages were then selected from values in which the predicted efficiencies fit well with the experimental efficiencies; selected a values are plotted in Supplementary Figure S5. Without Al3þ, the modeled harvesting efficiencies were less than 10% no matter what a was assumed because too few collisions occurred between bubbles and individual small algal cells (cf. Supplementary Fig. S1). At intermediate dosages (20–40 mg g1 Al3þ in Fig. 5), there was substantial floc formation, but only moderate harvesting efficiency suggesting a less efficient attachment efficiency than is obtained at higher dosages. This trend is reflected in Supplementary Figure S5. At higher dosages, the probability of attachment became high with the biomass mostly due to the larger flocs size; the observed harvesting efficiencies suggested attachment efficiencies in the range 0.4–0.6. Previous authors have also reported that attachment efficiency increased with Al3þ concentration. For example,
Figure 5. Influence of Al3þ dosage and attachment efficiency (a) on DAF harvesting efficiency of nitrogen-depleted C. zofingiensis. Harvesting efficiencies at different Al3þ dosages were calculated using the measured floc size distribution at each Al3þ dosage, and a series of assumed a values (open symbols). The a value at a given Al3þ dosage was selected from values in which the modeled efficiencies (red solid circles) approximate the measured harvesting efficiencies (black pluses). The solid curve is a regression of the logistic function using the modeled efficiencies at different Al3þ dosage and selected a values (red solid circles) as inputs.
Liers et al. (1996) reported that a ranged from 0.3 to 0.4 and Haarhoff and Edzwald (2004) used a values between 0.5 and 1 for optimum coagulation. Bubbles had a negative charge ranging from 25 to 150 mV at neutral pH without coagulant, as measured by several different methods (Fukushi et al., 1998; Han et al., 2007; Okada et al., 1990), and the algal cells had negative charges as well. When Al3þ was added, the negative charges of the algal surface were neutralized, and floc–bubble attachment increased with Al3þ dosage as the floc surfaces became more positively charged. It should be noted that algal properties and medium chemistry affect algal coagulation using Al3þ (Zhang et al., 2012), and thus different attachment efficiency-Al3þ dosage relationships were observed for nitrogen-depleted and nitrogen-replete culture (Supplementary Fig. S5). Nonetheless, with the developed Al3þ dosage-attachment efficiency relationship and the measured floc size distribution, the model is capable of reproducing the DAF harvesting efficiency at any given Al3þ concentration. As both the attachment efficiency and floc size change with increased Al3þ dosage, a method to separate the effects of the size distribution and the attachment efficiency is to alter the floc surface with a surfactant. These measurements were conducted using the nitrogen-replete algal culture where it is easier to observe reduced harvesting efficiencies at intermediate Al3þ dosages. For the Al3þ coagulated suspension (50 mg Al3þ g1), the harvesting efficiency was 22%, with the calculated floc–bubble attachment efficiency of 0.04. Adding 50 and 100 mg g1 of the surfactant CTAB increased 0.14 and 0.38 more on attachment efficiencies and thus raised the harvesting efficiencies to 57% and 69%, respectively (Fig. 6). This change in efficiency occurred without a change in the
Figure 6. Increase in DAF harvesting efficiencies as a result of enhanced floc– bubble attachment efficiency (a) by addition of CTAB to Al3þ coagulated nitrogenreplete culture. CTAB addition (at the dosage of 50 and 100 mg g1) did not change the flocs size distribution generated by Al3þ.
floc size distribution (data not shown). To check the independent influence of CTAB, harvesting efficiencies without prior Al3þ coagulation were also measured. Less than 5% harvesting efficiency was obtained when 50 and 100 mg g1 CTAB was used alone as CTAB formed few flocs at these concentrations. The results confirmed that, for DAF algal harvesting, floc generation was insufficient to achieve a high harvesting efficiency, and the floc surface should also be favorable for floc–bubble attachment. DAF Harvesting Model Evaluation-Influence of Recycle Ratio and Biomass Concentration The key extension in the model presented here relative to previous models is the incorporation of the influence of a bubble limitation. This aspect of the model was evaluated by comparing the predicted harvesting efficiencies with the experimental data at different recycle ratios and different biomass concentrations. In Figure 7 the predicted and measured efficiencies with different recycle ratio is examined with initial algal concentrations of 1.5 and 7.4 g L1, respectively. For both biomass concentrations low efficiencies were found at low recycle ratios (equivalent to lower bubble concentrations, cf. Equation 3), and increasing the recycle ratio improved the harvesting efficiencies. The trends were predicted by the model with reasonable accuracy. As the recycle ratio increased, more bubbles were predicted to be attached to the larger flocs and there were still not enough bubbles for the smaller size flocs, which limited the further increase of harvesting efficiencies. Figure 8 shows the predicted and measured influence of the biomass concentrations on the harvesting efficiency at a fixed recycle ratio along with the model predictions. At the highest biomass
Figure 7.
Influence of recycle ratio on DAF harvesting efficiency in nitrogendepleted cultures. The algal suspension at the initial concentration of 7.4 g L1 was DOM-free. Al3þ dosages were 68 and 48.5 mg g1 for algal suspensions with an initial concentration of 1.5 and 7.4 g L1, respectively.
Zhang et al.: DAF Algal Harvesting Modeling Biotechnology and Bioengineering
2483
Figure 8. Influence of initial dry weight on DAF harvesting efficiency at recycle ratios of 20% and 40%. An attachment efficiency of 0.5, and a size distribution obtained from coagulation of nitrogen-depleted C. zofingiensis with 68 mg g1 Al3þ were used in the model.
concentrations, it is apparent from both the model and the experimental data that additional strategies might be required to attain high harvesting efficiencies. Such strategies might include enhanced flocculation through alternate floc generation to generate large and dense flocs, enhanced bubble production through greater recycling ratio or greater air saturation, efficient bubble distribution among the flocs, and the transition from flotation to sedimentation at highest biomass loads, etc. In general, the ratio of the bubble concentration to the number of flocs containing the majority of the biomass is a critical parameter for high efficiency harvesting, and this parameter is introduced through the model developed in this article. The model can make a number of further predictions with other parameters, and the predictions of the model sensitivity to the bubble diameter, algal cell diameter, and floc fractal dimension are provided in Figures S6, S7, and S8 of the Supplementary Information as examples. Many of these predictions are inherited from the earlier forms of the white water blanket model and the reader is referred to those references (Edzwald, 2010; Haarhoff and Edzwald, 2004).
Discussion When an algal suspension is destabilized, coagulation occurs and flocs form (Bratby, 2006). In the present work coagulation has a dual effect. First, the destabilization and flocculation of the algal cells also makes those cells more amenable to cell-bubble adhesion since both cells and bubbles generally have negative charges, and the coagulant tends to neutralize these charges. In the absence of coagulation, the ratio of cells to bubbles is too large and the collision frequency is too low to achieve high
2484
Biotechnology and Bioengineering, Vol. 111, No. 12, December, 2014
harvesting efficiencies. Collision efficiencies are higher between larger flocs and bubbles allowing the larger flocs to be more easily captured by bubbles (Han et al., 2007), as also indicated in Supplementary Figure S1A. The combined importance of floc generation and surface modification in enhancing floc–bubble collision and attachment is supported by CTAB addition after Al3þ coagulation (Fig. 6). When Al3þ was used to develop flocs and its dosage was not high enough for efficient floc–bubble attachment, the addition of the surfactant CTAB increased the attachment efficiency, and thus significantly enhanced harvesting efficiency. To achieve greater than 90% harvesting efficiency, the floc size must be similar to or larger than the bubbles (55 mm here), and the attachment efficiency should exceed 0.4 (Supplementary Fig. S9). A second effect of the coagulation process is the reduction in the total number of particles. In DAF processes with high solids loading like algal harvesting, the number of algal cells can exceed the number of generated bubbles by orders of magnitudes. For example, in algal biomass harvesting, a typical cell number was 8.6 1010 cells L1 (dry weight of 1.5 g L1, at an average size of 3.0 mm) compared with 1 108 bubbles L1. Coagulation reduces the number of algal cells by a factor of hundreds and serves to reduce the number of required collisions to that range. Further, the majority of the biomass is associated with a relatively small number of large flocs (Supplementary Fig. S2). This allows high efficiency harvesting at reasonable recycle ratios (i.e., 20%) for typical algal biomass concentrations of a 1–2 g L1, although at higher biomass concentrations the harvesting efficiency is reduced. The model developed in this work helps to explain and predict this process as indicated in Figure 7. Higher recycle ratios can extend the effective high-efficiency harvesting range, but there are limits as indicated in Figures 7 and 8. An important consideration with higher recycle ratios is the energy associated with generating additional bubbles; Hanotu et al. (2012, 2014) suggest alternate bubblegeneration strategies, though results are reported only at bubble concentrations about three orders of magnitude less than the present work. To predict these newly identified limiting factors a DAF algal harvesting model was developed by introducing a distribution of algal floc sizes and a limit to the available bubbles into the existing white water blanket model. The model can predict the influence of algal concentration and recycle ratio on harvesting efficiency with satisfactory results while it inherits the existing capabilities of the original models (i.e., predicting effects of bubble sizes, Supplementary Fig. S6, predicting effects of algal cell sizes, Supplementary Fig. S7, predicting effects of floc fractal dimension, Supplementary Fig. S8). Floc size distribution can be measured experimentally; however, another input—floc– bubble attachment efficiency—needs to be determined by fitting measurements. Further development of coagulation models for the prediction of algal floc size distributions, floc structure, and floc surface properties is necessary to better expand the application of the modified white water blanket model to DAF algal harvesting.
Conclusions In this study, the impacts of type and dosage of coagulant, algal culture conditions, initial cell dry weight, and recycle ratio on DAF algal harvesting of C. zofingiensis were critically evaluated. A model was developed by extending the previous white water blanket model and was applied to the DAF harvesting process. The model extensions focus on the incorporation of the floc size distribution, the preferential interaction of bubbles with larger flocs and the limited availability of bubbles relative to the number of flocs. The harvesting efficiencies predicted by the model agree reasonably with experimental data obtained at different Al3þ dosages (the effect of different floc size distributions), initial algal concentrations, and recycle ratios. On the basis of the experimental data and model predictions, three critical conditions necessary to achieve a high DAF harvesting efficiency were identified: (i) flocs of similar or larger sizes than the DAF bubbles (55 mm here) should be generated to improve bubble-floc collision efficiency; (ii) the surface of algal flocs after coagulation should be favorable for bubble attachment; and (iii) a sufficient number of bubbles is necessary to capture the flocs generated when harvesting algal biomass with a higher initial concentration, which can be achieved by increasing the recycle ratio or altering the floc size distribution. The authors thank Dr. James Edzwald for many constructive comments in the review of a draft of this article. This work was partially supported by the Laboratory Directed Research and Development program at Sandia National Laboratories and by the Arizona Center for Algae Technology and Innovation at Arizona State University. Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL85000.
References Ahmad AL, Yasin NHM, Derek CJC, Lim JK. 2011. Optimization of microalgae coagulation process using chitosan. Chem Eng J 173(3): 879–882. Bratby J. 2006. Coagulation and flocculation in water and wastewater treatment. London: IWA Publishing. Christenson L, Sims R. 2011. Production and harvesting of microalgae for wastewater treatment, biofuels, and bioproducts. Biotechnol Adv 29(6):686–702. Conti ALR, Guerrero JM, Regueira JM. 2005. Levels of microcystins in two argentinean reservoirs used for water supply and recreation: Differences in the implementation of safe levels. Environ Toxicol 20(3):263–269. Del Campo JA, Rodriguez H, Moreno J, Vargas MA, Rivas J, Guerrero MG. 2004. Accumulation of astaxanthin and lutein in Chlorella zofingiensis (Chlorophyta). Appl Microbiol Biotechnol 64(6):848–854. Divakaran R, Pillai VNS. 2002. Flocculation of algae using chitosan. J Appl Phycol 14(5):419–422.
Edzwald JK. 1995. Principles and applications of dissolved air flotation. Water Sci Technol 31(3–4):1–23. Edzwald JK. 2010. Dissolved air flotation and me. Water Res 44(7): 2077–2106. Edzwald JK, Wingler BJ. 1990. Chemical and physical aspects of dissolved-air flotation for the removal of algae. Aqua (Oxford) 39(1):24–35. Fukushi K, Matsui Y, Tambo N. 1998. Dissolved air flotation: Experiments and kinetic analysis. J Water Services Res Technol-Aqua 47(2):76–86. Gregory R, Edzwald JK. 2010. Sedimentation and flotation. In: Edzwald JK, editor. Water quality and treatment. 6th edn. New York, NY: McGraw Hill. Haarhoff J, Edzwald JK. 2004. Dissolved air flotation modeling insights and shortcomings J Water Supply Res Technol-Aqua 53(3):127–150. Han M, Kim T-I, Kim J. 2007. Effects of floc and bubble size on the efficiency of the dissolved air flotation (DAF) process. Water Sci Technol 56(10): 109–115. Hanotu J, Bandulasena HCH, Zimmerman WB. 2012. Microflotation performance for algal separation. Biotechnol Bioeng 109(7):1663–1673. Hanotu J, Karunakaran E, Bandulasena H, Biggs C, Zimmerman WB. 2014. Harvesting and dewatering yeast by microflotation. Biochem Eng J 82:174–182. Hu Q, Sommerfeld M, Jarvis E, Ghirardi M, Posewitz M, Seibert M, Darzins A. 2008. Microalgal triacylglycerols as feedstocks for biofuel production: Perspectives and advances. Plant J 54(4):621–639. James HA, Smith CJ, Sutton A, Caley J, Franklin O, Parr W. 1994. Survey of the concentrations of algal toxins in water supplies. Final report to the Department of the Environment, UK. Liers S, Baeyens J, Mochtar I. 1996. Modeling dissolved air flotation. Water Environ Res 68(6):1061–1075. Liu J, Huang J, Fan KW, Jiang Y, Zhong Y, Sun Z, Chen F. 2011. Production potential of Chlorella zofingienesis as a feedstock for biodiesel. Bioresour Technol 101(22):8658–8663. Ma J, Lei G, Fang J. 2007. Effect of algae species population structure on their removal by coagulation and filtration processes—A case study. J Water Supply ResTechnol-Aqua 56(1):41–54. Malley JP, Edzwald JK. 1991. Concepts for dissolved-air flotation treatment of drinking waters. Aqua (Oxford) 40(1):7–17. Okada K, Akagi Y, Kogure M, Yoshioka N. 1990. Analysis of particle trajectories of small particles in flotation when the particles and bubbles are both charged. Can J Chem Eng 68(4):614–621. Tambo N, Matsui Y, Fukushi K. 1986. A kinetic study of dissolved air flotation, Proceedings of World Congress III of Chemical Engineering, 3, 200–203, Tokyo, Japan. Teixeira MR, Rosa MJ. 2006. Comparing dissolved air flotation and conventional sedimentation to remove cyanobacterial cells of Microcystis aeruginosa Part 1: The key operating conditions. Sep Purif Technol 52(1): 84–94. Vandamme D, Foubert I, Meesschaert B, Muylaert K. 2010. Flocculation of microalgae using cationic starch. J Appl Phycol 22(4):525–530. Wyatt NB, Gloe LM, Brady PV, Hewson JC, Grillet AM, Hankins MG, Pohl PI. 2012. Critical conditions for ferric chloride-induced flocculation of freshwater algae. Biotechnol Bioeng 109(2):493–501. Zhang X, Amendola P, Hewson JC, Sommerfeld M, Hu Q. 2012. Influence of growth phase on harvesting of Chlorella zofingiensis by dissolved air flotation. Bioresour Technol 116 477–484.
Supporting Information Additional supporting information may be found in the online version of this article at the publisher’s web-site.
Zhang et al.: DAF Algal Harvesting Modeling Biotechnology and Bioengineering
2485
Department of Applied Sciences and Mathematics, Arizona State University, Polytechnic Campus, Mesa, Arizona 85212; telephone: þ1 480 727 5294; fax: þ1 480 727 1475; e-mail: [email protected] 2 Sandia National Laboratories, Albuquerque, New Mexico 3 School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, Georgia 4 Institute of Hydrobiology, Chinese Academy of Sciences, Wuhan, Hubei, China
Introduction ABSTRACT: In this study, Chlorella zofingiensis harvesting by dissolved air flotation (DAF) was critically evaluated with regard to algal concentration, culture conditions, type and dosage of coagulants, and recycle ratio. Harvesting efficiency increased with coagulant dosage and leveled off at 81%, 86%, 91%, and 87% when chitosan, Al3þ, Fe3þ, and cetyl trimethylammonium bromide (CTAB) were used at dosages of 70, 180, 250, and 500 mg g1, respectively. The DAF efficiency-coagulant dosage relationship changed with algal culture conditions. Evaluation of the influence of the initial algal concentration and recycle ratio revealed that, under conditions typical for algal harvesting, it is possible that the number of bubbles is insufficient. A DAF algal harvesting model was developed to explain this observation by introducing mass-based floc size distributions and a bubble limitation into the white water blanket model. The model revealed the importance of coagulation to increase flocbubble collision and attachment, and the preferential interaction of bubbles with larger flocs, which limited the availability of bubbles to the smaller sized flocs. The harvesting efficiencies predicted by the model agree reasonably with experimental data obtained at different Al3þ dosages, algal concentrations, and recycle ratios. Based on this modeling, critical parameters for efficient algal harvesting were identified. Biotechnol. Bioeng. 2014;111: 2477–2485. ß 2014 Wiley Periodicals, Inc. KEYWORDS: flotation; model; coagulant; flocs; bubble; Chlorella
Correspondence to: X. Zhang Contract grant sponsor: The Laboratory Directed Research and Development program, Sandia National Laboratories and by the Arizona Center for Algae Technology and Innovation at Arizona State University, U.S. Department of Energys National Nuclear Security Administration Received 28 February 2014; Revision received 2 May 2014; Accepted 23 May 2014 Accepted manuscript online 31 May 2014; Article first published online 14 July 2014 in Wiley Online Library (http://onlinelibrary.wiley.com/doi/10.1002/bit.25300/abstract). DOI 10.1002/bit.25300
ß 2014 Wiley Periodicals, Inc.
Dissolved air flotation (DAF) has been widely used to separate microalgae from natural water and wastewater (Edzwald, 2010; Ma et al., 2007), and has often proven more efficient than sedimentation for algae-water separation (Teixeira and Rosa, 2006). DAF has increasingly been used to harvest algae from raceway ponds and photobioreactors for biofuel applications (Christenson and Sims, 2011; Zhang et al., 2012). An important factor in the economic application of DAF to algal harvesting is the energy costs associated with bubble formation, and recent work in efficient bubble generation has been reported by Hanotu et al. (2012, 2014). The application of DAF to algal harvesting differs from algal separation in water and wastewater treatment in that the cell concentration in algal mass culture is typically thousands of times greater than in water and wastewater (Conti et al., 2005; James et al., 1994). Further, the biochemical and chemical properties of algal cells and culture media change considerably depending on algal strains and culture conditions, which may in turn affect coagulant demand and flotation efficiency. Operational protocols for DAF algal harvesting are usually developed on the basis of empirical considerations combined with jar testing. There is a pressing need to critically evaluate and understand the influence of algal properties, bubble properties, and coagulation conditions on DAF algal harvesting. Modeling of DAF algal harvesting would help to identify key parameters involved in the process and thus be valuable in efforts to achieve high harvest efficiencies. The white water blanket model, initially developed by Edzwald and colleagues (Edzwald, 1995; Edzwald and Wingler, 1990; Malley and Edzwald, 1991) has been widely used for modeling the separation of water and particles, including algal cells, in water and wastewater treatment. The term “white water blanket” describes a rising air bubble suspension, a “blanket” in a DAF tank that captures Biotechnology and Bioengineering, Vol. 111, No. 12, December, 2014
2477
particles above as it moves toward the water surface. In this model, the bubble number had not been considered to be a limiting factor, as the ratio of particles to bubbles in water or wastewater typically is not large. The particle number-based efficiency has also generally been calculated using the average particle size instead of a size distribution. However, with higher initial cell concentration in algal culture, the bubble number may be limiting. After coagulation or flocculation, the flocs number distribution is dominated by the smallest particle sizes, while most of the biomass is associated with a few large flocs. It will therefore be necessary to consider the floc size distribution when modeling DAF algal harvesting, as the collection of larger flocs contributes more to the overall harvesting efficiency. In this study, the effects of culture conditions, cell concentration, coagulant type and dosage, and recycle ratio on DAF harvesting efficiency were investigated using the oleaginous green microalga C. zofingiensis as a model organism. A DAF algal harvesting model was developed to predict the mass-based harvesting efficiency by introducing a bubble limitation and mass-based floc size distribution into the white water blanket model. The critical parameters for effective DAF algal harvesting were identified. The objectives of this study are to (1) evaluate the influence of algal cell properties, coagulants and DAF operation conditions on algal harvesting efficiency, and (2) extend the white water blanket model to understand and predict the outcome of the DAF algal harvesting process by incorporating the effects of floc size distribution and limits to the number of bubbles.
Materials and Methods Algal Culture and Characterization The freshwater microalga C. zofingiensis (LRB-AZ-1201) isolated and maintained in the Laboratory for Algae Research and Biotechnology at Arizona State University (Arizona, USA) was grown in both nitrogen-depleted and nitrogen-replete conditions, as nitrogen starvation favors lipid accumulation (Hu et al., 2008). C. zofingiensis was selected as a model organism because it is widely used as a production strain for biofuels and high-value bioproducts (Del Campo et al., 2004; Liu et al., 2011) and has a spherical shape. The nitrogen-depleted cells were cultured indoors in 15 L panel reactors with a 5 cm light path and continuous fluorescent lighting. The nitrogen-replete culture was grown in 50 L panel reactors with the same light path and natural dark-light cycles. The light intensity, culture pH, initial and final NO3-N concentration, and initial and final dry weight are summarized in Table I. The culture was harvested after six days of cultivation, and the dry weights of the nitrogen-depleted and nitrogenreplete culture reached approximately 1.5 and 2.0 g L1, respectively.
2478
Biotechnology and Bioengineering, Vol. 111, No. 12, December, 2014
Table I. Summary of culture conditions for both nitrogen-depleted and nitrogen-replete C. zofingiensis. Parameters
Nitrogen-depleted
Nitrogen-replete
Indoor or outdoor Culture volume (L) Light path (cm) Initial algal concentration (g L1) Initial NO3-N (mg L1) Final NO3-N (mg L1) Algal concentration before harvesting (g L1) Culture period (d) Growth phasea
Indoor 15 5 0.2 33.6 NDb 1.5
Outdoor 50 5 0.5 128 23.5 2.0
6 Stationary
Light intensity (mmol m2 s1) pH range Culture temperature ( C)
220 7.0–8.2 26
6 Exponential growth 1900c 7.0–8.5 15–30
a
Growth phases were identified by changes in cell number (data not shown). b Not detectible. c Daily maximum.
DAF Algal Harvesting C. zofingiensis cultures were harvested using a DAF jar tester (DBT6, EC Engineering) using the procedure previously described (Zhang et al., 2012). In this study, chitosan (Sigma– Aldrich 417963-100G, practical grade), aluminum sulfate (Al2(SO4)3 18H2O), ferric sulfate (Fe2(SO4)3), and cetyl trimethyl ammonium bromide (CTAB) were tested for their efficacy for DAF pretreatment, and aluminum sulfate was then selected for further study. After rapid injection of the coagulant, the pH was adjusted immediately to optimal coagulation conditions (pH of 7.0 0.2, 6.2 0.2, 7.5 0.5, and 8.0 0.2 for chitosan, aluminum, CTAB and iron, respectively) by adding pre-determined amounts of 1 M NaOH. After coagulation, the algal suspension was flocculated for 10 min at 30 rpm (velocity gradient G ¼ 30 s1) before bubbles were injected to float the algal cells. The harvesting efficiency was calculated on the basis of the mass of the algae in the subnatant and in the original media. Dissolved organic matter (DOM)-free cultures were used to exclude the influence of DOM on Al3þ consumption and floc size distribution when harvesting algal cells with different biomass concentrations at a 20% recycle ratio and when harvesting algal cells at the same biomass concentration (7.4 g L1) with different recycle ratios. DOM in the nitrogen-depleted culture was removed by centrifugation (6,000 rpm for 10 min), and cells were re-suspended into fresh BG11 culture medium prior to harvesting. The Al3þ dosage of 48.5 mg g1 was used in these parameter studies based on preliminary experiments.
Floc Size Distribution Measurement Before bubble injection, a sample of the algal suspension was transferred to a microscope slide for floc size measurement.
Samples were carefully taken to avoid disturbing the flocs using a 1 mL transfer pipet cut at the tip to provide a larger opening. A minimum of three samples were taken and at least 2,000 algal cells and flocs for each sample were captured via bright field microscopy. The area of each floc in the images was determined using a pixel-counting algorithm (ImageJ software, version 1.43u), and the equivalent diameter of each floc was calculated from the square root of the fine-grained area. The number of flocs in each size range (1.0–2.0, 2.0– 3.0… 30–40… 390–400 mm) was then counted to obtain a number-based size distribution as a function of the floc diameter, df. The size measurements were verified using 3.0 mm-, 30 mm-, and 100 mm-SiO2 monodisperse particle standards (Microspheres-Nanospheres, New York). As algal biomass harvesting efficiencies are typically expressed in terms of dry weight rather than cell number, the number-based size distribution (dFn) was then converted to a mass-based size distribution (dFm) for model input using Equation (1): idF n dF m ¼ R1 idF n
ð1Þ
dp
where dp is the diameter of individual cells and i is the number of algal cells in a floc. A detailed discussion of Equation (1) can be found in Section S1 (Supplementary Information).
Modeling of DAF Algal Harvesting White Water Blanket Model for DAF Particle Separation To better understand the DAF algal harvesting processes, the white water blanket model was employed to describe floc-bubble interactions and subsequent separation. This model predicts the number of flocs removed relative to the initial number (Edzwald, 2010; Edzwald and Wingler, 1990): nf ðtÞ ¼ expðltnb Þ nf ;0
ð2Þ
where nf is the floc concentration, nb is the bubble concentration, t is the floc-bubble contact time determined by the DAF unit flow characteristics and l ¼ 0:25ahT vb pd 2b is a rate expression in inverse concentration units. In the rate expression, hT is the sum of various dimensionless flocbubble collision efficiencies (Haarhoff and Edzwald, 2004), a is the floc-bubble attachment efficiency, vb is the bubble rise velocity and db is the bubble diameter. Further information on the floc-bubble collision efficiencies employed is provided in Section S2 and Figure S1 of the Supplementary Information. The total number of bubbles is a function of the bubble diameter, recycle ratio R, air mass concentration in the water at saturator pressure (ys), and at ambient pressure (y0), and the combined saturator and nozzle efficiency, k, as described in Equation (3). A detailed
derivation of Equation (3) is provided in Section S3 (Supplementary Information). nb ¼
6kðys y0 ÞR pd b 3 ð1 þ RÞrair
ð3Þ
DAF Algal Harvesting Model With Floc Size Distribution and Bubble Limitation In the DAF algal harvesting process, the floc mass is distributed over a range of floc sizes (Supplementary Fig. S2). In addition, the bubble number may be limiting as the floc number is much higher than in typical DAF water and wastewater treatments. Thus the above model was extended to account for limited bubble availability; these extensions to the model are explained in somewhat greater detail in the Section S4 of the Supplementary Information. As no simple analytical solution jointly solves the evolution of bubbles and flocs over a range of floc sizes, a simpler approach was used based on the greater collision coefficients of bubbles with large flocs (Supplementary Fig. S1A). Bubbles are allowed to interact with flocs of a given size class at the probability given by the original model. Because floc size is often comparable to or exceeds bubble size, once bubble-floc collisions occurred the bubble-floc agglomerates were assumed to not sweep up further algal flocs, although further bubble collisions were allowed up to a maximum number of bubbles determined by the floc size. Bubbles are assumed to interact with the largest flocs first and the prediction of bubble–floc interaction was carried out by integrating across the size distribution from the largest to the smallest flocs; that is, the remaining bubbles were allocated to possible interactions with the next smaller floc size until all the bubbles were assigned or the entire floc size distribution was covered. This assumption that the interaction occurs sequentially from the largest to smaller flocs is certainly imperfect; simulations with greater resolution of floc–bubble interaction dynamics could help resolve the limitations of the model but are beyond the present work scope. The maximum number of bubbles that can attach to a floc of a given size (nmax) is calculated using Equation (4), assuming each bubble occupies a square (db db) on the surface of a floc (Tambo et al., 1986): nmax ¼ p
2 df db
ð4Þ
The minimium number of bubbles needed to float a floc of a given size (nmin) is calculated using Equation (5), assuming the floc–bubbble can be floated if its density (rfb) is lower than the density of water (rw) rfb ¼
rf d f 3 þ nmin rb d b 3 < rw d f 3 þ nmin d b 3
ð5Þ
The total number of bubbles available for interaction with flocs smaller than df depends on the number of bubbles that
Zhang et al.: DAF Algal Harvesting Modeling Biotechnology and Bioengineering
2479
have attached to larger flocs. Similar to Equation (2), an expression for the bubble-floc collision rate gives the number of bubbles that stick to flocs of size df: dBn ¼ minfBn ½1 expðlt nf dFn Þ;
DðdFn Þnmax g
ð6Þ
where Bn is the bubbles available to interact with flocs of size df, and dBn is the number of bubbles that stick to flocs of size df. dBn is limited to the maximum number of bubbles that attach to a given floc, D(dFn)nmax. Note that we have left Bn in concentration dimensions to simplify the notation; it has not been normalized to unity. The number of bubbles available for interaction with smaller flocs is found by integrating dBn over the floc size distribution: 2 3 Zdf ð7Þ dBn5 Bn ¼ max40; nb;0 1
Based on Equation (2), the number fraction of flocs of a given size collected by bubbles is 1exp(ltBn) where Bn varies according to Equation (7). The fraction of these that actually float will be determined by the floc-bubble rise velocity, vfb, depending on rfb in Equation (5). The number fraction of flocs that are actually harvested is: DðdFn Þ ¼ r fb ½1 expðltBn Þ
ð8Þ
where r fb ¼ min 1; max 0; vfbht r , which denotes the fraction of flocs floated, tr is the allowed rise time and h is the water depth. For the experiments in this study, the allowed rise time was effectively infinite so that rfb is approximately unity as long as the floc-bubble velocity is positive. The total mass harvested by bubbles (mH) is the integral over floc distribution and the fraction harvested times the initial dry mass, m0,
Equation (8) and Equation (9) according to floc-bubble collision efficiency and attachment efficiency, using the available bubble number at the given floc size range. Small flocs are collected at reduced (or zero) efficiency as bubbles are depleted. The final mass-based harvesting efficiency was then calculated using Equation (10), the integral of the mass of the flocs collected in all floc size ranges. A schematic diagram of the modeling algorithm is shown in Supplementary Figure S3.
Results Influence of Coagulants on DAF Harvesting Efficiencies Coagulants differ in their capacity to generate flocs and in their ability to enhance floc-bubble attachment efficiency. The influence of different coagulants and their dosages on measured DAF harvesting efficiencies is shown in Figure 1. Harvesting efficiency increased with coagulant dosage and leveled off at 81%, 86%, 91%, and 87% when chitosan, Al3þ, Fe3þ, and cetyl trimethylammonium bromide (CTAB) were used at dosages of 70, 180, 250, and 500 mg g1, respectively. Different efficiency-dosage relationships were obtained for different coagulants with the lowest dosage requirements observed for chitosan, which agrees with the literature (Ahmad et al., 2011; Divakaran and Pillai, 2002; Vandamme et al., 2010; Wyatt et al., 2012). The different efficiencydosage relationships obtained for different coagulants may be due to their different floc formation properties and the different influences on floc-bubble interactions, possibly as a result of the characteristics of charge density of different coagulants. Larger flocs were generated using chitosan, whereas smaller flocs were generated using CTAB
Z1 mH ¼ m0
DðdFn ÞdFm
ð9Þ
dp
The total harvesting efficiency can be obtained by comparing the total mass harvested, mH, with the initial dry mass, m0: hmDAF ¼
mH ¼ m0
Z1 DðdFn ÞdFm
ð10Þ
dp
In summary, the mass-based fraction (dFm) of flocs in each size range was first calculated using Equation (1). Bubbles were allocated to the flocs using collision efficiencies based on the original models, from larger to smaller sizes, until all bubbles were assigned. The number and mass of flocs in a given size range collected by bubbles was then calculated with
2480
Biotechnology and Bioengineering, Vol. 111, No. 12, December, 2014
Figure 1.
DAF harvesting efficiencies for nitrogen-replete C. zofingiensis as a function of chitosan, Al3þ, Fe3þ, and CTAB dosages.
(Supplementary Fig. S4). Since similar trends as a function of relative dose are obtained for all coagulants, subsequent experiments were conducted using Al3þ. Influence of Culture Conditions on Coagulant Dosage Requirements Harvesting efficiency depends on the surface properties of the algae and the ability of the coagulants to alter algal cell surface properties to favor flocculation and floc–bubble adhesion. Physiological stress, as occurs under nitrogen-depleted culture conditions, accelerates algae aging and changes their surface properties. C. zofingiensis cultures were grown in nitrogen-depleted and nitrogen-replete media as indicated in Table I, and used to evaluate the influence of surface properties on DAF harvesting. At the time of harvesting, the nitrogen-depleted culture had reached the stationary growth phase with a dry weight of 1.5 g L1 and the nitrogen-replete culture was still in the exponential growth phase with a dry weight of 2.0 g L1. Figure 2 shows the DAF efficiency-Al3þ dosage relationships for the cultures under a fixed recycle ratio of 20%. Harvesting efficiencies increased with Al3þ dosages, peaking near 90%. However, to achieve the same harvesting efficiencies, the nitrogen-replete culture required more Al3þ than the nitrogen-depleted culture. This result agrees with previous work that showed algae surfaces differ depending on culture conditions, and that aged C. zofingiensis required less Al3þ for harvesting (Zhang et al., 2012).
fixed 20% recycle ratio (Fig. 3A). The harvesting efficiency was greater than 92% when the initial algal cell concentration was 1.2 g L1. The harvesting efficiency dropped to 81% as the initial cell concentration increased to 3.1 g L1, and it further decreased to 41% at a 7.4 g L1 initial cell concentration. The typical range of recycle ratios used for algal cell separation in water and wastewater treatments is between 6% and 12% (Gregory and Edzwald, 2010), but the cell concentrations in these cases are typically less than 107 cells L1 (Conti et al., 2005; James et al., 1994). With a concentration of 1.5 g L1, C. zofingiensis cell number was approximately 8.6 1010 cells L1, with 2.2 108 flocs L1 of
Influence of Initial Algal Biomass Concentration and Recycle Ratio on Harvesting Efficiency The influence of the initial algal biomass concentration on harvesting efficiency was tested with DOM-free cultures at a
Figure 2. DAF harvesting efficiency-Al3þ dosage relationships for both nitrogenreplete (2.0 g L1) and nitrogen-deplete (1.5 g L1) C. zofingiensis at the recycle ratio of 20%.
Figure 3. Influence of initial algal dry weight (A) and recycle ratio (B) on DAF harvesting efficiency of C. zofingiensis. The recycle ratio was 20% in (A), and the initial cell concentration was 7.4 g L1 in (B). DOM in the nitrogen-depleted culture was removed, and algal pellets were suspended in fresh culture medium to make different concentrations.
Zhang et al.: DAF Algal Harvesting Modeling Biotechnology and Bioengineering
2481
different sizes after coagulation. As the initial algal dry weight increased, the number of flocs also increased, thus requiring more air bubbles for their removal. To confirm the impact of the bubble limitation on DAF algal harvesting, the influence of recycle ratio on the harvesting at a high biomass concentration, 7.4 g L1, was tested (Fig. 3B). Bubble concentration increased with recycle ratio as shown in Equation (3). The harvesting efficiency increased with recycle ratio, reaching 72% when the recycle ratio increased to 50%. The results in Figure 3A and B both suggest that the available number of bubbles is a limiting factor at high biomass concentrations: more bubbles are required when harvesting algal biomass with a higher concentration. This behavior has not traditionally been predicted with the white-water blanket model for DAF separations and motivated the model developed in Equations (6–10). In the following section we evaluated the model in the context of the data presented here. DAF Harvesting Model Evaluation—Influence of Floc Size Distribution and Attachment Efficiency Figure 4 shows images of algal cells and flocs after Al3þ flocculation. Without Al3þ, C. zofingiensis cells were uniform in size, averaging 3 mm. When Al3þ was added, flocs were generated. Increased dosages of Al3þ resulted in larger floc sizes and fewer single cells surrounding the flocs. A reduction in the number of single cells was also observed with increased dosages of chitosan, Fe3þ, and CTAB. Both number-based and mass-based size distributions at different Al3þ dosages are given in Supplementary Figure S2, while distributions for different coagulants are compared in Supplementary Figure S4 for relevant respective high-end dosages (Supplementary Information). Although the number distribution was dominated by the smallest particle sizes, most of the
Figure 4. Micrographs of flocs of nitrogen-depleted C. zofingiensis at different Al3þ dosages. The length of the scale bar is 100 mm.
2482
Biotechnology and Bioengineering, Vol. 111, No. 12, December, 2014
biomass was associated with a few large flocs after coagulation, so that a high harvesting efficiency can be achieved if relatively few large flocs are generated through coagulation and then collected by bubbles. The addition of coagulant alters the algal surface allowing floc formation and easier attachment between flocs and bubbles, represented by the floc–bubble attachment efficiency, a. The influences of both Al3þ dosage (or floc size distribution) and a on the predicted DAF harvesting efficiencies of nitrogendepleted C. zofingiensis are shown in Figure 5 along with measurements from Figure 2. Dashed lines connect symbols representing predicted harvesting efficiencies for varying a for a given size distribution (a function of the Al3þ dosage from Supplementary Fig. S2). Values of a at different Al3þ dosages were then selected from values in which the predicted efficiencies fit well with the experimental efficiencies; selected a values are plotted in Supplementary Figure S5. Without Al3þ, the modeled harvesting efficiencies were less than 10% no matter what a was assumed because too few collisions occurred between bubbles and individual small algal cells (cf. Supplementary Fig. S1). At intermediate dosages (20–40 mg g1 Al3þ in Fig. 5), there was substantial floc formation, but only moderate harvesting efficiency suggesting a less efficient attachment efficiency than is obtained at higher dosages. This trend is reflected in Supplementary Figure S5. At higher dosages, the probability of attachment became high with the biomass mostly due to the larger flocs size; the observed harvesting efficiencies suggested attachment efficiencies in the range 0.4–0.6. Previous authors have also reported that attachment efficiency increased with Al3þ concentration. For example,
Figure 5. Influence of Al3þ dosage and attachment efficiency (a) on DAF harvesting efficiency of nitrogen-depleted C. zofingiensis. Harvesting efficiencies at different Al3þ dosages were calculated using the measured floc size distribution at each Al3þ dosage, and a series of assumed a values (open symbols). The a value at a given Al3þ dosage was selected from values in which the modeled efficiencies (red solid circles) approximate the measured harvesting efficiencies (black pluses). The solid curve is a regression of the logistic function using the modeled efficiencies at different Al3þ dosage and selected a values (red solid circles) as inputs.
Liers et al. (1996) reported that a ranged from 0.3 to 0.4 and Haarhoff and Edzwald (2004) used a values between 0.5 and 1 for optimum coagulation. Bubbles had a negative charge ranging from 25 to 150 mV at neutral pH without coagulant, as measured by several different methods (Fukushi et al., 1998; Han et al., 2007; Okada et al., 1990), and the algal cells had negative charges as well. When Al3þ was added, the negative charges of the algal surface were neutralized, and floc–bubble attachment increased with Al3þ dosage as the floc surfaces became more positively charged. It should be noted that algal properties and medium chemistry affect algal coagulation using Al3þ (Zhang et al., 2012), and thus different attachment efficiency-Al3þ dosage relationships were observed for nitrogen-depleted and nitrogen-replete culture (Supplementary Fig. S5). Nonetheless, with the developed Al3þ dosage-attachment efficiency relationship and the measured floc size distribution, the model is capable of reproducing the DAF harvesting efficiency at any given Al3þ concentration. As both the attachment efficiency and floc size change with increased Al3þ dosage, a method to separate the effects of the size distribution and the attachment efficiency is to alter the floc surface with a surfactant. These measurements were conducted using the nitrogen-replete algal culture where it is easier to observe reduced harvesting efficiencies at intermediate Al3þ dosages. For the Al3þ coagulated suspension (50 mg Al3þ g1), the harvesting efficiency was 22%, with the calculated floc–bubble attachment efficiency of 0.04. Adding 50 and 100 mg g1 of the surfactant CTAB increased 0.14 and 0.38 more on attachment efficiencies and thus raised the harvesting efficiencies to 57% and 69%, respectively (Fig. 6). This change in efficiency occurred without a change in the
Figure 6. Increase in DAF harvesting efficiencies as a result of enhanced floc– bubble attachment efficiency (a) by addition of CTAB to Al3þ coagulated nitrogenreplete culture. CTAB addition (at the dosage of 50 and 100 mg g1) did not change the flocs size distribution generated by Al3þ.
floc size distribution (data not shown). To check the independent influence of CTAB, harvesting efficiencies without prior Al3þ coagulation were also measured. Less than 5% harvesting efficiency was obtained when 50 and 100 mg g1 CTAB was used alone as CTAB formed few flocs at these concentrations. The results confirmed that, for DAF algal harvesting, floc generation was insufficient to achieve a high harvesting efficiency, and the floc surface should also be favorable for floc–bubble attachment. DAF Harvesting Model Evaluation-Influence of Recycle Ratio and Biomass Concentration The key extension in the model presented here relative to previous models is the incorporation of the influence of a bubble limitation. This aspect of the model was evaluated by comparing the predicted harvesting efficiencies with the experimental data at different recycle ratios and different biomass concentrations. In Figure 7 the predicted and measured efficiencies with different recycle ratio is examined with initial algal concentrations of 1.5 and 7.4 g L1, respectively. For both biomass concentrations low efficiencies were found at low recycle ratios (equivalent to lower bubble concentrations, cf. Equation 3), and increasing the recycle ratio improved the harvesting efficiencies. The trends were predicted by the model with reasonable accuracy. As the recycle ratio increased, more bubbles were predicted to be attached to the larger flocs and there were still not enough bubbles for the smaller size flocs, which limited the further increase of harvesting efficiencies. Figure 8 shows the predicted and measured influence of the biomass concentrations on the harvesting efficiency at a fixed recycle ratio along with the model predictions. At the highest biomass
Figure 7.
Influence of recycle ratio on DAF harvesting efficiency in nitrogendepleted cultures. The algal suspension at the initial concentration of 7.4 g L1 was DOM-free. Al3þ dosages were 68 and 48.5 mg g1 for algal suspensions with an initial concentration of 1.5 and 7.4 g L1, respectively.
Zhang et al.: DAF Algal Harvesting Modeling Biotechnology and Bioengineering
2483
Figure 8. Influence of initial dry weight on DAF harvesting efficiency at recycle ratios of 20% and 40%. An attachment efficiency of 0.5, and a size distribution obtained from coagulation of nitrogen-depleted C. zofingiensis with 68 mg g1 Al3þ were used in the model.
concentrations, it is apparent from both the model and the experimental data that additional strategies might be required to attain high harvesting efficiencies. Such strategies might include enhanced flocculation through alternate floc generation to generate large and dense flocs, enhanced bubble production through greater recycling ratio or greater air saturation, efficient bubble distribution among the flocs, and the transition from flotation to sedimentation at highest biomass loads, etc. In general, the ratio of the bubble concentration to the number of flocs containing the majority of the biomass is a critical parameter for high efficiency harvesting, and this parameter is introduced through the model developed in this article. The model can make a number of further predictions with other parameters, and the predictions of the model sensitivity to the bubble diameter, algal cell diameter, and floc fractal dimension are provided in Figures S6, S7, and S8 of the Supplementary Information as examples. Many of these predictions are inherited from the earlier forms of the white water blanket model and the reader is referred to those references (Edzwald, 2010; Haarhoff and Edzwald, 2004).
Discussion When an algal suspension is destabilized, coagulation occurs and flocs form (Bratby, 2006). In the present work coagulation has a dual effect. First, the destabilization and flocculation of the algal cells also makes those cells more amenable to cell-bubble adhesion since both cells and bubbles generally have negative charges, and the coagulant tends to neutralize these charges. In the absence of coagulation, the ratio of cells to bubbles is too large and the collision frequency is too low to achieve high
2484
Biotechnology and Bioengineering, Vol. 111, No. 12, December, 2014
harvesting efficiencies. Collision efficiencies are higher between larger flocs and bubbles allowing the larger flocs to be more easily captured by bubbles (Han et al., 2007), as also indicated in Supplementary Figure S1A. The combined importance of floc generation and surface modification in enhancing floc–bubble collision and attachment is supported by CTAB addition after Al3þ coagulation (Fig. 6). When Al3þ was used to develop flocs and its dosage was not high enough for efficient floc–bubble attachment, the addition of the surfactant CTAB increased the attachment efficiency, and thus significantly enhanced harvesting efficiency. To achieve greater than 90% harvesting efficiency, the floc size must be similar to or larger than the bubbles (55 mm here), and the attachment efficiency should exceed 0.4 (Supplementary Fig. S9). A second effect of the coagulation process is the reduction in the total number of particles. In DAF processes with high solids loading like algal harvesting, the number of algal cells can exceed the number of generated bubbles by orders of magnitudes. For example, in algal biomass harvesting, a typical cell number was 8.6 1010 cells L1 (dry weight of 1.5 g L1, at an average size of 3.0 mm) compared with 1 108 bubbles L1. Coagulation reduces the number of algal cells by a factor of hundreds and serves to reduce the number of required collisions to that range. Further, the majority of the biomass is associated with a relatively small number of large flocs (Supplementary Fig. S2). This allows high efficiency harvesting at reasonable recycle ratios (i.e., 20%) for typical algal biomass concentrations of a 1–2 g L1, although at higher biomass concentrations the harvesting efficiency is reduced. The model developed in this work helps to explain and predict this process as indicated in Figure 7. Higher recycle ratios can extend the effective high-efficiency harvesting range, but there are limits as indicated in Figures 7 and 8. An important consideration with higher recycle ratios is the energy associated with generating additional bubbles; Hanotu et al. (2012, 2014) suggest alternate bubblegeneration strategies, though results are reported only at bubble concentrations about three orders of magnitude less than the present work. To predict these newly identified limiting factors a DAF algal harvesting model was developed by introducing a distribution of algal floc sizes and a limit to the available bubbles into the existing white water blanket model. The model can predict the influence of algal concentration and recycle ratio on harvesting efficiency with satisfactory results while it inherits the existing capabilities of the original models (i.e., predicting effects of bubble sizes, Supplementary Fig. S6, predicting effects of algal cell sizes, Supplementary Fig. S7, predicting effects of floc fractal dimension, Supplementary Fig. S8). Floc size distribution can be measured experimentally; however, another input—floc– bubble attachment efficiency—needs to be determined by fitting measurements. Further development of coagulation models for the prediction of algal floc size distributions, floc structure, and floc surface properties is necessary to better expand the application of the modified white water blanket model to DAF algal harvesting.
Conclusions In this study, the impacts of type and dosage of coagulant, algal culture conditions, initial cell dry weight, and recycle ratio on DAF algal harvesting of C. zofingiensis were critically evaluated. A model was developed by extending the previous white water blanket model and was applied to the DAF harvesting process. The model extensions focus on the incorporation of the floc size distribution, the preferential interaction of bubbles with larger flocs and the limited availability of bubbles relative to the number of flocs. The harvesting efficiencies predicted by the model agree reasonably with experimental data obtained at different Al3þ dosages (the effect of different floc size distributions), initial algal concentrations, and recycle ratios. On the basis of the experimental data and model predictions, three critical conditions necessary to achieve a high DAF harvesting efficiency were identified: (i) flocs of similar or larger sizes than the DAF bubbles (55 mm here) should be generated to improve bubble-floc collision efficiency; (ii) the surface of algal flocs after coagulation should be favorable for bubble attachment; and (iii) a sufficient number of bubbles is necessary to capture the flocs generated when harvesting algal biomass with a higher initial concentration, which can be achieved by increasing the recycle ratio or altering the floc size distribution. The authors thank Dr. James Edzwald for many constructive comments in the review of a draft of this article. This work was partially supported by the Laboratory Directed Research and Development program at Sandia National Laboratories and by the Arizona Center for Algae Technology and Innovation at Arizona State University. Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL85000.
References Ahmad AL, Yasin NHM, Derek CJC, Lim JK. 2011. Optimization of microalgae coagulation process using chitosan. Chem Eng J 173(3): 879–882. Bratby J. 2006. Coagulation and flocculation in water and wastewater treatment. London: IWA Publishing. Christenson L, Sims R. 2011. Production and harvesting of microalgae for wastewater treatment, biofuels, and bioproducts. Biotechnol Adv 29(6):686–702. Conti ALR, Guerrero JM, Regueira JM. 2005. Levels of microcystins in two argentinean reservoirs used for water supply and recreation: Differences in the implementation of safe levels. Environ Toxicol 20(3):263–269. Del Campo JA, Rodriguez H, Moreno J, Vargas MA, Rivas J, Guerrero MG. 2004. Accumulation of astaxanthin and lutein in Chlorella zofingiensis (Chlorophyta). Appl Microbiol Biotechnol 64(6):848–854. Divakaran R, Pillai VNS. 2002. Flocculation of algae using chitosan. J Appl Phycol 14(5):419–422.
Edzwald JK. 1995. Principles and applications of dissolved air flotation. Water Sci Technol 31(3–4):1–23. Edzwald JK. 2010. Dissolved air flotation and me. Water Res 44(7): 2077–2106. Edzwald JK, Wingler BJ. 1990. Chemical and physical aspects of dissolved-air flotation for the removal of algae. Aqua (Oxford) 39(1):24–35. Fukushi K, Matsui Y, Tambo N. 1998. Dissolved air flotation: Experiments and kinetic analysis. J Water Services Res Technol-Aqua 47(2):76–86. Gregory R, Edzwald JK. 2010. Sedimentation and flotation. In: Edzwald JK, editor. Water quality and treatment. 6th edn. New York, NY: McGraw Hill. Haarhoff J, Edzwald JK. 2004. Dissolved air flotation modeling insights and shortcomings J Water Supply Res Technol-Aqua 53(3):127–150. Han M, Kim T-I, Kim J. 2007. Effects of floc and bubble size on the efficiency of the dissolved air flotation (DAF) process. Water Sci Technol 56(10): 109–115. Hanotu J, Bandulasena HCH, Zimmerman WB. 2012. Microflotation performance for algal separation. Biotechnol Bioeng 109(7):1663–1673. Hanotu J, Karunakaran E, Bandulasena H, Biggs C, Zimmerman WB. 2014. Harvesting and dewatering yeast by microflotation. Biochem Eng J 82:174–182. Hu Q, Sommerfeld M, Jarvis E, Ghirardi M, Posewitz M, Seibert M, Darzins A. 2008. Microalgal triacylglycerols as feedstocks for biofuel production: Perspectives and advances. Plant J 54(4):621–639. James HA, Smith CJ, Sutton A, Caley J, Franklin O, Parr W. 1994. Survey of the concentrations of algal toxins in water supplies. Final report to the Department of the Environment, UK. Liers S, Baeyens J, Mochtar I. 1996. Modeling dissolved air flotation. Water Environ Res 68(6):1061–1075. Liu J, Huang J, Fan KW, Jiang Y, Zhong Y, Sun Z, Chen F. 2011. Production potential of Chlorella zofingienesis as a feedstock for biodiesel. Bioresour Technol 101(22):8658–8663. Ma J, Lei G, Fang J. 2007. Effect of algae species population structure on their removal by coagulation and filtration processes—A case study. J Water Supply ResTechnol-Aqua 56(1):41–54. Malley JP, Edzwald JK. 1991. Concepts for dissolved-air flotation treatment of drinking waters. Aqua (Oxford) 40(1):7–17. Okada K, Akagi Y, Kogure M, Yoshioka N. 1990. Analysis of particle trajectories of small particles in flotation when the particles and bubbles are both charged. Can J Chem Eng 68(4):614–621. Tambo N, Matsui Y, Fukushi K. 1986. A kinetic study of dissolved air flotation, Proceedings of World Congress III of Chemical Engineering, 3, 200–203, Tokyo, Japan. Teixeira MR, Rosa MJ. 2006. Comparing dissolved air flotation and conventional sedimentation to remove cyanobacterial cells of Microcystis aeruginosa Part 1: The key operating conditions. Sep Purif Technol 52(1): 84–94. Vandamme D, Foubert I, Meesschaert B, Muylaert K. 2010. Flocculation of microalgae using cationic starch. J Appl Phycol 22(4):525–530. Wyatt NB, Gloe LM, Brady PV, Hewson JC, Grillet AM, Hankins MG, Pohl PI. 2012. Critical conditions for ferric chloride-induced flocculation of freshwater algae. Biotechnol Bioeng 109(2):493–501. Zhang X, Amendola P, Hewson JC, Sommerfeld M, Hu Q. 2012. Influence of growth phase on harvesting of Chlorella zofingiensis by dissolved air flotation. Bioresour Technol 116 477–484.
Supporting Information Additional supporting information may be found in the online version of this article at the publisher’s web-site.
Zhang et al.: DAF Algal Harvesting Modeling Biotechnology and Bioengineering
2485
