Effectiveness and Kinetics of Ferrate As a Disinfectant for Ballast Water
By Jessen, Andrea Randall, Andrew; Reinhart, Debra; Daly, Luke
ABSTRACT: This study examined whether ferrate could meet the international standards for successful ballast water treatment, including final concentrations of less than 1 CFU/mL of Enterococci, less than 2.5 CFU/mL of Escherichia coli, and less than 1 CFU/100 mL of Vibrio cholerae. Pure cultures of E. coli, Klebsiella pneumoniae, and V. cholerae, and a mixed culture of Enterococcus faecium and E. faecilis were grown in saline solution to simulate ballast water and were treated with dosages of ferrate ranging from 0.25 to 5.0 mg/L. A ferrate dose of 5 mg/L resulted in complete disinfection of all organisms tested, and smaller dosages were also very effective. Tailing was consistently observed, and the Hom’s model (1972) appeared to most accurately represent the action of ferrate on these organisms. Salinity and pH did not adversely affect results, and regrowth was not a problem. Ferrate shows good potential as an effective disinfectant in the treatment of ballast water.
Water Environ. Res., 80, 561 (2008).
KEYWORDS: disinfection, ferrate, ballast, kinetics, Chick- Watson, Hom’s model, salinity.
The introduction of aquatic nuisance species (ANS) and bacterial pathogens from the discharge of ballast water by seafaring vessels is an ongoing problem that threatens ecosystems and human health. Ballast water provides stability to oceangoing vessels, and its uptake and discharge allows ships to compensate for cargo loads. American waters receive more than 79 million tonnes of foreign ballast water per year (Carlton et al., 1995), and the number of species carried in these waters is vast. All major marine trophic groups were found in ballast water from Japanese cargo ships in an Oregonian port, such that ballast water acts as a phyletically and ecologically nonselective transport vector (Carlton and Geller, 1993). Many recent biological invasions have been attributed to ballast water release, the most famous of which is Dreissena polymorpha, the zebra mussel, which was most likely introduced via the release of larvae contained in ship ballast water to the Great Lakes in the late 1980s (Hebert et al., 1989). Zebra mussels continue to expand their range, and their tendency to form dense mats has led to pipe fouling at power plants and other facilities (Kovalak et al., 1993).
Bacteria are also among the aquatic nuisance species introduced by ballast water discharge. One study found Vibrio cholerae 01 in ballast water from 5 of 19 cargo ships sampled in Alabama and Mississippi (McCarthy and Khambaty, 1994), causing the Food and Drug Administration (Rockville, Maryland) and U.S. Coast Guard to recommend that ship crews exchange their ballast water in midocean, before entry into United States ports. Viable counts of V. cholerae were approximately 106 cells/mL, in salinities ranging from 12 to 32 ng/kg (12 to 32 ppt), again emphasizing the potential environmental effects of ballast water. However, open-ocean exchange proved inadequate, and even freshwater flushing leaves many viable organisms in tanks (Hulsmann and Galil, 2001).
Because of the growing urgency of this problem, voluntary international guidelines for ballast water management have been established by the Maritime Environment Protection Committee of the International Maritime Organization (IMO) (London, United Kingdom) (Raaymakers, 2001), and performance specifications have been set for the successful testing and approval of ballast water management systems, including the following:
* The final concentration of Enterococci should be less than 1 CFU/mL,
* The final concentration of E. coli should be less than 2.5 CFU/ mL, and
* The final concentration of Vibrio cholerae should be less than 1 CFU/100 mL.
An exact standard of reduction for coliforms has not been set, but the IMO specifications state that they should be measured in any proposed treatment method.
Disinfection of potable water commonly has two goals-(1) immediate destruction or inactivation of pathogens, and (2) persistent inactivation of pathogens through residual action (Xie, 2004). An ideal disinfectant should be toxic to microorganisms, yet nontoxic to higher forms of life. Treatment of ballast water (and other applications to prevent the spread of invasive species) has slightly different goals than that of drinking water, because the water is typically treated and then released into the local environment. Because the local species are quite similar to the alien species, the disinfectant should not persist in the environment, so that the local biota is not harmed. Also, some higher organisms, such as crustaceans, are considered invasive, so an ideal disinfectant for ballast water would have a broader spectrum of action.
One agent that holds promise for ballast water disinfection is ferrate. Ferrate is iron with a valence of +6, which confers upon it a tendency to accept electrons, to attain its preferred valence of +2 or +3. Therefore, it acts as an oxidant, like many disinfectants, with a reduction potential of +2.20 V in acidic solution and +0.72 V in alkaline solution (Wood, 1957). This reduction potential makes it a more powerful oxidant in acid than hypochlorite, ozone, hydrogen peroxide, or permanganate (Lee et al., 2004). Because of its steric resemblance to phosphate anion, ferrate binds to enzyme phosphoryl groups and therefore site-specifically inactivates phosphatases, dehydrogenases, phosphorylase b, phosphoglucomutase, and other proteins with this functional group. It also specifically oxidizes amino acid residues on muscle phosphorylase, pancreatic ribonuclease, triose phosphate isomerase, and E. coli DNA polymerase- I (Basu et al., 1987). Ferrate can also be used as a coagulant (Jiang and Lloyd, 2002) and for other uses, but its cost of production has hindered its commercial usage.
Three general methods of synthesizing ferrate have been used-wet oxidation, dry oxidation, and electrolysis (Lee et al., 2004). Electrolysis involves the anodization of a pure iron electrode in concentrated alkaline solution. In dry oxidation, iron oxides are treated with oxidants at high temperature and pressure to generate ferrate salts. Dry oxidation has the advantage of a one-step process (Perfiliev and Sharma, 2004) and shows potential as a green technology to recycle iron oxide wastes from steel manufacturing processes. Wet oxidation shows good potential for ferrate production, but the requirement for relatively pure chemicals has led to high costs (Lee et al., 2004). Fortunately, technologies are now in development to lower the cost of wet-oxidation ferrate synthesis, to produce large quantities of ferrate on-site, and to address issues of stability, packaging, handling, and transport, with the immediate goal of on-board treatment of ballast water.
Ferrate decomposes to nontoxic forms of iron in aqueous solution, according to the following equation:
2FeO^sub 4^^sup -2^ + 3H^sub 2^O -[arrow right] 2FeO(OH) + 1.5O^sub 2^ + 4OH^sup -^ (1)
The rate of decomposition depends upon many factors, including pH, temperature, and initial ferrate concentration (Schink and Waite, 1980). According to a Pourbaix diagram, ferrate is most stable in an alkaline, highly aerobic environment. It will convert to aqueous ferric and ferrous iron in acidic, aerobic environments; to nonaqueous Fe(OH)^sub 3^ and Fe(OH)^sub 2^ precipitates in alkaline, aerobicto-somewhat-anoxic conditions; and to solid iron in highly anoxic (i.e., reducing or low oxidation-reduction potential) conditions across the entire pH range. More dilute solutions result in a decreased driving force towards precipitation (Western Oregon University, 2006). Seawater is typically slightly alkaline; thus, depending on the oxygen levels, ferrate would be expected to precipitate out as a hydroxide or as solid iron, as long as the pH does not become acidic. Precipitation of ferrate might be a useful method to remove it from treated ballast water, before discharge, to meet effluent iron standards. These standards vary widely by state, country, and water classification. Within Florida waters, the maximum iron standard ranges from 0.3 to 1.0 mg/L (Florida Department of Environmental Protection, 2006).
Many oxidants have the disadvantage of producing disinfection byproducts, including trihalomethanes (THMs), and their concentration in drinking water is regulated because of health concerns (Metcalf & Eddy Inc., 2003). Even if the breakdown products of the oxidant are not in themselves toxic, they can react with constituents in the water to produce environmentally harmful products. The THMs, for example, are formed by the reaction of natural organic matter with chlorine or other oxidants (Metcalf & Eddy Inc., 2003). The presence of natural organic matter also reduces the effectiveness of oxidants. In the case of chlorination, this produces a stepwise phenomena referred to as breakpoint chlorination chemistry, in which the chlorine first reacts with readily oxidizable organic and inorganic material, then with ammonia to form chloramines, and then with less readily oxidizable molecules, including, presumably, the target species, producing a chlorine residual after all oxidizable matter is consumed (Matheickal et al., 2004). Likewise, one would expect the action of ferrate to vary with respect to organic content, which creates an additional oxidant demand, but ferrate has the advantage that its breakdown products are some form of iron, which is nontoxic to humans, except in very large doses. The effectiveness of ferrate must be compared with that of free chlorine and chloramines and other disinfectants, using similar techniques and models. The contact time approach, used commonly in drinking water regulations (AWWA, 1999), examines combinations of concentrations and contact times. It assumes a linear relationship between contact time and degree of disinfection. This is rarely observed, however. Often, there are tailing and shoulder (or lag) effects, resulting from disinfectant, environmental, and bacterial characteristics, and numerous modelers have sought to explain these phenomena and predict disinfection rates from them. The Chick-Watson law relates the survival rate of the target organism to the concentration of disinfectant and a dieoff constant (Metcalf & Eddy Inc., 2003). It is expressed as follows:
dN^sub t^/dt = -k’C^sup n^N^sub t^ (2)
N^sub t^ = number of organisms at time t;
k’ = dieoff constant;
C = concentration of disinfectant (mg/L); and
n = coefficient of dilution, which determines the relative importance of disinfectant concentration.
The integrated form of this equation is as follows:
In(N^sub t^/N^sub 0^)) = -k’C^sup n^t (3)
N^sub 0^ = starting bacterial concentration (organisms/L).
When n is greater than 1, concentration is more important than time. Deviations from this law, such as tailing or shoulders caused by media interactions, disinfectant decay, and other factors, result in a nonlinear graph of time versus survival (N^sub t^/N^sub 0^). A log-log plot can still result in a linear graph for these deviations, provided C^sup n^t is a constant (AWWA, 1999). To better describe deviations from Chick-Watson behavior, eq 2 was modified by Horn to also consider the effect of time (Horn, 1972). The integrated form of Hom’s equation is as follows:
In(N^sub t^/N^sub 0^) = -k’C^sup n^t^sup m^ (4)
m = a coefficient of importance for time.
The rate of inactivation decreases with time when m is less than 1. Adding the m coefficient permits further characterization of the effect of time on disinfection efficiency, but might also lead to overparameterization.
A general differential rate law of disinfection states that the rate of inactivation, dN/dt, relates to its components as follows (Gyurek and Finch, 1998):
dN/dt =-kmN^sub x^C^sup n^t^sup m-1^ (5)
k = experimental reaction rate constant, and
x = empirical constant.
This equation reduces to the Chick-Watson equation when m = 1 and x = 1 and to the Hom’s equation when x = 1. Both the Chick-Watson and Horn’s equations assume that the concentration of disinfectant is constant for the test period or for a given contact time. This approach assumes that disinfectant demand is negligible or that the error resulting from disinfectant decay is acceptable for the objectives of the study. In this study, we have assumed that disinfectant decay was negligible. It should be noted that, even if it was not insignificant, the demand was approximately the same in the experiments being compared, because a synthetic saline solution that was identical in each experiment was used. This means that if there was any error in parameter estimates resulting from disinfectant demand, it would be the same in the experiments being compared (i.e., E. coli versus K. pneumoniae), and the conclusions based on relative comparisons would still be valid. In addition, if there was demand that was not accounted for, the log kills reported would be conservative (i.e., underestimates rather than overestimates). Finally, in the literature, most of the batch experiments with significant demand were wastewater or high- dissolved-organiccarbon (DOC) experiments (Haas and Karra, 1984b). In contrast, this study looked at very low DOC water, that is, distilled deionized water with Instant Ocean (Aquarium Systems, Mentor, Ohio) (inorganic salts) added to it. If disinfectant demand was significant and followed first-order decay, it would be necessary to use the incomplete gamma function for an exact solution of the Horn’s model, or an approximation of the incomplete gamma function (Haas and Joffe, 1994). However, many studies have found good correlation coefficients using the Chick-Watson or Horn’s models for disinfection of various bacteria by free chlorine and other disinfectants, with demand assumed to be negligible, implying that the assumption of negligible demand was valid over the contact times studied (Haas and Karra, 1984a). As a result, demand was assumed to be negligible in this study.
An equation recommended by the Water Environment Federation(R) (Alexandria, Virginia) Wastewater Disinfection Manual of Practice (WEF, 1996) has been used to describe inactivation of a dinoflagellate algae in ballast water (Oemcke and van Leeuwen, 2005) in the following form:
C = oxidant concentration after initial demand has been met (mg/ L), and
b and d = experimental regression coefficients.
A steeper slope (i.e., higher b) indicates a higher kill achieved with relatively low doses, and a higher ;c-intercept indicates a relatively high initial oxidant demand. This is similar to the Chick- Watson model, but does not consider time as a variable, so the coefficients must be determined for each experimental contact time. This equation is referred to as the oxidant demand model.
Preparation and Quantification of Bacteria. Heterotrophic and coliform bacteria were represented by E. coli and K. pneumoniae (American Type Culture Collection [ATCC] numbers 11303 and 13833, respectively). The Enterococcus group was represented by Enterococcus faecium and E. faecilis (ATCC numbers 19434 and 19433, respectively), which were grown as a mixed culture. Vibrio cholerae was grown from ATCC number 51352 stock. Instant Ocean (Aquarium Systems) was added to deionized water to produce salt water at 36 ng/ kg (36 ppt), per the manufacturer’s instructions. All bacteria were cultured under sterile conditions at room temperature (22[degrees]C), with shaking at 25 rpm, in a 5:1 mixture of Instant Ocean:nutrient broth (8 g/L), for 3 to 4 days. At the time of testing, the bacterial cell suspension was centrifuged for 10 minutes and resuspended in 36 ng/kg (36 ppt) Instant Ocean (unless otherwise indicated), under sterile conditions. Bacteria were washed twice more in this manner, resuspended in 36 ng/kg (36 ppt) Instant Ocean, and quantified. E. faecium, E. faecilis, Escherichia coli, and K. pneumoniae were quantified by the IDEXX Quanti-Tray (IDEXX Laboratories, Inc., Westbrook, Maine) enumeration procedure. V. cholerae was quantified by plate culture using nutrient agar (Becton Dickinson and Company, Franklin Lakes, New Jersey).
Experimental Procedure and Enumeration. The suspension of washed bacteria was subdivided into a number of flasks or 50-mL centrifuge tubes; ferrate was added at a range of dosages with stirring or shaking; and bacterial enumeration was performed again at a number of time intervals for each dosage. For bacteria quantified by the IDEXX method, an appropriate amount of liquid from each vial was transferred at the sampling time to a 100-mL coliform test vial, with the sodium thiosulfate pellet already in solution, to stop the oxidation reaction immediately, in a final volume of 100 mL. Colilert-18 substrate for quantification of K. pneumoniae and E. coli was added before transfer to Quanti-trays, which were heatsealed with an IDEXX Quanti-Tray Sealer and incubated at 35[degrees]C for 18 hours. K. pneumoniae was quantified by color change; E. coli was quantified by fluorescence; and Pseudomonas aeruginosa was run as a negative control. Combined cultures of E. faecilis and E. faecium were quantified by the IDEXX Enterolert Quanti-Tray enumeration procedure. Blue fluorescence was detected with a UV light, after incubation at 41[degrees]C for 24 hours. For V. cholerae, aliquots for all time points were added to disposable glass tubes containing sodium thiosulfate at a final concentration (after adding treated specimen) of 0.81 mM. Serial dilutions were performed in dilution buffer containing 2.00 mM magnesium chloride (MgCl^sub 2^-6H^sub 2^O) and 0.312 mM potassium phosphate (KH^sub 2^PO^sub 4^), and 100 to 1000 [mu]L was transferred to nutrient agar plates (Remel, Lenexa, Kansas) and spread using a glass spreader. The plates were incubated for 7 days at 35[degrees]C, at which time, colonies were counted and expressed as colony-forming units per milliliter.
Calculations and Statistics for Kinetics. For both QuantiTray enumeration procedures, the most probable number (MPN) per 100 mL was calculated from the MPN tables provided by the manufacturer and converted to MPN per milliliter using the appropriate dilution. When no wells were positive, a value of 1 was used for calculations. Because testing was performed on saltwater specimens, the procedure for marine water samples was followed, and specimens were diluted at least 10-fold with sterile distilled water (i.e., 10 mL of sample in the final volume of 100 mL was the lowest dilution). Untreated cultures of these organisms were run as positive controls, and tap water was run as a negative control. For V. cholerae, CFU per milliliter was used, and again a value of 1 was used when no colonies were observed, multiplied by the appropriate dilution. For kinetics, four experiments were run for each organism, and the Chick- Watson kinetic model coefficients were obtained for each experiment by the method of Metcalf & Eddy (2003), using Microsoft Excel spreadsheets (Microsoft Corporation, Bellevue, Washington) to graph log-log the negative natural log of N^sub t^/N^sub 0^ versus contact time. The time needed for a 4-log reduction for each dosage was obtained from each of these log-log graphs, and this 4-log reduction time versus ferrate dosage was then graphed log-log. The Chick- Watson n coefficient is then equal to the negative inverse of the slope and k = -ln(N^sub t^/N^sub o^)/exp^sup (n ln(y-int))^. The Horn’s model coefficients were calculated by the method of the American Water Works Association (Denver, Colorado) (1999) using multiple linear regression and SPSS statistical software (SPSS Inc., Chicago, Illinois). Three columns of values were generated for each time (t) and dosage (concentration, C) of ferrate: ln(t), ln(C), and ln(-ln(N^sub t^/N^sub o^)). These were then entered into an SPSS input file, and a linear regression was run, with ln(t) and ln(C) as independent variables and ln(-ln(N^sub t^/N^sub o^)) as the dependent variable. From this, the coefficient for ln(t) = m, the coefficient for ln(C) = n, and exp^sup (coeff. for consant)^ = k.
The oxidant demand coefficients were determined by combining the experimental data for each organism and calculating the coefficients in eq 6 by linear regression for contact times of 1, 5, 15, and 30 minutes, using Microsoft Excel (Microsoft Corporation). The x- intercept (initial oxidant demand) is solved from the equation for the linear regression, with appropriate conversions from the log values. Slopes were compared by hypothesis testing in regression by the method of Helsel and Hirsch (2002), using a standard Student’s t table.
Salinity, pH, Antidumping, N^sub 0^, and Regrowth Experimental Procedures and Statistics. The effect of salinity was investigated by dividing a washed E. coli suspension and bringing it up to equal volumes of either 36 or 10 ng/kg (36 or 10 ppt) Instant Ocean. Various dosages of ferrate ranging from 0.25 to 5 mg/L were added and sampled at various times. Untreated controls were sampled at the same times for the two salinities, to control for differences in mortality resulting from salinity without the addition of ferrate, and sodium thiosulfate was also added to untreated controls. The delta log removals were calculated by subtracting untreated controls from the treated controls for each salinity over time. Similarly, the effect of pH was tested by preparing a washed E. coli suspension in Instant Ocean, as described above, and measuring the initial pH (pH = 8.0 to 8.1). One portion of the suspension was then adjusted to a pH that was 1.5 units higher than the initial pH, with 1 N sodium hydroxide (pH = 9.5 to 9.6), and one portion was adjusted to a pH that was 1.5 units lower than the initial pH, with 1 N hydrochloric acid (pH = 6.5 to 6.6). Two ferrate dosages (0.75 and 1.5 mg/L) were tested at various times for the three pH values, and untreated controls were also sampled for each of the conditions (initial pH, acidic pH, and alkaline pH) to adjust for any disinfection effects resulting from the changes in the starting pH alone. Delta log removals resulting from the effect of ferrate were calculated for each of the pH conditions and dosages, by subtracting the change in MPN for the untreated control from the corresponding treated condition. Salinity and pH experiments were done in duplicate with separate ferrate batches. The statistical significance of pH and salinity effects was determined using a fixed- effect analysis of variance (ANOVA) of a multifactorial design in S- PLUS 2000 (Insightful Corporation, Seattle, Washington), using time, dosage, and pH or salinity as independent variables. Antidumping effects were investigated by washing E. coli three times with Instant Ocean, as described above, and then either including a low- speed centrifugation (170 x g for 5 minutes) and resuspension of the supernatant in an appropriate volume of Instant Ocean, or intense vortexing after addition of ferrate and just before each sampling event. These were done as two separate experiments and compared simultaneously, with normal preparation and treatment (i.e., washing the bacterial culture three times with Instant Ocean and resuspension in Instant Ocean and shaking the specimens before sampling). The statistical significance of antidumping measures was determined by again using a fixed-effect ANOVA of a multifactorial design in S-PLUS 2000, using time, dosage, and antidumping measure (low-speed centrifugation or vortexing versus normal preparation) as independent variables. The effect of N^sub 0^ was investigated by performing a serial dilution on a washed suspension of K. pneumoniae and dosing these dilutions for 5 and 15 minutes with 1 mg/L ferrate. Regrowth was measured by growing and resuspending E. coli in Instant Ocean, as above, dosing with 0, 5, or 10 mg/L ferrate for 1 minute, then quenching with sodium thiosulfate and quantifying at 15 minutes. Glucose was added to aliquots from each of these conditions, to a final concentration of 0.2 mg/L, and bacteria were requantified at 48 hours.
Results and Discussion
Log Reductions and Kinetics. The average log reductions achieved at 5 minutes for a 5-mg/L dosage for the test organisms are summarized in Table 1. The n in parentheses for all tables is the number of observations used to obtain the given value. The MPNper- milliliter values at various times and dosages for a typical experiment are shown in Figure 1. The corresponding log-log graph of contact time versus survival is shown in Figure 2, to obtain the contact times necessary to obtain a 4-log reduction for each dosage, following Metcalf & Eddy’s (2003) procedure for determining the Chick-Watson coefficients, as described in the Methods section. The resultant coefficients for the Chick-Watson and Horn’s models for the four organisms are summarized in Table 2, with the correlation coefficients. The fact that m does not equal 1 for any of the organisms indicates deviations from Chick’s Law. The shape of the dosage-response curves and the observation that all of the Horn’s m coefficients are less than 1 indicates that tailing is occurring. Furthermore, a Chick-Watson n coefficient of greater than 1 indicates that dosage is more important than time, which is what is observed for all organisms.
The coefficients and correlations obtained from the oxidant demand equation are summarized in Table 3. The fact that the initial oxidant demand (x-intercept) values are highest for 1-minute contact times indicates that the initial oxidant demand had not been met until the later contact times. Our slope or b values (2.1 to 6.5) are considerably higher than those of Oemcke and van Leeuwen (2005) (0.2 to 0.5), which theoretically indicates greater inactivation at lower dosages, but it is not valid to compare rates of inactivation of bacteria with a dinoflaggelate algae, and the uncommon use of this equation limits other comparisons. The b coefficient is fairly consistent for each organism at the various contact times. Comparing slopes for each organism at the different contact times, at the p = 0.05 level, the 1-minute contact time is different from at least one other time for each organism, perhaps suggesting that the initial oxidant demand had not been satisfied yet for this contact time. The only other slopes that differ significantly are the 5- and 30- minute contact times for E. coli, suggesting again the importance of dosage. This equation is not a rate equation, as is Chick-Watson and Horn’s, and in fact only relates survival as a function of dosage for a given contact time. Also, because it is necessary to calculate coefficients for each contact time, this approach seemed more cumbersome than the Horn’s model. Furthermore, the consistently higher correlation coefficients obtained with the Horn’s equations, summarized in Table 4, indicate that the 3-parameter model is necessary to adequately model the data. Therefore, the following equations best predict the effect of ferrate on the test organisms:
E. coli ln S =-5.4c^sup 0.5^t^sup 0.2^ (7)
K. pneunwniae ln S =-2.9^sup 0.8^t^sup 0.2^ (8)
Enterococci ln S =-3.0C^sup 0.7^t.^sup 0.3^ (9)
V. choleras ln S = -4.0C^sup 0.4^t^sup 0.2^ (10)
t = time (minutes),
C = initial ferrate concentration (mg/L), and
S = survival of the organism (N^sub t^/N^sub o^).
These equations result in the times required to obtain a 4-log reduction over a range of ferrate dosages, presented in Table 5. K. pneunwniae demonstrates the greatest dependence on dosage, as evidenced by the relatively large ratio of times required to achieve 4-log reductions at the low and high dosages in Table 5, summarized in Table 6.
As mentioned, the observation that all of the Horn’s m coefficients are less than 1 and all of the Chick-Watson n coefficients are greater than 1 indicates that dosage is more important than time. Haas and Karra (1984a) reviewed a number of disinfection studies conducted under demand-free conditions, determined by chemical analyses of the residual, and found that free chlorine generally produced Chick-Watson n values greater than 1 (i.e., tailing), whereas n values of less than 1 (i.e., shoulders) were observed with combined chlorine. However, most of the studies they examined were at lower (2- to 3-) log reductions than this study, which also was not done under demand-free conditions. They observed that the Horn’s model did not significantly increase predictive power for most of the experiments, but also note that a less satisfactory agreement with Chick-Watson is likely for higher log reductions, because tailing is a more common occurrence with greater inactivation values. In addition, the k coefficients are experimental rate constants-the higher the k value, the lower the survival. The Chick-Watson k’ values cited by Haas and Karra for experiments with E. coli range from 0.3 to 30.6 for free chlorine, for pH values ranging from 7.0 to 10.7 and from 2.5 x 10^sup -5^ to 1.0 L/mg-min for pH values from 7.0 to 10.5. The Chick-Watson n values range from 0.8 to 1.5 for free chlorine and 0 to 13.3 for combined chlorine. Thus, this study’s n value of 2.1 and the k value of 0.9 for E. coli is comparable with the values of either of the other disinfectants, given the wide ranges, but a direct comparison is difficult, given the differences in experimental design. pH and Salinity. The transferability of these equations to ballast water applications will depend on several factors, including pH and salinity. The pH of seawater is close to that of the bacterial suspensions in Instant Ocean (approximately 8, based on our field and laboratory data), but the pH of ballast water might differ significantly, as a result of the products of corrosion or metabolism of bacteria and other organisms present. The results for a 0.75-mg/L ferrate dosage, shown in Figure 3, indicate that a slightly acidic pH (6.5 to 6.6) enhanced disinfection. Results for 1.5 mg/L ferrate (not shown) were similar (p 0.05 for both dosages). However, at the 0.75-mg/L dosage, the p value of alkaline versus starting pH was 0.07, and log reductions appeared to be increasing with time relative to the neutral pH. The effect of the pH change itself was minor for the acidic conditions (a 0.09-log change over the experimental time of 30 minutes) but slightly higher (a 0.25-log reduction) for the alkaline condition. This effect has been subtracted from the effect of ferrate for each time point, as described in the Methods section, to determine the log removals shown in Figure 3. These results are consistent with the higher reduction potential of ferrate in acidic solution, but the effects of the increased stability of ferrate in alkaline solution are less clear and perhaps are being confused by the higher mortality from alkaline pH and the small sample size (two experiments). Salinity can also differ, as a result of natural and manmade influences, for example, in ports where incoming river water reduces salinity, but it did not have a clear effect on log reductions (p > 0.05), at a variety of ferrate dosages, as shown in Figure 4.
Initial Bacterial Concentration (N^sub 0^). Because the starting bacterial concentration represents an oxidant demand, it also has a potential effect on disinfectant dosage, but the literature on this subject is somewhat contradictory. Rincon and Pulgarin (2004), for example, found that, as N0 increased, the time for inactivation of E. coli with illuminated titanium dioxide (photocatalysis) increased, or, in other words, efficiency of disinfection decreased. Likewise, Greets and Fomichev (1985) found that the disinfection efficiency of ozone on E. coli and P. aeruginosa (and their phages) improved with a decrease in their initial concentration and that disinfection seemed to be a function of the concentration of ozone per bacterial cell. On the other hand, Haas and Kaymak (2003) found that, as N^sub 0^ decreased, the disinfection efficiency of ozone against Giardia muris decreased, which the authors hypothesized could be the result of a form of quorum sensing in these protozoa, similar to chemical signaling observed in bacterial populations. Figure 5 indicates that a reduction in the starting concentration of K. pneumoniae corresponded with an increase in log kill by a 1-mg/L ferrate dose. The initial bacterial concentration (N^sub 0^) was 5.3 x 10^sup 7^ organisms/mL, and there was a particularly significant increase in log kills with the first log dilution at 5 minutes. After this, the log kills seemed to stabilize, although the results at the final time of measurement (10 minutes of additional kill, for a final time of 15 minutes) increased in a linear fashion with dilution until the 1:1000 dilution. This is probably because the final dilution of 1:1000 N^sub 0^ resulted in no bacteria detected at the end time, and this was counted as 1 organism, according to the procedure outlined in the Methods section, and multiplied by 1000 for the dilution, resulting in the log reduction shown. Thus, the decrease in log reduction for the final dilution appeared to be attributable to the decrease in bacteria available to kill (i.e., the log kill began to decrease simply because there were fewer bacteria to kill). Similar results, with respect to the effect of initial bacterial concentration, were obtained for E. coli (data not shown). The data indicate that the log kill achieved with a given contact time and dosage decreased as initial bacterial concentration increased. The initial values used in this study were conservative, in that they were quite high (approximately 107 MPN/mL for E. coli and K. pneumoniae and 10^sup 5^ MPN/mL for Enterococci). Initial studies by this research group found, for E. coli, less than 1 MPN/ mL in two locations in waters of the Port of Cape Canaveral (Florida), approximately 7 orders of magnitude lower than our initial concentrations. Therefore, our estimated dose would be higher than the actual requirement to inactivate the organisms to the required standard.
However, in other countries, researchers are finding higher concentrations of organisms in ballast water, with a great deal of variability. For example, specific oligonucleotide probe analysis of six ballast waters in Singapore Harbour found concentrations of eubacteria, Enterobacteria, Vibrio spp, and E. coli ranging from 10^sup 4^ to 10^sup 6^ cells/mL for each of these organism classes (Joachimsthal et al., 2004). It is possible that holding water promotes breeding of bacteria, but the same team of researchers found slightly lower concentrations in ballast water relative to seawater and hypothesized that ballast water creates an environment that favors facultative anaerobic bacteria, which might also favor pathogens (Joachimsthal et al., 2003). Therefore, our starting bacterial concentrations might be comparable with, or perhaps slightly higher than, typical ballast water concentrations. Furthermore, the ferrate demand in ballast water resulting from oxidant-demanding compounds, including inorganics, such as corrosion products, and organics, needs to be assessed to translate results into the field, with field testing in a variety of locations.
Antidumping. The observation of tailing indicates that clumping of cells might have occurred in the experiments. To investigate this possibility, two antidumping procedures-a lowspeed centrifugation to remove clumps and intense vortexing-were compared with the standard procedure. The results are presented in Figures 6 and 7. Neither procedure eliminated the observation of tailing, but both procedures appeared to improve the log reductions (p
Regrowth from a Viable but Not Culturable State. Lastly, possible regrowth of E. coli was investigated. Figure 8 summarizes the average results of two experiments. The short contact time of 1 minute with a high ferrate dosage of 10 mg/L resulted in irreversible kill, with no regrowth, even with the addition of glucose. The same contact time with a ferrate dose of 5 mg/L did not produce a complete kill, but the number of bacteria continued to decline following the exposure to ferrate, even with the addition of glucose, indicating irreparable damage to the organisms. However, the population with glucose declined at a slower rate than did the population without glucose, indicating at least some metabolic activity. The data imply that, at the higher doses, the bacteria were probably killed rather than temporarily impaired and not metabolically active in the enumeration method, but, at 5 mg/L, there were still some potentially viable organisms. Neither of the disinfected samples showed a rebound in E. coli numbers, whether substrate was added or not.
Comparison with Alternative Ballast Water Treatment Processes. The attractiveness of ferrate as a disinfectant for ballast water requires a comparison with other disinfectants, and the problem of aquatic nuisance species has spurred recent studies of disinfection of seawater. Azanza et al. (2001) measured the decimal reduction times (D-values), which is the time to achieve a 1-log reduction of E. coli in seawater in response to treatment with UV radiation and with chlorine. At 35 ng/kg (35 ppt) salinity and 25[degrees]C, Azanza et al. (2001) obtained a D-value of 2.0 to 2.7 minutes for a 0.5-mg/L (0.5-ppm) dosage of chlorine and a D-value of 0.23 to 0.37 minutes for UV radiation at a dosage of 16 mWs^sup -1^cm^sup -2^. Substituting this dosage of ferrate and this level of survival into our Horn’s equation for E. coli (eq 7), we obtain a D-value of 0.1 minute, which is superior to their values. Oemcke et al. (2004) obtained 2- to 3-log reductions at somewhat higher UV doses. The hydroxyl radical, generated by ionization of seawater using a strong electric field, produced at least 2-log reductions of microorganisms (Bai et al., 2003), and heating of ballast water also shows early potential as a ballast water disinfectant (Rigby et al., 2002). Ferrate shows potential as an effective oxidant for ballast water and merits additional study and cost-effectiveness comparisons with the alternatives, after required dosages for all candidates are established. Conclusions and Recommendations
A dosage of 5 mg/L ferrate was sufficient to achieve the international standards of less than 1 CFU/100 mL for V. cholerae, less than 2.5 CFU/mL for E. coli, and less than 1 CFU/mL for Enterococci, with our experimental initial concentrations of at least 104 bacteria/mL and very short contact times, and 4-log removals could be achieved for all organisms with a ferrate dosage of 2 mg/L and a contact time of at least 20 minutes. Both the Chick- Watson and Horn’s models indicate that dosage is more important than time, and the Horn’s model was superior in predicting the action of ferrate. The tailing effect that was consistently observed in the dosage-response curves suggests that ferrate decomposes rapidly, a desirable trait for subsequent release into the receiving water body, and the natural alkalinity of seawater favors precipitation of the remaining iron for later removal with the ballast tank sludge, if needed to comply with iron-discharge standards. The disinfection properties of ferrate need to be validated under a variety of field conditions, including a range of initial bacterial concentrations and organic content, but they appear to be significant with various pH values and salinities. Ferrate compares favorably with other candidates for ballast water treatment and shows promise as an environmentally friendly, effective disinfectant.
Funding for this project was from the National Oceanic and Atmospheric Administration (Washington, D.C.) and Ferrate Technologies, LLC (Orlando, Florida), and an internal 1-4 grant from the University of Central Florida (Orlando, Florida).
Submitted for publication August 30, 2006; revised manuscript submitted manuscript submitted December 18, 2007; accepted for publication December 18, 2007.
The deadline to submit Discussions of this paper is September 15, 2008.
American Water Works Association (1999) Water Quality and Treatment: A Handbook of Community Water Supplies, 5th ed.; McGraw- Hill: New York.
Azanza, M. P. V.; Azanza, R. V.; Gedaria, A. I.; Sententa, H. G.; Idjao, M. V. (2001) Decimal Reduction Times of Pyrodinium bahamense var compressum and Escherichia coli in Chlorine- and Ultraviolet- Treated Seawater. Lett. Appl. Microbiol., 33 (5), 371-376.
Bai, M.; Zhang, Z.; Bai, X.; Zhou, X.; Deng, S.; Wang, N. (2003) The Killing of Microorganisms in Ship as Ballast Water Using Hydroxyl Radicals. Oceanologia et Limnolgia Sinica, 34 (5), 484- 489.
Basu, A.; Williams, K. R.; Modak, M. J. (1987) Ferrate Oxidation of Escherichia coli DNA Polyrnerase-I. J. Biol. Chem., 262 (20), 9601-9607.
Carlton, J. T.; Geller, J. B. (1993) Ecological Roulette: The Global Transport of Nonindigenous Marine Organisms. Science, 261 (5117), 78-82.
Carlton, J. T.; Reid, D.; van Leeuwen, H. (1995) The Role of Shipping in the Introduction of Nonindigenous Aquatic Organisms to the Coastal Waters of the United States (Other than the Great Lakes) and an Analysis of Control Options (Tech. Rep. No. CG-D11-95, U.S. Coast Guard, Washington D.C.). In Global Spread of Microorganisms by Ships, Ruiz, G. M., Rawlings, T. K., Dobbs, F. C., Drake, L. A., Mullady, T., Huq, A., Colwell, R. R. (2000).
Florida Department of Environmental Protection (2006) Surface Water Quality Standards, Chapter 62-302, http://www.dep.state.fl.us/ legal/Rules/shared/62-302.pdf (accessed March 2006).
Greets, N. V.; Fomichev, A. Y. (1985) The Dependence of Ozone Antibacterial and Antiphage Action on the Concentration of Cells and Phage Particles in the Reaction Medium. Mickrobiologiya, 54 (3), 410- 413.
Gyurek, L. L.; Finch, G. R. (1998) Modeling Water Treatment Chemical Disinfection Kinetics. J. Environ. Eng., 124 (9), 783-792.
Haas, C. N.; Joffe, J. (1994) Disinfection Under Dynamic Conditions: Modification of Horn’s Model for Decay. Environ. Sci. Technol., 28 (7), 1367-1369.
Haas, C. N.; Karra, S. B. (1984a) Kinetics of Microbial Inactivation by Chlorine: Review of Results in Demand-free Systems. Water Res., 18 (11), 1443-1449.
Haas, C. N.; Karra, S. B. (1984b) Kinetics of Microbial Inactivation by Chlorine-II: Kinetics in the Presence of Chlorine Demand. Water Res., 18 (11), 1451-1454.
Haas, C. N.; Kaymak, B. (2003) Effect of Initial Microbial Density on Inactivation of Giardia muris by Ozone. Water Res., 37 (12), 2980-2988.
Hebert, P. D. N.; Muncaster, B. W.; Mackie, G. L. (1989) Ecological and Genetic Studies on Dreissena polymorpha (Pallas): A New Mollusc in the Great Lakes. Can. J. Fish. Aquat. Sci., 46 (9), 1587-1591.
Helsel, D. R.; Hirsch, R. M. (2002) Statistical Methods in Water Resources, Chapter 3. In Techniques of Water-Resources Investigations of the United States Geological Survey, Book 4, Hydrologic Analysis and Interpretation, http://water.usgs.gov/pubs/ twri/twri4a3 (accessed January 2006).
Horn, L. W. (1972) Kinetics of Chlorine Disinfection in an Ecosystem. J. Sanit. Eng. Div. Am. Soc. Civ. Eng., 98 (SA1), 183- 194.
Hulsmann, N.; Galil, B. S. (2001) The Effects of Freshwater Flushing on Marine Heterotrophic Protists-Implications for Ballast Water Management. Mar. Pollut. Bull., 42 (11), 1082-1086.
Jiang, J.-Q.; Lloyd, B. (2002) Review: Progress in the Development and Use of Ferrate Salt as an Oxidant and Coagulant for Water and Wastewater Treatment. Water Res., 36 (6), 1397.
Joachimsthal, E. L.; Ivanov, V.; Tay, J.-H.; Tay, S. T.-L. (2004) Bacteriological Examination of Ballast Water in Singapore Harbour by Flow Cytometry with FISH. Mar. Pollut. Bull, 49 (4), 334-343.
Joachimsthal, E. L.; Ivanov, V.; Tay, J.-H.; Tay, S. T.-L. (2003) Flow Cytometry and Conventional Enumeration of Microorganisms in Ships’ Ballast Water and Marine Samples. Mar. Pollut. Bull., 46 (3), 308-313.
Kovalak, W. P.; Longton, G. D.; Smithee, D. (1993) Infestation of Power Plant Water Systems by the Zebra Mussel (Dreissena polymorpha pallas). In Zebra Mussels: Biology, Impacts, and Control, Nalepa, T. F.; Schloesser, D. W. (Eds.); Lewis Publishers: Boca Raton, Florida.
Lee, Y.; Cho, M.; Kim, J. Y.; Yoon, J. (2004) Chemistry of Ferrate (Fe(VI)) in Aqueous Solution and Its Applications as a Green Chemical. J. Ind. Eng. Chem., 10 (1), 161-171.
Matheickal, J.; Selvakumar, P.; Weitz, F.; Mahmud, H.; Daly, L.; Reinhart, D. R.; Sharma, V. K.; Ching Thin, L. L. (2004) Singapore’s R&D Efforts in Developing Innovative Ballast Water Treatment Technology with a Special Emphasis on Fe-Vi (Ferrate) as a Potential Secondary Disinfection Chemical. Presented at the Ballast Water Treatment Symposium, Singapore, May; IMO: London, United Kingdom.
McCarthy, S. A.; Khambaty, F. M. (1994) International Dissemination of Epidemic Vibrio cholerae by Cargo Ship Ballast and Other Nonpotable Waters. Appl. Environ. Microbiol., 60 (7), 2597- 2601.
Metcalf & Eddy Inc. (2003) Wastewater Engineering: Treatment and Reuse, 4th ed.; McGraw-Hill: New York.
Oemcke, D.; Parker, N.; Mountfort, D. (2004) Effect of UV Irradiation on Viability of Micro Scale and Resistant Forms of Marine Organisms: Implications for the Treatment of Ships’ Ballast Water. J. Mar. Environ. Eng., 7 (3), 153-171.
Oemcke, D.; van Leeuwen, J. (2005) Ozonation of the Marine Dinoflagellate Alga Amphidium sp-Implications for Ballast Water Disinfection. Water Res., 39 (20), 5119-5125.
Perfiliev, Y. D.; Sharma, V. K. (2004) Ferrate (VI) Synthesis: Dry and Wet Methods. Proceedings of the International Symposium Innovative Ferrate(VI) Technology in Water and Wastewater Treatment, Prague, Czech Republic, May 31; ICT Press, Institute of Chemical Technology: Prague, Czech Republic, 32-37.
Raaymakers, S. (2001) 1st International Ballast Water Treatment Standards Workshop Report, March 28-30, IMO: London, United Kingdom, http://globallast.imo.org (accessed March 2006)
Rigby, G. R.; Hallegraeff, G. M.; Sutton, C. (2002) Novel Ballast Water Heating Technique Offers Cost-Effective Treatment to Reduce the Risk of Global Transport of Harmful Marine Organisms. Mar. Ecol. Prog. Ser., 191, 289-293.
Rincon, A-G; Pulgarin, C. (2004) Bactericidal Action of Illuminated TiO^sub 2^ on Pure Escherichia coli and Natural Bacterial Consortia: PostIrradiation Events in the Dark and Assessment of the Effective Disinfection Time. Appl. Catal. B: Environ., 49 (2), 99-112.
Schink, T.; Waite, T. D. (1980) Inactivation of f2 Virus with Ferrate (VI). Water Res., 14 (12), 1705-1717.
Water Environment Federation (1996) Wastewater Disinfection, Manual of Practice FD-10; Water Environment Federation: Alexandria, Virginia.
Western Oregon University (2006) Pourbaix Diagrams. Western Oregon University: Monmouth, Oregon, http://www.wou.edu/las/physci/ ch412/pourbaix.htm (accessed March 2006).
Wood, R. H. (1957) The Heat, Free Energy and Entropy of the Ferrate (VI) Ion. J. Am. Chem. Soc., 80 (9), 2038-2041.
Xie, Y. F. (2004) Disinfection Byproducts in Drinking Water: Formation, Analysis, and Control. Lewis Publishers: New York.
Andrea Jessen1*, Andrew Randall1, Debra Reinhart1, Luke Daly2
1 University of Central Florida, Orlando, Florida.
2 Ferrate Technologies, LLC, Orlando, Florida.
* 44238 Forest View Rd., Deland, FL 32720; e-mail: andrea8888@ embarqmail.com.
Copyright Water Environment Federation Jun 2008
(c) 2008 Water Environment Research. Provided by ProQuest LLC. All rights Reserved.