Water Research 36 (2002) 825–834

The use of hydrous iron (III) oxides for the removal ofhydrogen sulphide in aqueous systems

Simon W. Poultona,*, Michael D. Kroma, Jaap Van Rijnb, Robert Raiswella

aSchool of Earth Sciences, University of Leeds, Leeds, LS2 9JT, UKbDepartment of Animal Science, Faculty of Agriculture, The Hebrew University of Jerusalem, P.O. Box 12,

Rehovot 70100, Israel

Received 23 January 2001; received in revised form 7 June 2001; accepted 13 June 2001

Abstract

The potential for iron (hydr)oxides to remove dissolved hydrogen sulphide from seawater has been examined under

flow-through conditions. Ferrihydrite (a hydrous iron (III) oxide) was stabilised by precipitation onto zeolite pellets,and rates of sulphide removal were determined under laboratory conditions at pH 8.5. Sulphide removal kinetics weredependent on the initial sulphide concentration, substrate mass and flow rate. The experimental data suggest that theseparameters can be optimised to result in the rapid and effective removal of hydrogen sulphide. The results from

laboratory experiments compared favourably with sulphide removal kinetics determined in a series of experimentsperformed online in a recirculating mariculture production system. However, the presence in solution of ligands such asphosphate may also significantly affect reaction rates; a 50% reduction in sulphide removal rate for substrate removed

from an online system was partly attributed to phosphate adsorption. The formation of a more crystalline, less reactiveiron (hydr)oxide in recharged substrate was the likely result of FeS oxidation, which may also have contributed to theobserved reduction in sulphide removal rates. Ferrihydrite-coated zeolite would appear to provide an efficient, low-cost

method for sulphide removal, which is particularly suited to relatively small-scale aqueous flow-through systems. Thereaction of iron (hydr)oxides with dissolved sulphide is also accompanied by a distinct colour change due to theformation of black FeS(s) which, under appropriate conditions, may be used as a rapid indicator of sulphidic conditions.

r 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Hydrogen sulphide; Ferrihydrite; Flow-through; Kinetics; Iron oxides; Zeolite

1. Introduction

Hydrogen sulphide is a highly toxic compound thatcan form in any aqueous system which contains bothorganic matter and sulphate, and is limited in its supply

of dissolved oxygen. Such conditions are found innatural systems (e.g. the Black Sea) and can arise as aresult of anthropogenic activities, including aquaculturalpractices [1] and the production of anaerobic sewage and

industrial wastewaters [2]. Hydrogen sulphide is formedas a by-product of anaerobic organic matter oxidation

during bacterial sulphate reduction:

ðCH2OÞ106ðNH3Þ16H3PO4 þ 53SO2�4

) 53CO2 þ 53HCO�3 þ 53HS�

þ 16NH3 þ 53H2O þ H3PO4 ð1Þ

The generation of hydrogen sulphide in such systems isusually undesirable due to the toxic and odorous nature

of the gas, in addition to the potential for corrosionfollowing biological oxidation to sulphuric acid [3].Hydrogen sulphide is highly toxic to aquatic organisms

at low concentrations [4], and has been the cause of massfish mortality in aquaculture systems [5].

*Corresponding author. Tel.: +44-113-233-5237; fax: +44-

113-233-5259.

E-mail address: s.poulton@earth.leeds.ac.uk

(S.W. Poulton).

0043-1354/02/$ -see front matter r 2002 Elsevier Science Ltd. All rights reserved.

PII: S 0 0 4 3 - 1 3 5 4 ( 0 1 ) 0 0 3 1 4 - 1

Hydrogen sulphide can be controlled in aqueousenvironments by oxidative removal. Several oxidants

have been proposed for this purpose, including oxygen[6], hydrogen peroxide [7], hypochlorites, chlorine andpotassium permanganate [8], and more recently ferrate

(VI) [9]. These methods generally require a separatemonitoring stage to identify undesirable levels ofsulphide. However, routine monitoring using standardchemical techniques tends to be time consuming and

provides only intermittent temporal coverage, whilst thelong-term efficiency of electrochemical detection meth-ods in continuous, flow-through mode remains to be

established [10]. Clearly, scope exists for the develop-ment of an effective, low-maintenance system, capable ofboth detecting and then removing the dissolved sulphide

produced.Iron (hydr)oxides have significant potential for such a

system. The reaction between dissolved sulphide and

iron (hydr)oxides is well-documented, and occurs in avariety of natural, anaerobic environments [11,12]. Themechanism consists of a complex sequence of reactionswhich ultimately result in the formation of metastable

iron monosulphides [13,14]. The reaction is accompa-nied by a marked (and thus potentially diagnostic)colour change, from the characteristic red, orange, or

yellow colours of iron (hydr)oxides to black FeS.Furthermore, iron monosulphides are unstable in thepresence of oxygen, and thus an iron oxide removal

system for dissolved sulphide has the potential forregeneration following sulphide uptake.

The use of iron minerals for the removal of hydrogensulphide is not a novel concept. However, most previous

studies relate to the treatment of dry gas, either atelevated (e.g. [15,16]) or ambient temperatures (e.g. [17]).In the present study, we focus on the potential for iron

(hydr)oxide to remove dissolved sulphide in aqueous,flow-through systems. The kinetics of dissolved sulphideremoval by ferrihydrite, the most reactive Fe (hydr)

oxide [18], are examined under various conditions via aseries of laboratory flow-through experiments. Theseresults are then evaluated in a series of experiments

performed online in a recirculating mariculture produc-tion system.

2. Background

The reaction of dissolved sulphide with iron (hydr)

oxides has been proposed to occur via a reductive-dissolution mechanism based on the following sequenceof reactions [19,20], whereby sulphide initially forms a

complex at the oxide surface, followed by electrontransfer between the sulphide and Fe3+:Surface complex formation:

> FeOH þ HS�3 > FeIIIS�þH2O ð2Þ

Electron transfer:

> FeIIIS�3 > FeIIS � ð3Þ

The S � � free radical is then released, followed by Fe2+

dissolution:

> FeIIS�þH2O3 > FeIIOHþ2 þS � � ð4Þ

> FeIIOHþ2 ) new surface site þ Fe2þ ð5Þ

In the case of sulphide oxidation by ferrihydrite(Fe5HO8 � 4H2O), each S � � free radical rapidly reduces

one Fe3+ to form elemental sulphur, and the dissolvedFe2+ produced is able to react with additional dissolvedsulphide to form black FeS(s):

8S��þ8 > FeIIIOH ) S08þ8Fe2þ ð6Þ

Fe2þðaqÞþHS�3FeSðsÞþHþ ð7Þ

At constant temperature the major parameters affectingthe reaction kinetics for a particular iron (hydr)oxide are

mineral surface area, initial sulphide concentration, andpH [14,19–21]. Reaction rates may also be significantlyaffected by organic and inorganic solutes [22]. Phos-phate in particular may be extensively sorbed to the

oxide surface (e.g. [23]), thus decreasing the availabilityof reactive surface sites and hence reaction rates.Although ferrihydrite is generally considered to be the

most reactive iron (hydr)oxide [18], the kinetics of thereaction with dissolved sulphide have not previouslybeen established.

3. Methods

3.1. Experimental

Ferrihydrite was prepared by forced hydrolysis atelevated pH according to the method of Schwertmann

and Cornell [24]; a 50 mM solution of Fe(NO3)3 � 9H2Owas titrated to a pH of 6.5 with 1 M KOH. Theferrihydrite was stabilised for use in a flow-through

system by precipitation onto zeolite (crystalline sodiumalumino-silicate, 1/8 in pellets, Merck #540046S). Priorto coating, the zeolite was washed and dried, and then

added to the ferric nitrate solution before the forcedhydrolysis of ferrihydrite. After hydrolysis, the ferrihy-drite-coated zeolite was washed thoroughly to removetraces of nitrate, and dried at 601C. An Fe content of

0.41 wt% was determined after digestion (see below) ofduplicate samples of ferrihydrite-coated zeolite. Mea-sured surface areas (by BET) of uncoated and coated

zeolite were similar (within analytical error) anddetermined to be 56.3 and 58.8 m2/g, respectively.

Laboratory and online experiments were performed

at 251C in a water-tight plexiglas cylindrical cartridge(1 L volume; 10 cm diameter). Water was pumped into

S.W. Poulton et al. / Water Research 36 (2002) 825–834826

the cartridge at the base (via a narrow plexiglas tubeattached centrally to the top cover of the cartridge),

passed through the ferrihydrite-coated zeolite anddischarged through a port on the top cover of thecartridge. Air-tight 3-way sample valves were attached

to the input and output ports of the cartridge to allowsample collection via a syringe. The laboratory experi-ments consisted of an air-tight 5 L glass vessel with portsfor degassing and stock sulphide input, and an outlet

tube to allow the transport of water from the vessel tothe cartridge. Four litres of 20 ppt artificial seawaterprepared from seasalt (Red Sea pHarm Ltd., Israel)

were taken directly from a recirculating mariculturesystem (see below), and bacteria were removed byautoclaving and filtering (0.02 mm). The seawater was

then degassed in the acid-washed glass vessel for 1 h withultra high purity N2. Stock sulphide solutions wereprepared daily by dissolving rinsed Na2S � 9H2O crystals

in N2-purged deionised water. An aliquot of stocksulphide solution was added to the vessel via a syringeand the appropriate pH was achieved by the addition ofdegassed 0.1 M HCl. The sulphide spiked seawater was

then pumped from the glass vessel through the cartridgevia a peristaltic pump.

Dissolved sulphide and Fe2+ were measured in

triplicate at the input sample port of the cartridge atthe start of each experiment, and were measured as atime series at the output. The dissolved sulphide

concentration was also measured at the cartridge inputport towards the end of each experiment, in order toexamine whether the experiments were carried out underanaerobic conditions. No detectable decrease in dis-

solved sulphide was found in any of the experiments.Two series of experiments were performed online in a

pilot mariculture recirculating system (see [25]) in order

to evaluate the performance of the reaction cartridgeunder operational conditions. The pilot plant was basedat the Faculty of Agriculture, Hebrew University of

Jerusalem, Rehovot, Israel. In the first online experi-ment, a reaction cartridge containing 500 g of ferrihy-drite-coated zeolite was attached to an anaerobic loop in

the system, in a position where dissolved sulphide isgenerated only occasionally. Water was pumpedthrough the reaction cartridge at a rate of 100 ml/minvia a peristaltic pump. Subsamples (160 g) of zeolite

were removed from the cartridge at 68, 145, and 238 hfor subsequent use in the laboratory experiments. Theoutflowing water was also periodically monitored for

dissolved sulphide and Fe2+ in order to evaluatethe potential importance of dissolved Fe2+ release, inthe absence of sulphide, as a result of bacterially-

mediated Fe reduction.In the second series of online experiments, a

cylindrical cartridge containing 500 g of ferrihydrite-

coated zeolite was placed in an anaerobic loop of thesystem, in a position where dissolved sulphide is

continually produced. This ferrihydrite-coated zeolitehad already been used in laboratory experiments, and

thus corresponded to substrate that had already beenrecharged once (see below). These experiments wereperformed at a flow rate of 100 ml/min and were

monitored regularly for dissolved sulphide and Fe2+

over a period of up to 27 h.Phosphate adsorption experiments were carried out

using a 150 ml volume flow-through reaction cylinder

containing 20 g of ferrihydrite-coated zeolite. Seawaterwas prepared to the same ionic strength (20 ppt) andphosphate concentration (300 mmol/L) as the system

seawater used in the sulphide removal experiments, andthe pH adjusted to 8.5 with 1 M NaOH. The initialphosphate concentration was determined in triplicate

and the seawater was then pumped through the reactioncartridge at a known flow-rate. Dissolved phosphate wasmeasured as a time series at the output, up to the point

of ferrihydrite saturation (i.e. when input phospha-te=output phosphate).

3.2. Analytical

Dissolved sulphide was measured by the methyleneblue method [26]. Both filtered (0.2 mm) and unfiltered

samples were initially analysed by this method todetermine whether any solid phase FeS was dischargedwith the outflowing water. Solid phase FeS is soluble in

dilute acid and is thus dissolved by the 6 N HCl presentin the methylene blue reagents [14,21]. However, notrace of solid phase FeS was found in any samples taken

at the cartridge output, and subsequently only thesulphide present in unfiltered samples was determined.Replicate determinations of a stock sulphide solutiongave a mean of 1088712mmol/L (2s, n ¼ 8). Dissolved

Fe2+ and Fe3+ were determined by the revised ferrozinemethod of Viollier et al. [27]. Dissolved Fe3+ was foundto be negligible in all experiments. Replicate determina-

tions gave a mean of 7.170.3mmol/L (2 s, n ¼ 6)dissolved Fe2+.

Solid phase Fe was determined by flame AAS,

following a buffered (pH=4.8) sodium dithioniteextraction on crushed samples [12,28]. This method(reproducible to o4%) quantitatively extracts ferrihy-

drite, in addition to lepidocrocite, goethite and hematite[28]. The same technique was used to determine theconcentration of phosphate associated with the ironoxide. Dissolved phosphate was measured by spectro-

photometer using the phosphomolybdate-blue method[29]. The dithionite solutions were first diluted to acitrate concentration of less than 2.4 mM to prevent

interference between the dithionite and citrate in theextraction solutions and the molybdate reagent [30].Surface area was measured by the multi-point BET

method using a Beckman/Coulter SA 1300 analyser witha Beckman/Coulter SA-PREP outgasser unit. Electron

S.W. Poulton et al. / Water Research 36 (2002) 825–834 827

microscope analyses were performed on a CamscanSeries 4 SEM with an Oxford Microanalytical Link

10/25 s detector.

4. Results

4.1. Kinetic experiments

In Fig. 1, the concentrations of dissolved sulphide andFe2+ are plotted versus time for four representative

laboratory runs. The experiments were performed underthe same conditions of ferric oxide mass (150 gzeolite�615 mg Fe3+), pH (8.5) and flow rate (100 ml/

min), but with varying initial total sulphide (Stot)concentrations. The results highlight the initial rapidremoval of dissolved sulphide, with a gradual increase in

outflow concentrations as the availability of reactiveferric oxide surface sites decreases. Fig. 1 demonstratesthe influence of Stot on the removal of dissolvedsulphide; initial sulphide removal rates increased with

higher initial sulphide concentrations. Dissolved Fe2+

concentrations initially increase with time (input con-centrations were zero), followed by a gradual decrease as

the rate of sulphide removal decreases. In 16 out of the

21 reaction runs, maximum Fe2+ concentrations wereobserved 2–4 min after the initial maximum removal rate

for dissolved sulphide. This observation is consistentwith the proposed reaction mechanism, whereby therelease of dissolved Fe2+ is the rate determining step,

occurring only after the initial removal of sulphide at theferric oxide surface. The transformation of orangeferrihydrite to black FeS was observed in all experi-ments, although the time-scale for a visible colour

change depended on the specific reaction conditions.Specific rate parameters for dissolved sulphide re-

moval by ferrihydrite were evaluated by the initial rate

method. At constant pH, the overall rate equation canbe expressed by

Ri ¼ �d½S2�=dt ¼ k½Stota½Fe3þðsÞ

b; ð8Þ

where Ri is the initial rate, [Stot] is the initial concentrationof dissolved sulphide, [Fe3+

(s) ] is the mass of Fe3+ presentas ferrihydrite, a and b are the reaction orders withrespect to Stot and Fe3+

(s) respectively, and k is the overall

rate constant. In logarithmic form, Eq. (6) becomes

log Ri ¼ log kþ alog½Stot þ b log½Fe3þðsÞ ; ð9Þ

and thus a logarithmic plot of Ri versus [Stot] has a slope

of a; the order with respect to reduced sulphur. This

Fig. 1. Dissolved sulphide and Fe2+ concentrations as a function of time for four representative reaction runs. Filled squares indicate

initial sulphide concentration.

S.W. Poulton et al. / Water Research 36 (2002) 825–834828

method was used to derive the reaction order withrespect to each individual component (Fig. 2). The

sulphide removal rate (Ri) for each experiment wasdetermined by regression analysis of the initial linearphase of sulphide removal (to give an initial output

concentration in M/L), which was subsequently com-bined with the appropriate flow rate to give a removalrate in terms of M/min. It should be noted that Ri

approximates the total concentration of sulphide re-

moved from the water released at the cartridge output

during the first minute of discharge. It does not reflectthe concentration of sulphide removed during a 1 min

exposure time to the ferrihydrite-coated zeolite, since thedischarged water may have been in contact with thesubstrate for less than or greater than 1 min, depending

on the flow rate and substrate mass used. Removal rateswill be affected by the flow characteristics within thecartridge and will vary as a function of the overallporosity and cartridge design. Nevertheless, the results

presented here can be used to provide an indication ofthe relative effects of different parameters on sulphideremoval kinetics and on the general conditions required

for effective treatment of sulphidic waters in differentsystems.

4.2. Reaction order with respect to dissolved sulphide,ferrihydrite mass and flow rate

Fig. 2 highlights the relationships between initialsulphide removal rate and initial sulphide concen-

tration, ferrihydrite mass (as Fe3+), and flow rate.Linear regression analysis of the initial sulphide datagave a slope of 1.1170.12, which is consistent withprevious studies [14,21] indicating a first order reac-

tion with respect to the total sulphide concentration(reaction orders determined are rounded to thenearest 0.5).

The effect of ferrihydrite mass on sulphide removalrates was determined at a pH of 8.5, with a flow rate of100 ml/min and an initial sulphide concentration of

approximately 430 mmol/L. A slope of 0.4770.004 wasobtained, indicating a square root order with respect toferrihydrite mass (Fig. 2). This observation contrastswith the first order dependency on surface area

determined under batch experiment conditions[14,20,21]. However, this is to be expected given thatthe batch experiments are well-mixed, allowing the

entire mineral surface to react with the sulphide in thesolution. In contrast, the flow-through system consistsof a column of dense substrate which is not mixed by the

inflowing water. Since the sulphide removal rate dependsdirectly on the sulphide concentration in solution,removal rates decrease as the sulphide depleted water

progresses through the additional substrate in thecolumn.

The effect of flow rate on the initial removal rate ofdissolved sulphide was determined under conditions of

constant ferrihydrite mass (615 mg Fe3+), initial sul-phide concentration (480mmol/L) and pH (8.5). Theconcentration of sulphide in the outflowing water was

higher at faster flow rates, due to the shorter exposuretimes to ferrihydrite. However, an overall increase insulphide removal rate occurs at increased flow rates,

with a regression slope of 0.6370.02 indicating a squareroot order with respect to flow rate (Fig. 2).

Fig. 2. Logarithmic plots of initial reaction rate (Ri) versus

initial sulphide concentration (Stot), substrate mass (as Fe3+),

and flow rate.

S.W. Poulton et al. / Water Research 36 (2002) 825–834 829

4.3. Effects of recharging and online ageing on reactionrates

Table 1 details the initial sulphide removal rates forrecharged and online-aged ferrihydrite-coated zeolite,under constant conditions of ferrihydrite mass (615 mgFe3+), initial sulphide concentration (450mmol/L), pH

(8.5) and flow rate (100 mL/min). The ferrihydrite-coated zeolite was recharged at the end of eachexperiment by washing with deionised water, followed

by drying at 601C. Recharging results in a progressivedecrease in the initial sulphide removal rate, withdecreases of 21%, 29% and 43% for the first 3 recharges

respectively. Ferrihydrite-coated zeolite taken from theonline cartridge also showed a marked decrease inreactivity towards dissolved sulphide with an average50% decrease in Ri: However, we found no evidence for

a progressive decrease in reactivity with time online(Table 1). The slight decrease in the Fe3+ content of therecharged and online-aged ferrihydrite (0.36–0.38 wt%)

relative to the fresh ferrihydrite (0.41 wt%) is withinanalytical error, and insufficient to explain both themagnitude and progressive nature (for recharged zeolite)

of the decrease in reactivity.

4.4. Online experiments

Dissolved sulphide concentrations at the input andoutput of an online ferrihydrite cartridge (containing

500 g zeolite�2.05 g Fe3+) are demonstrated in Fig. 3(flow rate=100 mL/min). On average the cartridgeremoved 87713% of the inflowing dissolved sulphide.The ferrihydrite-coated zeolite close to the water inlet

(the central point at the base of the cylindrical cartridge)reacted more rapidly than the visible ferrihydrite at theedges of the cartridge. Nevertheless, a colour change

from orange to black was observed on the outside layersof substrate after approximately 20 min, with an almostcomplete colour change of the entire ferrihydrite within

about 5 h. Despite this almost complete colour change,the cartridge remained effective at removing sulphide

over the remainder of the experiment (Fig. 3), with noloss in the relative efficiency of sulphide removal.

Dissolved Fe2+ concentrations were found to be

below detection limits (o1 mmol/L) in the downstreamonline cartridge designed to determine the potential forbacterially mediated iron reduction. This does not in

itself provide direct evidence for an absence of bacterialiron reduction within the cartridge. However, the lack ofany significant losses in the dithionite-soluble Fe

contents of coated zeolite from this online system (Table1), suggests that bacterially mediated iron reduction isunlikely to greatly affect reaction rates in onlinecartridges over a period of at least days to weeks.

4.5. Phosphate adsorption experiment

The adsorption of phosphate to ferrihydrite-coatedzeolite (20 g) at pH ¼ 8:5 is plotted as a function of thetotal water flux through the substrate in Fig. 4. An

initial rapid drop in dissolved phosphate occurs,followed by a gradual increase in outflow concentrationsup to the point of ferrihydrite saturation. An exponen-

tial function was fitted to the data giving an integratedtotal phosphate removal of 32.6mmol. Thus, at pH ¼8:5 the ferrihydrite was able to adsorb only 3.8% of thetotal phosphate flux before equilibrium was established.

5. Discussion

5.1. Kinetics of sulphide removal

Determination of the reaction orders with respect toStot; ferrihydrite mass and flow rate allows a rateexpression for the removal of dissolved sulphide at

pH ¼ 8:5 to be defined of the form:

RiðM=minÞ ¼ k½Stot½Fe3þðsÞ

0:5½flowrate0:5 ð10Þ

Table 1

Composition and initial reaction rates (Ri) for recharged and

aged ferrihydrite-coated zeolite

Fe3+ (wt%) Ri (mmol/min)

Initial run 0.41 32.37

Recharged#1 0.38 25.43

Recharged#2 0.38 22.86

Recharged#3 0.38 18.57

Online for 68 h 0.36 15.20

Online for 145 h 0.36 16.28

Online for 238 h 0.38 16.70

Fig. 3. Input and output dissolved sulphide concentrations for

an online cartridge (Fe3+=2.05 g, flow rate=100 mL/min).

S.W. Poulton et al. / Water Research 36 (2002) 825–834830

The rate constant, k; calculated as the average of 12reaction runs for system seawater at 251C is0.30270.042 L0.5 g�0.5 min�0.5 when Stot is expressed inM/L, Fe3+

(s) is expressed in g, and flow rate is expressed

in L/min. The experimental data were determined overthe ranges 200–800 mM Stot; 160–925 mg Fe3+ (equiva-lent to 40–225 g substrate), and 40–350 mL/min flow

rate. The applicability of this rate equation to themodelling of sulphide removal rates under non-experi-mental conditions can be evaluated relative to the online

data. Eq. (10) was used to calculate an observed rateconstant (kobs) for each online measurement (samplingat the outflow of the column was delayed according tothe flow rate, such that output measurements corre-

sponded to those determined at the input). The value forkobs was determined as 0.21370.019 L0.5 g�0.5 min�0.5,suggesting an approximate 30% reduction in reaction

rates for the recharged substrate used in the onlinesystem, relative to the fresh substrate used in thelaboratory experiments. This 30% reduction in reaction

rate for the online system corresponds well with thelaboratory study of the effects of recharging thesubstrate (Table 1), and suggests that the rate equation

can be used to accurately predict sulphide removalkinetics in online systems.

Eq. (10) can be used to model the effect of flow rate onsulphide removal rates (Fig. 5; based on a substrate mass

of 100 g�0.41 g Fe3+). Clearly, faster flow rates result inincreased sulphide removal rates; however, the relativeeffect of flow rate gradually lessens at higher flow rates.

Thus, an optimum flow rate may be defined relative tothe requirements of a particular system. However, Fig. 5specifically relates to the kinetics of sulphide removal

rather than removal efficiency. The effect of flow rate iscomplicated by the fact that although faster flow ratesresult in faster sulphide removal rates, the actualconcentrations of sulphide remaining in the outflowing

water increase due to shorter exposure times to the iron

(hydr)oxide. In relation to the efficiency of sulphideremoval, it is perhaps more appropriate to consider thedata in terms of the mass of substrate required to reduce

the outflow sulphide concentrations to a required level.Fig. 6 demonstrates the substrate mass required toreduce outflow sulphide concentrations by 100%, 90%

and 50% as a function of flow rate (based on Eq. (8)).The required substrate masses are independent ofinflowing sulphide concentrations, and thus for a

particular system the appropriate substrate mass maybe determined based on the flow rate and the requiredefficiency of sulphide removal. For 100% sulphide

removal, the relationship between the substrate massand flow rate is given by

substrate mass ðkgÞ ¼ 2:67 ðflow rateÞ ðL=minÞ: ð11Þ

At faster rates of flow, relatively large substrate masses

are required to remove all of the inflowing sulphide(Fig. 6). Furthermore, the relationships demonstrated in

Fig. 4. Phosphate concentration versus total water flux at pH

8.5 (Fe3+=82 mg). Open square represents input phosphate

concentration.

Fig. 5. Predicted sulphide removal rate as a function of water

flow for three initial sulphide concentrations (for 100 g

substrate�0.41 g Fe3+).

Fig. 6. Calculated substrate mass required to remove 100%,

90% and 50% of inflowing sulphide as a function of flow rate.

S.W. Poulton et al. / Water Research 36 (2002) 825–834 831

Fig. 6 relate to initial rates of sulphide removal, whichgradually decrease as reactive surface sites are consumed

(see Fig. 1), and hence an excess of substrate would berequired to continue to effectively treat additionalinflowing water. However, required substrate masses

are reduced considerably as the removal efficiencydecreases (Fig. 6), and in many systems 100% sulphideremoval may be unnecessary. At slower flow rates,sulphide is effectively removed by relatively small

substrate masses, and thus a balance may be achievedbetween the need for efficient sulphide removal (slowerflow rates) and the potential requirement for the

cartridge to act as a warning system via a rapid colourchange as black FeS is formed (faster flow rates andhence greater rates of sulphide removal).

5.2. Online ageing and recharging

The significant decrease in sulphide removal rates forboth aged and recharged substrate (Table 1) may

potentially be attributed to three factors: dissolution ofFe3+ from the substrate, phosphate adsorption, andmineralogical transformations. As previously discussed,the release of dissolved iron from the substrate does not

appear to be a significant process, and can thus bediscounted as the major cause of the reduction inreaction rates.

The adsorption of phosphate to iron (hydr)oxides cansignificantly inhibit reaction rates [22,21], particularly atthe relatively high concentrations of dissolved phosphate

(approximately 300mmol/L PO4) present in the seawatersystem. However, the potential for iron (hydr)oxides toadsorb dissolved phosphate varies as a function of pH,with relatively reduced uptake from seawater as the pH

increases [23]. The extent of phosphate adsorption inonline-aged ferrihydrite could not be determined byleaching techniques, as the zeolite itself contained

significant traces of leachable phosphate (SEM analysisrevealed the presence of apatite (Ca5(PO4)3(OH,F,Cl))within the zeolite structure) which masked any addi-

tional phosphate adsorbed to the ferrihydrite surface.However, the total concentration of phosphate able tobe adsorbed by 20 g of ferrihydrite-coated zeolite (82 mg

Fe3+) at pH ¼ 8:5 was determined experimentally as32.6mmol. Dzombak and Morel [31] estimate thereactive site density of ferrihydrite to vary between0.1–0.9 moles per mole of Fe. Phosphate forms binuclear

surface complexes (i.e. two metal centres from thesurface lattice are bridged) with ferrihydrite [22,32].Thus, at equilibrium 5–44% of the ferrihydrite surface

sites can be estimated to be occupied by phosphate atpH ¼ 8:5: Clearly phosphate may occupy a significantproportion of the reactive surface sites in the online-

aged ferrihydrite, and may thus contribute substantiallyto the decrease in reaction rates (Table 1).

Electron microscope images of unused, recharged andonline-aged substrate are shown in Fig. 7. The ferrihy-drite forms a continuous coating, penetrating to a depth

of 3–4 zeolite grains. However, in recharged and online-aged substrate a more crystalline iron oxide mineral isalso apparent, which is likely to be the product of FeSreoxidation. The new mineral phase clearly coats the

existing ferrihydrite, thus preventing further reaction of

Fig. 7. Backscatter SEM images of unused (a), online-aged (b),

and recharged (c) substrate.

S.W. Poulton et al. / Water Research 36 (2002) 825–834832

the ferrihydrite with dissolved sulphide. Electron back-scatter diffraction techniques were unable to resolve

the actual mineralogy of this new phase. However,the product of ferrous iron oxidation is usually eithercrystalline lepidocrocite (g-FeOOH) or magnetite

(Fe3O4), depending on the conditions of formation[24]. The reactivity of these minerals towards dissolvedsulphide is considerably reduced relative to ferrihydrite[18], and thus FeS reoxidation may contribute signifi-

cantly to the observed decrease in rates of sulphideremoval.

The decrease in reactivity of online-aged and re-

charged substrate may be due to a combination ofphosphate adsorption and mineralogical transforma-tions during the reoxidation of FeS. Phosphate satura-

tion of ferrihydrite occurs relatively rapidly (Fig. 4),which accounts for the absence of a progressive decreasein reactivity with time online (Table 1). Thus, ferrihy-

drite-coated zeolite may be expected to effectively treat apulse of sulphide-rich water even after a considerableperiod of time online. In contrast, the mineralogicaltransformation to a more crystalline iron (hydr)oxide

with recharging, suggests that the substrate wouldrequire replacing more regularly in systems wheresulphide is produced frequently or continually.

6. Conclusions

Hydrous iron (III) oxides may be effectively used forthe treatment of hydrogen sulphide in aqueous, flow-

through systems. Ferrihydrite is the preferred ironmineral for sulphide removal, due to its reactivityrelative to other iron (hydr)oxides [18], ease of prepara-tion, and apparent stability when precipitated onto a

substrate such as zeolite. The technique is particularlysuited to relatively small-scale systems, where sulphide isproduced occasionally or intermittently and the con-

tinuous addition of chemical oxidants is unwarranted.The presence in solution of ligands such as phosphatemay significantly inhibit reaction rates (dependant on

pH) and thus require consideration relative to the designand operational parameters of an iron oxide removalsystem for dissolved sulphide. Following sulphidation,

the iron oxide cartridge may be regenerated by exposureto air or oxygenated water. However, reoxidation resultsin a decrease in sulphide removal rates due to theformation of more crystalline, less reactive iron (hydr)o-

xides, and thus it may be preferable to recoat thesubstrate with additional ferrihydrite following sulphi-dation.

The reaction of hydrogen sulphide with iron (hydr)o-xides has the advantage of a distinct colour change dueto the formation of black FeS(s), which may be used as a

warning signal for sulphidic conditions. In larger scalesystems, ferrihydrite may be used as a warning device for

sulphide and as a means of treating an initial flow ofwater, prior to the larger scale removal of sulphide by

the addition of chemical oxidants. The effectiveness ofhydrous iron (III) oxides as a sulphide warning systemincreases at faster flow rates due to increased rates of

sulphide removal, although the general requirement isfor the formation of FeS over a relatively small volumeof substrate, thus allowing the rapid identification of acolour change. Work is in progress to develop an

automated warning system for dissolved sulphide basedon this colour change.

Acknowledgements

We are grateful to Eric Condliffe and Geoff Lloydfor SEM assistance. This study was carried out withfinancial support from the Commission of the European

Communities, Agriculture and Fisheries (FAIR) specificRTD programme, CT98-4160 ‘Development of recircu-lating mariculture production systems designed tominimize environmental impact’.

References

[1] Blackburn TH, Lund BA, Krom MD. C- and

N-mineralization in seawater fish pond sediments. Mar

Ecol Prog Ser 1988;44(3):221–7.

[2] Thistlethwayte DK. The control of sulphides in sewerage

systems. Sydney: Butterworth, 1972.

[3] Kotronarou A, Hoffmann MR. Catalytic autoxidation of

hydrogen sulfide in wastewater. Environ Sci Technol

1991;25(6):1153–60.

[4] Colt J, Tchobanoglous G. Chronic exposure of channel

catfish, Ictalurus punctatus, to ammonia: effects on growth

and survival. Aquaculture 1978;15(4):353–72.

[5] Krom MD, Porter C, Gordin H. Causes of fish mortalities

in semi-intensively operated seawater ponds in Eilat, Israel.

Aquaculture 1985;49:159–77.

[6] Tomar M, Abdulah THA. Evaluation of chemicals to

control the generation of malodorous hydrogen-sulfide in

waste-water. Water Res 1994;28(12):2545–52.

[7] Millero FJ, LeFerriere A, Fernandez M, Hubinger S,

Hershey JP. Oxidation of H2S with H2O2 in natural

waters. Environ Sci Technol 1989;23(2):209–13.

[8] Cadena F, Peters RWJ. Evaluation of chemical oxidizers

for hydrogen sulfide control. J Water Pollut Control Fed

1988;60:1259–63.

[9] Sharma VK, Smith JO, Millero FJ. Ferrate (VI) oxidation

of hydrogen sulfide. Environ Sci Technol 1997;31(9):

2486–91.

[10] Schmidt E, Marton A, Hlavay J. Determination of the

total dissolved sulphide in the pH range 3–11.4 with

sulphide selective ISE and Ag/Ag2S electrodes. Talanta

1994;41(7):1219–24.

[11] Berner RA. Sedimentary pyrite formation. Am J Sci

1970;268:1–23.

S.W. Poulton et al. / Water Research 36 (2002) 825–834 833

[12] Canfield DE. Reactive iron in marine sediments. Geochim

Cosmochim Acta 1989;53:619–32.

[13] Rickard DT. Kinetics and mechanism of the sulfidation of

goethite. Am J Sci 1974;274:941–52.

[14] Pyzik AJ, Sommer SE. Sedimentary iron monosulfides:

kinetics and mechanism of formation. Geochim Cosmo-

chim Acta 1981;45:687–98.

[15] Tamhankar SS, Hasatani M, Wen CY. Kinetic studies on

the reactions involved in the hot gas desulfurization using a

regenerable iron oxide sorbent. Chem Eng Sci 1981;36:

1181–91.

[16] Lee Y-S, Kim H-T, Yoo K-O. Effect of ferric oxide on the

high-temperature removal of hydrogen sulfide over ZnO-

Fe2O3 mixed metal oxide sorbent. Ind Eng Chem Res

1995;34:1181–8.

[17] Boon AG, Vincent AJ, Boon KG. Avoiding the problems

of septic sewage. Wat Sci Tech 1998;37(1):223–31.

[18] Canfield DE, Raiswell R, Bottrell SH. The reactivity of

sedimentary iron minerals towards sulfide. Am J Sci

1992;292:659–83.

[19] Dos Santos Afonso M, Stumm W. Reductive dissolution

of iron(III) (hydr)oxides by hydrogen sulfide. Langmuir

1992;8:1671–5.

[20] Peiffer S, dos Santos Afonso M, Wehrll B, Gachter R.

Kinetics and mechanism of the reaction of H2S

with lepidocrocite. Environ Sci Technol 1992;26(12):

2408–13.

[21] Yao W, Millero FJ. Oxidation of hydrogen sulfide by

hydrous Fe(III) oxides in seawater. Mar Chem 1996;

52:1–16.

[22] Biber MV, dos Santos Afonso M, Stumm W. The

coordination chemistry of weathering: IV inhibition of

the dissolution of oxide minerals. Geochim Cosmochim

Acta 1994;58(9):1999–2010.

[23] Hawke D, Carpenter PD, Hunter KA. Competitive

adsorption of phosphate on goethite in marine electrolytes.

Environ Sci Technol 1989;23(2):187–91.

[24] Schwertmann U, Cornell RM. Iron oxides in the

laboratory: preparation and characterisation. Cambridge:

Weinheim, 1991.

[25] Van Rijn J. The potential for integrated biological

treatment systems in recirculating fish cultureFa review.

Aquaculture 1996;139:181–201.

[26] Cline JD. Spectrophotometric determination of hydrogen

sulfide in natural waters. Limnol Oceanogr 1969;14:454–8.

[27] Viollier E, Inglett PW, Hunter K, Roychoudhury AN, Van

Cappellen P. The ferrozine method revisited: Fe(II)/Fe(III)

determination in natural waters. Appl Geochem

2000;15:785–90.

[28] Raiswell R, Canfield DE, Berner RA. A comparison of

iron extraction methods for the determination of degree of

pyritization and the recognition of iron-limited pyrite

formation. Chem Geol 1994;111:101–10.

[29] Koroleff F. Determination of nutrients. In: Grasshoff K,

Ehrhardt M, Kremling K, editors. Methods of seawater

analysis. Ottawa, Canada: Verlag, 1976. p. 117.

[30] Yuan G, Lavkulich LM. Colorimetric determination of

phosphorus in citrate-bicarbonate-dithionite extracts of

soils. Soil Sci Plant Anal 1995;43:387–95.

[31] Dzombak DA, Morel FMM. Surface complexation

modelling: hydrous ferric oxide. New York: Wiley, 1990.

[32] Fuller CC, Davis JA, Waychunas GA. Surface chemistry of

ferrihydrite: Part 2. Kinetics of arsenate adsorption and

coprecipitation. Geochim Cosmochim Acta 1993;57:2271–82.

S.W. Poulton et al. / Water Research 36 (2002) 825–834834