Reaction of hydrogen sulfide with oxygen in the presence of sulfite

Commonly, abatement of hydrogen sulfide emission from a geothermal powerplant requires that hydrogen sulfide dissolved in the cooling water be eliminated by chemical reaction. Oxidation by atmospheric oxygen is the preferred reaction, but requires a suitable catalyst. Nickel is the most potent and thereby cheapest catalyst for this purpose. One mg/L nickel in the cooling water would allow 99% removal of hydrogen sulfide to be attained. A major drawback of catalytic air oxidation is that colloidal sulfur is a major reaction product; this causes rapid sludge accumulation and deposition of sulfur scale. We studied the kinetics and product distribution of the reaction of hydrogen sulfide with oxygen, catalyzed by nickel. Adding sodium sulfite to the solution completely suppresses formation of colloidal sulfur by converting it to thiosulfate. The oxidation reaction is an autocatalytic, free radical chain reaction. A rate expression for this reaction and a detailed reaction mechanism were developed. Nickel catalyzes the chain initiation step, and polysulfidoradical ions propagate the chains. Several complexes of iron and cobalt were also studied. Iron citrate and iron N-hydroxyEDTA are the most effective iron based catalysts. Uncomplexed cobalt is as effective as nickel, but forms a precipitate of cobalt oxysulfide and is toomore » expensive for practical use. 33 figures, 9 tables.« less

Chaper 2 rcvicr: the literature on inorganic sulfur chcmistry, with cmphasis on sspccts rclcvant to orr rorlc Cbapcr 3 describa our study of the kinctics and product stoichionrctry of thc niclel catalyzcd rcaction of hydrogen sulfide with oxygcn in the pacncc of gulfitc ion.
In many studi€s, Il;S and ILS arc not distinSuished in thc prcscntation or intcrprctation of the data. "Total sulfidc' or'H S " is used in thc scnsc of "Il$ + ,tS-+ S -'in many places in this revicw as well.
AZ ImrStlic sulfur coqotla& Thc nrct stablc snd abundant forms of sulfur in naturc arc sulfatc mincrals, sulfidc mincra.ls, and "organic sulfur' in buried organic rnattcr. Dcposits of clcrncntal sulfur also cxist. Hydrogcn sulfidc and sulfatc ion are thc dominant forms of sulfur in natural Euids, Thcy ccur in rcducing and oxidizing environments, respaivcly. Be,causc rmt undcrgrornd *atcrs and gases arc forrncd in reducing environrnents, they contain mctly hydrogen sulfide, in which thc velcncc of sulfur is -2. When such fluids are brought to the surface and exposcd to of,ygcn thc It $ ultimatcly is con. verted to sulfate ion in which thc valencc is *6. The oxidation r€actions are vcry complcr, and gcn. erally involvc rmny interrnedietc oxidation statcs of sulfur. Many compounds and ions in which sulfur has an intcrrncdiate valcncc are knowr.
Hydrogen sulfide and bisulfide. The bisulfidc ion (llS J can also bc oridizcC to e ndical ion. but not as casily as sulfite; in bisulfide, therc arc no orygen atorrs to which the elecron deficit can be spread. In practicc, rapid oxidation of l{S rcquircs the prcacncc of polyeulfidat to catalyzc the reaction. Bccause hydrogen sulfide has no bosic frcc clcctron pairs, is much less rcsctivc. As with sulfites, lorering pH increascs the rcistancc of the sulfidc systcm to oridstion. The oxidation of H 13 is discusscd further in Scction 15.
Thc oxidation of polythionatcs probably bcgins with tbcir decompooition to sulfate, colloidal sulfur, and cither thisulfate or sulfite. The stability of polythionatcs tends to decrease with increasing numbcr of sulfur ators. The qualitative chcmistry of thc polythionates and other sulfur oro acids was rcviercd by ttatr (1970).
Meny plants and microorganisms utilize sulfatc as a sourcc of sulfur for bicynthcsis, and somc micrurgrnisrm ue it as an clcctron acoeptor in their respiration, reducing it to clenrntal sulfur or H f (kstgstc, 1968). Thcc orpnisrm "clcc thc sulfur cyclc' within thc bicphcrc. by reducing sulfatc produccd by oxidation, both inorganic and orgsnic.
Redustion of sulfate by inorganic rcection in aqucons solutiqr has not bccn detmnstrsted in the laboratory or docurnentcd in nature. If inorgonic reduction dca occur, it pmbably is a vcry slow reaction, evcn under the conditions that crist in a gcotbcrmel rccrtoir.
Triplet oxygen readily attaches itself to othu radicals: This typc of rcaction plays an important rolc in combustion proccsscs.
Thc rarc of a frec radical chain rcaction initiatcd by a catalyst commonly is proportional to thc squsrc root of catalyst conccntration. However. the rclationship bctwecn conocntration and rate may bc compticated by hydrolysis or prccipiution of the catalyst. If thc oxidation subotrate is involvcd in thc initiation stcp, its apparent kinctic ordcr in thc ovcrall ratc of oxidation rnay be one-half or thrc+halvcs.
To scwc as a catalyst, a transition nrctal ion must usually bc complcxed by something other tben watcr. In water, thc rmrc oxidizcd form comrmnly hydrolyzcs, and may precipitatc if not 16othenisc complexcd. Tbis is perticularly truc of iren(lll). Uncomplexcd iron is an cffcctive calslyst for IJ aS oridatim, but its usc results in the poduction of a noxious and concive fcrric li3,f^*rdr sludge. Cunplexed with HEDTA or citrate, iron is a powcrful catalyst for H $ oxida-':on at lotp conccnEation and des not prccipitatc (Scction 4.5).
C.oppcr is a very potcnt catdyst for the oxidation of sulfitc. but its cffective conccntration in solution is limitcd by the small solubility of cupric hydroxidc at pH > 7 (Section 2.6.1). C.omplexing agents mry incrcsse thc catalytic potency of coppcr for this resction, but this has not bcen t€stcd. Complexation by ammonia is lnown to inctessc thc tatalytic potency of coppcr in thc oridation of thiculfatc (Scction 2.7).
Many sulfur compounds are good ligsnds; for examplc, sulfite. rllS -, and thiculfate. Thcre, fore, the reaction subetrate or poduct rnay serve as the ligand nccded for catalysis by a transition rrctal ion. Complexation of cobdt by sulfitc must be postulated to explain the catalysis of sulfitc oxi&tion by cobalt (Scction 2.6.2). Sotm complcring agcnts may inactivatc I catalytic ion by sequestering it, thcrcby prcventing thc formetion of catdytically active complexes. For examplc. EDTA blocks catalysis of H 25 oxi. dation by nickel (our own obccrvation). lvlany transition metal complexes involving oxygcn as a ligand are known (Valcntinc, 1973;Martell, l9t2). If an oxidation subctratc is also prcscnt as a ligand, caulyzcd transfcr of clcctrons to the oxygcn rnolcculc may occur. This form of catalysis can initiatc chain rcactions, or dircctly inducc oxidation. The nrctal ion cxpcdites the rcaction by l) bringing the rc8ctants toSethcr, 2) sta' bilizing the peroxidc or superoxide radical that is produccd, or 3) providing its low-lying empty 3d or 4s orbitals as conduits for elcctron transfcr. In thc sccond and third cas€s, thc catslyst is dircctly involved in thc oxidation of each rnolcculc of the substratc.
Armng the first scric transition mctals, cobalt complexes with oxygcn particularly wcll (tt{artell, t982). Mct cobalt-oxygen complexcs arc binuclear. Their propcrtics suggcst that they contain a lrroxidc ion, which is complcxed by cobal(lll) at either cnd. Thc formation of such a complcx is bclievcd to bc the fint stcp in thc oxidation of compleres of cobalt(ll) by oxygcn (Cotton and Wilkinson, 1965, pp. 867-t). A fcw l:l cobalt-oxygcn complcxcs also arc known. In thcsc, onc cnd of the oxygen npleculc is attachcd to thc singlc cobalt ion; thc electronic structur€ rnay be that of supcroxidc stabilized by complcxing cobalt(ltl). Thc bcst known example is the oxygcn compler of vitamin B,rr. The iron complexes oxyhemogfobin and oxymyoglobin also bclong to this catcgory (Valcntinc,-1973). If an oridation substrate is included in the coordination sphere, its one electron oxidation rnay bc favored by stabilization of the supcroxide ion produccd. This may bc the rnechanism by which c€rtain complexcs initiatc chain rcactions.
In all wcll charactcrizcd transition mtal complexcs of oxygen, ligands other than orygen also arc prB€nt. The nature of thcsc other ligrnds has a largc cffcct on thc stability of thc complex and iu rcsctivity.

It.
Theunere two rmximr in the resction r&tc as a function of pH, onc in thc rangc pH 6.7'8.0, depcnding m initial sulfidc conccntration, and thc othcr at ebout pH ll. The rcaction did not pro€cd bclow pH 5. Thc pcak in rcaction rate at about pH 7 rras corrslated with thc formation of polysulfides as r€action intcrmpdiatcs. It was Fopoccd that the rcaction is catalyzcd by polysulfidcs; i.c., it is autocatdytic. fui induction priod is charactcristie of autocatalytic reactions; an 8utocat8. lytic raction proccrds vcry slowly until a significant conccntmtion of thc autocatalyst has accumu. latcd. A frcc radical chain lrcchanism was pmpmed, onsistent with the apparcnt fractional kinetic orders of the rcactants.
The pcscncc of polysulfidcs is readily dctected by rneasuring absorbancc at 285 nrn ottly polysulfidcs and othcr chain rnoleculcs abaorb at this wavclcngth. Typically, thc optical dcnsity at 285 nm would pcak aftcr about thrcc hours, snd \rould bc greatcst at pH 7.0. At pH < 6' polysulfides are protonated to sulfancs which decompoee to colloidal sulfur and I/ zS ; thcrefore, the autocatdytic rcaction cannot procecd. At pH > 8.5, the reaction proccods, but the major reaction product is thimulfatc. At high pH polysulfrde arc not producrd, and the relativcly rapid autocatalytic pothway is not availablc. Thc maximunr, lransicnt conversion of If f to polysulfides was 6tirnatcd to bc about t5%, which was attained with initial sulfidc: I mM and pH near ?. At highcr initial conccntrations, forrration of colloidsl sulfur was favorcd, while lower concentrations favored the formation of tbiculfate and sulfitc. Chen and Cupta ( 1973) confirmcd and cxtcndcd thc rcsufts and conclusions of Chen and lrloris (1972).
Axidation of H rS in seo *cter. l"Io catalyst was intentionally addcd to thc water in thesc studies, but thc rcaction probably w8s catalyz€d by mctal ions naturally P,rcacnt.
Clinc and Richards (l%9) used water from Lake MtinEt, an anoric fiord on thc cost of Vancouver Island. Typically, about 60pM sodium sulfidc was added. Sulfidc. sulfite, and thimulfatc were determined by colorimctric mcthods. No turbidity, which would havc indicated production of cotloidal sulfur, was obncrvcd. Thc rcsults wcre consistcnt with a rate law that is first order in both sutfide and oxyg€n. Thc half-life of It $ was about fiftecn houn, and somc expcrimcnts exhibitcd induction periods. Thc major oxidation products wcrc sulfite, thiculfate, and, prcsumably, sulfate, which was not actually mcasurcd, Usually, thc product distribution was 3G35%thimulfate, lGl5% sulfite, and the rest prcsurnably sulfate. Thc conccntration of sulfite tcnded to levcl off early, suggesting furthcr rcaction to sulfate nas uking placc. Mding 5rrM iron(ll) to the water accelerated the reaction somcwhat. and incrcascd thc fraction of thiosulfate among thc products to 82Vo.
The r€action catalyzcd by Co TSP was studied in detail. Thc kinctic orden arc onc in both sulfide and CoTSP, The aparent kinctic ordcr in O 2 was dctcrmincd as cither zcno or tnuthirds, depending on the O zl H S ratio. TSP complexcs rcscmblc c€rtain cnzymcs and, in fact, thc kinetics of this reaction rnay bc fittcd wcll using thc bisubstratc Michealis-Mcnton law uscd by biochcnr ists. The initid stcp of the reaction wcrc infcrrcd to bc: Thc 6rst two rcsctions arc rcvcrsible pre.cquilibria, and thc third is ratc dctcrmining. Subscqucnt rcactions which producc sulfate and thiculfatc are rapid and do not affect thc overall rate. Incrcasing ionic strcngth incrcascd thc rate of rcaction, consistcnt with a mechanism that i;,volvcs a reaction bctwccn two anions. Bctwccn pH 5 and E thc rate of rcaction increased with incrcasing dissociation of //25 to ItS-. Bctwecn pH l0 and 12 it increased again, apparently with increasing dcpdonation of the catalyst. Dcprotonation is bclievcd to stabilizc the more active, monomeric form of the catalyst. 2loridetion oulpt with enhrnccd activity. Tbcy hypohesizcd tbat attachncnt to a polynrcric subtrate cnbancod activity by prcvcnting pairing of thc catalyrt rmlcculcs (thc bridga bctrvccn thcm tnuld bc an O 2 rnolccule mrdinatcd by a mbalt ion at cither cnd).
Cobalt tctruulfo or teFasminoprthdocyaninc bound to polyrncric or solid subsiratcs is bclicvcd to bc uscd in ccrtain proprlctary poccsscs for I/ 25 rtmoval from liquids and gascs.
Tbc autqtalytic rcaction is catalyzcd by transition nrtal ions, and rnay or mry not mc|lr in thcir ahcnce, Thc dcpcndencc on nickel conccntration rcportcd by Snarrcly and Blount (1969) sug' g6ts one.hatf kinetic ordcr in nickel, onsistcnt with a frcc-radical chain rcaction in which nickcl participatcs in the chain initiation step. If so, the rnechanism probably rescmblcs that propoccd by Steijns er al. (1976a), cxcrpt that dl spccies arc in aqucous solution, and H2.S (ot tIS -) is directly involvcd. Catalysis by ccruin organic reducing agens which easily form frcc radicals is consistent with this interprctation (Chcn and Morris, 1972a). Snavcly and Blount establishcd that the rcaction is zcro order in O r, and that above pH 6 the rate is indcpcndcnt of pH.

Ftcc rdcrl cLb rcdu rlth oxygca
Thc rtection of sulfite witb oxygen is catalyzcd by light and ccrtain transition mctal ions. It is a fest rcaction, with e timc scale of scconds to minut6.
Biclstrbm proposcd a frcc radical chain mcchanism. ln the reaction catalyzed by coppcr, the initiatim stcp is: Fullcr and Crist (l%l) studicd the oxidation of sulfitc catalyzed by coppcr. Adding 0.0lpM coppcr sullate greatly accelerated thc rcaction. Incrcasing the conccntration of coppcr further, up to l(DpM, had no furthcr effect on thc ratc. This was cxplaincd in tcrms of hydrolysis: thc conccntra. tion of cupric ion is limitcd by thc srnall solubility of cupric hydroxidc. After saturation with cupric hydroxide has bcen rcached, adding rmre copper das not increasc thc conccntration of copper in solution. Koganovshi and Taran (1955) confirmcd that the rate maximum for sulfite oxidat'on falls near pH 7, and found that thiculfate, phenol, and aniline inhibit the rcaction.
A rcaction product caulyzcs thc rcaction, in this case ali a cocatalyst scting togcther with nickel. This causcs thc chcmical proprties of the rcaction medium to dcpnd on its previous history. This introduccs a large, sdditiond variancc into thc hnctic data.
In a contact condcnser cquippcd geothermal Unit at Tbe Gcyscrs, thc conccntration of H sS in the mling \rarer teaving thc condenscr typically is about l5O pM (spprn). In a surfacc con' dcnscr cquippcd Unit, this typically is about 70 pM. For rcasons of cct and safcty, thc conentra' tion of nickei should bc lm pM (6ppm) or less, Accordingly, m6t rate data wtrc gcncratcd with 20 or 100 sM nickel, and 70 to 150 sM initial sulfidc.
The combination of unstable reactants in low conccntration, short rcaction timc, catalysis, and autocatalysis greatly increascd experimcntal difficultics. In particular, autocstalysis made thc use of a straighiforward "-once-through" kinetic apparatus (like that daqibcd in Chaptcr 4) inappropriate. Ultimaiely, we dccided to use a simple experimcntal tcchniquc which would allow rnany data points to be gentratcd, and allow experimcntal scatter to bc ovcrwhclmed by avcraging. Each day a batch of "synthctic cmling water" (SCW) containing nickcl was sct up, and air and Il 25 wcrc added to it continuously. Afier some time (typically I or 2 hours) the SCW would rcach chcmical stcady state. Then iliquots would be withdrawn from it and uscd for individual kinetic or other expcri' ments.
Repatability iilrong experirncnts using diffcrcnt diquots rcmoved from the flask during a given run usually was gmd. Repcatability bctwern different runs was psrcr, especially in regard to ieactivity. This ptaccd obvious constraints on the erpcrimcnts that could bc pcrformed. Reproduci' bility improved with expcrience.
Mct of the erpcrirnents u/'erc pcrformcd at pH 7.8. Becausc the SCW is naturdly wcll buffercd by lVll r* and B (OH )l atnvc pH 7.5, additional buffcring mmpounds usually wcrc not necded. Erperinrcnts with SCW that contained amrmnium chloridc or boric acid only dcmonstrated that tbc only role of thc rnajor ions in thc SCW *rs to buffcr itl therefore, thc particular recipc uscd had no cffcct on the reculs obtaincd.
The output of the pcristaltic prmp casscttes varied from day to day, probably due to minute variatims in thc poeitions of thc tubes in the casscttcs. Thc cassettcs wcre calibratcd daily by running the output of cach into a graduated cllinder for ien rr,inutcs. Using tbis calibration data. thc pump spoed and conccntrations of the reagent solutions wcrc adjusted to give thc chcmical fccd ratcs desircd.

32-
pH was lowcrcd to 7.3 with sulfuric acid, and the solution madc up to volumc with dcioniz€d water. When first ncutralized, this solution was opoque ycllow, but bcnanrc clear ycllow afier about 30 minutes.

3.2.X Worh rt lower pH
Erperiments bclow pH ?.8 requircd additional buffcring compounds, A combination qf l0mM each of phcphate and malcatc was used. A conccntratcd buffcr stock solution was preparcd from malcic acid and moncodium orthophsphate, and its pH adjustcd to that daired with NaOll . With this buffcr added, the SCW was wcll buffcrcd from pH 2 to 10.
This buffcr (likc othen) slowcd the developmcnt of reactivity if prcscnt from the start of the reaction. Thcrefore, the reaction was always initiated at about pH ?.8 without thc buffer. and the bufrer added at t : 60 min. or later. The buffcr was added as I conccntrsled solution, the pH of which had bcen ad.iustcd to a value slightly lowcr than that desired in the resulting buffered SCW. Once cstablished, reactivity is unaffected by thc addition of thc buffer (Scction 3.3.4).

Iltcrdnrdoo of n*drlty
"Rcactivity' was uscd as an empirical measurc of thc solution's ability to destroy I/ 15 . To determine the reactivity, a 25 ml diquot of SCW wes rcrnovcd from the flask and acrated by blowing air through it for a few scconds. This also reduccd thc anrount of IJ f and sulfite in the SCW bcfore measuring the resctivity. A rnagnetic stirring bar was addcd to thc beakcr containing thc aliquot, and it was sct upon a rugnctic stirrcr. Thcn a small arnount of Nc $ solution was added to thc bcaker using an adjustablc micropipttc. (Gilson Pipttcman, lvlodcl P20 or P2fi), dcpnding on volumc.) Usually, enough rVa $ was addcd to introduce a total sulfidc conccntration of ?0 pM. Bccause thc conc€ntration of the sulfidc stock solution was about 0.2 M. about 9 sl would typically bc added. Aftcr fiftccn scconds, 25 ml of sulfidc antioridant buffcr (SAOB) was added to the bcaker with rapid stirring, and this qucnched thc rcection. Thcn clcctrodcs wcrc introduced to rn€asure thc residual sulfrdc conccntralion.
lmmcdiately after the elcctrodcs were put into the mixture of tcst aliquot and SAOB, thc voltage reading droppd rapidly for sevcral seconds, indicating incrcasing appar€nt sulfidc concentration. dfter about 20 scconds thc voltage reachcd a minimum and bcgan to incrcase slowly. The minimum voltage ob'scrvcd was recorded and used to calculate the conccntration of sulfide.
Finally, the measurcd sulfide conccntration was divided by that initially introduccd to givc the fraction remaining aftcr l5 scconds. The numerical value of this fraction is the quantitative expression of what we call "reactivity': if the fraction of I/ f remaining is small, reactivity is said to be good. and vice-versa.
The preparation and handling of thc sulfide stock, SAOB, and othcr solutions is described in Appendix 3. l.
An Orion Model 70lA potentiometcr, Model 95 I printcr, and Model 605 electrode srvitch were used for all measurements of sulfide concentration and pH. The sulfide electrode used was the Orion Model 94 16, which is of the.4g$ membranc type. The refercnce electrode uscd was the Orion Model 9G02, with Orion 90m02 filling solution in the inncr compartmcnt and lM WO 3 in the outer. Each day before use, the electrodcs were standardizcd with freshly preparcd standards in 50/50 SAOB/D.1. water, which corresponded to sample conc€ntrations of 7 and 70 rrM. Typically, a sulfidc electrode would last sevcral months in daily service bcfore failing by leakage through the .49 25 membrane.
In their studics of //25 oxidation, Gtlund and Alcxander (1963), Algrcn and Hagstrijm (1974), Hoffmann (1977), and Hoffrnann and Lim (1979) measured the conc€ntration of //S directly with an ,{g 15 electrode, In all cases, the time scale of the reaction was minutes to hours (Section 2.5). The response timc of the lg 2.S eleetrode is about l0 seconds. In our work, an Ag 75 ele*trode could not bc used to mcasure /lS directly, bccause the timc scale of the reaction was comparable to thc electrode reponsc time.
Throughout, turbidity is rcportcd in Ncphelonrtric Turbidity Units (NTU). Turbidity is apparent to the nakcd cyc at 20 NTU, and a colloidal sulfur sol of 150 NTU lmks like dilutcd skimrned milk.

Deteruinrtioo of rstioa products
The SCW used in erpcriments to study rcsction product distribution differcd from that described above in that ac.tat€ was uscd in the placc of sulfate. Also, acctic acid was uscd to ncutralize the lya S . This changc allowcd the amount of sulfatc produccd by thc reaction to bc detcrmincd. "Acltatc based SCW" was not uscd in thc othcr erpcrimcnts bcsusc thcrc was evidencc that accute decgeascd the ratc of oxidation somcwhat. In a few expcriments "perchlorate based SCW" was uscd.
In the stoichiornetric cxperiments aliquots were pcriodically withdrawn for the detcrmination of colloidal sulfur, thisulfate and trithionate. Sulfatc was dctermined only at thc end of the expcrimcnt. Colloidal sulfur was determined only if the solution was visibty turbid, and sulfite was ditermincd rarely. a Tk ccctrtc ht d SCW containcd only half as much amrmnium acctatc at it rhould have; i.e., l.65mM rathcr than 7.30tnlL Thc clrE error wa! rmdc in onpqrnding tbc pachtorrtc be!.d SCW. Thir potably hsd no cfet on thc recult! obuincd. 3a-@llcidal sufur. Thc technique of Bartlctt and Skoog (1954) was used in modificd form. A 50 ml diquot of SCW was filtered through a nrembrane filter under pressure with nitrogen (Millipore typc VC, 0. lpm porc sizc, 47mm diameter). Thc filter rvas removed from thc filtcr housing and dissolved in l0 ml of 95% soetonswater. This was mixcd with l5 rnl of the sodium cyanide reagent spccified by Bsrtl€tt and Skoog. Aftcr a few minutes, a prccipitate of filtcr rnstcrisl formed. The mixturc rvas filtcred through fine filter paper to rernove this precipitate. The rest of the procedure procccded as dcscriH by Bartlctt and Skmg, Thiasulfate and trithionate. The rpthod of Kelly, et al. (1969) was used lor the simultaneous dctermination of thisulfate, tFithionste, and teuathionate. Tetrathionate was never detected in rnore than tra€ amunts, and thesc values ar€ not rcport€d hare.
Sulfate. A variant of the stendard gravirnctric proccdure was used. A 200 ml aliquot of SCW was mcmbrane filtercd to remoyc colloidal sulfur, if present. The pH of thc aliquot was adjusted to about 2.5 with 0.1 N HCl. To rernovc thiculfate, one pcrccnt tincturc ol iodine was slowly addcd with stirring until I pcrsistcnt, faint yellow color formed. Twenty ml of 0.1 N SaCl 2 was added, and the solution was covcred 8nd sct aside at room tcmperature until the nert day. Boiling was avoidcd to rcducc intcrfcrencc by trithionate. Finally, thc prccipitate was filtcrcd, ashed, and weighed thc usual way.
Sulfite. The technique of Wcst and Craeke (1956) was used in rnodified form. Thc sodium tetrachloromcrcuratdll) reagcnt was elirninated from the proccdurc. Instead, a l0 ml aliquot of SCW was directly mixcd with the prcaniline and formaldehydc r€sg€nts.
Dithionate is the only likely rcaction product that thcs€ proccdurcs would not have detected. Dctermination of dithionatc is here, as in rnct contexts, an unsolvcd problem.

$fety corldcndoos
Work with H 25 , solid sodium su!fide, and conccntrated solutions of ffa rS csn b€ hazardous. Most of the work of preparing Na 2.S stock solutions was donc under a fume hood, and these solutions werc handled outside of the hood in small containcrs only. Spills of solid sodium sulfide and lfo 15 stock were cleaned up promptly, Conccntrated iVa 1S solutions wcre never poured down the drain. They, along with waste solid sodium sulfide, were dispmcd of by reaction with an excess of ferric chloride solution. The resulting prccipitate of iron oxysulfidc was allowcd to age, filtered out of solution, and thcn allowed to age again. Ultimately, it was dispccd of as solid waste, and the residual fenic chloride solution was poured down the drain. The ffask of SCW was sct up and used in a water bath inside a fumc hmd. Thesc precautions ensurcd safety, and largely elirn' inated complaints from others working in the same rmrl

Developrrnt of reectirlty rnd tu6idity
In these experimens the reaction was allowed to prmcd while turbidity and reactivity were measured periodically. Usually the experiment was started with addition of lVa 25 , acid, and air only, and the addition of lVa lSO 3 commenc€d 20 minutes later.
Typical data are presented in Figures 3.1 and 3.2. Before the addition of sodium sulfite commenced at t : 20 min. turbidity incrcascd rapidly due to formation of colloidal sulfur.

Reectivity lrd rutocrtdysis
In all cases, reactivity wits poor at first (i.e., the residual fraction was large), but improved rapidl;r, reaching a steady state value about an hour after the start of the experiment. This is characteristic of an autocatalytic reacticn.
Reactivity develop slowly or not at dl if the addition of sodium sulfite is commenced at the very start of the reaction (not shown). This suggcsts that the unkrrown cocatalyst contains chains of zero valent sulfur atoms; sulfite destroys such molccules by converting the zero valent sulfur in them to thimulfate. Most likely, othc cocatalyst" is a mixture of polysulfide ions (-S" -J, polysulfidomonmulfonate ions (-Sn SO l-), and their corrcsponding radical ions. The yellow color of the SCW (when reactive but not turbid) suggests the prescncc of such molecules, and they are   Howcvcr, reactivity uould dcvelop with sulfite addcd from thc very start, if the sulfidc and sulfite fccds were combined to form one reegent stream bcforc bcing addcd to the SCW. lo*tring rhc pH of this mixturc by mixing it with thc SCW cscntially would poducc Wackcnrodcr's Solution, which contains polysulfides, polythionates, etc. In thcsc expcrimcnts the SCW bccamc ycllow. but never turbid.
Thc effect of adding Wackenroder's solution as such to thc SCW was tcstcd. At the very start of the experirncnt, enough was added to introducc a total conccntretion of sulfur of 4tr pM. This accclerated the development of reactivity (Fig. 3. l).
Formally, the SCW was supersaturated with nickel sulfide by scvcral orders of magnitudc. Dspite this, there was usually no sign of its prccipitation. Whcn $crc was, it was limitcd to thc appearance of a bronze color in the otherwise blue and ycllow SCW . This suggcsts thst thc nickcl is complexed by polysulfides, and thcreby kept from precipitating. Thc formation of this complcx rnay bc an aspect of autocatalysis.
Buffcring compounds probably retsrd the dcvelopmcnt of reactivity (Scction 3.2,2) by chclating nickel, and rendcring it inactive. That buffcrs do not affect reactivity after it has bccn cstablishcd suggcts that the nickel is tightly chelated by polysulfides. tTb.

Tbe efiect of sodiun sulffte
When the addition of sodium sulfite commcnc€s, the formation of colloidal sulfur is slowcd. Under the propr conditions, the colloidal sulfur formed earlicr is destroycd. The destruction of colloidal sullur incrcases with increasing sulfite:sulfide ratio (compare Figs. 3.1 and 3.2).
This data sug,gcsts that at pFI 7.8 a sulfite:sulfrde mole ratio somewhere between 0.5 and 0.?5 is needed to ensure good clarification. This is discused furthcr in Chaptcr 5.
Five mM sodium thimulfate was addcd at the beginning of some expriments. At sulfrte:sulfide -0.5, thiculfate inhibited the "clarification reaction'. This may be a case of inhibition by the reaction product. The parasitic reaction of sulfite with thicullate to give trithionate may also play a role: SzOr-+4SOj-+6H+-2SrOe-+HzO At sulfite:sulfide : 0.75 (Fig. 3,2). thioulfate has the oppcite effect: it enhances the clarification reaction. Pcrhap, bcing a rcducing agent in its own right, thiculfate is able partially to protect sulfite from oxidation, thereby making rnore of it available for reaction with colloidal sulfur. In any case, there must bc two oppming effccts, either of which may predominatc under different conditions.
The initial prcsenc€ of thiculfate and the sulfite:sulfide nrcle ratio havc no effcct upon the development of reactivity.

Tbc efiect of pH
The pH value 7.8 was originally choscn bccausc it is approxirnately the "natural pH" of cmling water at The Gcyscrs. lt is also near the lowest pH at which the SCW has adcquatc buffcring capacity. In the ficld, the cooling water pH would probably be nuintaincd sorpwhcrc bctween 6.2 and 7.0 in order to reducc the amount of NaOH rcquircd. Thereforc, sornc cxpcriments wcrc also made at pH values in this range. In thcse experinrcnts, thc SCW was buffcred with phmphatc end nuleate (Section 3.2.2). Fig. 3.3 depicts one such experimcnt. At the initial pH of 7.t, a mole ratio of 0.?5 sufficcd to prevcnt the formation of colloidal sulfur, but turbidity inccased again aftcr pH was lowcred to 7.0. lncreasing the sulfite:sulfide ratio to 1.0 caused to turbidity to decrcasc again. This behavior prob ably is due to accelerated oxidation of sulfitc at the lower pH (Section 3.6). Ncither pH nor sulfite:sulfide ratio affects reactivity.

Reectioo kiaetics
Reactivity was detcrmined under various conditions. Thesc data arc prescnted in Figures 3.4 to 3.9. The solid lines in Figures 3.4 to 9 werc calculated using formula (3. l), with thc paramctcrs given in Table 3. l. The computer program used for this purposc is listed and documcnted in Appndix 3.3. The variancc of the data is discussed in Appendix 3.4.
There was no correlation of reactivity with eithcr pH or sulfite:sulfide ratio. Accordingly, data generated at various values of pH and sulfitc:sulfide ratio were plotted and analyzcd together in Fig.  3.4 and the following Figures.
The effect of temperature on residual H lS suggests a small, positive energy of activation (Fig. 3.7). With 957o confidence (Student's t test), the energy of activation is bctween 0.9 and 2.9 kcal; 2.4 kcal is our bcst estirnatc.
(2) The kinctic order in sulfide decreascs from two to onc with incrcasing sulfide conccntrstion.
The only significantly diffcrcnt, but plausiblc altcrnativc to (3. l) is; This expression was rejected, because it gave a visibly poorcr fit, ln particular, the fittcd curves in Figure 3.4 were horizontal, straight lincs, and the dropoffin the fitted curves in Figures 3.5 and 3.6 was much tm rapid. Likewise, assuming that kinetic order in nickcl : I is inconsistent with Figures 3.4 and 3.5: it causes the predicted rcsidual fraction to drop off much tm rapidly with increasing nickel concentration.
Minor variations from the form (3.1) are not cxcluded by the data; for example, the details of the change from second to first kinctic order may bc differcnt, or the kinetic order at low concentration might bc three.
Forpurpces of fitting, expicssion (3. l) was convcrted to integral form. The values of /< 1 and & 2 at 35"C were simultaneously estimated by a bivariant least-squares fitting proccdure, using all data points in Figures 3.4, 3.5, and 3.6. The value of thc activation cnergy 4 *as cstimated from the data in Fig. 3.7 only with the assumption that & 1/& 2 dm not changc with tcmperaturc. The data in Figures 3.E and 3.9 werc not fittcd at all.
The statistical uncertainty in thc calcurlated value of t I is about l0% we acc€pt this as the uncertainty of the reaction rate at 35"C, predictcd using (3.1).

tlstribution of rcectioe produce
Mrmally, 7r9M of thc sulfur put into the SCW was recovercd. The sulfur recovery improved with increasing conccntration of nickel, but dcteriorated with dercrcasing pH. This sug-Bests that the rnain source of the clcurc crror was the lcs of Il 25 to thc atrncphcre. Incomplctc recovery of sulfate and nondetermination of dithionate also may have contribut€d to it. Fig. 3.10 illustratc the reaction product distribution with nosulfitc added. The major reac. tion products are thimulfate and colloidal sulfur, in that order, and there was a small armunt of sul. fate present at the end of the cxperinrcnt. Only a tracc of trithionate formcd, if any. The curvc of colloidal S vs. timc is concavc. Thc accumulation of colloidal S is limited bv its furthcr oridation to thioulfate and sulfate. Fig. 3.ll illustrates reaction product distribution at sulfite:sulfide :0.75 and two diffcrent values of pH. At pH 7,9 thiculfate is the major product, and smaller arnounts of sulfatc and trithionate also are produccd. l.to colloidd S is produced at this pH. At pH 6.5, thiculfate is still the major reaction product, but rnore sulfate is produccd than at pH 7.9. There is somc colloidal S at pH 6.5, but little or no trithionate. Figure 3.12 illustratcs reaction product distribution with l00rM Ni. Sulfur recovery is noticably bctter than with 20pM Ni, but thc distribution of products is otherwise unchangcd (compare to Figure 3. l l). Fig. 3. l3 illustrates the effcct of varying sulfite:sulfidc ratio at constant pH. Thiculfate is the major reaction product in all cascs, but thc yield of sulfate and trithionate increascs rapidly with added sulfite. This suggcsts that sulfate and tritionate are poduccd mctly by reactions involving sulfite. Colloidal S is significant only when no sulfite is added.
Figures 3.14 and 3.15 illustrate thc cffe.t of varying pH at sulfitc:sulfide :0 and 0.75, r€spectively. ln both cases, decreasing pH definitely favors the formation of sulfate at the expcnse of thiosulfate. With no added sulfite (Fig. 3, 14), the arnount of colloidal sulfur forrned is not affected by pH. With sulfitc added (Fig. 3.15), colloidal S is present only at the lowcst pH.

Tbc rcrction mcbnisn
A proposed mcchanism that is consistent with the data is presented in Table 3.2. Thc reactions are approximate and may bc changed somewhat without materially affecting the conclusions. Which species are protonated is not known with certainty, and polysulfidomonmulfonates may take the place of polysulfides. Reactions (la to c) are a cycle that adds frcc radicals to the system. Reactions (la) and (lc) are irrcversible, and ( lc) determincs the overall rate of radical generation. Reaction ( lb) is a rapid, reversible equilibrium with equilibrium constant K 15 Reaction (ld) rapidly converts all I/S ' pro duced to relatively stable polysulfido radicals, and does not affect the overall rate of reaction. With these assumptions, the rate of generation of frec radicals is: All of the nickcl is complexed which prevens the precipitation of nickel sulfidc.
The total conccntration of radicals varies slowly. The approximate balancc between reactions (l) and (3)  From the values of ,t 1 and,t 2 given in Table 3,l, we estinute K ra : l.rl4x l0l0 Mr (3.?t) The limiting forms (3.6a and b) are dso thme of the empirical rate law (3.1). However, the rate liw derived here is nrcrc complicated than (3.1). It can be intcgratcd, but the integral form is clumsy and ill-suited for convenient analysis of kinetic data. This is why the similar but simpler form of (3. l) was used instead.
Neither (la) nor (2a) is rate determiningt therefore, the rate of reaction is independent of oryg€n concentration, This mechanism does not explain why the rate is independent of pH. Perhaps, reactions (2b and c) can utilize Il 15 and HS at the same rate, or else changing pH has different effects that cancel out. No fully satisfactory explanation has becn found.
Reactions (5a and b), which convert zero valent sulfur to thimulfate, are side reactions which do not affect the rate of disappearance of H f .
The stoichiornetric data suggest that the oxidation of sulfite is the rnajor source of sulfate.
In particular: -Very little sulfate is produced without addition of sulfite. - The amount of sulfatc produccd increases with thc arnount of sulfite added. - The arrnunt of sulfate produced at pH 7.8 is less than at pH 6.5 or 7.0.
The oxidation of sulfite was rcviewcd in Scction 2.6. and rcactions (?a, b and c) are taken form Table 2.4.
The arnount of trithionate formcd also increases with thc amount of sulfite added. This is consistent with reaction (8) which is discussed in Section 2.2.4.

[Iscussioo
All earlier studies utilized a "one-shot" experimental design: a solution of known sulfide concentration, nickel concentration, etc., was prepared, and the decline of sulfide or oxygen conccntration was monitored as a function of time. Experiments like this yield an empirical r€action rate that largely rcflects the increasing concentration and cffect of the cocatalyst. We used solutions which already had attained steady state cocatalyst conc€nlration. Thus, the effect of varying cocatalyst concentration was eliminatcd.
Conse4uently, only qualitative comparison with the results of previous studies is pcsible. Overall, our conclusions agre€ with thme of Snavely and Blount ( 1959) (Sections 3, l and 2.5).
The exact nature of the cocatalyst has not becn determined, but the molecule certainly contains a chain of zerovalent sulfur atorrs. Most probably, 'the cocatalystn consists of a complex mixture of polysulfidcs, polysulfidorrnnmulfonates, and the corresponding radical ions. which inter-@nv€rt and change chain length in the course of the reaction.
Contact condensers steam strip oxygen from the cmling water, rcndering it anoxic between condenser and cooling tower. This is where the fI 25 must bc oxidized, if emision to the atmo sphcre is to be prevented. The oxidation reaction is able to proceed in the absence of oxygen, bccause the t/ f is removed by reaction with oxygcn-containing polysulfide molecules, rather than by reaction with oxygen itself (reactions 2b and 2c in Table 3.2).
The rate law presented in Table 3. | fits the rate data to within the scatter of those data. Although it is incomplete in some respects, the reaction mechanism presented in Table 3.2 49satisfactorily explains rnct of the data, both kinetic and stoichiometric. Why the rate of reaction is independent of pH over a fairly wide range (5.5 to E.5) is the only major question that remains unanswered.
The mistaken decision to conc€ntrate on l5 seconds rcaction time seriously affected the choice of experimental technique, practically forcing us to use "aliquot tests", and quench the reaction with sulfide antioxidant buffer bcfore measuring residual sulfide concentration. If we were now to repeat this work, we would use a different technique. 'Synthetic cooling water" would be prepared as in this work, but kinetic measurements would be made using a 'flowing' apparatus like that described in Section 4.2 and Appendix 4. l. A two channel, precision syringe sump would bc used, otre channel would pump sodium sulfide solution, and the other, proprly "prcaged" SCW. Reaction times of 30 ro 150 seconds would be provided by varying pumping rate and/or the length of the delay coil, and the residual sulfide in the solution would be dctermined directly, by flowing it p6st an ,.{g 25 electrode. Reaction tcmperature would be controlled by immersing the delay coil and other tubing in a thermctatted water bath. Mct work would be performed using nickel concentration no Sreater than 20pM. Such is the wisdom of hindsight.
Appeudix 3.1. PreFntioo and stor4e of solutios Sulfide antioxidant buffer (SAOB) and solutions of sodium sulfide and soldium sulfite arc sus-c€ptible to oxidation by air, and mesns to control this were developd.
Preparation of anoxic solutions. A simple apparatus was developed for preparing and transferring anoxic solutions (Figure 3.16). A large graduated cylinder or a large flask was used as thc deoxygenation vessel. This vessel was placed upon a magnetic stirrer. Frequ€ntly, a flexible plastic bag of the kind used to store intravancous pharmaceutical solutions was used to store anoxic solutions (see below), A magnetic stir bar and the solution (or water) to bc deoxygenated are put in the deoxygcnation vesel and the stopper with tubcs is put in place. Thc T-valves are initially set to allow gas flowing out of the deoxygenation vessel to go by way of the exhaust tubc, and to leave the system by way of the gas exhaust line. The valve on the nitrogen cylinder is opened to allow moderate bub bting; no bubbles should reach the exhaust tube. The solution is stirred. Aftcr frve minutes, the nitrogen flow is interrupted, the stopper removed, and chemicals are added (e.9,, solid sodium sulfide). The stoppcr is replaced and nitrogen flow resumed briefly to purge any oxygen that rnay have been introduccd.
At this time a clean lV bag is attached to the delivery tube, and the exhaust line is connected to the hous€ vacuum, T-valve #l is turned to evacuate the bag using house vscuum. (The vacuum flattens the bag,) Then this valve is turned to connect thc bag to the deorygenator and the nitrogen valve opened to fill the bag with nitrogen. The bag is again evacuated. This process is repeated onc€ more to purge the bag of traces of oxygen. Finally, T-valve #2 is turned and the nitrogen valve opened again to drive the solution from the deoxygenator through the solution withdrawal tub€ into the collaped bag. When no further solution can be driven out, the nitrogen is shut off, the bag removed from the tube, squeezed to eliminate the remaining nitrogen, and stoppered.
Storage of anoxic solutiots. One liter IV bags proved ideal for the storagc of anoric solutions. Bccause they are collapible, they allow small amounts of solution to bc withdrawn without expcing the rest to air. The bags used were distributed by Travenol Laboratories, and their material of construction was stated to be "PLI46 TM" plastic. They were obtained from a hmpital and cold sterilized with l0% hydrogen pcroxide before usc. These bags proved very durable and showed no signs of deterioration after many cycles of filling and rinsing. Dilute Na 25 solutions (<0.04 M) and solutions of ferrous salts were always stored in such bags.
Storage in an lV bag was frequently unnecessary. Because the diffusion of oxygen into an anoxic solution is very slow unless the solution is stirred or aerated, relatively conc€ntrated solutions could be stored and used in open containers for up to several hours with little degradation. For example, the Na 25 stock solution used to determine reactivity was generally kept in an opn bottle while work was in progress. BIH Fit" 3.16. Apporatus for prqring enoric solutions. a. nitrogcn from cylindcr. b. nitrogcn intakc tubc and diffuscr. c. gas outlet tubc. d. liquid outlct tubc. c-line to bousc va$um up. f. intravancous solution bag. g. stining pllct. h. nugnctic stirrcr.
Prepantion of sulfide antioxidant bufer (SAOB). Sulfide antioxidant buffer is used to quench oxidation reactions and converl all sulfide species present to S -, which is what the sulfide electrode actually measures. To measure the conc€ntration of sulfide in a solution, nn aliquot of it is first mixcd with an cqual volume of SAOB. The SAOB recipc used by us was that given by Orion Research in their manual for the sulfide electrode, modified by replacing tetrasodium EDTA with acid EDTA and an equivalent arrxrunt of NaOH.
To prepare one liter of SAOB use: 95.94 g NaOH 5E.24 g EDTA 35.00 g ascorbic acid First dissolve the NaOH in about 0.25 I of deionized water. After cooling this solution, slowly add the EDTA to it while stirring, taking care that lump do not form. After this has dissolved completcly, add and dissolve the ascorbic acid, and add deionizcd wat€r to give a final volume of I liter.
Preparation of sdium sulfde stock solution. BEFORE ATTEMPTING THE FOLLOWING PROCEDURE, REVIEW THE SAFETY PRECAUTIONS IN SECTION 3.2.6. The stock solution of No 15 contained about 0.2 Msulfide. It was prepared from reagent grade Na 2.S '91l20 (Mallinckrodt). This material came out of the bottle in large lump which were first broken with a ham-m€r to pieccs that would fit into a volumetric flask. These pieces were rinsed with deionized water to remove oxidation products and other superficial contaminants. (There was usually a white material on the surface of the lumpprobably elemental sulfur.) A 2-liter volumetric flask was used as the deoxygenation vessel. Somewhat less than 2 I of deionized water was deoxygenated, about 100 grams of the washed lump of Na 25 were added to 5tit, and the total volume was made up to 2 I with additional deionized water. Nitrogen ffow was restarted to finish deoxygenating the solution, and continued with stirring until the crystals disolved completely. Finaliy, the resulting solution was transferred by nirogen pressure to prcpored 62-ml polypropylene bottles with screw cap. The bottles were squeczcd and carefully closcd so that no sir would rernain in them. These bottles were rinsed of spilled sulfide solution, dricd and thc top wrapped with Parafilm. Unopened bottles of ffo 25 stock solutions could be stored indcfinitcly with no measurable decline in sulfide concentration. orrce or twice a week a fresh bottle of this stmk solution was opned. The exact concentration of sulfide in the stock solution was determined by titration with lead nitrate. using a sulfide electrode to determine thc end point. After the bottle had been unscaled, the screw cap only was used to reclos€ it, and the sulfide in it slowly oxidized to polysulfides. Spot checks indicated that the rate of decline in sulfide conctntration after the bottle had been unsealed was about l% pcr day. lVa 1S solutions would be discarded when they turned yellow or after a week had elaped since unsealing thc bottle.
Preparation and storage ol sdium sulfte stek solutions. Ordinarily, the sodium sulfite resgent solution of conccntration equal to that of the Na 25 stock solution was prepared each day from a 1.0 M Na 2SO 3 stock solution which was made up each Monday. In experimcnts that might have bccn sensitive to the exact conccntration of sulfite, the sulfite solution was prepared each rnorning from crystalline reagent grade sodium sulfite. These solutions were prepared using water that had not been deoxygenated, and kept in an open container while in use. In retrospect, thcy probably should have bcen treated as were the sodium sulfide solutions. Howcvcr, our data did not appcar to have been affected by oxidation of the sodium sulfite solutions.
Appeodix 3.2. W rbsorptim specm Bisulfide ion has a pronounccd absorption maximum at about 230nm in the near UV ( Figure  3. l7). Ellis and Golding ( 1959) measured the aboorption of Il ;S solutions at 23(hm as a function of pH, and extracted values of the first acid dissociation constant from this dau. At 230nm, absorption by thimulfate is severalfold weaker, while sulfatc, sulfite and hydrogen pcroxide hardly ab'sorb at all. This suggests that W ab'sorption spectrophotorn€try rnay be an adequate analytic method when no other sulfur species are prcsent. For example, this nuy bc thc casc in the condensate line of a surface condenser-type Unit. bcfore it joins the cmling water line.
When hydrogen pcroxide is added to the condensate line for secondary abatement, H 25 abatcment cfficiency could be monitored by nte,asuring the conccntration of I/S in the condcnsate just before it mixes with the omling water. The emission rate determined this way would bc somewhat high, because it would not allow for pcsible further oridation in the cmling watcr line, ctc. There would be some interference by thiculfate. This could bc corrected for by also measuring absorbance at 2l5nm where thiculfate has maximum absorbancc, and numerically corrccting the measured concentration of I{S .
The concentration of polysulfides, etc., in the condensate is unknown. [f they are present, they may interfere strongly, making the technique impractical. This would have to be dctermined eithcr expcrimentally or in the field.
Bisulfide probably could not be measured in the cooling water in this way, bccause there its conccntration would be thirty-fold lower, and that of thiosulfate five-fold higher.
Once reactivity has been established, the SCW is practically opaque below 250nm (Fig. 3. lE). This is why UV absorption spectrophotometry is practically useless for measurement of bisulfide when nickel is being used. The gradual increase of optical density with decreasing wavelength in the visible c{luses the SCW to appear yellow by transmitted light. Thc two peak at about 340nm and 390nm are always present, but their intensity relative to each other and the background is highly variable. There appears to be some correlation between their size and the reactivity of he SCW; pcrhap they are associated with the cocatalytic species.
The absorption sp€ctrum of the SCW is associated with polysulfides and similar comp,unds. Polysulfides absorb strongly in the violet and ultraviolet (Gggenbach, 1972), and Wackenrodcr's solution is bright yellow. The yellow color and strong LIV absorption havc becn correlated with the rate of Il$ oxidation by others (Chen and Morris, l9?4 Chcn and Gupta, 1973;Hoffrnan, 1977).

Ancdix 11 Coqutcr codc OXTAB
Comprter mde OXTAB calculates the rate and extent of H 25 oridation, and outputs thcsc rcsuls in tabular form. A listing of OXTAB and a sample input deck are presented in Table 3.3. lt is writtcn in CDC FORTRAN.
The subroutine KINOX pcrforms the actual calculation, and may bc used separately from the main program Q)CIAB. KINOKs input arguments TC, RT, CM, and CSI are, rcspcctivety, thc tempcraturc in "C, the rcaction time, the totd conccntration of nickel in solution, and the total initial conccntration of It !S , including HS -. ( ts) (?F10.2) Card I spccifics the numbclof tablcs to bc gencratcd during this job, and each subsequcnt card specifies a particular table. Card 2 is rcpcatcd NCAR-D timcs. The first four clcrncnts of VI contain the values of TC, RT, CNl, and CSl, respcctively, to bc uscd in gcnerating the givcn tablc, exc€pt that one of thcsc ficlds is left blank or contains a zero. The blank ficld or zcrc concsponds to thc input variable which is to b€ varicd in generating the table. Thc initial and maximum r.alues of this input variable are spccificd in VI(5) and VI(6), rupectivcly. The valuc by which it is to be incrcmcnted bctwcen lines of output is spccificd by VI (7).
The sampte input dcck at the end of Tablc 3.3 will spccify one table, for 35oC, l00l M Ni, and 70pM initid sulfide, with tirnc varying from l0 to 60 scconds, in incrcmcnts of l0 seconds.
If TC, RT, CM and CSI are all spccified, a single linc of output colculated using these values will bG gcnerated. The remaining clcmcnts of VI need not bc specified in this case. If scvcral cards that specify single lincs of output are groupcd totcther, thc resulting outprt will be combincd in a single table. Mixing input cards that specify single lincs with input cards that specify tablcs is not rccommcndcd. The output from OXTAB is self explanatory. Mct of the point scattcr is associated with variation anrcng experiments, rather than armng points from any givcn experiment. This indicatcs that variations in developmcnt of reactivity are responsi. ble for nrcst of the point scattcr, rather than variations in the actual reactivity rneasurements; aute catalysis sc€ms responsible for rnct of the point scatter. Variation in electrode properties or experi. mental proccdure from day-today would have affected the data similarly, but close examination of laboratory rccords lcnt no support to this hypothesis.
The scatter and deviation (from the frtted curves) of data from a given experiment increases with increasing complexity of that cxpriment. For example, each rcactivity test in Figure 3  Fitting thc data using (3.2) tavc an cstimate of mcan k r : 17.7 and STD -3.0 (STD/mcan : 0.28). Each data point yiclds an estimate of ,t 1 in this case, and thesc arc averaged. Thus, using (3.1) improvcs quality of thc fit as mcasured by the valuc of STD/mcan only slightly over that obtained with (3.2). The choicc of (3.1) was ultimatcly adhered to bccausc it gives a bctter fit in a qualiutive scnsc (Scclion 3.4. I ). 35.

56-
The 158 data points prcscnt€d in Figurcs 3.4, 3.5 and 3.6 are an interna.lly consistent subset of the 230 data points collrctcd in the conesponding series of expcriments, hta from four of the fiftccn experiments in this series were grcsly inconsistent with data from the other elevcn, and anrcng themselves. Thc apparently bad data from these four experirnents was neither frttcd nor included in the Figures. Fourtecn of the data poins from the 'good" exprimcnts wcrc prturbcd in an obvious way by spoilagc (accidental oxidation) of thc sodium sulfide solution used in rcactivity measurem€nts. These points also were dcleted.
Values of mean t 1 and STD wcre calculated for cach of the fifteen data sets with & 2/t I fixed as described above. Anrcng the eleven data scts included in the final analysis, STD/mean ranged from 0. ll to 0.21. Armng the four data scts drop@ from the analysis, STD/nrcan rangcd from 0.33 to 0.41. Thrcc of the four nran ,t r values calculated for these data s€ts deviate by a large anrcunt from thc other €levcn. Also, whcn plotted, these four data scts lmked "funnyn. All this confirmcd the dccision to drop these four data scts from the analysis of kinctic data.

Forrrrd
Unchelated nickel was choscn as the mmt promising catalyst after various transition metal complexes had becn screcned. This screening work, described here, was completed bcfore the intensive study of the nickel catalyzed reaction bcgan (Chapter 3).
Bccause of autocatalysis, a 'once througho, flowing kinetic apparatus is unsuited for studying the nickel catalyzed reaction. Bcfore the role of autocatalysis was appreciated, such an appsratus had bccn developd and applied to studying thc oxidation reaction catalyzed by iron compounds. lt is adequatc for this purp6e, because autocstalysis is much less pronounccd with iron.
Cobalt also is a potent catalyst. Although cobalt is not a serious practical contender for geothermal applications because of cost, the kinetics of the cobalt catalyzed reaction were studied briefly, using the same methods as with nickel.

Experimntd mthods
In "continuous recharge experiments" reactivity, turbidity and reaction product distribution were determined as with nickcl.

t$cthod for scrccairy crtalysb
The experimcntal method used in the early screening work was the precursor of the continuous rechargc method dcscribcd in Section 3.2.2.
Tenth-molar catalyst stock solutions were prepared. The sodium sulfide stock solution was about 0,2 M, and was always standardized before use. Qre-quarter liter of synthetic cooling water was prepared in an Erlenmcyer flask and prcheated to 45'C in a thermctatted water bath. ln this work, the SCW contained 1.70 mM /VH 3 plus lVfil.+ , 0.786 mM HCO j-. 14.813 mM I (OIf )3 plus H zBO f , 1.70 mM Cl -, and 1.572 mM lVa +. lts pH was between 7.8 and 8.0.
At the beginning of the experiment there was no catalyst in the SCW. With stirring, ffa $ stock solution was added to the SCW to give 62pM total sulfide (typically, about 80pl of stock solu. tion). A 25-nrl aliquot of SCW was withdrawn by pipette and mixed with 25 ml of SAOB, and the conc€ntration of sulfide was measured with a sulfide electrode. This procedure vcrified propr sulfide stock concentration and electrode calibration.
To the remaining 225-nl of SCW catalyst stock solution was added to give the catalyst con-c€ntration desired, typically betwecn l0 and 100 FM. The SCW was acrated for about l0 scc, and allowed to stand lor about 3 min. With stirring, sulfide stock solution was again added to give 62 pM sulfide, allowing for the now smaller volume of SCW. A 25 rnl aliquot was withdrawn by pipctte. One minute after addition of sulfide, 25 ml of SAOB was added to this aliquot of SCW, and the sulfide content measured. The pH of the remaining SCW was measured. This cycle was repeated six times, with decreasing amounts of No 25 stock solution to compcnsate for the decreasing volume of SCW.
ln some experiments, aliquots of l/a 2SO I solution were added to the SCW in altcrnation with aliquots of No f . These experiments dcmonstratcd that a nickcl bearing solution could be 'clarified", and that gmd reactivity without accumulation of colloidal sulfur could bc achicved only within a ccrtain range of Na 2SO 3 to lVa f mole ratio. At that point, the method describcd in Section 3,2.2 for substituted for the one described here.

CONTINUOUS FLOW KINETIC SYSTEM
The flowing solutions arc mixcd in plastic "mixing blmks" (Figurc 4. l0). Each blak contains a miniaturc magnetic stirring pcllet, and rests upon a submersible, air driven magnetic stirring pad. (Unstincd "T-block" were found to Sive inadequate mixing duc to laminar flow through them.) The SCW strcam is preheated in a coil of thin-walled, lmm I.D. Tcflon tubing. All mixing blocks and coils are imnrcrsed in the water bath.
The catalyst solution is added to the aeratcd SCW in thc first mixing block. The first delay coil (lmm I.D. Teflon) providcs time for reaction of thc catalyst with dissolved oxygen. With chelated iron(ll), this allows the effect of "catalyst oxidation tirn€" to bc studied. lt was found that changing this time interval had no effect, and that thc initial oxidation statc of complcxed iron did not influence its catalytic potency. After this was realizcd, the catalyst was put into the SCW, and the first mixing block and dclay coil were climinated.
Sodium sulfide stock solution was added to thc SCW in the second mixing block, typically introducing 62pM sulfide. The lcngth of the second dclay coil determined the time available for thc oxidation reaction. Either reaction time could bc varied by substituting a dclay coil of different lcngth. Thc oxidation reaction was qucnched in thc third mixing block by adding an equal volume of SAOB to the SCW. Thcn the mixture flowed past the reference and ,{g $ electrodes, which were mounted in suitable "electrode blocks". The pH was measured by temporarily switching the flow of SCW to a pH electrode. x8L 827.t591

59-
Thc pcristaltic pump is the rnajor source of troublc in this system. The tubcs gradually distort or even collape, causing the outFlt of the pump channcls to vary. This forccd tcdious daily calibration and frequcnt tubc rcplaccmcnt. Numerous cxprimcntal runs wcrc ruincd by a suddcn change in pumping ratc in one or anothcr channel. This problem could bc climinatcd by substituting a prF cision multichannel syringe pump for the peristaltic pump, Aside from this problem, the systcm is basically a gd onc. The intrinsic modularity of thc various blocks and coils allows thcm casily to bc reanangcd to modify thc expcrimcnt. The timc rcsolution of the system could bc incrcascd by using miniaturc clcctrodcs instead of standard sizcd ( l-cm diameter).
Dctails of construction, solution formulation, and calibration arc reviewed in Appendix 4, 1.
,Ll Scrccd4 of potcrtid crldysb Using the rncthod describcd in Section 4.2. l, rnst of the first row transition metals wcre scrccned for catalytic activity. The rcsults of thesc tests art summqrized in Tablc 4. 1. Cstalytic activity is indicatcd if the rcsidual sulfide concentration stabilizes after a few cyclcs. The lowcr this limiting conccntration value, the stronger the catalyst. l-ack of catalytic activity is indicatcd by a continuing increasc in rcsidual sulfide conc€ntration, which indicates the accumulation of unrcactcd It tS from cycle to ryclc. Evcn in thc worst casc (zinc), thc incresse from cycle to cycle is much lcss than 62gM, bccause I/ 25 is lct to the atrncphcrc. Prccipitation of metal sulfides may also rcducp the sulfidc conccntration, The infcrrcd ordcr of catdytic activity is: Nr' +2 > b +2 > Fccitratc > Cr +J,cu +2 The othcr spcies testcd arc weakly catalytic or not at all. The nickel data clearly show auto caulysis. and autocatalysis is suggcsted with cobalt as well. In tests with nickel the SCW was ycllow after the fint cyclc, and bccam€ turbid with further cycles.
Hydroquinone at a conoentration of 2mM is also a potent catalyst, inducing an autocatalytic reaction (not shown). The solution turns vcry dark as reactivity dcvelop. Hydroquinone is known readily to form rclatively stable frce radicals. Probably, these react with orygcn and H $ as do€s the cocatalytic spcies inferred in the case of nickel, and polysulfido chains grow out from thcm. fu these radicals accumulate, thcy Sive the solution its dark color.

4.,L Wrclcnrodcr's rctction
Hydrogcn sulfide, in which sulfur has the valencc -2, and sulfur dioride, in which the valcncc is *4, rcact to produce a mixture of compounds of intermcdiatc valcncc state. For examplc, if the ratio of SO 2 to fl $ is 2, the main product will be thiculfate, in which the forrnel valcncc o[ sulfur is t2. At lower ratim of SO 2 to It rS , thcre will also bc formcd polysulfides, colloidal sulfur, and higher potythionates (above tetra-), in all of which the average formal valcncc is < 2. This rcaction is called Wackenrodcr's reaction, its product is called Wackcnrodcr's solution, and it is uscd industrially in the Claus process. Wackenroder's resction itself dastroys It $ in competition with oxidation. When working with weak catalysts (e.9. iron compounds), it is conccivable that Wackcnroder's reaction will make a significant contribution to the overall ratc of disappearancc of // f , thcreby confusing interpretation of thc cfrect of the catalyst.
To test this hypothesis an expcriment was run with no catalyst at all. The concentrations of rcactants and products vs. time arc pcscnted in Figurc 4.2. The solution was clear yellow, indicat. ing the prcsencc of polysulfidcs and/or polythionat€s, and the abscncc of colloidal sulfur. Undcr thcse conditions thc cxtcnt of reaction is limitcd by the armunt of sulfite introduced. Most of the sulfite is convcrted to thimulfatc by reaction with part of the H 25 , part of the unreactcd f/ 25 accurnulates in solution, whilc the rcst is lct to the atrncphere. The trithionate probably was formcd by reaction of sulfite with thimulfate. Some sulfate probably was fornrcd by oxidation of SO r-, but it was not dctcrmincd.
Clearly, this reaction's ability to destroy ll $ under these conditions (SO z:f/ aS < l) is tm small to mimic rapid oxidation, and tm snull for it to serve as a means of /J 25 emission abate-rn€nt by itself.
Figurc 4.3 comparcs thc ratcs of rcaction catalyzed by iron citrate and iron EDTA lron citrate is the strongcr catalyst, and thc initial oridation statc of iron makes no differcncc.
Thc catalytic action of iron conplexa rnay includc cycling bctwecn the two oxidation states; for cxample, oxygen reacts with iron(ll) to givc iron(lll), and then iron(lll) rcacts with I/S-to give f/S ' and iron(ll). If so, the r8te of the rcaction of iron(ll) with oxygcn may affect thc ovcrall rate of // f oxidation. This was tartcd by varying the armunt of time allowcd for reaction of the catalyst, initially in the fcrrous statc, lvith oxygen bcfore adding No 25 (Figure 4.4). Varying the ocatalyst oxidation time" has no eflbct, and this argues against such a cycle.
Precipitation of iron(lll) hydroxidc was rrKrre comrmnly obnerved with complexcs of iron(ll) than with iron(lll). Consequcntly, work with iron(ll) complexes was tcrminat€d once it was established that initial oxidation state makcs no differcncc. Apparcntly, thc iron (ll) complcxes are weak cnough for part of the iron(ll) to bc unchelated, This unchelated iron(ll) oxidizes to unchelated iron(lll), which precipitates as iron(lll) hydroxide bsfore it can rcact with the fre€ chelating agcnt. Apparently, iron(lll) complexcs arc stable enough to pr€vent the prccipitation of iron(III) hydroxidc.
The data in Figure 4.5 were obtained using the methds describcd in Sections 3.2,1 and 3.2.3. There is little evidcnce for autocatalysis, and thc catalytic activities of iron citrate and iron HEDTA appcar e4ual. The reactivity data in Figure 4.5 is consistent with the citrate data in Figure 4.3. This supports thc validity of the two cxperimental mcthods, and argues against a significant role for autocatalysis, since there is no opportunity for it to dcvelop in the continuous ffow experiments. The diffcrence in turbidity#ay havc becn caused by the precipitation of iron hydroxide from the SCW containing Fe HEDTA 'HEmAt$hydrorycthylcncdiaminctriacctic acid, EDTA -cthylcncdiaminctctraa€tic acid, DTPA : dicrhylcnrtri. amj ncpe.ntsacctic aci d ('Si anrsc twi ns' of EDTA).
This was accidental. ln this crprirrnt, thc catalyrt rcchargc rvas in thc sulfuric acid solution, as had bccn rhc prsctioc with unchclEtcd nickcl. ln thc acid solution, thc iron(l1l) wa: not chclatcd, and thus availablc for prccigritarion whcn it entcrcd thc SCW. In thc crprirtlnt with ciiratc, thc catalyst was addcd as a ncutral solurion of ir alonc, through a rcpratc prmp charncl. "l/2 SzOl= "l/3 SgOe= 6,1 -In general, the rcpcatibility of reactivity and turbidity data from experiments with iron corn' plexes was less than with nickel. Probably, this was because the lcs of H :S to the atmosphere was relatively more significant with the nnre weakly catalytic iron complexcs. Cnod kinetic data could be obtained with the continuous flow kinetic system, because it is clced and does not l6e ff $ to the atmcphere, and because autocatalysis is not a significant factor with iron. Figure 4.6 illustrates the distribution of resction products with iron citrate and no sulfite added. Thiqullate is the major reaction product, and little colloidal sulfur is produced (although the SCW is turbid).
These limited data do not allow a detailed reaction mechanism to be inferred.

Crtrlysis by cobdt
Uncomplexed cobalt is nearly as potent a catalyst for f/ f oxidation as uncomplexed nickel, while cobalt citrate is more potent still (Figure 4.7: compare with Figure 3.4). However, the variation of residual fraction with initial H 15 concentration is different. That the apparent kinetic order in ff 25 is < I suggests that the catalytic effect of cobalt can be saturated.
The reaction catalyzed by cobalt is very strongly autocatalytic (Figure 4.8), and the SCW has a very dark, reddish-black color. Bccause the SCW is nearly opaque, turbidity cannot meaningfully bc mcasured. and the color of the SCW interferes with colorimetric determination of colloidal sul. fur, thiculfate, and trithionate. Thiosulfatc is the main reaction product (not shown).
Adding a modest amount of magncsium chloride to the SCW causes a dark precipitate to form, leaving a clear yellow supernatant. This destroys the reactivity of the solution. Thus, both the color and reactivity of the SCW arc associated with colloidal particles suspnded in it. These probably consist of amorphous cobalt oxysulfide, and the oxidation reaction probably is catalyzed by their surface. This would explain catalyst saturation phenomena. The beneficial effect of citrate ion has not bcen explained.  As wirh nickel, the conccntration of dissolvcd oxygen docs not affcct reactivity (Figurc 4.9), nor docs thc pH (not shown). Ovcrall, this data suggcsts that with cobalt thc radical initiation step takcs placc on the surface of colloidal particles, while the rcst of the rcaction nrcchanism is similar to that with nickcl.
Snrall changes in tubc configuration in the cassctte, combincd with thc gradual fatiguc of the plastic causcd thc pump rati6 to vary from day to day. Bccausc of this, thc system necded to bc rccalibratcd daily. A solution of methyl orange with an optical density of 1.2(X) at 450 nm was fcd succcssively into cach pump channcl, whilc the othcrs pumped watcr. The actual pumping ratic were detcrmined by mcasuring the optical dcnsitics of thc rcsulting mixturcs. To rcduce pump tubc deformation (flattening), the cassettcs were unclampcd at the end of cach day and rc'clamped thc following morning.
The miring and elcctrode blaks were machincd from Lucite, and thc rnachined surfaccs polished to make them transparcnt. Thc inlct and outlct ports of the blocks wcrc plastic scrcws fittcd into threadcd holes in the Lucite. The scrcws had holes drilled through them, into which thinwalled lmm l.D. Tcflon tubing fit snugly. Figure 4.10 is a scalc drawing of a mixing block. The rnagnctic stining pellct was dcsigncd for use insidc spectrophotornctric cuvettes. It was a slott'd, Teflon cstcd cylindcr 9mm diarnctcr x 6mm tall. The cavity in the mixing blak was clced off above the pcllct with a mcchincd Tcf,on plug.
Thc electrode block used with thc /€ tS and rcfercnce electrodcs is diagrarurrd in Figurc 4.11. Thc pH elcctrode block was similar, execpt that its cavity had a conical bottom to povidc rmm for the electrode's glass bulb (not shom). The anulus around thc elcctrodc burcl was tightly sealed with two Grings. A similar but larger blak was made for thc dissolvcd orygcn probc (not shown). The electrodcs were calibratcd each day after the pumping ratic had bc€n d€tcrmined. A special solution of SAOB with a small, known conccntration of ffa 1S was prepared. Deaeratcd water was pmpcd through the othcr thrcc channels. Knowing the pumping rati6, it was possiblc to rclate thc conccntration of sulfide in thc SAOB to a virtual concentration of sulfide in the reaction strcam (i.e.. SCW+catalyst*iVa rS ), which would givc the same conccntration in the electrodc chamber and. thereby, the sarne elcctrode response. In practicc, thc amount of sulfidgadded was calculated so that the virtual sulfide conccntration in the reaction stream would be l0'/'x 62 tM or 196 rrM. A sccond calibration standard was prepared by diluting part of this solution ten-fold with additional SAOB. Flowing thcsc tso standards through the SAOB channel and recording the corresponding electrode responscs allowcd thc clcctrodc calibration equation to bc determined.
Using the mcasured pumping ratic, the l{a 25 and catalyst solutions were prcpared to give the conc€ntrations of sulfide and catalyst desircd in the reaction stream.
fuioxic solutions of No 25 , iron(ll) compounds, and SAOB wcre preparcd as described in Appcndix 3. 1. The oxygen ontcnt of solutions prcparcd in this way was sometimcs measured using a dissolved oxygen probe fitted with a suitable, flow-through electrodc block, Dsolved oxygcn con. entrstions less than 0.2 ppm (the scnsitivity of thc probc) could routinely be achievcd. Tess with deaerated water in all channels showed about lppm orygen in the water leaving the system, probably duc to diffusion of oxygcn through the thin walls of the Teffon tubcs that werc used as conncctors and delay coils.

CI{APTER 5 SI.'MMARY AI{D TXSCUSSTON
The rate of the nickel catalyzed reaction of H 25 with oxygen has bcen determined over a range of conditions, and an empirical rate expression that fits the data well has bccn derived. This rate expression allows meaningful extrapolation to other conditions, including thce expected in the ficld, The reaction products also have been determined and explained well cnough to allow estimates to bc rnade for other conditions. Because iron complexes are much weaker catalysts than nickel or cobalt, the iron catalyzed reaction could not propcrly be studied using our methods. However. we found little evidence that the iron catalyzed reaction is fundamentally different from that catalyzed by nickel. fuide from much lower catalytic activity, the major difference observed was that less colloidal sulfur formed with iron (compare Figures 4.5 and 6 to Figures 3. I and 3. l0). In the absence of sulfite, formation of thisulfate and sulfate protrably involves oxidation of colloidal sulfur initially formed, The accu. mulation of colloidal sulfur will be determined by the balance between the reactions that produce and rernove it. Because the oxidation of f/ $ (which produces colloidal sulfur) is much slower with iron than with nickel, less accumulation of colloidal sulfur is to be erpected.
In the field, the oxidation reaction takes place in the anoxic part of the cooling water lmp, and forrnation of colloidal sulfur probably is favored with iron, tm. ln fact, opcration of Gcysers Unit ll with iron HEDTA but without SO 2 caused extensive depcition of sulfur scale in the watcr distribution trays, Ers had happned with nickel in Unit 2 (PG&E DER private communication). Overall, we believe that there is little practical difference betwecn iron HEDTA and nickel in this application, and that what is learned about one system can, in large mcasure, bc applied to the other.
In contact condenser equipped Units at The Geysers, the Il 25 in the condenser vent-gas will be convcrted to SO 2 by a burner-scrubber or equivalent device, and this SO 2 added to the cooling water. Enough caustic soda will bc added to the cooling water to maintain pH at about 7.0. At this pH value, about half the H 25 will go with the vcnt-8,as and bc convertd to SO 2, while the othcr half dissolves in thc cooling water. This portion will necd to be eliminatcd by catalyzed air oxidation in the cmling water.
Field conditions diffcr from our experiments in several resp€cts: (l) Typically, about 90 seconds is availablc for reaction bctween condenser and cmling tow€r, not the 15 seconds used in our work.
(2) The cmlinqwater tempcrature between condenser and cooling towcr typically is 48oC, rather than the 35"C used in our reactivity determinations. (3) With contact condensers, the coolinS water is anoxic bctween condenser and cooling tower.
Using OXTAB, we estimate that in 90 seconds at 48oC, lOpM nickcl in the cooling water (0.6 ppm) will eliminate 9EVo of the Il $ dissolved in it. This amounts to about 99% abatement overall. That the cooling water is anoxic bctween condenser and cmling tower should not,affect this conclusion, as the rate of reaction is independent of oxygen conc€ntration (Section 3.4).
That the coling water is anoxic between condenser and cooling tower will affect the distribution of reaction products. Mmt of the sulfate produced in our experiments comes from oxidation of sulfite (Section 3.6). This parasitic oxidation reaction is favored, bccause the SCW is fuily aerated at all timcs. By consuming sulfite, it incrcases the sulfite:sulfide ratio needed to prevent accumulation of colloidal sulfur. In the field, the SO 2 probably will be added to the cooling water in the anoxic part of the system, where it is necded to prevent the formation of colloidal sulfur. There it will have the opportunity to react with the sulfur produced by oxidation of H 2.S before being oxidized to sulfate itself.
The reaction producs formed determine the amount of alkali that must be addcd (if any) to maintain an acceptable cooling water pH. The less sulfate produced, the less alkali necded. Reaction of bisulfite with thiculfate to produc€ trithionate also will hclp decrease the pro<iuction of 70sulfate, and the need for alkali.
Test work at Ccyscrs Unit l, using iron HEDTA has shown that acc€ptsble mling water pH can sometiln€s be maintained without addition of caustic soda (Yancey,l98l). This suggests that thimulfate and trithionate are, in fact, the major reaction products, with little or no sulfatc pro duced. lf 0.6 ppm Ni is indecd sufficient and little caustic soda necd bc added, thc chcmical cost of nickel/So 2 prccess will be negligible. That the prms using iron HEDTA works very well suggcsts that the prooess utilizing nickel also will work, probably with much lower chemical costs.
fu far as could bc detcrmined, there is no diffcrcncc at all bctween iron HEDTA and iron citrate. Because iron citrate would be cheaper to usc than iron HEDTA iron cirrate is to be preferred.
The reaction catalyzcd by cobalt app€ars to involve rcsction on the surfacc of colloidal particles, probably eompced of cobalt oxysulfide. For this rcason as well as cct. cobalt is unsuited for use in power plant cooling watcr. Howcver, the catalytic particles are easily rcrnovcd by flocculation, and this suggests use of cobalt in cascs where thc cstalyst can and must bc rccovcred from thc treated water.