P R E P R I N T – ICPWS XV Berlin, September 8–11, 2008
A Fundamental Study of Flow-Accelerated Corrosion in Feedwater Systems D.H. Listera, L. Liua, A. Feichta, M. Khatibia, W. Cooka, K. Fujiwarab, E. Kadoic, T. Ohirac, H. Takiguchic and S. Uchidad a
University of New Brunswick, Fredericton, New Brunswick, Canada Central Research Institute of the Electric Power Industry, Yokosuka, Japan c Japan Atomic Power Co., Tokyo, Japan d Japan Atomic Energy Agency, Tokai, Japan Email:
[email protected]
b
A once-through water loop operating under typical power reactor feedwater conditions is being used to determine the effects of water chemistry and flow on flow-accelerated corrosion (FAC). Electrochemical corrosion potential (ECP) and corrosion rate are measured with on-line probes and mechanisms are indicated by detailed surface analyses. The program so far has investigated FAC at 140°C in both neutral and ammoniated water over a range of flow rates. In neutral water, FAC was correlated well with fluid shear stress at the surface and was stifled with oxygen concentrations of about 40µg/kg (ppb). Chromium in the steel reduced FAC significantly. Corroded surfaces developed thin magnetite films and some overlaid typical “scalloped” textures. Scallops were influenced by the microscopic oxide structures that developed from underlying metal grains; in particular, pearlite grains formed lamellar oxides that predominated on scallop crests. In ammoniated water at a room temperature pH of 9.15-9.35, FAC was extremely sensitive to traces of oxygen and apparently proceeded by a “front” of protective oxide, based on haematite, moving downstream as the less protective magnetite was progressively oxidised. The subsequent reduction of a more oxidised, stifling film back to magnetite and the resumption of FAC when oxygen levels were reduced also proceeded from the upstream ends of probes.
Introduction Flow-accelerated corrosion (FAC) of carbon steel is a common problem in many types of steamraising plant. The catastrophic failure of a suction line to the main feedwater pump at the Surry-2 PWR in 1986 and of a feedwater line at the Mihama-3 PWR in 2004 resulted in fatalities; similar accidents at fossil plants have been reported [1,2]. Not surprisingly, further insights into the mechanisms of FAC are being sought in the search for mitigating techniques. To that end, a collaborative research program between Japan and Canada is investigating the details of how dissolved oxygen reacts with oxide films on carbon steel undergoing FAC under PWR feedwater conditions. In a laboratory loop at the University of New Brunswick (UNB) in Canada, on-line probes indicate electrochemical corrosion potentials and corrosion rates while analyses of exposed specimens are carried out at UNB and at the Central Research Institute of the Electric Power Industry (CRIEPI) in Japan. The results are reviewed by the whole team, which includes representatives from the Japan Atomic
Power Company (JAPCo) and the Japan Atomic Energy Agency (JAEA). The mechanisms leading to a threshold concentration of oxygen, above which FAC is inhibited or “stifled”, are being sought in order to specify a practical value that may be used in operating power plants. Experiments The experimental water loop has been described before [3]. It is made mostly of Hastelloy-C and stainless steel and can operate at temperatures up to 310°C and pressures up to 11MPa. Although designed as a once-through system, its coolant is recirculated from the pressure-reducing valve to the positive-displacement pump via a cooler, ionexchange columns and a controlled-atmosphere tank where the chemistry is adjusted with ammonia, hydrazine, etc. The pump can deliver up to 3.5 L/min of coolant to the test section, which is fitted with by-pass piping to enable probes to be inserted or removed without shutting down the whole loop. Oxygen is added as required via a pump injecting solutions in water to the low-pressure coolant return line before the tank, where an on-line Orbisphere
Figure 1. Schematic diagram of FAC loop
about three days after measured oxygen had fallen to zero. Three runs are described here. Run 1 used a carbon steel (CS) with 0.019 wt % Cr, the other two
measures its concentration continuously. A schematic diagram of the loop is presented in Figure 1. Rates of FAC are monitored on-line with tubular probes made of the carbon steel of interest. The tubes are typically 90 mm long with bores of 1.6 mm, 2.4 mm or 3.2 mm and they have the middle length of about 12 mm turned down to a wall thick ness that gives an electrical resistance that can be measured accurately; changes in resistance of 10 µΩ (±4%) can be detected reliably (see Figure 2, which shows a typical resistance probe with its restraint to avoid disassembly under pressure). The rate of increase of resistance as measured with a multimeter then gives a measure of the FAC rate as the wall thins. Electrochemical corrosion potential (ECP) of a resistance probe relative to a hightemperature Ag/AgCl reference electrode is also measured. Two or three probes are typically installed in series in the test section and tubes of similar sizes are installed downstream to be removed as required for surface analysis. Stainless steel probes designed solely to measure ECP were installed in TS 1. It was noted that during the commissioning of the loop, when the coolant initially was saturated in air, ECP finally levelled off at about – 215 mV (SHE), but only
Figure 2. Diagram of resistance probe runs used a carbon steel specially made with 0.001 wt% Cr. All runs were at 140 °C, the temperature of the feedwater line that ruptured at Mihama 3 and close to that known to produce the maximum rate of FAC. Runs 1 and 2 were at neutral chemistry
2
range of temperatures using a temperaturecontrolled furnace, the measurements of electrical resistance were readily converted to wall thickness and the inner radius calculated. The slopes of plots of radius versus time indicated FAC rate. Similar plots were obtained from the other runs. The radius plots from Run 1 shown in Figure 3 indicate that the response to changing flow rate was rapid. In some experiments under similar conditions, where FAC rates are high, the plots tend to become less steep with time as the probe bores increase and the coolant linear velocity decreases. This effect is not evident in Figure 3. The effect of coolant temperature is evident, however, since shifts in the data occurred with inadvertent heating or cooling. The slopes of the plots – and therefore the indicated FAC rates – were unaffected. The effect of added oxygen is also evident in Figure 3, which shows the stifling of FAC for about two days when the ion-exchange column was replaced with a fresh, air-saturated one at Day 14. Measurements of ECP were less clearly defined. Ingress of oxygen into the loop, particularly during runs under ammoniated conditions, controlled the ECP. The ammonia affected many of the polymer seals in the loop; they degraded and became permeable to oxygen and had to be replaced with more resistant materials. Under neutral conditions, oxy-
conditions while Run 3 had ammonia added to give a pH25°C between 9.15 and 9.35. In Run 1, probes with three internal diameters of 2.4, 1.6 and 3.2 mm (in order downstream) were installed in each of Test Sections (TSs) 1, 2 and 3. Those in TS 1 were resistance probes to measure FAC rate on-line, those in TS 2 were for surface analysis at the end of the run and those in TS 3 were for surface analysis after each change in loop parameter such as flow rate (the TS 3 probes were replaced at each change). The effects of flow rate were studied. In Run 2, probes of the lower-Cr steel with internal diameters of 2.4 and 1.6 mm were installed in TSs 1, 2 and 3. The effects of concentration of added oxygen on FAC were studied and indications of flow-rate effects were obtained to compare with the results from the higher-Cr steel in Run 1. In Run 3, probes of the lower-Cr steel with internal diameters of 2.4 and 1.6 mm were again installed in TSs 1, 2 and 3. At an average pH25°C of 9.2 with ammonia, the effects of concentration of added oxygen on FAC rate were studied. Results Since the steels used for the probes were separately calibrated for electrical resistivity over a
Figure 3. Plots of probe radius versus time for Run 1.
3
Figure 4. SEM pictures of probe inner surfaces after Run 1.
cementite and ferrite lamellae in the pearlite form oxides of different composition. As FAC proceeds by the accepted mechanism of the protective magnetite film’s being thinned by dissolution in turbulent coolant undersaturated in dissolved iron, so that it reaches a steady-state thickness at a constant FAC
gen ingress was much easier to control. Seals were effective and the high FAC rates, driven by high iron solubility, appeared to act as a sink for oxygen. On the CS resistance probes, even when the oxygen may have been too dilute to affect the FAC, it had a measurable effect on the ECP. In general, the ECP plots for the CS probes varied in concert and fell with increasing FAC rate – as expected. Values between -610 and -950 mV (SHE) were measured in Run 1. The examination of the probe surfaces after Run 1 revealed a distinct effect of coolant flow on the morphology (see Figure 4). Oxide “stringers” were aligned axially and scallops were evident – the latter in particular on the more corroded surface of the 1.6 mm probe. The oxides were identified as magnetite with laser-Raman microscopy. A more detailed scrutiny of the oxides indicated that the FAC mechanisms had promoted unusual oxide formations. As indicated in Figure 5, “corallike” growths were associated with the stringers and with the scallop crests. Metallography of a probe cross-section and its oxide showed that the growths occurred over pearlite grains in the metal, so it is assumed that the
Figure 5. Oxide on scallop crest in Run 1 rate, the oxide over the ferrite lamellae will dissolve at about the same rate as that over the surrounding ferrite metal grains, leaving the oxide over the cementite to form the “coral”. Presumably, as these oxide features develop, the flow patterns of the coolant at the surface are modified so that the local turbulence perpetuates the scallops. The magnetite 4
ammoniated conditions described later. The onset of stifling and the resulting removal of an oxygen sink caused the oxygen concentration to jump to 120 ppb. It should be noted that oxygen had to be completely removed – to zero concentration – for a day or so before FAC could be resumed. This hysteresis in the effect of oxygen no doubt reflects the interaction with the oxide film on the probe surface. It is postulated that for stifling of FAC to occur, the oxygen has to penetrate through the pores of the magnetite to the metal-oxide interface, where it must achieve some minimum concentration in order to react with the Fe2+ ions as they are produced and form a more protective oxide based on Fe2O3 (maghaemite or haematite) rather than Fe3O4 (magnetite). In diffusing through the magnetite, it will also react with the Fe2+ in the oxide surface and may never reach the metal if the driving force is too low. Conversely, when a surface is stifled with a passive metal-oxide interface protected with Fe2O3 and the coolant becomes depleted in oxygen, the oxide has to be reduced to some extent before the production of Fe2+ ions can recommence at the metal. Laser-Raman studies of the surface of a probe removed during the period of oxygen injection but before stifling (when the concentration was about 20 ppb) in fact indicated the presence of haematite. During Run 2, the ECP varied roughly as the oxygen concentration, with the 2.4 mm probe showing the largest variation (between about -1 V vs. SHE at zero oxygen to about -170 mV at the peak oxygen level); a somewhat higher potential of the 1.6 mm probe was attributed to its different location in the loop relative to the reference electrode. Curiously, although the FAC rates during the non-oxidising phases were higher than in Run 1, scallops were not seen on the probes. Moreover, the oxides showed little tendency to align with the flow direction, though SEM examination at highmagnification showed coral-like oxides similar to those associated with underlying pearlite grains that were prevalent in the scallop crests in Run 1 (see Figure 7). This lack of tendency to produce scallops cannot be attributed to a different metal grain structure of the low-Cr steel, since metallography indicated that it was similar to that of the higher-Cr steel. In particular, the pearlite grains were distributed similarly, although those of the lower-Cr steel in Run 2 were slightly more elongated. In any case, as described later, localised severe scalloping was found
in the scallop troughs was very fine-grained (see Figure 6). On average, the base oxide was 0.5 – 1.0 µm thick, which is customary for FAC situations.
Figure 6. Oxide in scallop trough in Run 1 Energy-dispersive X-ray analysis (EDX) indicated that trace elements from the underlying metal were concentrated in the oxides on the probes exposed in Run 1. In particular, Cr was concentrated to a ratio with Fe of 3.35×10-3 in the crest oxide and to 2.01×10-3 in the valley oxide; by contrast, the underlying metal had a Cr:Fe ratio of 1.92×10-4 (EDX of such thin oxides has an inherent uncertainty because of interference from the underlying metal; however, the interference is expected to produce low values if anything, so that the concentration effect should be real). Note that most of the Cr in the oxides probably came from the underlying metal, since the main loop surfaces that might release Cr to the coolant (i.e., the stainless steel and Hastelloy components) should also release Ni, yet no (or only small amounts of) Ni were found on the probes. Chromium is known to impart protective properties to carbon steel in FAC environments [4], presumably by making the oxide more compact and less soluble. In Run 2, the two probes (2.4 and 1.6 mm – in order downstream) produced radius plots similar to those of the higher Cr steel in Run 1, except that the FAC rate was about 2.5 times higher. This comparatively rapid increase in radius lowered the linear velocity of the coolant in the probes significantly during the run, so that the FAC rate declined with time. An average FAC rate for a period at an average velocity is therefore quoted here. This allowed velocity/diameter effects to be examined in more detail for this run than would otherwise be possible with only two nominal probe sizes and a constant volumetric flow rate. Part-way through the run, oxygen was injected in increments to find the concentration at which the FAC was stifled. At about 40 ppb, the upstream probe stifled to be followed by the other, downstream, probe about four hours later. This apparent movement of an oxidized front moving downstream was seen in Run 3 under 5
point has been included. By contrast, Run 1 had 32 flow conditions and Run 2 had four. The correlation is based on the premise that the FAC of carbon steel is mass-transfer controlled, which in turn is based on the postulate that there are two processes in series – the corrosion at the metaloxide interface creating the oxide film there by half the iron that enters solution precipitating as magnetite (the rest diffuses through the film and is lost to the bulk coolant), followed by the dissolution of the film at the oxide-coolant interface by coolant undersaturated in dissolved iron. This leads to the equation for FAC rate R [5]: 2hk d ΔC (1) R= (h + 2k d )
on this material exposed under ammoniated conditions, when general FAC rates were lower.
Figure 7. Oxide feature on low-Cr steel in Run 2
where ΔC is the undersaturation in iron, kd is the dissolution rate constant and h is the mass transfer coefficient. If mass transfer controls, h is small compared with kd and Equation (1) reverts to: (2) R = hΔ C so that R varies as h for a given chemistry condition. For a given flow condition, the differences in R for different materials (e.g., the lower- and higherCr steels in these experiments) are then caused by the different equilibrium concentrations of dissolved iron (i.e., solubilities of the oxide films) in
In Run 3, under ammoniated chemistry at pH25°C of 9.2 with no oxygen, the FAC rates of the two resistance probes (2.4 and 1.6 mm diameter – in order downstream) of the lower-Cr steel were about half of those under neutral conditions. A flow correlation that has been applied before [3] to Runs 1 and 2 has been applied to Run 3 (see Figure 8). Note that only two probes were installed in this run, so only two flow conditions were obtained. However, since the correlation is expected to pass through (or close to) the origin, the zero
Figure 8. Flow correlation for Runs 1, 2 and 3 6
of oxygen is well-established [9]; stifling concentrations here were within 1-2 ppb for both the 1.6 and the 2.4 mm probe, determined during additions to 10 ppb over four weeks. The upstream, 2.4 mm, probe stifled about two days before the downstream one. After the resistance probes had stifled, downstream surface analysis probes were removed from the loop and examined. It was significant that the more upstream probe exhibited a “front” of haematite-rich oxide part-way along its length. Upstream of the front the probe surface had a thick, red oxide while downstream the oxide was the characteristic shiny black of FAC-induced magnetite. LaserRaman microscopy and SEM confirmed that the upstream, red oxide was haematite, comprising a thick layer of plates and needles several µm across. Presumably, these were Fe2O3 crystals, and they were on top of what appeared from their octahedral shape to be Fe3O4 crystals, again up to several µm in size. These formations are in Zone 1 in Figure 9. Downstream, in the FAC zone (Zone 4 in Figure 9), the by-now familiar coral-like formations of oxide associated with pearlite grains in the metal appeared with the surrounding thin layer of magnetite.
the ΔC term. The effect of Cr on creating more insoluble oxides was commented on earlier. Correlations for mass transfer usually relate the dimensionless quantities Sherwood Number and Reynolds Number (Re) to link the mass transfer to the flow characteristics. For constant physical properties of the system, this leads to the expression for FAC rate: Rd = A Re p (3) where d is pipe diameter and A and p are constants. For many applications, the Re exponent p falls between 0.6 and 0.9 [6,7]. However, the data for Runs 1 and 2 gave values of 1.2 and 1.3 with correlation coefficients of 0.83 and 0.98, respectively. These exponents seem rather high. The three points (including the zero) for Run 3 gave a very high exponent of 4.8. An alternative approach leading to Figure 8 considered the Reynolds analogy of transport parameters in turbulent flow. The Stanton number for mass transfer (Sh/Re/Sc, where Sh is the Sherwood number and Sc is the Schmidt number) is equated to the friction factor (τ/(ρ.u2), where τ is the fluid shear stress and ρ is the fluid density). If the (constant) property terms are neglected and it is again assumed that R is proportional to h, Equation (4) is obtained: Ru = Bτ (4) where B is a constant. The least-squares regressions for the data in Figure 8 indicate that for Runs 1 and 2 the relationships are virtually linear, with exponents of shear stress of 1.06 and 0.97, respectively (the corresponding correlation coefficients are 0.98 and 1.00). Equation (4) apparently holds for FAC under these conditions of neutral chemistry, when magnetite solubility is high and corrosion is rapid. For Run 3, the exponent of 1.52 is rather high to attribute the correlation to Equation (4). This might suggest that the assumption of mass-transfer control is doubtful for high-pH chemistry, when magnetite solubility is comparatively low (about 14 ppb at 140°C under ammoniated conditions at pH25°C of 9.2, in comparison with about 119 ppb under neutral conditions [8]). A similar suggestion might arise from the fitting of the Run 3 data to the Reynolds number correlation, which gave an exponent of 4.8 (mentioned earlier). However, with only two data points, such a conclusion remains speculative. The sensitivity of the loop oxygen control under ammoniated chemistry was mentioned earlier. This was also reflected in the probe behaviour when oxygen was injected in Run 3. That FAC is stifled under ammoniated conditions at low concentrations
Figure 9. Inside surface of 2.4 mm probe after exposure to oxidizing chemistry in Run 3 These observations of surface features are consistent with an oxidation front that moves downstream as the corroding surface becomes progressively protected by the conversion of its magnetite film to haematite or maghaemite by the dissolved oxygen. The thickness of the oxide on the stifled portion of the probe surface suggests that the plates and needles precipitated on top of the magnetite, which itself oxidised rather slowly. Also, the front must have allowed magnetite crystals to form as haematite precipitated on top. Presumably, the oxygen at the surface varied from close to measured bulk concentration in Zone 1 to below the threshold level for stifling in Zone 4. Further downstream, the concentration probably approached zero as oxygen interacted with the FAC process (the “sink”). 7
Figure 10. Upstream (2.4 mm) probe after exposure to oxygenated then low-oxygen coolant in Run 3 (left-to-right: upstream end, middle, downstream end)
Figure 11. Downstream (1.6 mm) probe after exposure to oxygenated then low-oxygen coolant in Run 3 (left-to-right: upstream end, middle, downstream end) creased towards the downstream end. The oxide film on the downstream 1.6 mm probe gave a stronger signal than that on the other probe, suggesting it was in fact thicker, although there seemed to be more haematite on the upstream probe. The reason why neither probe had indicated the resumption of FAC was clear; the severely corroded and scalloped length at the upstream end (about 4 cm for both probes) had not extended far enough into the machined and monitored length in the middle of the tube to affect the electrical resistance. These observations are significant. They indicate that FAC resumed on both probes but only to a certain distance downstream from the entrance. Presumably, the fluid-flow turbulence due to the entrance effect augmented the effect of the morereducing chemistry under low-oxygen conditions to modify the passivating film based on haematite and make it less protective. This might indicate that the reduction of the oxide is mass-transfer controlled; however, both the 1.6 mm and the 2.4 mm probe had the section of resumed FAC extending about the same distance downstream, so the reduction “front” would have progressed at the same rate downstream, even though the coolant velocity and Reynolds number in the 1.6 mm probe were higher than in the 2.4 mm probe. In any case, the mechanism cannot be likened to the one that generated the oxidation front that was described earlier for the oxygen addition experiment, where there was a reactant (dissolved oxygen) that was transported
This observation of an oxidising front moving downstream opens the possibility of adding oxygen to the upstream end of a coolant system undergoing FAC in amounts sufficient to passivate the surface progressively and to be consumed to the threshold stifling level at the downstream end. The passivated state could then be maintained with oxygen concentrations close to the threshold. In order to find out how long the stifled probes would take to resume FAC, Run 3 was continued under deoxygenated conditions (zero-0.5 ppb, with an occasional spike to about 20 ppb when a fresh ion-exchange column was valved in). After running in this mode for about five months, with no resumption of FAC, the two resistance probes were removed from the loop, sectioned and analysed with SEM and laser-Raman microscopy in the same way as surface-analysis probes were examined. Figures 10 and 11 present low-magnification SEM pictures of the inside surfaces. Both were matt-black, although the upstream (2.4 mm) probe had a tinge of brown/red and the 1.6 mm probe seemed to have a thicker film. Surprisingly, both probes displayed severe corrosion and scalloping at the upstream end, visible to the naked eye, in which the metal glinted through the oxide. A rough comparison indicates that the inlet of the 1.6 mm probe corroded to a diameter similar to that of the inlet of the corroded 2.4 mm probe. Laser-Raman analysis along the lengths of both probes indicated mainly magnetite with an admixture of haematite that in8
numerous to mention individually – who have contributed to the laboratory work and to the ongoing technical discussions.
and depleted downstream by reaction with surfaces (separate calibration studies measured gradients in dissolved oxygen around the loop; see [3], for example). Rather, it would have been the slow conversion of a protective, ferric-based oxide at the metal-oxide interface, presumably controlled by the very low solubility of that oxide. More experiments are required to identify the mechanism precisely.
References [1] Czajkowski, C., “Metallurgical evaluation of an 18-inch feedwater line failure at the Surry Unit 2 power station”, NUREG/CR-4868. Brookhaven Nat. Lab. (1987). [2] NISA, “Final report on Mihama-3 secondary system piping failure”, Nuclear and Industrial Safety Agency, Tokyo, Japan (2005). [3] Lister,D.H.,Feicht,A.,Cook,W.,Khatibi,M.,Liu,L., Ohira,T., Kadoi,E., Takiguchi,H. Fujiwara,K, and Uchida,S. “Effects of dissolved oxygen on flow-accelerated corrosion in feedwater systems”. Proc. 13th Internl. Conf. on Environmental Degradation of Materials in Nucl. Power Systems, Whistler, BC, Canada. (2007 Aug. 19th-23rd). [4] Bouchacourt, M. “Identification of key variables: EdF studies”EPRI Workshop on Erosion-Corrosion of Carbon Steel Piping, Washington, DC, USA (1987 April). [5] Berge, P., Ducreux, J. and Saint-Paul,P., “Effects of Chemistry on Erosion Corrosion of Steels in Water and Wet Steam”, Proc. 2nd. BNES Internl. Conf. on Water Chem. of Nucl. Reactor Systems, Bournemouth, UK. (1980). [6] Berger, F.P. and Hau, K-F. F-L., “Mass Transfer in Turbulent Pipe Flow Measured by the Electrochemical Method”, Internl. J. Heat and Mass Trans., 20, 1185 (1977). [7] Sydberger, T. and Lotz, U., “Relation Between Mass Transfer and Corrosion in a Turbulent Pipe Flow”, J. Electrochem. Soc. 129 (2), 276 (1982). [8] Tremaine, P.R. and LeBlanc, J.C., “The Solubility of Magnetite and the Hydrolysis and Oxidation of Fe2+ in Water to 300oC”, J. Solution Chem., Vol. 9 No. 6 (1980). [9] Woolsey, I.S., Bignold, G.J., De Whalley, C.H. and Garbett, K., “The Influence of Oxygen and Hydrazine on the Erosion-Corrosion Behaviour and Electrochemical Potentials of Carbon Steel under Boiler Feedwater Conditions”, Proc. 4th BNES Internl. Conf. on Water Chem. of Nucl. Reactor Systems, Bournemouth, UK. (1986).
Summary and conclusions Experiments on the flow-accelerated corrosion (FAC) of carbon steel have been carried out using on-line electrical resistance probes installed in series in a laboratory loop operating under simulated PWR feedwater conditions at 140 °C. Results have indicated the effects of chromium content of the steel, of coolant flow, of ammonia additions at pH25°C of 9.2 and of oxygen additions to the coolant, on FAC rate. At neutral chemistry, steel containing 0.001% Cr corroded at about 2.5 times the rate of steel containing 0.019% Cr. FAC was apparently controlled by mass transfer, although standard masstransfer correlations linking the rate with Reynolds number did not describe the results as well as one derived from the Reynolds analogy linking the rate with fluid shear stress. Scalloping, influenced by curious oxide formations that were associated with the grain structure of the metal, was seen only on the higher-Cr steel. An oxygen concentration of 40 ppb stifled FAC of the lower-Cr steel. Adding ammonia approximately halved the FAC rate of the lower-Cr steel. Mass-transferbased relations did not describe the FAC rates under these conditions very well, although the data were too few for the conclusion to be definitive. The high-pH chemistry made the loop very sensitive to dissolved oxygen. A concentration of 1 – 2 ppb was sufficient to stifle FAC, apparently via a front of oxide based on haematite (rather than magnetite, which is commonly found on steel undergoing FAC) that progressed downstream. The subsequent resumption of FAC when dissolved oxygen was removed occurred only at the inlet sections of probes, where pronounced corrosion and scalloping were seen. Acknowledgement The authors would like to thank their respective organisations for supporting this research program. In addition, they are grateful to the people – too 9
