Determining water chemistry conditions in nuclear reactor coolants

Transcription

1 Journal of Nuclear Science and Technology ISSN: (Print) (Online) Journal homepage: Determining water chemistry conditions in nuclear reactor coolants Derek Lister & Shunsuke Uchida To cite this article: Derek Lister & Shunsuke Uchida (2015) Determining water chemistry conditions in nuclear reactor coolants, Journal of Nuclear Science and Technology, 52:4, To link to this article: Published online: 28 Oct Submit your article to this journal Article views: 1964 View Crossmark data Full Terms & Conditions of access and use can be found at

2 Journal of Nuclear Science and Technology, 2015 Vol. 52, No. 4, , REVIEW 50th Anniversary Invited Review Determining water chemistry conditions in nuclear reactor coolants Derek Lister a and Shunsuke Uchida b a Department of Chemical Engineering, University of New Brunswick, PO Box 4400, Fredericton, NB E3B 5A3, Canada; b Japan Atomic Energy Agency, 2-4 Shirane, Shirakata, Tokai-mura, Naka-gun, Ibaraki , Japan (Received 8 April 2014; accepted final version for publication 24 September 2014) The chemistry of the process and coolant systems in water-cooled nuclear reactors is tightly controlled to minimise material degradation and, for some systems, to regulate reactor power. Tight control entails monitoring the systems and making appropriate adjustments. Online monitoring can be utilised where instruments are available but otherwise samples must be taken and measurements made offline. This paper reviews the current technologies for monitoring and sampling. Keywords: water chemistry; nuclear reactor coolant; online instruments; sampling water; high-temperature measurement 1. Introduction Controlling the chemistry of the coolants of watercooled nuclear reactors is vital for reactor control as well as for the integrity of the systems. Thus, pressurised water reactors (PWRs) rely on boron additions to the primary coolant to control core reactivity throughout a fuel cycle but have to balance the acidic tendencies of boron with lithium hydroxide to minimise corrosion and corrosion-product transport. Pressurised heavy water reactors (PHWRs i.e., CANDUs) have a rapid shutdown system that involves injecting gadolinium into the heavy water moderator but to ensure that the shutdown will be effective the moderator has to be maintained chemically neutral or slightly acidic. Boiling water reactors (BWRs) control the chemistry of the reactor water; originally, pure water with very low conductivity was specified but now several plants use additives such as hydrogen gas to minimise the oxidising potential and the propensity to crack materials, and zinc to impede the absorption of radionuclides by oxides on system surfaces and consequently to minimise activity transport (the growth of radiation fields around out-core components). Effective control entails knowing the chemical conditions of the systems, which in turn involves monitoring the key chemical parameters. Instruments for remote, online monitoring are convenient; they allow operators to follow system behaviour closely and to keep chemistry under tight control. Unfortunately, not all systems are amenable to online instrumentation. For example, probably the most critical system in the reactor, the primary coolant, operates under conditions too harsh to accept commercial instruments such as a ph meter, yet the ph at temperature is an important operating parameter for PWRs and CANDUs; it must be deduced by extrapolation from measurements of boron and/or lithium concentrations in samples of coolant at low temperature, and this entails employing a chemistry model based on the chemical equilibria of the numerous species involved. Similarly, the concentration of hydrogen peroxide, H 2 O 2, is an important parameter in BWR reactor water, yet the molecule is too unstable to be sampled and in situ measurements have only been made in the laboratory; it is deduced by applying radiolysis models. The ability to sample coolant systems is clearly important. In PWR and CANDU primary coolants and in BWR feedwater, the concentration of corrosion products determines fouling of the fuel and subsequently activity transport, and in secondary coolant feedwater the concentration determines fouling of the steam generators (SGs). To control corrosion products, operators need to know the concentrations under operating conditions, and this entails sampling from strategic places in the system. Difficulties then arise because samples interact with the sampling system itself. This paper describes the chemical parameters of the various nuclear systems and the instruments available for monitoring them, and goes on to review the technology of sampling high-temperature water circuits with Corresponding author. dlister@unb.ca C 2014 Atomic Energy Society of Japan. All rights reserved.

3 452 D. Lister and S. Uchida Table 1. Major targets of water chemistry in NPPs and determining phenomena. (1) Major targets Determining phenomena (1) Reactivity control [thermal hydraulics, chemistry, refuelling] (2) Structural integrity [corrosion, thermal hydraulics, chemistry] (3) Fuel integrity [corrosion, thermal hydraulics, chemistry, fouling] (4) Dose rate reduction [activity transport, thermal hydraulics, chemistry, fouling] (2) Categories of phenomena Determining major parameters (1) Thermal hydraulics [temperature, pressure, flow velocity, voidage, heat flux] (2) Corrosion [species conc., ph, conductivity, temp., flow, neutron flux] (3) Activity transport [radioactivity and metal conc., temp., flow, neutron flux, fouling] (4) Fouling [ph, conductivity, metal conc., temp., flow, heat flux] (5) Chemistry [ph, conductivity, species conc., temperature] reference to developments in techniques for obtaining optimum values of parameters. 2. Instrumentation 2.1. Purposes of water chemistry control The major targets related to water chemistry control in nuclear power systems are summarised in Table 1 [1 3]. Water chemistry control procedures to achieve the targets are supported by understanding the key phenomena, which are affected by many parameters as shown in Table 1. In order to quantify the effects of each parameter on the phenomena and to establish suitable control for the parameters to achieve the targets, water chemistry parameters should be quantified as accurately as possible. A major phenomenon affected by chemistry is corrosion. Nuclear power plants are operated under high temperature and since corrosion is usually exacerbated by temperature quantifying the major parameters under operating conditions is essential for its understanding and control [4] Determination of key parameters related to water chemistry in nuclear power systems The key parameters of water chemistry in nuclear power plants are divided into four kinds, i.e., (1) physical parameters (temperature and pressure), (2) complex chemical parameters that determine corrosive conditions (ph, conductivity), (3) elemental chemical isokinetic sampling (steam or two-phase systems), memory effects Figure 1. Determination of properties of cooling water in NPPs.

4 Journal of Nuclear Science and Technology, Volume 52, No. 4, April Table 2. Control items and chemistry parameters to be monitored. Reactor PWR BWR CANDU Item Primary Secondary Reactor Feedwater Primary Secondary Moderator Temp. ( C) Press. (MPa) ph 25 C (at T Neutral Neutral = 300 C) Reactivity control [B] Voidage Refuelling [B], [Gd] Structural integrity ph, [O 2 ], conductivity [H 2 ], [Cl], [SO 4 ] [D 2 ], [O 2 ] Fuel integrity Activity transport ph T,[B],[Li], [H 2 ], conductivity [Li], [Cl], [Fe] ph T,[ 131 I] ph T,[B],[Li], [ 58 Co] [ 60 Co] ph, [O 2 ], [SO 4 ], [N 2 H 4 ] conductivity Note: DN, delayed-neutron detection; γ, feeder gamma-scan. ph, [Li] [D 2 ], [Cl] [SO 4 ], [F] ph, [ 131 I] [O 2 ] [Cl], [Fe] ph, [ 131 I] DN, γ [O 2 ], [ 58 Co] [ 60 Co] [Fe], [Co] ph, [Fe] [Zn] [O 2 ] [ 60 Co] ph, [O 2 ], [SO 4 ], [N 2 H 4 ][Cl] conductivity parameters that determine corrosive conditions ([O 2 ], [H 2 O 2 ], [H 2 ]) and (4) parameters resulting from corrosion (concentrations and radioactivity of corrosion products), as shown in Figure 1 [2]. There are four approaches to determining the key parameters of water chemistry under high temperature, which define four groups of parameters: (1) Direct determination with high-temperature water chemistry sensors. Electrochemical corrosion potential (ECP) as well as physical parameters, e.g., temperature, pressure and flow velocity, are categorised as parameters of Group (1). (2) Determination with online monitors at ambient temperature. Conductivity, ph, [O 2 ], [H 2 ] and concentrations of some metallic ions and anions are categorised as parameters of Group (2). (3) Determination from sampled water at ambient temperature. Concentrations and radioactivity of corrosion products are categorised as parameters of Group (3). (4) Indirect determination with theoretical tools. The distributions of concentrations of radiolytic species and ECP throughout the cooling system are determined by water radiolysis codes and ECP codes. Any to be determined by the theoretical tools are categorised as parameters of Group (4). Each approach currently has both advantages and disadvantages. Direct determination is restricted to physical properties, e.g., temperature, pressure and flow velocity. An ECP sensor with sufficient lifetime and reliability is still to be developed, while other sensors are still under development for plant application. Determination with online monitors at ambient conditions and from sampled water (Groups (2) and (3)) has problems associated with reducing temperature from the original points of interest to measuring points at lower temperature. The details of the problems and countermeasures are discussed in the section on sampling. Even when the key parameters based on Groups (1) (3) are determined, there are additional requirements to interpolate or extrapolate the data to locations where the measurements cannot be taken. Also, for application to BWRs in particular, some radiolytic species are so unstable at high temperature that determining their distribution and the resulting ECP throughout the cooling systems can only be achieved with extrapolation based on theoretical tools. There are many such theoretical tools for predicting corrosive conditions, but their reliability is often open to question. On the other hand, several instruments are available to indicate corrosion by measuring the interaction with metal components directly, rather than by quantifying water chemistry parameters. The key parameters to be monitored in water-cooled nuclear plants are compiled in Table Instruments for determining the key parameters of water chemistry Proven instruments The reactor control items and the key parameters to be monitored are listed in Table 2 (after [3]). Determination of physical parameters and macroscopic parameters is based on traditional instruments. The reliability of the reference electrode is the key issue in ECP measurement. The determination of macroscopic parameters is based on traditional polarography instruments, while that of H 2 O 2 concentration is based on newly developed luminor luminescence spectroscopy [5]. Boron concentration is determined by neutralisation titration and Li concentration is determined by atomic absorption spectroscopy

5 454 D. Lister and S. Uchida (AAS) or ion chromatography (IC) [6 8]. Metallic ion concentration is measured with AAS or inductively coupled plasma - atomic emission spectrometry (ICP-AES or, with mass spectrometric analysis, ICP-MS), while in Japan X-ray fluorescence spectroscopy (XFS) has been applied to samples collected on membrane filters. Stripping polarography has been used at some plants in the US. Membrane filters can also be used to determine radioactive nuclides with Ge (Li) or high-purity Ge semiconductor detectors [9 11] High-temperature sensors High-temperature water chemistry sensors are divided into those determining water chemistry directly and those determining corrosive conditions indirectly by analysing the interaction between water and materials [12]. Major high-temperature water chemistry sensors are listed in Table 3. The purposes of high-temperature water chemistry sensors are divided into three categories: Category 1: Direct determination of water chemistry parameters at elevated temperature instead of theoretical extrapolation from data obtained at room temperature. Changes in data at room temperature and elevated temperature are estimated but the extrapolation procedures may contain uncertainties (e.g., relating to ph and conductivity). Category 2: Indirect determination of water chemistry parameters when the water chemistry index cannot be determined at room temperature. Some chemical species disappear due to thermal Table 3. High-temperature water chemistry sensors. Measured items Detector Principle Appl. Ref. (1) Water chemistry (a) [O 2 ] Pressure-balance type Membrane + polarography Lab [13] ECP detector ECP [O 2 ] relationship Lab [1,14] (b) [H 2 ] Pressure-balance type Membrane + polarography Lab [13] Pd-wire type H 2 absorption + ER Lab [15] (c) [H 2 O 2 ] Sensor array ECP/FDCI couple Lab [16] (d) Conductivity Coupled triple electrodes Complex impedance Lab [17] (e) ph ZrO 2 type Ion electrode Lab [18] TiO 2 type Flat-band potential Lab [19] (2) Interaction water-materials (a) ECP External (Ag/AgCl) Ion electrode NPP [4] Internal (Pt) Ion electrode NPP [20] Internal (Ag/AgCl) Ion electrode NPP [21] Internal (Fe/Fe 3 O 4 ) Ion electrode NPP [22] Internal (Ni/NiO) Ion electrode Lab [23] (b) Corrosion rate DC current ER change due to thinning Lab [14] DC scanning Tafel plot Lab [14] AC scanning Bode plot Lab [14] HEPro H 2 effusion NPP [24] FOLTM Ultrasonic gauging NPP [25] (c) Oxide-film properties CDE-arrangement Contact electrical resistance Lab [26,27] FDCI sensor Complex impedance Lab [16] (d) Crack initiation Electrochemical potential detector Potential noise analysis Lab [28,29] (e) Crack growth CT specimens Potential drop NPP [30] DCB Potential drop NPP [31] Note: Lab: laboratory experiments only. NPP: applied in nuclear power plants. ER: electrical resistance. FDCI: frequency-dependent complex impedance. CT: compact tension. DCB: double cantilever beam. CDE: controlled distance electrochemistry. FOLTM: feeder online thickness measurement. HEPro: measures FAC via atomic hydrogen generated.

6 Journal of Nuclear Science and Technology, Volume 52, No. 4, April decomposition during the cooling-down process ([H 2 O 2 ] in BWR reactor water) and chemical reactions in the sampling line ([O 2 ] in PWR feedwater). Category 3: Direct measurement of interaction of water and materials at elevated temperature (ECP, corrosion rate, crack growth). An international collaborative programme on hightemperature water chemistry sensors has been carried out by the International Atomic Energy Agency, IAEA (WACOLIN [32]: , WACOL [33]: and DAWAC [34]: ). As a result of compiling the literature survey and internationally exchanging information and experience on high-temperature sensors, it was concluded that many instruments had been developed for direct measurement of water quality, and some of them had been successfully applied in laboratory tests, but only a few had been applied to operating power plants [4]. The instruments reviewed appear in Table 3. The requirements for applying high-temperature sensors in operating plants are as follows: (1) high accuracy: application without cooling to determine water chemistry index directly; (2) quick response: direct connection between sensor and data acquisition system; and (3) high reliability and safety: easy calibration, maintenance-free at least for a light-water reactor (LWR) operation period (1 1.5 years). Of great importance as a transient species in BWR reactors, H 2 O 2 unfortunately is unstable in hightemperature water so determining its concentration by sampling is difficult. Almost all of the H 2 O 2 disappears during the sampling process. The procedures to determine [H 2 O 2 ] are the application of theoretical radiolysis models or the direct or indirect measurement at elevated temperature [12]. One measurement technique for [H 2 O 2 ] employed in the laboratory is frequency-dependent complex impedance (FDCI), which records the impedance of an interface as a sinusoidal potential difference is applied. To measure [H 2 O 2 ], the semicircles of Cole-Cole (Nyquist) plots of the real part of the impedance against the imaginary part have been correlated with the concentration. As [H 2 O 2 ] was reduced from 100 to 5 ppb in a BWR environment with stainless steel electrodes, the radii of the low-frequency semicircles on the Cole-Cole plots increased continuously, while the ECP remained constant at concentrations down to 10 ppb and then decreased a little to 5 ppb [35]. The increasing radii meant an increasing resistance from formation of oxide on the electrode. Both dependences of ECP and FDCI semicircle radii on [H 2 O 2 ] were not affected by co-existent [O 2 ] with the same level of oxidant concentration. As a result of water radiolysis evaluation for BWR reactor water, it was concluded that 100 ppb H 2 O 2 co-existed with 200 ppb O 2 under normal water chemistry (NWC), while 10 ppb H 2 O 2 existed without measurable concentrations of O 2 under hydrogen water chemistry (HWC). Coupled with the FDCI measurements, this meant that corrosive conditions of HWC were the same as those of NWC from the viewpoint of ECP, even as [H 2 O 2 ] was varied over a wide range. The corrosion of structural materials could be affected. By contrast, in situ experiments of the general corrosion of AISI 316L stainless steel under BWR conditions with varying concentrations of H 2 O 2 and H 2 with no added O 2 indicated that as NWC was switched to HWC so that the ECP fell from V (vs. saturated hydrogen electrode, SHE) to 0.56 V, the corrosion rate increased sharply, but when the fall was from to 0.42 V there was no change in corrosion rate. It was pointed out that the equilibrium potential for the transition between α-fe 2 O 3 and Fe 3 O 4 is 0.47 V, so the difference in behaviour between the two experiments was ascribed to changes in the structure of the oxide film at the lowest potential [36]. We note that inferring high-temperature ECP, corrosion rate and crack growth rate by extrapolation from the data obtained at room temperature causes a large uncertainty due to insufficient data bases to support the extrapolation. High-temperature sensors applied at operating plants are listed in Table 3. Most of them are sensors for structural material integrity. High-temperature reference electrodes for ECP measurements and in situ potential drop detectors for crack propagation measurements on compact tension test specimens have been applied during the benchmark tests of HWC in operating BWRs. Flow-accelerated corrosion (FAC) of the feeders in CANDU primary coolants has been monitored with FOLTMs and HEPros, and the latter have been applied to feedwater piping in CANDUs and thermal plants and to water-wall tubing in thermal plants. In order to ensure a reductive environment and thus mitigate secondary side corrosion of SG tubing, the optimum hydrazine content in the secondary system of dual-cycle reactors can be assessed from ECP measurements. However, ECP measurements are carried out only in a very few units where, once the optimum hydrazine condition is defined, the plant staff need only monitor hydrazine routinely and ECP measurements can then be discontinued. Instead of direct ECP measurements, a combined approach of concentration measurements of anions and cations by ion chromatography and empirical calculations based on crevice concentration factors and ph evaluation may be successfully applied to determine the corrosive conditions at the tubing and in the crevice between the tubing and the tube support plates. The value of ECP control for secondary coolant systems has been demonstrated by Turner and Guzonas [37], who presentedthe safeoperating zones for common SG tube materials in terms of the

7 456 D. Lister and S. Uchida Figure 2. ECP-pH zones to minimise risk of degradation of SG alloys at 300 C; after [38]. Figure 3. Major water chemistry monitoring items related to safe and reliable NPP operation.

8 Journal of Nuclear Science and Technology, Volume 52, No. 4, April Table 4. Gaps between desired information and measured data. Desired information to understand phenomena Measured WC data in plants Major measures to bridge the gaps Corrosive conditions [H 2 O 2,O 2,H 2 ] Measured [O 2,H 2 ] Theoretical models for water radiolysis HT O 2 sensors ECP sensors Crack propagation rate Crack growth rate under simulated conditions HT crack growth rate sensors Theoretical and empirical models for crack propagation rate High temperature ph ph of cooled water Theoretical evaluation HT ph sensors Properties of oxide film on sampled specimens Characterisation of oxide film Theoretical oxidation models HT impedance sensors ECP versus high-temperature ph in the SG crevices (see Figure 2 from [38]). High-temperature sensors for water chemistry, namely ph, conductivity and [O 2 ], have provided valuable data in laboratory tests. However, their application in operating plants is still at the final stage of being accepted. In order to facilitate their application, the following steps should be taken. (1) Clear presentation of their necessity and benefits. (2) Improvement of their reliability during operation: sufficient integrity of sensors themselves with straightforward calibration procedures at suitable intervals. (3) Ensuring protection against their failure during application: suitable procedures for avoiding release of loose parts and chemical contamination. (4) Provision of information from case studies involving plant diagnoses of systems based on high-temperature sensors. Much experience with developing high-temperature sensors in laboratory tests and then improving their reliability and ease of handling suggests opportunities for their application not only in nuclear plants but also in thermal power plants. However, significant barriers to their installation in primary coolants in nuclear plants are the existence of unstable and often aggressive radiolytic species which degrade the sensor materials, as well as the irradiation damage of the sensors themselves, and in thermal plants the barriers are the higher temperatures and chemical treatments which again degrade the sensor materials The gaps between the desired information and the measured data Major gaps between the information needed to understand the phenomena and the measured water chemistrydataareshownintable 4, along with measures to bridge the gaps [4]. Some mismatching between the information desired to understand plant conditions at elevated temperature and the water chemistry data measured at room temperature has been reported [1]. From the viewpoint of water chemists, procedures for bridging the gaps as well as provision of quality assurance for acquiring water chemistry data are essential for optimum operation of the plant. The major measures to bridge the gaps are twofold: the theoretical approach and the direct determination by applying high-temperature water chemistry sensors. 3. Sampling 3.1. General The extent of sampling capabilities is exemplified by the list of main sampling points in a PWR of early design [39]. The list has 10 major systems with sampling capability: primary coolant; water-filled part of pressuriser; steam-filled part of pressuriser; reactor-water clean-up system; gas volume regulation; main steam; condensate; feedwater; SG blowdown; steam extraction from turbine. Not all of these would be used routinely. In fact, as indicated in Figure 3 [40], major chemistry parameters are usually determined by sampling regularly from five systems in PWRs (reactor primary coolant, SG water via the blowdown, low-pressure and high-pressure feedwater, and condensate), six in CANDUs (as in PWRs with the addition of the moderator) and five in BWRs (recirculating reactor water, reactor-water clean-up system, low-pressure and high-pressure feedwater, and condensate). PWRs may sample the primary coolant from the hot leg and/or the chemical and volume control system (CVCS); different results are often obtained, particularly for transition metal concentrations, probably because different sets of heat exchangers are involved in cooling the samples. Secondary coolant samples may also be taken from the moisture-separator-reheater. Steam systems are not sampled routinely since they carry generally saturated steam at relatively low

9 458 D. Lister and S. Uchida pressure and temperature and contain few impurities. This is in contrast to fossil plants that have steam at higher temperature and pressure and can carry-over species such as silica, sodium, copper, chloride and sulphate that are harmful to the turbine (they promote erosion and deposition on turbine blades, localised corrosion of turbine blades and discs [especially environmentally assisted cracking], and corrosion of crossover piping and steam extraction lines). It is assumed in nuclear plants that control of the chemistry of the feedwater and SG blowdown will adequately control steam conditions, which will be reflected in condensate samples [41]. As indicated in Figure 3, a major consideration in the provision of sampling systems has been the monitoring of chemistry parameters such as ph, conductivity, concentrations of radionuclides, hydrazine, amine, oxygen, etc., depending on the coolant to be sampled. Also, in PWR primary coolant, besides the concentration of hydrogen gas, the concentrations of boron and lithium are important control parameters. The chemical species constituting these parameters are assumed to be unaffected by the design of the sampling system itself (although oxygen and hydrazine concentrations may be affected by delays at temperature in long sample lines [42] and, as pointed out elsewhere in this article, hydrogen peroxide, an aggressive species in reactor coolants [in BWRs especially], decomposes before a reliable sample can be obtained) Sampling corrosion products General The concentration of corrosion products in coolant systems is an important quantity in operating power plants and has to be monitored also. The corrosion products may be in the form of particulate and colloidal metal oxides along with dissolved and hydrated metal ions. In the reactor coolant, they control activity transport and are responsible for fouling of the fuel and the potential for subsequent fuel failure and release of fission products and actinides (as in the crud-induced localised corrosion, or CILC, of BWR fuel sheaths); in modern high-duty PWRs such fuel fouling can cause CIPS (crud-induced power shifts also known as AOA, axial offset anomaly) as boron is sequestered in deposits and affects the neutron balance. In CANDUs, corrosion products have fouled the primary side of SGs, lowering their heat transfer efficiency and raising the temperature of the coolant at the entrance to the reactor core. In secondary coolants, corrosion products in feedwater deposit in the SGs, fouling the tubes and tube supports, creating sludge piles on the tube sheets and harbouring impurities that promote corrosion. In order to measure corrosion-product concentrations, reactor operators must take samples of the coolant in question. Since a sampling system is normally made of materials such as stainless steel that constitute much Fe 3 O 4 solubility (ppb Fe) Magnetite solubility (μg/kg or ppb) vs. tempera- Figure 4. ture [44]. At ph 7 At ph Temperature ( C) Fe 3 O 4 solubility (ppb Fe) of the coolant piping and components that are being sampled, the possibility then arises that the corrosion products in the sample will be affected by the sample system itself. Even with continuous turbulent flow the balance between suspended and deposited particles will change along the sample line and through the temperature gradient imposed by the cooler, where the drop in temperature changes oxide solution properties and promotes dissolution or precipitation, affecting the concentration of particulate and dissolved species. Surface properties such as zeta potential and sorption capacity are also changed, again affecting the concentration of particulate oxides as well as dissolved species. Because of these interactions of corrosion products with the sample system itself, obtaining representative samples is fraught with difficulties. Figure 4 shows the change in solubility of magnetite, Fe 3 O 4, a common corrosion product in power systems, with temperature at two ph values. Sampling a coolant at 300 ºC and ph 25ºC of 10.0 obtained with a strong base (such as the lithiated primary coolant of CANDU reactors) would reduce the iron solubility continuously from 1 ppb (μg/kg) in the coolant to 0.1 ppb at the sample outlet at 25 ºC. If the concentration of iron in the coolant were greater than the latter value (and in a rapidly recirculating system values between the solubility at the high and low temperatures of the reactor system are expected [43]), precipitation on the walls of the system and on particles would occur. Such precipitation has been observed on nominally isothermal high-temperature filters such as the silver membrane of 0.45 μm pore-size operating at 300 ºC in an experimental stainless steel loop containing simulated CANDU primary coolant (see Figure 5). On the other hand, sampling a coolant at neutral ph (such as the reactor water of a BWR) would first increase the solubility on cooling from 5 ppb at 300 ºCto 127 ppb at 120 ºC and then decrease it to 63 ppb at the

10 Journal of Nuclear Science and Technology, Volume 52, No. 4, April Figure 5. Magnetite crystals formed on a 0.45 μm pore-size silver membrane at 300 C. sample outlet. Overall, iron-based oxides would be expected to dissolve from the walls of the sample system andfromparticles Laboratory studies Experiments in a high-temperature loop operating under PWR primary coolant conditions [45] showed that gold-plating the loop sample system of stainless steel in order to make it inert did not eliminate the holdup of corrosion products. After the coolant was filtered hot through a 0.45 μm silver membrane (with the intention of leaving only dissolved species for the experiments), the radioactive tracer 58 Co was seen to deposit along the hot section of the system by concentrating in defects in the gold plating. In the normal sampling system of AISI 316 stainless steel, the deposition of 58 Co followed approximate first-order kinetics throughout a 2 m long section and was largely mass-transfer controlled. Sharp increases in deposition occurred at the two discontinuities in the section and there was little difference in deposition behaviour in a cooled length (in which the temperature fell from 300 to 240 C) and in an isothermal length. Subsequent attempts to apply the same deposition kinetic model to the stainless steel sampling system at the Doel 1 PWR were unsuccessful [46]. Elemental and 58 Co data did not reflect the previous laboratory experience, indicating fundamental differences in the behaviour of the sample systems and suggesting that the deposition in the PWR system was surface-reaction controlled rather than mass-transfer controlled. Moreover, results for 58 Co were incompatible with those for 60 Co. It should be noted that the laboratory system filtered the coolant sample hot while the reactor system had the filter installed after the cooler. No doubt the interaction between corrosion-product precipitation and particles was different in the two sample systems, but it should be pointed out also that the coolant systems being sampled were radically different in other ways the laboratory loop was a simple once-through injection system while the reactor primary coolant was a PWR hot leg. Note also that sampling reactor water requires a delay in transit, usually provided by a delay coil in the sample line, to allow 16 N (half-life 7.13 s), arising from the neutron activation of 16 O in the water molecules in the core, to decay in a minute or so to levels at which the radiation at the sample cabinet is innocuous. Recent laboratory experiments on sampling coolant nominally saturated in iron at 300 ºC in a static autoclave and measuring the iron held up in the sample system made of a relatively inert material (titanium) indicated not only that precipitation greatly changed the concentration in the sample from that in the coolant, but also that particulate iron confounded the results in spite of a high-temperature filter of 0.5 μm pore-size at the sample system entrance in the autoclave [47]. This indicates that particles within the size range that normally defines dissolved material (i.e., <0.45 μm) can have a profound effect on sampling efficiency. It should be remembered that this range comprises colloids and actual oxide particles as well as ions and hydrated ions the proportions depending on the ph In-plant studies Several in-plant studies of the efficiency of sampling for corrosion products from reactor coolants have reported results that reflect the anomalies revealed in the laboratory studies. With an integrating sample system (filter membranes on top of ion-exchange membranes), concentrations of particulate and dissolved corrosion products in PWR and BWR feedwater were measured [48]. The deposition of undissolved corrosion products in the sample line (in coolers and valves in particular) was identified as a source of serious error. It was pointed out that systems must be made entirely of stainless steel and that best results are obtained when the system is left to run with continuous flow at a constant rate. Also in the studies, at 30 minutes before sampling, the whole system would be flushed for one minute with rapidly varying flow from zero to maximum rate. Sampling times varied between 15 and 23 hours in order to collect data from kg water. After operating transients, the flushing was extended to three minutes and accompanied with sharp rapping of the components to dislodge deposits. These techniques had been employed since the early 1970s at several sampling points in the feedwater at the Beznau II PWR and the Mühleberg BWR. The sampling studies identified locations of FAC and highlighted the advantages of condensate polishing, oxygen addition and high ph in reducing corrosion-product concentrations [49]. Samples have been extracted from the hot leg of the Ringhals 3 primary coolant system and, after

11 460 D. Lister and S. Uchida cooling, the concentrations of particulate and dissolved corrosion-product radionuclides (respectively via a filter and ion-exchange membranes) have been measured [50]. The sampling flow was always turbulent, and sampling flow rate affected the concentrations of dissolved species the higher the flow rate, the lower the concentration. Concentrations of particulates peaked dramatically after flow transients before returning to their previous values after 2 hours. A first-order depositionrelease model of the radionuclides distributed along the sample line indicated that the release constant was a maximum at 120 C and it appeared that the processes were under mass-transfer control. The time for the concentrations at the end of the sample system to reach so-called equilibrium concentrations was estimated to be 1 year. Overall, it was clear that samples were not representative of the coolant; in fact, concentrations of dissolved material did not, in general, bear any useful relationship to the entrance concentration for soluble corrosion products sampled at normal operating temperatures. The capabilities for sampling corrosion-product radionuclides from the hot legs of the primary coolant at the soon-to-be-commissioned Sizewell-B PWR were assessed, and a number of experimental findings as well as the results from Ringhals 3 were cited [51]. The report suggested that 50% 75% of particulate species could dissolve during transit down the sample system and that attenuation of dissolved species by interaction with the sample-line walls could be characterised by delay times of the order of weeks or months. From studies of sample systems on boilers, it was concluded that, to avoid gravitation settling of particles in 4 mm diameter lines, cold sampling required velocities >1 m/s and hot sampling >0.5 m/s. It was recommended that the sample lines be operated with the highest flow possible and that the capillary systems for pressure control be located downstream of the main coolers and not upstream. Later experience of sampling for corrosion products from operating systems has been more fruitful. Measurements at Sizewell B during the hot functional test (HFT) and Cycles 1 and 2 were reported by Garbett [52]. The sample system was a capillary line branched off the normal hot-leg sample line. Before each sampling campaign, the capillary was run continuously for two to three weeks to allow the surfaces to equilibrate. A 0.45 μm pore-size membrane filter attached to the end of the capillary was used to collect particulate corrosion products and the soluble fractions were obtained by ionchromatography analysis of the eluent that was collected in a vessel devoid of air by sparging with helium. The importance of collecting soluble corrosion products in such a non-oxidising environment was stressed, since it avoids the oxidation of iron to ferric hydroxide that precipitates and scavenges other transition metals and radionuclides from solution. It was observed that hydrogen peroxide additions to the system release deposited copper and chromium from capillary sampler walls [53,54]. In an attempt to correlate corrosion-product amounts in the coolant with the onset or severity of CIPS, a 2001 study [55] examined sampling results from 11 Westinghouse reactors in the US. Sampling techniques were not examined in detail, so there was no attempt to standardise sample extraction to enable strict comparisons to be made. Customary grab sampling was employed, but the sample bottles were carefully prepared and the samples were acidified with high-purity nitric acid in the same way for analysis. No precautions were taken to exclude air from the bottles, but since total corrosion products (dissolved plus particulate) were to be measured after acid dissolution this is not considered important. Samples taken directly from the hot leg of the reactor coolant system (RCS) were tabulated separately from those taken from the let-down line in the CVCS. Nickel and iron, the corrosion-product elements of interest, were analysed mainly with stripping polarography but ICP-AES and graphite-furnace AAS gave similar results. The scatter of results was considerable and it was difficult to conclude anything other than broad trends. In general, nickel concentrations were higher than the corresponding solubilities, indicating that particles had been entrained in the samples. There was no effect of sample flow rate and RCS samples tended to have higher metal concentrations than CVCS samples, although two plants showed opposite trends. Thus, average nickel concentrations from CVCS samples varied from 0.0 to 0.85 μg/kg and from RCS samples from 0.13 to 53 μg/kg. Corresponding values for iron were and μg/kg. The main connection between corrosion-product concentrations and CIPS was that the latter was associated with relatively high nickel values during start-up transients and at start of cycle. A concentration near 5 μg/kg was suggested as an indicator of increased risk of CIPS. No clear connection with iron values was evident. In 2006, the Electric Power Research Institute (EPRI) sponsored a review directed specifically at sampling in PWRs [56]. It compared results from corrosionproduct sampling of the primary coolants at Ringhals and Sizewell, where sample systems directly on the RCS hot legs were used, with those at Vandellos, where the system was connected to the let-down line in the CVCS. Although temperatures and flows in the CVCS are lower than those in the primary coolant, it was pointed out that all samples have to be cooled to room temperature anyway and precipitation/dissolution effects should overall be similar in both. Results from the Vandellos CVCS system were consistent over a 10-year period and significantly lower than previous results from grab sampling the RCS. Overall, the report indicated that with careful sampling practices and at steady operation, values for corrosion products were similar in all the plants. This contrasts with the more varied results in [55], possibly because of the less scrupulous sampling procedures in the latter. A table of elemental

12 Journal of Nuclear Science and Technology, Volume 52, No. 4, April concentrations for a number of PWRs and VVERs indicated that the dominant corrosion product in general was iron, with an approximate range of dissolved species concentrations of μg/kg and particulate concentrations of μg/kg; respective values of nickel were and μg/kg, and of cobalt were and μg/kg. Concentrations increased greatly at shutdown and recovered slowly during the first few months of the next cycle. The report concluded, among other things: that conventional grab sampling, in which the sample line is purged for up to 30 minutes and the sample collected in an open plastic bottle before filtering into soluble and particulate fractions some time later, is completely inadequate for corrosion products; that samples should not be exposed to air in order to avoid precipitation of oxidised iron species; that the best samples are obtained from continuously flowing lines; and, that transients in reactor operation can bias measurements of corrosion products high by a factor of 100. We note that a contemporary technical guidance document (TGD) from the International Association for the Properties of Water and Steam (IAPWS) for sampling fossil plants recommends practices that consider such factors [57]. Sampling for corrosion products from systems other than the primary coolant has no requirement for a delay in the sample traverse time to allow for radioactive decay. Using commercially available short sample systems, a study at the Koeberg PWR attempted to overcome some of the common anomalies when sampling the secondary coolant [58]. Systems sampled were SG blowdown, SG feedwater, heater drains and condensate extraction. Each system was installed very close to the main piping and thereby avoided long transit times and delays before collection, and connections to the sample cooler and the holder for a particulate filter (0.45 μm pore-size membrane) and ion-exchange membranes were short. Before sampling, flow through the system, bypassing the holder, was continuous to ensure that the sample line was at equilibrium when the filter holder was valved in, although for some time before the valve-in the line was blown-through at maximum flow to try to dislodge any deposits in the line; a waiting time was then imposed to allow equilibrium to re-establish. The sample measurements of corrosion-product concentrations were claimed to be representative of mainstream values, largely because of the continuous flushing of the lines and other precautions. The measurements allowed the secondary coolant chemistry to be adjusted continually to the optimum, resulting in decreasing amounts of sludge in the SGs. Low values of the WANO chemistry indicator (WCI) were also partly attributed to the success of the corrosion-product sampling and the ensuing improved control: WCI = (SG Cl /LV Cl + SG SO4 /LV SO4 + SG Na /LV Na + FW Fe /LV Fe + FW Cu /LV Cu )/5 (1) Figure 6. [58]. System for sampling high-temperature coolant where SG is the concentration of chloride, sulphate or sodium in the SG blowdown; FW is the concentration of iron or copper in the feedwater; and, LV is the limiting value of each parameter. Koeberg s LV values for chloride, sulphate, sodium, iron and copper are, respectively, 0.80, 1.70, 0.8, 5.0 and 3.3 μg/kg Sample system design Except for the delay coil, the design of nuclear sampling systems has varied little from those of traditional fossil systems. A typical sampling system for water or steam circuits in any type of industrial plant is illustrated in Figure 6 (for sampling water, the flow would be horizontal or vertically upwards; for steam, vertically downwards). The truncated-conical nozzle penetrating the coolant stream in Figure 6 is designed to be robust yet to have little impact on the flow field in its immediate vicinity [59]. A more common nozzle design as described in an ASTM standard [60] is presented in Figure 7. The opening at the nozzle tip is simply a 45º bevel cut across the penetrating tube that is oriented to face the oncoming flow. The tip should be positioned in the coolant pipe at a distance from the wall so that the stream velocity is close to the average coolant velocity. Other common nozzle designs have pitot-tube arrangements, with the tip opening facing the oncoming flow. According to the ASTM standard, the velocity of the sample in the nozzle tip should be equal to the velocity of the coolant approaching the tip. This is to assure isokinetic conditions, whereby corrosion-product particles are taken into the sample system at the same concentration as in the coolant. The isokinetic principle is illustrated in Figure 8 [42], which shows anisokinetic conditions as extra particles enter the nozzle by their inertia if the sample velocity is less than the local mainstream velocity ([b], leading to anomalously high concentrations) or as

13 462 D. Lister and S. Uchida Figure 7. Water sampling nozzle [59] μm pore-size may be installed before the cooler (bearing in mind that the high surface area may promote precipitation of corrosion products from solution, especially if the solubility temperature relationship is favourable as illustrated earlier). A typical arrangement of a filter in a sample line is shown in Figure 6. The components of sampling systems should be made of compatible materials; AISI 316L stainless steel is commonly used and carbon steel or copper alloys are scrupulously avoided since they contaminate the samples with their own corrosion products. Tubing is commonly of 4 mm internal diameter, although capillary tubing has been used in several plants to provide the required pressure drop [56]. The lines should be as short as possible, bearing in mind that nuclear primary coolant samples need to have a delay to minimise radiation fields, as pointed out earlier, and bends should not be sharp or have U-shaped configurations in the vertical plane to avoid pooling or voiding during transients in the flow. The sampling point in the main coolant pipe should be at least 40 pipe diameters downstream from any flow disturbance to ensure fully developed flow, and sample-line flow should be turbulent throughout to keep particulate corrosion products in suspension. Figure 8. (a) Isokinetic sampling; (b) and (c) anisokinetic sampling [42]. particles are diverted from the nozzle if the sample velocity is greater than the local mainstream velocity ([c], leading to anomalously low concentrations). It is of note, however, that a theoretical study [61] has indicated that isokinetic sampling is unnecessary in liquid water coolants, but in the relatively low densities and viscosities encountered in steam and supercritical water, particles deviate from the flow streamlines because of their inertia and must be sampled isokinetically. This verifies early measurements in feedwater at an operating nuclear plant that found no difference among samples taken over a wide range of sample velocities [48]. The IAPWS TGD describes the optimum systems and procedures for sampling corrosion products from fossil power plants; they apply broadly to nuclear systems as well [57]. The distinction between dissolved and particulate corrosion products is emphasised throughout and it is reiterated that particulate is usually defined as the component that is captured on a filter membrane of pore-size 0.45 μm, often with ion-exchange membranes placed below such filters to provide integrated measurements of dissolved material. Such membranes can be installed easily and operated with acceptable pressure drop in sample lines; they are used routinely in power plants for samples at room temperature, but high-temperature filter membranes also of Sample system flow The Reynolds number, Re, is a measure of the degree of turbulence in a flow system. It is the dimensionless ratio of inertial forces to viscous forces and is defined as duρ/μ, whered is a characteristic dimension of the system (in a pipe, usually the diameter), u is the average linear velocity of the flow, ρ is the fluid density and μ its viscosity. In a circular pipe, Re is correspondingly expressed conveniently as 4m/πdμ, wherem is the mass flow rate. This latter expression shows that as a sample flow is cooled ( conditioned ) from the process fluid conditions to room temperature, Re decreases as μ increases greatly. For a constant mass flow of water, cooling from 300 C at a pressure of 10 MPa to 25 Cdecreases Re by about 90% [57]. Since a critical Re value (Re crit ) of 2300 determines the transition from laminar flow to turbulent flow in a straight pipe, it follows that if the sample line were straight ensuring Re > 2300 at the sample outlet would assure turbulence throughout the sample line. However, the sample line inevitably traverses a cooler, which is usually a helical-coil heat exchanger, and the critical Re for turbulence is higher in bends than in straight pipe. Equation (2) is a correlation showing the variation of Re crit with dimension in a helical coil of tubing of diameter d and coil diameter D [62]: Re crit = 2300[ (d/D) 0.45] (2) This indicates that a water flow velocity greater than 1.9 m/s at 25 C is normally required at the outlet for turbulent flow in standard sample conditioning systems