Thermodynamic Study of the Chlorine Content of Stainless Steel


No thermodynamic data for the solution of Cl in ferrous alloys were found in the literature. This is in accord with recent Accelerator Mass Spectroscopy (AMS) analyses which showed that Cl contents in stainless steel (SS) are in the order of a few ppb. However, based on older chemical analyses of Cl in the order of 100 ppm, SS that has been irradiated with thermal neutrons in nuclear reactors is considered a major source of the long-lived 36Cl isotope in nuclear waste. In this study, the potential Cl contamination of SS originating from production and refinement processes is investigated. Unlike ferrous alloys, blast-furnace and steelmaking slags can dissolve significant amounts of Cl. The equilibrium distribution of Cl species between slags and gas phase was calculated for various steelmaking processes using the FactSage 7.2 software and databases. The results showed that despite the high volatility of metal chlorides at high temperatures, significant fractions of Cl can be retained in the slag phase even at 1600 °C. Chloride may also be incorporated in non-metallic inclusions originating from secondary refining. Based on these results and on several further assumptions, various scenarios for explaining, and also avoiding, Cl contamination of steel are discussed.


Due to its high environmental mobility, the presence of even trace amounts of 36Cl in low- and intermediate-level nuclear waste is of considerable concern.[1,2,3,4] In performance assessments of low- and intermediate-level waste, a major contribution to the 36Cl inventory originates from the large quantities of stainless steel (SS) that had been irradiated with thermal neutrons while being used as construction material in nuclear reactors.[5] These assessments have been supported by chemical analyses of SS samples that had resulted in Cl contents of up to 130 ppm.[6] However, these older results are in striking discrepancy with recent, much more sensitive measurements that have found only a few ppb Cl in SS.[7] A detailed discussion of the various analytical results of the determination of Cl in steel is presented in Appendix A.

The main objective of this work is to perform a high-temperature thermodynamic study to shed light on the distribution of chlorine species among the various solid, liquid, and gaseous phases involved in steelmaking processes. While, in agreement with the most recent analytical results,[7] no data exist for the solution of Cl in ferrous alloys, it is well established that chloride does dissolve in various slag phases,[8,9,10] and the corresponding thermodynamic data are available.[8] We therefore put forward the hypotheses that (1) Cl does not dissolve significantly, neither substitutionally nor interstitially, in the solid steel matrix, and (2) the above-mentioned disparity of Cl analyses in steel is due to either (i) exogenous inclusions of slag phases, (ii) primary endogenous non-metallic inclusions formed in secondary refining processes, or (iii) secondary non-metallic inclusions precipitated from the steel matrix during solidification and cooling. As will be elaborated in this paper, each of these types of inclusions may carry significant amounts of chloride. It is well known in the literature that inclusions in steel are common but can be minimized by careful control of slag compositions during production and refining processes, among other measures.[11] Such procedures might be a way forward to reduce the 36Cl content of low- and intermediate-level nuclear waste.

Theory and Calculations

Thermodynamic Modeling

The CALPHAD (CALculation of PHAse Diagrams) method aims at deriving reliable and consistent thermodynamic models for non-ideal multicomponent mixture phases over wide ranges of temperature and composition by optimization of Gibbs energies with respect to experimental thermodynamic property and phase diagram data.[12] In addition, numerous ab initio predictions of thermodynamic properties, such as structural energies and heats of formation, of various materials have been performed to inform CALPHAD-type evaluations.[13] The model parameters for the various phases relevant to steelmaking processes, as discussed below, were obtained by the CALPHAD method which was mainly based on experimental information and occasionally supported by first principles calculations.[14] A comprehensive and authoritative discussion of the thermodynamic models employed in the CALPHAD method was provided recently by Pelton.[15]

Ferrous alloys

Neither thermodynamic nor experimental data for (dilute) solutions of Cl in any of the Fe phases that are stable at ambient pressure, i.e., α (bcc, ferrite), γ (fcc, austenite), δ (bcc), and liquid, have been found in the literature. In contrast, it is well known that dilute solutions of other monoatomic non-metals like H, C, N, O, S, P, Si are readily formed in liquid ferrous alloys at equilibrium with either the elements (in the case of diatomic gaseous elements, Sieverts’ square-root law holds), various compounds or slags (liquid mixtures of silica and metal oxides).

In solid Fe phases, solubilities of non-metals are in general significantly lower than in the melt. The major factors that limit solid solubilities are atomic size restrictions and/or high electronegativity of the solute elements. Solid solubility can be either substitutional (solutes replacing host atoms at normal lattice sites) or interstitial (occupying octahedral of tetrahedral voids in the crystal lattice). In a γ-Fe (fcc, austenite) unit cell, there are 4 Fe atoms, 4 octahedral and 8 tetrahedral interstitials, which have radii of 126 pm, 52 pm, and 28 pm, respectively.[16] Consequently, tetrahedral interstitials can accommodate H atoms (37 pm radius), whereas N (75 pm) and C (77 pm) can occupy octahedral interstitials. Due to the size mismatch between octahedral voids and C atoms, the lattice strain energy leads to a distortion of the fcc structure so that only ca. 10 pct of the available interstitials are occupied at the maximum C solubility of 2.1 wt pct in γ-Fe.

It should be noted that O (73 pm) was recently found to occupy the octahedral interstitial positions of austenitic Fe-Ni alloys in minute concentrations before separate oxide phases were formed as a result of its high electronegativity.[17]

According to the Hume–Rothery rules, extensive substitutional solid solutions are possible if the size difference between host and solute atoms is less than 15 pct and the electronegativities of the two elements do not differ substantially. The non-metals forming substitutional solid solutions: Si (111 pm radius), P (106 pm), and S (102 pm) increasingly differ from Fe in size and electronegativity, and thus their solubilities in γ-Fe decrease in this order and all of them form compounds with Fe.

The even smaller size of Cl (99 pm) and its higher electronegativity (comparable to that of N), which favors the formation of ionic chloride compounds, suggest that its solubility in metallic Fe phases in the form of a substitutional solid solution is extremely small. The size of Cl also appears prohibitive to form interstitial solid solutions.

Non-metals often precipitate as carbides, nitrides, silicides, phosphides, and sulfides of Fe and alloying metals. Some of the resulting non-metallic inclusions tend to segregate to the grain boundaries during solidification, which frequently increases the propensity for intergranular corrosion of the alloy.

The thermodynamics of these systems is well established, see, e.g., optimizations of Fe-C-N,[18] Fe-Si,[19] Fe-P,[14] and Fe-S[20] systems. The evaluated Gibbs energy functions are routinely incorporated in large, consistent databases that have been developed, e.g., by SGTE (Scientific Group Thermodata Europe) and are also integrated in thermodynamic modeling packages such as FactSage[21] or Thermocalc.[22] Although the thermodynamic calculation of Cl solubilities in ferrous alloys is outside of current modeling capabilities, it cannot be ruled out that Cl actually dissolves in the metal phases at extremely low concentrations.

Liquid slag systems

Thermodynamic data do exist for Cl dissolved in oxide/silica slag phases and for numerous gaseous species containing Cl. It is therefore possible to calculate the distribution of chlorine species between these two phases when the thermodynamics of steelmaking processes is investigated. FactSage employs the Reddy–Blander ‘capacity model’ for halide ions in slags. The model has been optimized for a particular slag composition but otherwise gives a priori predictions.[8] A number of experimental studies have provided data for chloride capacities,

$$ C_{{{\text{Cl}}^{-} }} = \, \left[ {{\text{Cl}}^{-} /{\text{wt pct}}} \right]p_{{{\text{O}}_{ 2} }}^{1/4} /p_{{{\text{Cl}}_{ 2} }}^{1/2}, $$

which describe the exchange of oxide by chloride ions in the molten slag via redox reactions with the gaseous elements. There are some discrepancies that need to be resolved: Hirosumi and Morita[9] found that \( C_{{{\text{Cl}}^{-} }} \) increases with T and with the basicity of aluminosilicate slags, whereas Myoung et al.[10] found the opposite T dependence and deduced a strong affinity of Ca2+ with Cl in the molten slag from IR spectra.


FactSage is an integrated thermodynamic database package incorporating, among various other routines, a Gibbs energy minimizer for the calculation of multicomponent, multiphase equilibria.[21] It should be noted that for such equilibrium calculations, only the initial elemental composition of the system is required, i.e., the nature of the chloride compound entering the system is irrelevant. The FactSage databases contain critically evaluated standard and excess Gibbs energies that are described by thermodynamic models for a variety of non-ideal phases, including alloys, slags, mattes, salts, and aqueous solutions. Many of these models were developed by Pelton et al. and are discussed in detail in Pelton’s monograph,[15] so the rather complex model equations implemented in FactSage are not reproduced here. In addition to data for these non-ideal models, the FactSage databases also contain Gibbs energy functions for solid, liquid, and gaseous metal chlorides relevant to the present work. Given the high quality of thermodynamic data currently available,[21,22] reliable simulation results can be expected especially for conditions pertinent to pyrometallurgical processes, since reaction rates at high temperatures are generally fast so that equilibria are attained quickly. Consequently, the FactSage software package has been applied in numerous studies on the design of new materials and the optimization of metallurgical processes, including time-dependent models for various stages of iron and steel production.[23,24]

The potential contamination of stainless steel with traces of chlorine was investigated in this study using the thermodynamic modeling code FactSage 7.2, together with databases for ferrous alloys, slags, salts, and the gas phase. FactSage contains an oxide database optimized for the calculation of chloride solubility (see above).

Results and Discussion

Reactions of Chlorine with Stainless Steel: General Considerations

The production of stainless steel from cast iron involves very high temperatures. In SS making, both chromium and carbon oxidize when decarburization of the melt is done. The Ellingham diagram for oxide formation[25] indicates that carbon oxidation in preference to chromium oxidation can occur at temperatures greater than 1220 °C, when both elements are in pure state. Under all practical conditions, carbon oxidation can occur at temperatures above 1600 °C in the presence of chromium at atmospheric pressure.[25]

However, if these elements and oxides are components of mixed phases, e.g., a liquid alloy and a liquid slag phase, the number of degrees of freedom increases and the temperatures of oxidation change gradually as functions of composition.

As mentioned above, none of the FactSage databases for ferrous alloys (including a comprehensive dedicated database for steels which contains data for 31 elements in addition to Fe) contains data for Cl dissolved in the condensed phases. Of course, this does not imply that incorporation of Cl in solid phases of steel is not possible at all. However, it is very unlikely from a chemical point of view, given that chlorine is a very strong oxidant and, due to its high electronegativity mentioned above, tends to form ionic metal chlorides. In fact, a literature search for Cl plus steel gave results only on corrosion studies involving Cl(aq), Cl2(g) or HCl(g). In a recent modeling study by Schmid et al.,[26] coincidentally also performed with FactSage 7.2, the interplay of various corrosive gases, including Cl2(g), was investigated at 680 °C. The main conclusion of that study was that metal chlorides formed as corrosion products are volatile under these conditions.

This is even more likely under steelmaking conditions, where volatile chlorides are expected to be removed rather quickly by gas streams (e.g., in the basic oxygen furnace, BOF) or under vacuum. Even if traces of Cl dissolve in steel phases, not much of it will be left in the gas phase to equilibrate with the condensed phases.

Another strong oxidant, oxygen, slightly dissolves in the melt but is practically insoluble in solid ferrous alloy phases (see Section II–A–1). However, there are significant differences between the oxidation products (volatile iron chlorides vs condensed iron oxides) which are depicted in the following phase diagrams calculated with FactSage. Given the low solubilities of O and Cl in iron, pure condensed Fe phases are expected to be good approximations.

Figure 1 shows the stability of phases in the Fe-Cl system as a function of p(Cl2), indicating that the gas phase (consisting of volatile Fe chloride species) is stable at temperatures as low as ca. 300 °C. Austenite (γ-Fe) will be oxidized to a FeCl2-rich melt at ca. 1000 °C and p(Cl2) ≈ 10−8 bar. However, if it is ‘diluted’ with Ar(g), the gas phase becomes stable even down to ca. 200 °C (Figure 2). In contrast, the Fe-O system forms a liquid oxide (‘slag’) phase that is stable up to 2000 °C (Figure 3). The non-volatility of iron oxide phases is a prerequisite for all iron and steel production routes.

Fig. 1

Phase diagram of the binary Fe-Cl system. Fe and Fe(s2) denote bcc (ferrite) and fcc (austenite) phases, respectively. The salt melts are modeled as FeCl2/FeCl3 mixture phases

Fig. 2

Phase diagram of the ternary Fe-Cl-Ar system. Inert Ar(g) is present throughout

Fig. 3

Phase diagram of the binary Fe-O system. Spinel and monoxide denote magnetite and wüstite, respectively

It should be noted that these phase diagrams are of Type I (potential phase diagrams) in the Hillert–Pelton classification.[15,27]

The present FactSage calculations for molten SS 316 plus a deliberately high amount of 0.10 pct NaCl indeed showed that all metal chlorides are gaseous at 1600 °C, the main species being MnCl2(g), FeCl2(g), and CrCl2(g). A slight excess of oxygen forms Cr2O3(s) and then SiO2(s). With 0.10 pct Cl2(g) instead of NaCl but otherwise under the same conditions, no significantly different results were obtained. These calculations again show the high volatility of metal chlorides, but this does not exclude the possibility that traces of Cl dissolve in molten steel.

More realistic simulations of steelmaking processes include slag phases and will be discussed below. Chloride originating from impurities contained in feedstock or additives for iron- and steelmaking processes may dissolve in slags and subsequently be transferred to the final steel product. In the following, the chloride content of blast-furnace and steelmaking slags and their equilibria with other phases will be discussed.

Sources and Equilibria of Cl in the Blast Furnace

In 2018, the global production of crude steel exceeded 1.8 Gt, 70 pct of which were produced via the blast-furnace/basic oxygen furnace (BF/BOF) route, whereas the electric arc furnace (EAF) route accounted for the remaining 30 pct.[28] Chlorine impurities are likely to be introduced at various stages of these processes.

Reactions of chlorine species in the BF have received considerable attention recently. HCl is corroding gas pipelines and other stainless-steel components especially after changing from wet to dry de-dusting to save water, which would otherwise have absorbed the HCl. Dry dusts can contain up to 20 pct Cl.[29] Several sources of Cl have been identified:

  1. 1.

    Pulverized Coal Injection (PCI) of low-quality coal which usually contains <0.1 pct Cl in various inorganic and organic forms, but can reach 0.2 pct Cl,[30] and waste plastics injection to substitute expensive coke;

  2. 2.

    Ore after seawater beneficiation, with Cl contents of up to 0.1 pct,[30];

  3. 3.

    Sinter sprayed with CaCl2 solution to improve its mechanical strength (i.e., reduce disintegration) by delaying the reduction of hematite to magnetite.[31] An optimum concentration of 2 pct CaCl2 was found,[31] resulting in ca 0.2 pct Cl in the sinter after reduction at either 500 °C or 900 °C.[32] When the CaCl2 concentration is too high, incomplete volatilization will affect the reduction rate.[31] Another study reported that a sinter sample contained 0.022 pct Cl.[29]

  4. 4.

    Coke can contain up to 0.05 pct Cl, which improves its mechanical strength but is detrimental to reactivity.[30] Desorption of HCl(g) from coke at 500 °C to 900 °C has been reported.[33] Other studies have proposed cycling of KCl in the BF in the T range of 600 °C to 1000 °C.[34,35]

Given these numbers, an overall Cl content of 0.06 pct (corresponding to 0.10 pct NaCl) was selected as initial value in the following FactSage simulations:

The present modeling results indicated that cast iron at equilibrium with blast-furnace slag (280 kg/t of hot metal) + 0.10 pct NaCl retains 68 pct of Cl in the slag phase, corresponding to 0.15 pct Cl in the slag. It should be noted that this is a static equilibrium calculation, whereas a continuous gas flow in the blast furnace is expected to remove more Cl from the slag. Using their model, Myoung et al.[10] performed a similar simulation assuming Cl contents of ca. 0.06 pct in coke and coal only, which resulted in 0.14 pct Cl in the slag. A measured value for a BF slag from a Korean plant was 0.11 pct Cl.[10] Murav’eva and Bebeshko[36] found 0.037 pct Cl in a blast-furnace slag from a Russian plant using ion-selective electrodes after Na2CO3/ZnO sintering of the slag and subsequent leaching with water. In a European project[34] chlorine contents from <0.01 pct to 0.06 pct were determined in BF slags from four steel producers, depending on the amounts of Cl that were fed in the BF mainly via different types of coal.

Potential Sources of Cl in Steelmaking Processes

Processes performed on hot metal after the tapping of a BF usually comprise (i) hot metal pretreatment, essentially consisting of a desulfurization process that utilizes primarily Mg metal[37] which potentially contains Cl (see below) that may be transferred to subsequent process stages, followed by (ii) decarburization, e.g., in a BOF (see Section III–D), and (iii) secondary refining (Section III–D) which aims at reducing the contents of remaining impurities that are detrimental to steel quality. Among the most prominent reagents for this purpose are alkaline earth metals (Ca, Mg), Al, Si, Ti, etc.

In this context, it is interesting that various molten non-ferrous metals have been reported to dissolve their molten halides and vice versa. Earlier literature, including phase diagrams and solubility data, was reviewed by Bredig.[38] Common features of phase diagrams comprise (i) liquid–liquid and solid–solid demixing into metal- and salt-rich phases with associated monotectic and eutectic equilibria, respectively, and (ii) rapidly decreasing mutual solubilities with decreasing temperature, which lead to precipitation of salt inclusions in the metals during solidification and cooling. Despite being more reactive (electropositive) than Fe, alkali and alkaline earth metals are essentially retaining their metallic properties in these liquid mixtures rather than undergoing chemical reactions, although conductivity and freezing point depression data have indicated that a variety of species might be formed.[38] Ogasawara et al.[39] have measured the activity of Ca(l) in calcium–calcium halide melts up to CaO(s) saturation and have noted that such fluxes have excellent refining ability through the selective removal of P and several tramp elements.

Various Cl-containing solid phases have been reported to precipitate from relevant molten oxide systems when chloride is present. The binary CaCl2-CaO phase diagram has been determined by Neumann et al.[40] and Wenz et al.[41] who identified double salts having the compositions CaO·4CaCl2 and CaO·2CaCl2, respectively. Despite this discrepancy in composition, both double salts were found to solidify at ca. 835 °C.[40,41] The FactSage database contains data for CaO·4CaCl2(s) and for the liquid salt mixture, and moreover predicts that CaCl2(l) has a normal boiling point of 1934 °C. This implies that CaCl2(l) may be retained in condensed phases formed at molten steel temperatures during refinement, whereas it is likely to be removed completely under vacuum degassing conditions (calculated bp. of 1315 °C at 5 mbar, see Section III–D). Patsiogiannis et al.[42] described the incorporation of chloride in a solid phase, 11CaO·7Al2O3·CaCl2, which precipitated at (1048 ± 27) °C from a secondary steelmaking slag in the CaO-Al2O3-SiO2 system (albeit at relatively high initial CaCl2 contents) and persisted as a separate phase on cooling. Amorphous Cl-containing solid phases of varying compositions were found to form from silica-rich ironmaking slags.[42] Liu and Kobayashi[43] have measured sulfide capacities of CaO-CaCl2-CaF2 slags and observed their excellent desulfurizing ability, which was slightly reduced when CaF2 had been replaced by CaCl2 due to environmental concerns. When such slags are employed for hot metal pretreatment as suggested,[43] they represent another source of Cl which may contaminate steel. However, it was found that CaCl2 evaporation became significant in these slags above 1300 °C.[43]

Analogous results were reported for magnesium. In a recent study, Mg(l) was found to dissolve up to 18 ppm Cl at equilibrium with molten NaCl/MgCl2 mixtures at 735 °C.[44] Rosenkilde et al.[44] claimed that Cl dissolves in Mg(l) in “ionic” form and proposed the reaction ½ MgCl2(l) + ½ Mg(l) = MgCl(in Mg(l)) to explain their finding that the Cl solubility was proportional to the square root of the MgCl2(l) activity in the liquid salt mixture. It was not clarified by the authors[44] whether ‘MgCl’ resembles an ion pair with monovalent Mg, Mg+Cl, or an ionic subhalide, both of which would require redox reactions involving Mg. We like to emphasize that an equivalent thermodynamic analysis of the experimental data[44] can be performed by invoking the reaction ½ MgCl2(l) = ½ Mg(l) + Cl°(in Mg(l)), which implies that Cl dissolves in atomic form. This reaction is in accordance with the dissolution reaction of diatomic gases as described by Sieverts’ law (see Section II–A–1), ½ Cl2(g) = Cl°(in Mg(l)), which proceeds until formation of MgCl2(l) occurs at the solubility limit of Cl° in Mg(l), where p(Cl2) would be extremely low.

The implications of these studies for the present work are twofold:

  1. 1.

    Magnesium metal is known to contain chloride inclusions resulting from production (electrolysis) and recycling processes involving chloride melts.[44,45] Similar chloride inclusions are likely to be present in electrolytically produced Ca metal.[38,46] If such Cl-contaminated Mg or Ca are used in secondary refining, Cl may eventually be transferred to, and incorporated in, the resulting non-metallic inclusions in steel. This incorporation could result in, for instance, the Cl-containing solid phases discussed above, which might be present together with finely dispersed non-metallic inclusions in steel.

  2. 2.

    In analogy to the non-ferrous metal–chlorine systems discussed above, it appears plausible that molten steel also contains traces of dissolved monoatomic Cl at equilibrium with (i) Cl species in the gas phase, (ii) chloride and iron ions contained in molten slag phases, and/or with (iii) Cl-contaminated alloys employed in secondary refining. These traces of dissolved Cl may then react with steel constituents and precipitate, especially as MnCl2(s), on solidification and cooling due to a significantly lower solubility of such chlorides in solid steel (see Section III–E).

Thermodynamic Calculation of Cl Partitioning in Steelmaking Processes

Steels are generally produced in either of two processes:

  1. (i)

    an ‘integrated route,’ in which hot metal is transferred from a BF to a BOF, which employs either top (Linz–Donawitz or LD process) or bottom blowing of O2(g) resulting in the oxidation of C, P, Si, Mn, etc.[47];

  2. (ii)

    a ‘duplex route’ involving an EAF (using ferrous scrap forming an oxidizing slag as the oxidant, with additions of ferroalloys), followed by argon oxygen decarburization (AOD) converter process.[48] The AOD stage involves a ‘dilution effect’ resulting in the oxidation of <3 pct Cr to the slag. The BOF route with pure oxygen would lead to much higher Cr losses.

In subsequent ‘secondary steelmaking’ processes, deoxidation/desulfurization with alloys containing Ca, Mg, Al, Si, Mn, Ti, etc. and alloying with Cr, Ni, Mo, and other easily oxidable elements are carried out in a ‘ladle refining’ stage (usually under argon atmosphere) to produce stainless steels. Vacuum degassing (VD) may be added as a ‘cleaning’ stage which reduces, besides C, H, and N, the levels of volatile ‘tramp elements’ (impurities). All these stages involve slag phases whose compositions need to be adjusted carefully to achieve the desired alloy compositions.[49]

Endogenous inclusions inevitably result from reactions of the above-mentioned additives with non-metallic impurities like dissolved oxygen and sulfur, whose levels need to be minimized to improve the steel quality.[11] It is a general aim to minimize non-metallic inclusions, e.g., by optimizing the composition of top slags in the refining process to absorb inclusions, which has been supported by calculations with FactSage 7.1.[50] In special cases however, oxide inclusions are deliberately generated during deoxidation to act as nucleation seeds for achieving advantageous microstructures.[51] If finely dispersed non-metallic inclusions are not removed, they contribute to the total non-metal content of the alloy, in addition to the residual impurity levels dissolved in solid steel.[11] This would also apply to Cl if traces of this element were contained in non-metallic inclusions. Given that some deoxidation/desulfurization reagents, particularly metallic Ca and Mg, may contain small amounts of chloride, it is plausible that Cl eventually reports to the resulting non-metallic inclusions in steel (see Section III–C).

Applications of FactSage models and databases to inclusion engineering have been discussed by Jung et al.[52,53] There are numerous other studies, e.g., on inclusion evolution during refining and casting,[54] on slag–steel–inclusion reactions of 304 stainless steels[55] and on inclusion formation during the solidification of steel.[56] However, none of these workers considered chlorine in their models.

In this study, typical slag compositions for basic oxygen and electric arc furnaces[57] (see Table I) were employed in the following calculations.

Table I Chemical Composition of Steel Slag

Simulation of basic oxygen furnace

Detailed modeling of the BOF (O2 blowing) is rather complex and requires coupling of thermodynamic and kinetic models.[58] The present, static thermodynamic simulation of steel production in a BOF was set up to reduce the C content of crude iron (4 pct C, 0.4 pct Mn, 0.1 pct Si) to <0.1 pct C using appropriate amounts of O2(g) and fluxes to produce ca 500 m3/t gas phase (98.4 pct CO, 1.5 pct CO2, chloride species as given below) and 130 kg slag (of composition indicated in Table I) per ton of Fe(l). The calculation then predicts that ca 41 pct of Cl (from 0.10 pct NaCl) stays in the slag phase at 1600 °C, the rest is released to the gas phase. Calculations with HCl(g) and Cl2(g) give 63 pct and 68 pct in the slag phase, respectively. The reasons for these differences are the gaseous species that are formed: the partial pressure of NaCl(g) is 3.7 mbar, that of HCl(g) is 0.5 mbar, whereas in the third case, Mn, Fe, and Ca chlorides with partial pressures between 0.2 and 0.3 mbar are the predominant Cl species in the gas phase. As expected (Section II–A–2), there is an increase of Cl concentration in the slag with increasing basicity (MeO/SiO2 ratio) of the slag. As mentioned in Section III–A, lower Cl contents in the slag phase can be expected under dynamic conditions as a flowing gas stream would deplete the slag of Cl more efficiently. Thus, it is likely that the BOF process is not contributing significantly to the Cl content of steel (with the possible exception of exogenous inclusions of slags with high Cl content).

Simulation of electric arc furnace

If SS 304 is produced in an electric arc furnace (where ferrous scrap might be a Cl source) and undergoes argon oxygen decarburization (assuming the presence of 0.10 pct NaCl, ca 120 kg/t EAF slag of composition indicated in Table I, and 400 m3/t gas phase with ca 90 pct Ar), a smaller fraction of the metal chlorides (decreasing form 50 to 30 pct between 1400 and 1600 °C) is calculated to remain in the slag phase. Increasing the gas volume to ca 1200 m3/t reduces these values to 33 to 17 pct, respectively. The last value corresponds to 0.085 pct Cl in the slag. If the initial amount of NaCl is reduced to 0.01 pct, a larger fraction of the total Cl (47 pct) will remain in the slag phase at 1600 °C but the Cl concentration in the slag is reduced to 0.024 pct.

Our calculations for SS 316 gave almost identical results since Mo did not react with chlorides or chlorine at these high temperatures.

Simulation of secondary refining

The present calculations confirmed that in secondary refining involving sufficient Cl-contaminated Ca metal (i.e., in excess to the amount required for the reaction with other impurities), Cl impurities would be collected in a molten CaCl2-rich salt phase that is immiscible with deoxidation and desulfurization products. This liquid phase is predicted to solidify to CaO·4CaCl2 at ~830 °C which remains stable at lower temperatures. Depending on the Cl-contamination of the Ca metal, this scenario may explain significant Cl contents of steel.

Simulation of vacuum degassing

The present calculations have shown that the most efficient method of removing chlorine (and other impurities) from the condensed phases is the production of ‘clean steel’ by vacuum degassing: at a pressure of 5 mbar, which is easily achievable under industrially generated vacuum, >99 pct of Cl is predicted to be transferred to the gas phase (see Section III–C). If stable 35Cl is released to the gas phase during ‘clean steel’ production, it is then very improbable to have 36Cl in the structure of the solid steel after irradiation.

The results of these simulations are summarized in Table II. It should be noted that in such modeling studies, relative changes and trends are more significant than absolute values.

Table II Summary of Assumed Initial Amounts of Cl and Simulation Results for the Partitioning of Cl Among Condensed and Gas Phases in Steelmaking Processes

Possible Partitioning of Cl in the Solid State

The present FactSage calculations show that at very low free Ca levels in steel (i.e., after reaction of Ca(s) with dissolved impurities other than Cl during secondary refining), the most stable solid metal chloride forming at equilibrium between solid stainless steel and HCl(g) or Cl2(g) is essentially pure MnCl2(s), which contains traces of FeCl2(s) and NiCl2(s) in solid solution and starts precipitating at ca. 630 °C. Its formation mechanism might be in broad analogy with the precipitation of Mn sulfide which, in contrast to primary non-metallic inclusions produced during deoxidation, is formed as a result of solubility differences: sulfur dissolves in molten metal but is virtually insoluble in the solid alloy. A comprehensive thermodynamic modeling study of the Fe-S system[20] found a sulfur solubility in liquid and fcc-Fe of 13 pct and 0.05 pct, respectively, at the catatectic equilibrium fcc–bcc–liquid at 1356.4 °C and 1 bar (for comparison, typical sulfur contents of molten steel at equilibrium with steelmaking slags at 1600 °C are smaller by a factor of 100). In the binary Fe-S system, pyrrhotite (Fe1–xS) precipitates on cooling, and the sulfur solubility in liquid and fcc-Fe changes to 31 pct and 0.01 pct, respectively, at the eutectic equilibrium fcc–liquid–Fe1–xS at 988 °C and 1 bar. In SS, solid (Mn,Cr,Fe) sulfide phases will precipitate, resulting in a calculated S solubility in fcc-Fe of 5·10–6 pct (50 ppb) at 988 °C. Such Mn-rich solid sulfides usually give rise to localized (pitting) corrosion.[59,60]

As discussed in Section III–C, it appears plausible that molten steel can dissolve slight amounts of Cl which may either be transferred to the alloy phase from a slag phase during production or, more likely, during secondary refining via deoxidation agents such as Ca(s) and Mg(s) which may contain chloride inclusions. In view of the presumably much lower Cl solubility in solid steel structures, a precipitation mechanism analogous to Mn sulfide might result in secondary inclusions of MnCl2(s), which might then segregate to grain boundaries. This would be relevant for explaining the results of the stepwise dissolution of steel samples for the 36Cl AMS analyses of Winkler et al.[7] (see Section III–F).

Discussion of Analytical Results for Cl in Steel Samples

The recent analytical study of Winkler et al.[7] deserves some discussion. The authors used neutron activation of steel samples combined with AMS to lower the detection limit down to ppb levels (microgram per kilogram SS). The results of the first and the second acid leaching of the steels (drill shavings) are between 600 ppb to 1.2 ppm for the first and 200 ppb to 800 ppb for the second leaching. The authors[7] discuss this as removal of surface contamination that might have resulted from cutting fluids containing Cl or from reaction of the fresh metal surface with traces of HCl(g) in the atmosphere.[61] The third leaching, where the rest of the steel sample was dissolved and which the authors call the real analysis, resulted in chlorine contents of 9 to 25 ppb, whereas the blanks (not irradiated) resulted in 2 to 5 ppb chlorine. A discussion with the authors[62] enabled us to recalculate their results which were found to be in good agreement with Figure 1 of their paper.[7]

The calculated Cl concentrations for the three leaching steps allowed us to plot the percentage of leached Cl vs the percentage of steel dissolved in acid (Figure 4). It appears that 40 pct of Cl is dissolved immediately which might imply that it corresponds to real surface contamination that is removed at the first acid wash. Then, Cl and steel dissolution are correlated essentially linearly (trend line with slope = 1 shown in Figure 4) and, at 60 pct steel dissolution, almost no Cl is left. This may mean that either the ‘innermost core’ of 40 pct of the steel carries very little Cl (which raises the question about the origin of 60 pct of the Cl in the outer 60 pct of the shavings) or there is indeed some preferential leaching of Cl located along grain boundaries (resulting from primary non-metallic inclusions or secondary precipitation of MnCl2(s), see above) so that almost all Cl has dissolved in acid while there is still ca. 40 pct steel left. In either case, the actual Cl content of the samples would be significantly higher than that found in the third leach. However, it should be kept in mind that these values probably carry rather high uncertainties given that, in absolute values, the samples contained only 20 to 40 ng total Cl per 40 to 140 mg steel. In any case there seems that about 40 pct of the analyzed chlorine is indeed surface contamination, i.e., 72 to 220 ppb Cl out of the total analyzed chlorine of 180 to 550 ppb. This observation of high chlorine concentrations in the first acid wash is anyhow common for all analytical chlorine determination studies.[5,63]

Fig. 4

Percentage of leached Cl vs percentage of dissolved steel calculated from data of Winkler et al.[7]

Recent studies of the chloride attack of the passive layer of single-crystal SS (free of inclusions and grain boundaries) indicate chloride penetration of the passive layer and its accumulation beneath the passive layer.[64] If it is assumed that ‘surface contamination’ results from traces of HCl(g) and H2O(g) in the (laboratory) atmosphere, these two components would produce chloride ions when being ‘adsorbed’ on a fresh or passivated metal surface. One could even speculate further that external chloride contaminants penetrate the steel structure through microcrevices formed during initial stages of pitting corrosion initiated by sulfide inclusions[65] or via grain boundaries which are “well known [to] exhibit an irregular atom array yielding a loose structure that usually provide tunnels for species diffusion and transport.”[64] In any case, highly sensitive microscopic and surface analytical methods would be required to clarify these potential mechanisms.

Intergranular corrosion is a common problem encountered with stainless steels.[66] It usually requires that the alloy is ‘sensitized’ by some heat treatment which forms Cr-rich carbides, (Cr,Mo,Fe)23C6, along the grain boundaries and depletes the neighboring matrix of Cr, making it more prone to corrosion (in the present case, it is not known whether such heat treatment has been applied either deliberately during production (e.g., stress relieving heat treatments,[67]) or accidentally as a result of heat produced during drilling. In any case, structural changes due to work-hardening resulting from drilling have been reported, see, e.g.,[68] It may then happen that the entire ‘sensitized’ metal piece is penetrated by acid along the grain boundaries. Similar preferential acid attacks have been reported in the literature, including the extreme case that the metal even crumbles into single grains, e.g., in hot, concentrated HNO3.[66] It is also known in the literature that segregation of C, N, P, and S impurities to the grain boundaries is quite common—maybe this applies also to Cl impurities, as suggested above by the possible analogy of MnS(s) and MnCl2(s) formation. In the present case,[7] intergranular corrosion, or pitting corrosion initiated at non-metallic inclusions, during the stepwise dissolution in 2 mol L–1 HNO3 can neither be confirmed nor ruled out as the steel samples were not investigated metallographically.[61]

The drill shavings “of a steel sample with chlorine levels that were at or below the detection limit of mass spectrometry and neutron activation using 38Cl detection,”[7] which are discussed above, were apparently different from the drill core samples (cylinders) of “4 types of steel used for various parts of a nuclear reactor” (containing 2 to 6 ppb Cl, see Appendix A). These lower Cl contents may also be due to the different sampling method in the case of cylinders, which allowed taking samples with most of the surface not in contact with the environment before sampling. At the same time, the samples so produced had a much smaller surface than the drill shavings of the same metal mass.

The values of 2 to 6 ppb Cl determined for the drill core samples (cylinders) appear to be the ‘gold standard’ for Cl contents of ‘clean steels,’ whereas samples used in previous analytical studies may indeed have contained higher Cl levels, which might have been generated through the various mechanisms discussed in this paper.


The results of this study seem to find some support in the few attempts to determine 36Cl in steels which have been discussed in Appendix A. Although no unique mechanism for explaining potential Cl contamination of reactor steel could be found in this study, several possibilities were identified:

  • Although being unlikely due to high chemical reactivity and atomic size restrictions, trace amounts of Cl dissolved in solid steel structures cannot be ruled out. A case in point is oxygen, whose very low interstitial solubility in austenitic Fe-Ni alloys has been measured recently.[17] However, the atomic radius of Cl appears too large to fit in interstitial voids, and therefore there is no analogy with H, N, C, and O which form this type of solid solutions.

  • Based on the Hume–Rothery rules, substitutional incorporation of Cl atoms in steel crystal lattice also seems hardly possible: the atomic radius is too small (–22 pct) and the electronegativity of Cl is too high, and therefore Cl tends to form ionic metal chlorides in the solid state.

  • However, despite being more reactive (electropositive) than Fe, alkali and alkaline earth metals can dissolve significant amounts of halogens at equilibrium with their halogenides.[38] In recent measurements, Mg(l) was found to dissolve 18 ppm Cl (in monoatomic form) at equilibrium with MgCl2(l) at 735 °C.[44]

  • The present thermodynamic calculations have shown that Cl readily dissolves in slags formed during various steelmaking processes. Minute amounts of Cl might then be transferred to the molten crude iron or steel.

  • An upper limit of 0.1 pct “chloride ion content” has been specified for ground-granulated blast-furnace slag.[69] Similar values have been reported in the literature[10] and found in the present simulations. The chlorine content of a Russian blast-furnace slag was 0.037 pct,[36] while in a European project[34] chlorine contents from <0.01 pct to 0.06 pct were determined in BF slag, depending on the amounts of Cl fed in the BF.

  • The Cl content of steelmaking slags is also low: several sources (e.g., Souza et al.[70]) report Cl in EAF slag only below detection limit (< 0.1 pct by X-ray fluorescence). We found one study[71] which analyzed BOF slag and reported 0.01 pct Cl, whereas ladle slag apparently had a Cl content of <0.01 pct because no value was given.[71]

  • Thus, Cl may be present in large-scale exogenous inclusions, which mainly result from external sources such as various reoxidation processes, addition of alloying elements, or they are formed of entrapped refractory materials, mold flux, and covering slag due to poor de-slagging of the molten metal.[72]

  • The question remains whether some Cl values exceeding the detection limits of older analytical studies can be explained. Undetected, exogenous Cl-containing inclusions might have contributed to the high Cl contents of some steel samples found in those studies[6] (besides high detection limits and potential analytical errors and/or sample contamination).

  • In secondary refining processes,[11,49] finely dispersed, primary endogenous inclusions are generated by reaction of impurities with deoxidation/desulfurization agents, particularly Ca, Mg, Al, Mn, Si, and their alloys. Some of them, especially Mg and Ca, may contain chloride inclusions due to production (electrolysis) and recycling processes involving chloride melts, which partly dissolve in the liquid metals.[38,44,45,46] Cl may then dissolve in traces in the molten steel, or be transferred to, and incorporated in, the resulting non-metallic, endogenous inclusions in steel (Sections III–C and III–D).

  • Several analyses by SEM–EDX or SEM–WDS of the non-metallic inclusions (NMI) indicate chlorine levels below detection limit, i.e., the chlorine content in NMI is below 0.5 pct or 0.1 pct. According to Reference 73, “Non-metallic inclusions (NMIs) occur typically in low or very low volume fractions (from 10−2 in a high oxygen weld deposit to 10−5 in very clean bearing steels), but play an important role in many properties of steel.”

  • If it is assumed that the alloy contains only 0.001 pct non-metallic inclusions with a Cl content of 0.1 pct (evenly distributed in the matrix), this would result in a Cl content of 10 ppb overall, similar to values found in the most recent analyses (Appendix A). It is therefore important to undertake the chlorine analysis of non-metallic inclusions by methods that have lower detection limit than SEM–EDX, e.g., by nanoscale SIMS.[74]

  • Some non-metallic inclusions tend to accumulate at grain boundaries where they may be preferentially leached by acidic solutions, which are known to extract various elements from slags quite efficiently. NMIs of the sulfide type (most commonly MnS-rich) have been reported as sites where the pitting corrosion starts.[59,60] The role of NMIs, including oxide inclusions in the rather complex pitting corrosion mechanisms of SS, is discussed in recent review papers.[65,73]

  • There might be an analogy of Cl with S, which dissolves in the steel melt, but hardly at all in solid steel structures so that solid sulfides precipitate as secondary inclusions due to the solubility difference. Thus, besides occurring in slag or primary non-metallic inclusions, Cl might be contained in secondary precipitations, because the solubility of Cl is expected to be much lower in the steel matrix than in the melt. If the Cl content of such precipitations is sufficiently high, it might be possible to detect them by micro‐Raman spectroscopy.[75]

  • Segregation of chloride-containing inclusions to grain boundaries might explain the much higher Cl levels in the first and second leaching which was interpreted by Winkler et al.[7] to be due to contamination of the surface layer—see discussion in Section III–F.

  • Modern steelmaking technology results in ‘clean’ SS qualities that should be essentially free of chlorine. Vacuum degassing was shown to significantly reduce the number of oxidic micro-inclusions[76] and might provide a route to ensure that Cl contamination is efficiently removed from steel, as was confirmed by the present calculations.


  1. 1.

    H.J.M. Bowen: Environmental Geochemistry of the Elements, Academic Press, New York, 1979.

    Google Scholar 

  2. 2.

    H.W. Bentley, F.M. Phillips, S.N. Davis: 36Cl in the terrestrial environment, in Handbook of Environmental Geochemistry, Vol. 2b, P. Fritz and J.-Ch. Fontes, eds. Elsevier Science, New York. 1986, pp. 422–475.

    Google Scholar 

  3. 3.

    M.G. Zreda, F.M. Phillips, D. Elmore, P.W. Kubik, P. Sharma and R.I. Dorn: Earth Planet. Sci. Lett., 1991, vol. 105, pp. 94–109.

    CAS  Google Scholar 

  4. 4.

    S.C. Sheppard and M. Herod: J. Environ. Radioact., 2012, vol. 106, pp. 27–34.

    CAS  Google Scholar 

  5. 5.

    D.E. Robertson, C.W. Thomas, S.L. Pratt, E.A. Lepel and V.W. Thomas: Report NUREG/CR-6567 PNNL-11659. Pacific Northwest Laboratory, Richland, WA, 2000.

  6. 6.

    J.C. Evans, E.L. Lepel, R.W. Sanders, C.L. Wilkerson, W. Silker, C.W. Thomas, K.H. Abel, and D.R. Robertson: Report NUREG/CR-3474, Pacific Northwest Laboratory, 185 pp., 1984

  7. 7.

    S.R. Winkler, R. Eigl, O. Forstner, M. Martschini, P. Steier, J.H. Sterba and R. Golser: Nucl. Instr. Meth. Phys. Res., 2015, vol. B 361, pp. 649–53.

  8. 8.

    A. Pelton: Glastech. Ber. Glass Sci. Technol., 1999, vol. 72, pp. 214–26.

  9. 9.

    T. Hirosumi and K. Morita: ISIJ Int., 2000, vol. 40, pp. 943–948.

    CAS  Google Scholar 

  10. 10.

    K.J. Myoung, Y.S. Lee, D.J. Min, U.C. Chung and S.H. Yi: Steel Res. Int., 2004, vol. 75, pp. 783–791.

    CAS  Google Scholar 

  11. 11.

    L. Holappa and O. Weijk: Inclusion engineering, in Treatise on Process Metallurgy, Volume 3, Industrial Processes, Ch. 1.7, S. Seetharaman, ed., Elsevier, Amsterdam, pp. 347–372, 2014.

  12. 12.

    N. Saunders and A.P. Miodownik: CALPHAD (Calculation of Phase Diagrams): A Comprehensive Guide, Pergamon, Oxford, 1998.

    Google Scholar 

  13. 13.

    P.E.A. Turchi, I.A. Abrikosov, B. Burton, S.G. Fries, G. Grimvall, L. Kaufman, P. Korzhavyi, V. Rao Manga, M. Ohno, A. Pisch, A. Scott, and W. Zhang: Calphad, 2007, vol. 31, pp. 4–27.

  14. 14.

    H. Ohtani, N. Hanaya, M. Hasebe, S. Teraoka and M. Abe: Calphad, 2006, vol. 30, pp. 147–158.

    CAS  Google Scholar 

  15. 15.

    A.D. Pelton: Phase Diagrams and Thermodynamic Modeling of Solutions. Elsevier, Amsterdam, Oxford, Cambridge, 2019.

    Google Scholar 

  16. 16.

    G. Krauss: Steels: Processing, Structure, and Performance. ASM International, Materials Park, Ohio, 2005.

    Google Scholar 

  17. 17.

    D. Jullian, J. Zhang, D.B. Hibbert and D.J. Young: J. Alloys Comp., 2018, vol. 732, pp. 646–654.

    CAS  Google Scholar 

  18. 18.

    H. Göhring, O. Fabrichnaya, A. Leineweber and E.J. Mittemeijer: Metall. Mater. Trans. A, 2016, vol. 47, pp. 6173–6186.

    Google Scholar 

  19. 19.

    S. Cui and I.-H. Jung: Calphad, 2017, vol. 56, pp. 108–125.

    CAS  Google Scholar 

  20. 20.

    P. Waldner and A.D. Pelton: J. Phase Equilib. Diffus, 2005, vol. 26, pp. 23–38.

    CAS  Google Scholar 

  21. 21.

    C.W. Bale, E. Belise, P. Chartrand, S.A. Decterov, G. Eriksson, A.E. Gheribi, K. Hack, L.-H. Jung, Y.-B. Kang, J. Melancon, A.D. Pelton, S. Petersen, C. Rubelin, J. Sangster, P. Spencer, and M.-A. Van Ende: Calphad, 2016, vol. 54, pp. 35–53.

    CAS  Google Scholar 

  22. 22.

    J.-O. Andersson, T. Helander, L. Höglund, P. Shi and B. Sundman: Calphad, 2002, vol. 26, pp. 273–312.

    CAS  Google Scholar 

  23. 23.

    S. Arnout, H. Verhaeghe, B. Blanpain, P. Wollants, R. Hendrickx, G. Heylen: Steel Res. Int., 2006, vol. 77, pp. 317–323.

    CAS  Google Scholar 

  24. 24.

    I. Jung and M. Van Ende: Metall. Mater. Trans. B., 2020, vol. 51, pp. 1851–1874.

    CAS  Google Scholar 

  25. 25.

    A.K. Chakrabarti: Steel Making, PHI Learning, New Delhi, 2012, 240 pp.

    Google Scholar 

  26. 26.

    A. Schmid, G. Mori, E. Bucher and R. Haubner: Oxid. Met., 2019, vol. 91, pp. 1–10.

    CAS  Google Scholar 

  27. 27.

    M. Selleby, M. Hillert and J. Agren: Calphad, 2011, vol. 35, pp. 342–45.

  28. 28.

    Worldsteel Association:,, accessed 28/6/2020.

  29. 29.

    X. Liu, Q. Lu, and S. Zhang: Adv. Mater. Res. (Durnten-Zurich, Switz.), 2011, vol. 402, pp. 107–10.

  30. 30.

    B.S. Hu, Y.L. Gui, H.L. Guo, and C.Y. Song: Adv. Mater. Res. (Durnten-Zurich, Switz.), 2011, vol. 396–398, pp. 152–56.

  31. 31.

    X. Zhang, J. Zhang, Z.-W. Hu, H.-B. Zuo and H.-W. Guo: J. Iron Steel Res. Int., 2010, vol. 17, pp. 7–12.

    Google Scholar 

  32. 32.

    N. Taguchi, T. Otomo and Y. Omori: ISIJ Int., 1990, vol. 30, pp. 281–289.

    Google Scholar 

  33. 33.

    N. Tsubouchi, E. Kasai, K. Kawamoto, H. Noda, Y. Nakazato and Y. Ohtsuk: Tetsu-to-Hagane, 2005, vol. 91, pp. 751–756.

    CAS  Google Scholar 

  34. 34.

    R. Lin, U. Jahnsen, S. Widner, T. Bürgler, E. Lectard, D. Sert, P. Mannila, K. Heinänen, P. Tanskanen, S. Huttunen, S. Kallio, J. Härkki, L. Savov, O. Iwanov, D. Janke, and P. Drissen: Report EUR 20586, Brussels, 2003.

  35. 35.

    E. Lectard, E. Hess and R. Lin: Rev. Met. Paris, 2004, vol. 101, pp. 31–38.

    CAS  Google Scholar 

  36. 36.

    I.V. Murav’eva and G.I. Bebeshko: Steel Transl., 2017, vol. 47, pp. 287–90.

  37. 37.

    S.-Y. Kitamura: Hot Metal Pretreatment, in Treatise on Process Metallurgy, Volume 3: Industrial Processes, Ch. 1.3, S. Seetharaman, ed., Elsevier, Amsterdam, pp. 177–221, 2013.

  38. 38.

    M.A. Bredig: Report ORNL-3391, Oak Ridge National Laboratory, 59 pp., 1963.

  39. 39.

    T. Ogasawara, K.-I. Mori, T. Tominaga, F. Tsukihashi and N. Sano: ISIJ Int., 1996, vol. 36, pp. S30–S33.

    Google Scholar 

  40. 40.

    B. Neumann, C. Kröger and H. Jüttner: Z. Elektrochem., 1935, vol. 41, pp. 725–736.

    CAS  Google Scholar 

  41. 41.

    D.A. Wenz, I. Johnson and R.D. Wolson: J. Chem. Eng. Data, 1969, vol. 14, pp. 250–252.

    CAS  Google Scholar 

  42. 42.

    F. Patsiogiannis, U.B. Pal and K.-C. Chou: ISIJ Int., 1996, vol. 36, pp. 1119–1126.

    CAS  Google Scholar 

  43. 43.

    J. Liu and Y. Kobayashi: Metall. Mater. Trans. B, 2017, vol. 48B, pp. 1108–1113.

    Google Scholar 

  44. 44.

    C. Rosenkilde, L.H. Arnesen and O. Wallevik: J. Phase Equilib. Diffus., 2004, vol. 25, pp. 255–258.

    CAS  Google Scholar 

  45. 45.

    H.E. Friedrich and B.L. Mordike: Magnesium Technology. Metallurgy, Design Data, Applications. Springer, 2006.

  46. 46.

    J.J. Lukasko and J.E. Murphy: Report of Investigations 9315, U.S. Bureau of Mines, 1990.

  47. 47.

    H. Jalkanen and L. Holappa: converter steelmaking, in Treatise on Process Metallurgy, Volume 3: Industrial Processes, Ch. 1.4, S. Seetharaman, ed., Elsevier, Amsterdam, pp. 223–270, 2014.

  48. 48.

    J. Madias: Electric furnace steelmaking, in Treatise on Process Metallurgy, Volume. 3: Industrial Processes, Ch. 1.5, S. Seetharaman, ed., Elsevier, Amsterdam, pp. 271–300, 2014.

  49. 49.

    L. Holappa: Secondary steelmaking, in Treatise on Process Metallurgy, Volume 3: Industrial Processes, Ch. 1.6, S. Seetharaman, ed., Elsevier, Amsterdam, pp. 301–345, 2014.

  50. 50.

    Y. Ge, S. Zhao, L. Ma, T. Yan, Z. Li and B. Yang: Metals, 2019, vol. 9, p. 253.

    CAS  Google Scholar 

  51. 51.

    H.-N. Lou, C. Wang, B.-X. Wang, Z.-D. Wang, Y.-Q. Li and Z.-G. Chen: Metals, 2018, vol. 8, p. 534.

    Google Scholar 

  52. 52.

    I.-H. Jung, S.A. Decterov and A.D. Pelton: Metall. Mater. Trans. B, 2004, vol. 35, pp. 493–507.

    Google Scholar 

  53. 53.

    I.-H. Jung, S.A. Decterov and A.D. Pelton: ISIJ Int. 2004, vol. 44, pp. 527–536.

    CAS  Google Scholar 

  54. 54.

    X. Yin, Y.H. Sun, Y.D. Yang, X.F. Bai, X.X. Deng, M. Barati and A. McLean: Ironmaking Steelmaking, 2016, vol. 43, pp. 533–540.

    CAS  Google Scholar 

  55. 55.

    Y. Ren and L. Zhang: ISIJ Int., 2017, vol. 57, pp. 68–75.

    CAS  Google Scholar 

  56. 56.

    D. You, S.K. Michelic, C. Bernhard, D. Loder and G. Wieser: ISIJ Int., 2016, vol. 56, pp. 1770–1778.

    CAS  Google Scholar 

  57. 57.

    H. Yi, G. Xu, H. Cheng, J. Wang, Y. Wan and H. Chen: Procedia Environ. Sci., 2012, vol. 16, pp. 791 – 801.

    CAS  Google Scholar 

  58. 58.

    L. De Vos, I. Bellemans, C. Vercruyssen and K. Verbeken: Metall. Mater. Trans. B, 2019, vol. 50, pp. 2647–2666.

    Google Scholar 

  59. 59.

    G. Wranglen: Corr. Sci., 1974, vol. 14, pp. 331– 349.

    Google Scholar 

  60. 60.

    E.G. Webb and R.C. Alkire: J. Electrochem. Soc., 2002, vol. 149, pp. B272–B279.

    CAS  Google Scholar 

  61. 61.

    M. Martschini: Personal Communication, Universität Wien, Austria, 2019.

  62. 62.

    S. Winkler: Personal Communication, i-Themba Labs, South Africa, 2019.

  63. 63.

    S.J. Parry, B.A. Benet, R. Benzig, D. Redpath, J. Harrison, P. Wood and F.J. Brown: Anal. Chem., 1997, vol. 69, pp. 3049–3052.

    CAS  Google Scholar 

  64. 64.

    B. Zhang, J. Wang, B. Wu, X.W. Guo, Y.J. Wang, D. Chen, Y.C. Zhang, K. Du, E.E. Oguzie and X.L. Ma: Nat. Commun., 2018, vol. 8, 2559, pp. 1–9.

    Google Scholar 

  65. 65.

    J. Park and Y. Kang: Steel Res. Int., 2017, vol. 88, 1700130, pp. 1–26.

    Google Scholar 

  66. 66.

    M.A. Streicher: Corrosion, Intergranular. Encyclopedia of Materials: Science and Technology, Elsevier, pp. 1689–90, 2001.

  67. 67.

    P.C.S. Wu: Report ORNL/TM-6311, Oak Ridge National Laboratory, 100 pp., 1978.

  68. 68.

    S. Dolinšek: J. Mater. Process. Technol., 2003, vol. 133, pp. 63–70.

    Google Scholar 

  69. 69.

    Standards Australia: Australian Standard 3582.2, SAI Global, ISBN 978 1 76035 397 1, 2016.

  70. 70.

    L. Souza, E.R. Zolett and R. Carazedo: Ibracon Struct. Mat. J., 2016, vol. 9, pp. 572–586.

    Google Scholar 

  71. 71.

    I.Z. Yildrim and M. Prezzi: Adv. Civ. Eng., 2011, 463638.

  72. 72.

    S. Louhenkilpi: Continuous Casting of Steel, in Treatise on Process Metallurgy, Volume 3: Industrial Processes, Chap. 1.8, S. Seetharaman, ed., Elsevier, Amsterdam, 2014, pp. 373–434.

  73. 73.

    A.L.V. da Costa e Silva: J. Mater. Res. Technol., 2019, vol. 8, pp. 2408–22.

  74. 74.

    M.P. Ryan, D.E. Williams, R.J. Chater, B.M. Hutton and D.S. McPhail: Nature, 2002, vol. 415, pp. 770–774.

    CAS  Google Scholar 

  75. 75.

    S. Li and L. H. Hihara: J. Raman Spectr., 2017, vol. 48, pp. 137–142.

    CAS  Google Scholar 

  76. 76.

    C. Capurro, G. Cerrutti, and C. Cicutti: 9th International Conference on Clean Steel, Budapest, Hungary, September 2015.

  77. 77.

    N.E. Holden, T.B. Coplen, J.K. Böhlke, L.V. Tarbox, J. Benefield, J.R. de Laeter, P.G. Mahaffy, G. O’Connor, E. Roth, D.H. Tepper, T. Walczyk, M.E. Wieser and S. Yoneda: Pure Appl. Chem., 2018, vol. 90, pp. 1833–2092.

    CAS  Google Scholar 

  78. 78.

    S. Tolmachyov, S. Ura, N. Momoshima, M. Yamamoto and Y. Maeda: J. Radioanal. Nucl. Chem., 2001, vol. 249, pp. 541–545.

    CAS  Google Scholar 

  79. 79.

    J.C. Tait, R.J.J. Cornett, L.A. Chant, J. Jirovec, J. McConnell, and D.L. Wilkin: Materials Research Society Symposium Proceedings, vol. 465, Materials Research Society, Pittsburgh, PA, pp. 503–510, 1997.

  80. 80.

    S.C. Sheppard, L.H. Johnson, B.W. Goodwin, J.C. Tait, D.M. Wuschke and C.C. Davison: Waste Manage., 1996, vol. 16, pp. 607–614.

    CAS  Google Scholar 

  81. 81.

    Y. Pipon, N. Toulhoat, N. Moncoffre, L. Raimbault, A.M. Scheidegger, F. Farges and G. Carlot: J. Nucl. Mater., 2007, vol. 362, pp. 416–425.

    CAS  Google Scholar 

  82. 82.

    D. Calmet, N. Coreau, P. Germain, F. Goutelard, P. Letessier, C. Frechou, and D. Maro: Radiochemical Measurements Conference (47 BAER Conference), Honolulu, 2001.

  83. 83.

    I. Aitchison and P.H. Davies: J. Nucl. Mater., 1993, vol. 203, pp. 206–220.

    CAS  Google Scholar 

  84. 84.

    D. Charquet and F. Alheritiere: Proceedings Zirconium in Nuclear Industry, 7th International Symposium, ASTM STP 939, pp. 284–291, 1987.

  85. 85.

    X. Hou, L.F. Østergaard and S. Nielsen: Anal. Chem., 2007, vol. 79, pp. 3126–3134.

    CAS  Google Scholar 

  86. 86.

    W. Hummel: Nagra Technical Report NTB 17-05, Wettingen, Switzerland, 2017.

  87. 87.

    Nirex: Report 772, United Kingdom, 1997.

Download references


We are grateful to Elmar Schuster (Voestalpine Stahl Donawitz GmbH) and Gregor Mori (Montanuniversität Leoben) for sharing their insights about the role of chlorine in steelmaking and corrosion processes. We also thank Stephan Winkler (i-Themba Labs, South Africa) and Martin Martschini (Universität Wien) for discussing their 36Cl AMS analyses[7] with us. EK gratefully acknowledges a research grant by SKB to carry out this project.

Author information



Corresponding author

Correspondence to Erich Königsberger.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Manuscript submitted August 27, 2020; accepted December 12, 2020.

The original online version of this article was revised: This erratum is to correct the title of Appendix A. The title contains the chemical formula of a chlorine isotope, which is written as 36CL. This is incorrect and should be changed to 36Cl. It should appear as: APPENDIX A: POTENTIAL SOURCES, ANALYSES, AND ENVIRONMENTAL IMPACT OF 36Cl REPORTED IN THE LITERATURE.

Appendix A: Potential Sources, Analyses, and Environmental Impact of 36Cl Reported in the Literature

Appendix A: Potential Sources, Analyses, and Environmental Impact of 36Cl Reported in the Literature

Chlorine has 25 isotopes in total. These include two stable isotopes, 35Cl and 37Cl, with natural abundances of 75.76 pct and 24.24 pct, respectively, and 36Cl, which is the only long-lived radioactive chlorine isotope, with a half-life of (3.013 ± 0.015)·105 years. All the other radioactive isotopes of chlorine are short-lived (half-lives of minutes or less).[77]

Much attention has been paid to 36Cl in the environment due to its long half-life and its natural and anthropogenic origins.[1] Trace amounts of radioactive 36Cl exist in the environment, in a ratio of about (7 to 10)·10−13 to 1 with stable chlorine isotopes.[2,3] This corresponds to a concentration of approximately 1 Bq/(kg Cl).[4]

36Cl is produced in the atmosphere by spallation of 40Ar by interactions with cosmic ray protons and by the nuclear reaction 36Ar (n, p) 36Cl with cosmic ray neutrons. In the top meter of the lithosphere, 36Cl is generated primarily by thermal neutron activation of 35Cl and spallation of 39K and 40Ca.[3,78]

In a nuclear reactor, 35Cl that is present as an impurity in the cooling water, nuclear fuel, metallic materials, or in concrete is activated by thermal neutron capture to 36Cl. The reaction 35Cl + 1n → 36Cl + γ is characterized by a high cross section (43.63 barn); thus, even trace amounts of 35Cl in nuclear fuel and structural reactor materials are sufficient to produce significant amounts of 36Cl, in the presence of a high flux of thermal neutrons.

36Cl decays in one step to stable isotopes, mainly by beta minus decay (β, 98.1 pct) to 36Ar, and a small amount by electron capture/beta plus decay (ec/β+, 1.9 pct) to 36S. Hence, the decay products are mainly the noble gas argon and some sulfur.

Once scattered through the environment, chlorine quickly acquires the chloride anionic form Cl, the thermodynamically stable species of chlorine in water. Natural redox processes virtually never involve this form. Dissolved chlorine Cl2(aq) and most oxidized forms of chlorine such as hypochlorite ClO, chlorite ClO2, and chlorate ClO3 are highly reactive and are quickly reduced to chloride in the environment.

The only fairly stable form of oxidized chlorine in aqueous solutions is perchlorate, ClO4, due to the kinetic stability of its tetrahedral structure. Due to its relative stability, perchlorate can also occur as a contaminant in drinking water and groundwater; its isotopic composition could be used to identify the sources of contamination.[77]

The original chemical form of 36Cl in radioactive waste forms is unknown. However, in the long run and especially under reducing conditions, which are expected in most of the planned deep underground repositories for radioactive waste, all the oxidized forms of chlorine, including perchlorate, if present at all, will be reduced to chloride.

Chloride forms highly soluble salts with all major and minor components of ground and surface waters: all alkali and alkaline earth elements, iron, manganese, nickel, copper, aluminum. These salts are only important if rock salt is considered as host rock of a geological repository. The only sparingly soluble phases, also known as naturally occurring minerals, are chlorargyrite (AgCl) and calomel (Hg2Cl2). Other pure phases formed with “soft” metal cations like HgCl2 or PbCl2 are soluble in the gram per liter range.

In the case of a spent fuel repository, the anion Cl will not sorb on bentonite minerals due to electrostatic repulsion from the negatively charged clay mineral surfaces, and it will not be incorporated in any solid phase of the bentonite backfill or the host rock.

In fact, chloride is used as “conservative tracer” in diffusion experiments with clay or granite containing solids, as it is not retarded by any sorption or co-precipitation effects.

Considering the chemical behavior of the chloride anion in clay or granitic ground waters, acting as a “conservative tracer,” the large isotopic dilution of 36Cl by stable 35Cl and 37Cl in the near and far field of a deep geological repository, and the long half-life of 36Cl, no significant retardation effects are expected for 36Cl.

Hence, 36Cl is a dose-relevant nuclide in geological repositories for radioactive waste relying on the retention of radionuclides as a safety principle. Mitigating the dose of 36Cl by improving sorption in the engineered barriers does not appear to be feasible.

Therefore, an accurate estimate of the source term will be crucial for the post-closure safety assessment of radioactive waste repositories. However, the inventory of 36Cl is almost exclusively derived from calculations based on measured or merely assumed trace concentrations of its stable isotopes, which are impurities in nuclear fuel or structural reactor materials.

The only measurements of stable chlorine and 36Cl in nuclear fuel have been published by Tait et al.[79] They found that the total average Cl impurity level of four un-irradiated CANDU UO2 fuel samples was (2.3 ± 1.1) ppm, which is less than the 5 ppm initial Cl impurity concentration in spent fuel, assumed in the first safety assessment calculation for 36Cl.[80] The latter value is cited also in recent publications, e.g., Pipon et al.[81]: “Pristine chlorine … is present as an impurity in the nuclear fuel (< 5 ppm).”

No 36Cl is left in vitrified high-level waste from reprocessing. The volatile chlorine has evaporated during the reprocessing steps of the dissolved spent fuel; it is found in the surroundings of reprocessing plants.[82] Thus, even if fuel dissolution was involved during the production of the historical waste, no 36Cl from the fuel inventory is expected to remain in the waste.

The cladding also contains traces of stable chlorine. The only measurements published are for Zr-2.5Nb pressure tubes for CANDU reactors: Aitchison and Davies[83] report Cl impurities between 1 and 5 ppm. They state that the chloride in Zr-2.5Nb tubes originates from the Kroll process for producing Zr from crude Zr sponge. Zirconium is reacted with Cl2 to produce ZrCl4, which is purified by sublimation and then reacted with molten Mg to regenerate Zr and leave MgCl2.[83] By using Auger spectroscopy and EDX analysis, they detected an approximately 10 nm layer of complex zirconium-carbide-chloride precipitate in strips of low-energy fractures on Zr-2.5Nb tubes. Stringers containing Cl have also been observed in Zircaloy-2 and Zircaloy-4,[84] where stringer density in forged and rolled products depended on vacuum re-melting parameters.

Stainless steel is used extensively in nuclear reactors as a construction material and may be exposed to high neutron fluxes. Even if the 36Cl content in stainless steel is assumed relatively low, e.g., set to 1 ppm as the detection limit of the analysis method used, it has quite an impact on the results of the analysis because of the extremely high mobility of the nuclide. However, Robertson et al.[5] report that SS 304 used in reactor internal hardware has Cl concentrations ranging from less than 50 to 130 ppm Cl. Hou et al.[85] state: “No report on the analysis of steel for 36Cl is available.”

Nothing is known about the chemical form of 36Cl in any of these waste forms.

The report of Robertson et al.[5] is often cited as literature source for the content of <50 to 130 ppm chlorine in SS 304 (e.g., Hummel[86]). A closer look at that report[5] shows that these values are mentioned only in Table 2.3 of the report; at the bottom of the table also the sources are given where the results collected in the table were taken from (without specifying which reference is for which value). Their own measurements, reported in Table 5.1 for SS 304, result in chlorine values below detection limit for all 3 samples of SS 304 analyzed, and in the text (p. 49) they state that “Cl-36 was detectable in only the spent ion exchange resin samples…”.[5] Inspection of the data[5] revealed that the values <50 to 130 ppm chlorine content in SS 304 originate from Evans et al.,[6] which is listed at the bottom of Table 2.3 of the Robertson et al. report.[5] It is interesting to note that Evans et al.[6] analyzed un-irradiated steels by neutron activation, while Robertson et al.[5] analyzed irradiated samples using radiochemical methods.

In the Evans et al. report,[6] it is stated that they obtained SS samples from various reactors under construction and analyzed 36Cl by neutron activation. They calibrated the neutron activation method with standards, using NBS steel standards and the results are given in Table 4.3, p. 39 of their report. For 36Cl analysis, there are no values reported in the table for a total of 11 (5+6) NBS SS samples, while, e.g., for fly ash in the same table they report 44 ppm and 42 ppm for the fly ash standards (or previous measurements). From this we assume that (a) there is no detectable 36Cl in the 11 SS samples from NBS and (b) this is probably a double check, because NBS also does not report any chlorine in the samples, based on previous analysis.

The analysis of the steel samples from the various reactor sites is reported in Table 4.7 and the average composition of steel in Table 4.8 of the Evans et al. report.[6] In Table 4.7, one reads that the 36Cl data for the various steel samples were North Anna <50 ppm, Susquehanna <60 ppm, Enrico Fermi <60 ppm and <70 ppm, and Belle-Fonte 130 ppm. In Table 4.8, the average 36Cl in SS 304L is reported as 70 ppm, with range <50 to 130 ppm for the 5 samples analyzed. Thus, it appears that the only reason why they report above-detection-limit Cl content in SS 304 is the result of the analysis of the Belle-Fonte sample (130 ppm). In the report it is stated that 2 samples of SS 304 from Belle-Fonte were obtained according to Table 3.2, one from the reactor vessel and one from vessel cladding. We could not find which of these two (or both) gave 130 ppm. On the other hand, they have analyzed 11 NBS standard samples and 4 samples from other reactors which all resulted below their detection limit. No mention of these samples is made in the Robertson report,[5] which has 3 additional own steel samples analyzed resulting in chlorine contents below detection limit.

A much more intensive 36Cl analysis campaign in various ferrous materials was undertaken by Nirex, UK in the late 1990s. The results were summarized in a report.[87] The analytical method used was radiochemical neutron activation analysis (RNAA). According to this method, the fast dissolution of about 0.7 g of neutron activated ferrous material in the form of millings was followed by separation of chlorine and counting of 38Cl. In total, 176 SS and other ferrous material samples were analyzed. The measured chlorine concentrations in 170 of 176 samples were below detection limits, which were all low, in the range 0.06 ppm to 0.97 ppm, except for Cr/Mo/V steel for which the detection limits were between 0.26 ppm and 3.3 ppm. Other impurities in ferrous materials such as 34S and 39K are not significant sources of 36Cl.

In their conclusions, the authors of Reference 87 state, “These measurements confirm the view that the PNNL results, which were the only measurements available before this program of work, are suspect. Chlorine concentrations in excess of 100 ppm are considered implausible for a number of reasons: during the manufacture of steel blast-furnace temperatures are sufficiently high to cause volatilization of iron chlorides; it is important to produce steels with low chlorine concentrations because steels are known to corrode when exposed to chlorine. All the evidence suggests that the measurements of chlorine concentrations above the limits of detection by Pacific Northwest Laboratory were doubtful, and the limits of detection were so high that the measurements below these limits are of little value.”

The analytical method used in the Nirex report,[87] together with the results of the analysis of 30 steel samples for 6 different steel types, has been published in the open literature[63] and values below detection limit were reported for all samples.

Similar values below detection limit were reported by Hou et al.[85] for irradiated SS (below 0.2 ppm) from a Danish reactor.

The use of AMS (Accelerator Mass Spectroscopy) allows 36Cl measurements in steel with sensitivity better than 10−14 relative to stable chlorine or detection limits of the order of 1 ppb.[7] The measurements showed much higher chlorine levels in the first and second leaching of the ‘surface layer’ as compared to the levels of chlorine measured after the final, total dissolution of the sample (see discussion in Section III–F). The paper[7] also mentions the analysis of drill cores from 4 steel samples (cylinders of 1 mm diameter and 4 mm length) which are expected to give better reproducibility. The results of these analysis were delivered by O. Forstner, one of the co-authors of the paper,[7] to SKB and are reported here. The analysis of 6 doublet samples from 3 types of steel were below the detection limit of 2 ppb, while only one type of steel (58316) had 5.8 ppb and 6.0 ppb in the doublet samples, with an error of 0.5 ppb. The drill core samples were apparently taken from different steel types than the drill shaving samples of Winkler et al.[7]

It is evident that with more sensitive analytical techniques such as AMS, one can certainly exclude chlorine concentrations of the order of hundreds of ppm. With the hypothesis of chlorine contained in non-metallic inclusions or in slag inclusions and assuming a maximal content of 0.1 pct Cl in them, the steel of the Belle-Fonte sample discussed above should have contained 13 pct non-metallic or slag inclusions. This high percentage could have resulted more likely from large, undetected, exogenous inclusions rather than from the finely dispersed endogenous, non-metallic inclusions that usually originate e.g., from deoxidation processes in secondary refining; however, sample contamination or analytical errors might also have been important factors.

In summary, finding chlorine contents below 2 ppb in 75 pct of the analyzed samples gives a good ground to the claim that very low, if at all, levels of chlorine have to be expected in steel.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Königsberger, E., Spahiu, K. & Herschend, B. Thermodynamic Study of the Chlorine Content of Stainless Steel. Metall Mater Trans B (2021).

Download citation