DOI: 10.1080/02670836.2017.1309111 Niobium microalloying in rail steels Arunin Ray Depattuncul of Materials Scie w & Metallurgy, University of Cambridge, Cambridge CB3 0FS, UK email: arunimray@gmail.com, ar721@eam.ac.uk Abstract Rail steels rely primarily for their properties on wear and rolling-contact-fatigue resistance, These properties, together with toughness, can be optimised by implementing thermo-mechanical processing during rail man- uufacturing, assisted by controlled additions of strong carbide forming mieroalloying elements. Niobium ean be used at low concentrations, but the major difficulty is to ensure that niobium does not segregate during casting, especially when rail steels are high earbon, The solubility of niobium in ternary Fe-Nb-C austenite is critically assessed to address the issue of segregation. Thereafter, the effect of niobium on medium and high carbon (~0.2-0.8 wt%C) austenite grain-growth during reheating or reerystallisation kineties during hhot deformation and corresponding effect on pearlitic and bainitie microstructures and mechanical properties are reviewed,Introduction ‘There are several criteria which determine the suitability of a steel for rail track applications, the most im- portant being wear and rolling-contact-fatigue resistance. The general characteristics of rail steels have been reviewed by Bhadeshia {1} and it is noteworthy that the wear resistance of rails has inereased dramatically ‘over the decades, leanliness of the steels with respect to non-metallic inclusions has also been improved ‘over the years, leading to a reasonable improvement ia the rolling-contaet-fatigue resistance. ‘The structural integrity of rails can be compromised by localised plasticity during service, for example ‘on nomlinear sections of the track. Resistance to brittle fracture as defined by the toughness of the steel Is auotlier prime property requirement for rail steels. ‘The common way o Increase the wear resistance of a pearlitic steel is to reduce the inter-lamellar spacing, which hardens and strengthens the steel, It is well-known that an individual pentlite colony is an inter-penetrating bi-crystal of cementite and ferrite [2.3]. ‘Therefore, the erystallographie grain size, an aspect which determines toughness (1), is not necessarily refined by reducing the inter-lamellar spacing. It follows that pearlitic steels are hard but not particularly tough and therefore can be prone to brittle failure during service. Niobium is used as an microalloying element for precipitation strengthening and grain refinement in. low carbon high-strength steels [4-8] such as those used in manufacturing structural steels (pipes, plates, beams). Whether controlling austenite grain coarsening during reheating or inhibiting austenite recrystallisation during hot deformation by precipitate pinning, the theory and technology associated with thermo-mechanical processing is well established for these low carbon steels (0.01-0.2wt% C). But this aspect has rately been explored in the context of high earbon (~0.6-0.8 wi%C) steels like rails. ‘The major issue is the segregation during casting, leading to Nb and C rich liquid that is the last to solidify and resulting into coarse niobium carbides tat cannot be dissolved. ‘The lteracure on te microatloying of high-carbon steels is tmted [9,10], but the indications are that niobium additions would be beneficial in the context of rail steels [1] which also ssffer from the fact that the deformation experienced in different regions of the rail is not uniform because of their asymmetrical cross-section. The benefits are re ed to both in limiting the austenite grain size, and in modifying the subsequent development of pearlite ro aro a numnber of difficulties not addressed in the le literature: (i) The data reported are usually for alloys investigated in relatively small scale laboratory experiments. (ii) Full seale trials of the manufacture and testing of rails containing niobium do not seem to have been conducted, or alternatively, such trials have not heen reported in accessible literature.il) Issues due to the segregation of niobium added to large melts and the formation of coarse deposits of NDC (which rena us & permanent feature ia e steel), are seldom discussed. Alternative to pearlite have been developed, specifically banitie steels that are fee from cementite, and have ‘outstanding rolling-contact-fatigue performance [1]. To use niobium as an effective microalloying element in rail steels, the solubility of niobium carbide is an important parameter to determine the amount of effective mieroalloying in steel, To maximize precipitate pinning of grain boundaries, it is necessary to dissolve the enrbide ' or nitrides completely at the reheating temperature and precipitate them in the form of fine dispersions and stable particles during rolling, Though the decrease in niobium solubility in austenite with Inereasing carbon content is well-established, the possibility of an increased niobium solubility at high eaxbon contents has been reported by Balasubramanianet al. [11] and Obtanietal. (12). To addiess this anomaly, the description of niobium carbide solubility in ternary Fe-Nb-C austenite is reviewed and compared against thermodynamic databases, The effect of niobium on pearlitic and bainitie microstructures and mechanical properties are discussed thereafter, Description of niobium carbide solubility in Fe-Nb-C system There have been many experimental determinations as well as thermodynamic predictions of the solubility of niobium carbide precipitates in ternary Fe-Nb-C austenite [13-18,20-28}. The results have been reviewed ‘over time and sometimes reassessed to provide a more suitable thermodynamic description of NbC solubility in quotonite [11,12,22 25]. "The carly exporimentel mousuremente and thermodynamie caleulations of earbide solubility were based on the assumption that carhides (or nitrides) of transition elements such as Ti, Nb, V are stoichiometric in composition and that the anstenite is devoid of any ternary solute interaetions. The methodology considers the austenite-NbC equilibrium as follows: [Nb] + [c] = NbC () RT In [axe ac! °Gsvc re) ‘where [Nb] and [C] are elements dissolved in austenite, ays, and a are the activities of the elements, A°Gyie: is the standard free energy of formation of stoichiometric carbide from 1wt% austenite solution. Assuming dilute solutions in austenite, the solubility product ean be defined ass RT In (%Nb] [%C] = A°G yo (3)log [%Nb] [%C] = A+ B/T a where [%Nb] and [%C] are the weight percents of niobium and carbon dissolved in austenite respectively, A and B are approximate constants related to the enthalpy and entropy of carbide formation respectively. A summary of varlous investigations up to 1980s on NbC solubility 4s given in terms of A and B values in Table1. Some investigations showed an improved prediction with the experimental solubility data. for carbides by assuming the earbide compesition to be NbCosr instead of perfect stoichiometry. ‘Table 1: Solubility of niobium carbide in anstenite (11. Salnhility Proce’ Hemperstire og BANDLG] = A+ B/T Reference = A Ta-aR—2e0 1500 De Kasia, Toa] 1273 - 1573 K 3.18 ‘Mori, 1964 [14] ima isn a2 Nast, 1966 (3 ura-1s3K 308 Meyer, 196 16 wrs-1s73K 370 Sith, 196 (17) 1225 1328 K Johnsen, 1967 (5 1293 - 1523 K Irvine, 1967 [19] urs -1473K Nordberg, 1968 * [20 1323 - 1423 K Koyama, 1971 [21] 1273 - 1523 K Lakslummanan, 197 * [22 1173 - 473 K 7020 Sharma, 1984 * [23] * The precipitates were assumed to be Nbow Balasubramanian (11), reported two drawbacks in calculating NbC solubility by the above methodology. (i) Transition elements like ‘Ti, Nb, V ate strong carbide and nitride formers. So interactions between solutes need to be accounted for (ii) NDC is non-stoichiometric and hence its composition ean vary when precipitated in steels. This rules out a simplified analysis, An increase in Nb solubility in austenite has been reported at high-carbon concentrations by Balastib- ramanian (11). A solubility-minimum was observed (around 0.4-0.5 wt% ©) at 1273-1473 K. This was at- (ributed Lo the strong solute interactions int austenive and non-stoie miolry of Ue carbides. Th was argued that earlier investigations [13-23] were performed at Joy carbon concentrations (below 0.2 wt%) without addressing the solute interactions and carbide non-stoichiometry and hence are not capable of predicting the solubility at higher carbon levels where the effects of solute interactions become significant. ‘The elas- sical solubility product expression (equation 4) was modified by neorporating an additional term for theconcentration dependence on carbon [11]. log. DaNb] [KC] = A+B/T + (C/T — D).1%C} 6) ‘Thus, ralsing the carbon content Increases the solubility product of NbC and more nloblum ean be retalned in solution. Parallel but independent work by Ohtani ot ol. [12] also omggested an increase in NbC colubility at high-carbon concentrations hy taking account of strong interactions between niobium and carbon atoms in austenite, Table? and Figure compare the two separate thermodynamically caleulnted solubility data in 11] and (12), both reporting an inerease in Nb solubilicy at high-carbon concentrations. Both groups showed fair agreement at lower temperatures (<1373 K) but the difference widens with increasing temperatures. Different values for the free energy of dissolution of niobium in austenite are the cause of this anomaly in NUC solubility data, Table 2: Solubility limit and temary interaction in Fe-Nb-C system {11}. Composition and temperature are in wi% and K respectively. 98, is the activity co-efficient of dissolved niobium in austenite Solubility Limit ~(8080/7) + 889 + (1150/7) - 0.05}.%6C] Balasubramanian [11] log ANbILKC] -(4880/T) + 148 + [(1690/T) - 0.10].(%C] Ohtani et al. [12] Free energy of aissolution of 27990 + 18.007 Jmol * Balusubranantan [11] niobium in -88235 + 61.857" Jmol“! Ohtani et al. [12] (BSRAN/T) 4.95 -(87620/T) + 5.0 Obtani et al. [12]TR a TAK = = “ ‘Balasubramanion (11) gos ommiseain fg ox a onanicrantiay /) § 02 NY ae bes Bois] aimeminisy $7] Bou 17 Balastrarian (11) t\ 7s a CP catenrwes > (BY) catonrwes 7 CD? cxtonrwre ? Figure 1: Comparison between thermodynamically calculated solubility data by Balasubramanian [11] and Ohtani et al. [12): (a) 1273 K, (b) 1373 K, (¢) 1473 K, Complete original experimental data for NbC solubility from Ohtani’s work have not been possible to trace Only one instance of experimental data from Ohtan‘'s work (experimental data points up to1.wi%C) can be found from [26]. The corresponding complete original references ([27,28] cited in [26)) could not be obtained. Figure2 compares their experimental data with those of Balasubramanian [LI] (experimental data up to 2.01% C) and an earlier experimental investigation by Lakshmanan [22] which was limited to only 0.2wt%C. There is no apparent solubility-minimum in the experimental data from Ohtani et al. 12). ‘This contradicts the solubility minimum at about 04 0.5:8t%C in the th iodynamieally fitted curve on the same solubility data of Ohtani et al. from Figure te. Also, experimental data from Ohtani et al. agree well with those of Lakshmanan 2] for carbon. concentrations up to 0.2 wt%, But there remains a discrepaney with the experimental data from Balasubramanian [11] where a clear solubility-minimum can be observed (Figure2). The NbC solubility in Fe-Nb-C austenite was measured by Balasubramanian (11) from the deflection in equilibrated carbon content variation with Nb additions at constant carbon activity during dynamic gas-equilibration experiments [11]. The reported error is calculated by assuming negligible error in the determination of wt%C and equating the absolute error in the solubility product to the relative error in determination of wt%Nb. As shown in Figure3a, at high earbon levels (>1,0wt%C), the change of slope in the iso-aetivity curves is significant and hence the error in the determination of C and Nb compositions are small, However, at lower carbon levels (<1.0 wt%C)(FPigure3b-c), the error should be relatively higher due to a less pronounced! change in slope and error in measured wt% C should not be neglected. So in Figure 2, the experimental NbC solubility data at high carbon levels (2L0wt% C) due to Balasubramanian [11] may be accurate, but the solubility at lower carbon contents (<1.0wt% C) should contain a larger amount of error than reported,@) sw] BRK i) San tt : Omani etal (12) 4 Baonbromsran (11) , i ’ Balasubramanian [11] Poa 4 aaa on oo: * i" Huang (24 ee 001 pxmanan [22] Huns f Caxmanan 22]! 0 ° 0 2 ° as I Ts Carbon! wi% Figure 2: NbC solubility in austenite as a function of the carbon concentration: (a) experimentally measured at 1473 K, (b) thermodynamically calculated at 1273 K. in on On “ ost - ne 1473 K : 1473 K Figure 3: The carbon content of equilibrated Fe-Nb-C alloys at 1473 for different earbon potentials (a) high carbon potential [gas ratio (peut, /p},,)=1-34% 10"), (b) medium earbon potential (pex,/P}j,=0.61x10™), (©) low carbon potential (pon,/P%,0.28%10"). Open vizeles with error bars indicate Ue ineasured sole bility data at the corresponding carbon potential [1] Huang [24], optimized the thermodynamic description for Fe-Nb-C ternary system by using Nb-C and Fe-Nb data and the Fe-C description by Gustafson [28]. This work was reviewed by Lee [25], along with the other earlier experimental investigations on NbC solubility ineluding the anomaly between Ohtani etal. [12] and Balasubrat mn [LI], ‘The calculated solubility using Huang’s description [24] is too low, especially at low temperature (1273 K), when compared with data from (12) and {11} as shown in Figure 2h. In contrast, no inerease in solubility can be observed at large carbon contents in (24). These differences were attributed to @ more positive value of the interaction energy parameter Lyo,xy (-4784J mol) in Huang's calculations 24) than in [12] and [11]. ‘The value of the ternary interaetion parameter was calculated in [24] as eG), = -98276/7+3.1 whieh is more positive and does not correspond to the strong mutual soluteinteraction as compared with those of [12] and [11] (Table2). It was also argued by Huang [24] that the calculated activity of carbon from Bulasubramanian’s [11] experiaiental data un NUC solubility in auste could show a large error since the carbon activities of the Fe-C binary obtained by [1] show poor agreement with the well-established data by Gustafson [20], Balasubramanian et al, [30] also acknowledged this faet that some uncertainty stil exists in the description of Fe-Nb-C system. But the data from Huang’s calculation [24] were consistent with earlier investigations [13-18,20-23] in the ‘same range of temperature (1273 to 1473 K). Lee [25] developed a self-consistent thermodynamic description for the FeNb-Ti-C-N quinary system and on the way reassessed Huang’s work [24] on Fe-Nb-C ternary ystems in terms of its sub-binary aystems ie. Fe-Nb, Nb-C and Fe-C. In particular the Fe-Nb description was revised by making Lyaxp temperatuse dependent instead Keeping constant. A slight change in Nb-C description was also required but the revision was made in a way that, does not change the temperature dependence of the enthalpy of formation of NbC carbides, ‘Thus Fe-Nb phase becomes more stable in austenite at low temperatures which in turn improves the agreement between calculated and experimentally measured slopes of solubility produet for NbC. Commercial software like Thermo-Cale, MTDATA, MatCale, JMatPro [31] ete. have been developed, which, together with appropriate data, ean be used to calculate phase-equilibria and thermodynamic prop- ertics in multicomponent, multiphase systems. For etcel and Fe based alloys, TCFE ts the most common database which has been continuously developed ani improved over the years by Thermo-Cale Software AB [32]. It has also been translated and tested for use with MTDATTA [32]. MatCale has been developed with ‘me-fe’; ‘mne-steel” and ‘me-sample-fe’ thermodynamic databases for steel and Pe-based alloys [33]. De- tailed review of the database references reveals that the Fe-Nb-C description for TCFE and ‘me.sample-fe" databases have been developed based on the assessment by Lee [25] while for ‘me-fe-v2.009" and ‘me-stee!’, Chermodyuanie ealeulattons by Huang [24] have Docu taker tty account aldiion «9 Usat of Lee 25}. Fi tured compares the NbC solubility calculated by different commercial software and databases. There clearly is considerable uncertainty in calculating the appropriate niobium solubility for a given carbon concentration.