Influence of ferrite-austenite distribution in 2205 duplex stainless steel on high-temperature solution nitriding behaviour

Duplex stainless steel 2205 was subjected to annealing treatments in the range of 1000 – 1200 ◦ C to adjust the phase fractions of ferrite and austenite. Annealed steels were subsequently subjected to high-temperature so-lution nitriding, which transformed the surface-adjacent region into austenite under the influence of nitrogen ingress. Microstructural characterization of the pre-annealed state and the solution-nitrided cases were performed with X-ray diffractometry, light-optical microscopy, electron back-scatter diffraction and hardness indentation. The phase fractions of austenite and ferrite have a significant influence on the nitriding kinetics. A relatively high ferrite phase fraction results in finer grains in the nitrogen-stabilized austenite case. The developing austenite case and the evolution of the microstructure during solution nitriding are discussed in a computational thermodynamics and kinetics context. Slower grain growth in the austenite cases for higher ferrite phase fractions can be understood in terms of phase distribution, alloying element partitioning, and possibly, pinning effect by M 2 N nitrides.


Introduction
Duplex ferritic-austenitic stainless steels (DSS) are a family of stainless steels, relying on the co-existence of two phases, austenite (face-centered cubic lattice) and ferrite (body-centered cubic lattice) [1,2].DSS have significantly better strength and chloride stress corrosion cracking resistance than 300 series austenitic stainless steels, and exhibit higher ductility and toughness than ferritic grades.The properties of DSS strongly depend on the ferrite/austenite phase fractions.DSS 2205, containing 22 wt% Cr and 5 wt% Ni, is a second-generation nitrogen-alloyed DSS, which has excellent structural stability/strength, good fabricability/weldability and excellent corrosion resistance in marine environments.This steel grade is extensively used for gas pipes and other applications on offshore platforms, which account for nearly 60 % of DSS applications [3].
Over the past decades, a gas-based nitrogen (surface) alloying technique, i.e., high-temperature solution nitriding (HTSN), was developed [4], and has been successfully used to case harden martensitic, austenitic and duplex stainless steels [5][6][7].Particularly, the interstitial dissolution of nitrogen in DSS leads to a transformation from dual phase austeniteferrite to monophase austenite, yielding a high-nitrogen containing austenite case atop the duplex core [6][7][8][9][10][11][12][13][14][15].The development of the austenite case provides considerable incentive for a wider application of DSS.In addition to the effective solid solution strengthening effect [16], interstitially dissolved nitrogen significantly improves the pitting and crevice corrosion resistance.This is reflected by a more effective impact of nitrogen on the pitting resistance equivalent number (PREN) than substitutional alloying elements as chromium/molybdenum [17,18].If applied as key components in certain environments, e.g.sewage, the austenite case is more resistant against cavitation than a ferritic or duplex structure, while the duplex core offers excellent energy absorption capacity against shock-induced damage [8,9,11,12].Moreover, the nitrogen-stabilized case may enable the use of DSS beyond (both higher and lower) the recommended temperature range − 50-250 • C for 2205 grade that should prevent the embrittlement of ferrite [19,20], or the formation of detrimental intermetallic phases [1,21] near the surface.
The temperature for HTSN is in the range 1000-1200 • C, where N 2 readily dissociates at the steel surface.This temperature range overlaps with that for solution-annealing of DSS, which might suggest that HTSN is insensitive to the thermal history of DSS, or even, that surface HTSN and core solution-annealing can be performed simultaneously at the same temperature.On the other hand, solution-annealing in this temperature range results in a change in the phase fraction, the composition, the morphology, and the distribution of ferrite and austenite, without causing precipitation of nitrides [22,23].Diffusion of nitrogen atoms proceeds faster in ferrite than in austenite, while the solubility of nitrogen in ferrite is appreciably lower than in austenite [16,24].It is anticipated that differences in phase fractions of ferrite/austenite lead to differences in diffusion kinetics during HTSN, resulting in different growth kinetics of the austenite case.
A serious limitation of solution nitriding of DSS is coarsening of the grains in the nitrogen-enriched austenite case [6,7,11,13].In the dualphase core, grain growth is constrained by the necessity of partitioning of substitutional elements between ferrite and austenite.On the other hand, in the monophase austenite case, such partitioning is not necessary because the stabilizing effect of nitrogen on austenite applies for a broad range of substitutional compositions of austenite.Consequently, considerable grain growth would occur in the austenite case.As reported in Ref. [11], HTSN treatment performed alternately in highpressure N 2 and in vacuum led to cycling between sorption and desorption of nitrogen gas, such that smaller grains resulted in the austenite case.The ferrite ↔ austenite cyclic phase transformation that results from a cyclic variation of the dissolved amount of nitrogen, is responsible for the suppression of austenite grain growth.Accordingly, it is anticipated that a relatively high amount of ferrite, as obtained by preannealing at elevated temperature, could prevent the formation of the coarse austenite grains in subsequent HTSN treatments.In addition, the volume fractions of ferrite and austenite in the dual-phase core depend on the applied HTSN parameters.
In this work, DSS 2205 was subjected to annealing at various temperatures prior to high-temperature solution nitriding.For this reason, the influence of the ferrite/austenite phase fraction on the solution nitriding response was investigated.In particular, the microstructural evolution and nitriding kinetics were studied.

Experimental
DSS 2205 used in this study has the chemical composition listed in Table 1.The material was delivered in hot rolled and 1060 • C annealed condition as Ø 20 × 6 mm rods.Fig. 1a shows the equilibrium phase fractions as a function of temperature as calculated with Thermo-Calc 2017b, applying the TCFE steels/Fe v11.0 database [25].A two-phase region exclusively comprising ferrite and austenite occurs in the temperature range ~1375 • C to 940 • C. The phase fraction of ferrite/ austenite decreases/increases with lowering the temperature; equal fractions of austenite and ferrite are reached at approx.1122 • C.
In order to vary the phase fractions of austenite and ferrite, solution annealing was performed at five temperatures at increments of 50 • C in the temperature range 1000-1200 • C. To this end, steel rods were annealed in a Nabertherm furnace in air for 90 min, followed by water quenching.To exclude the influence of changes in the surface condition after annealing in air, e.g.oxidation or even nitriding, 2 mm thick discs for subsequent HTSN treatments were cut from the middle portion of the annealed rods.Prior to HTSN, the surfaces of the discs were ground with 1000# abrasive paper to remove the cutting-affected surface zone, followed by polishing with 1 μm diamond slurry.Solution nitriding was carried out at 1125 • C (1398 K) in 0.3 bar (total pressure) N 2 gas for 1.0 and 2.5 h in a horizontal tube furnace equipped with a gas quenching system.This temperature was chosen to obtain equal phase fractions of austenite and ferrite in the core region (ref.Fig. 1a).The nitrogen pressure was selected on the basis of the isopleth in Fig. 1b that shows the phase stability for the steel composition in Table 1 as a function of nitrogen content.Iso-activity lines corresponding to the indicated total N 2 pressures are superimposed onto the isopleth.Note that for a starting condition of each annealed case, due to the different metallic compositions of ferrite and austenite, different equilibrium conditions of gassolid interface are expected between austenite and ferrite.A representative example for the 1100 • C annealed condition, shows the isopleth with respect to the equilibrium compositions of ferrite (Fig. 1c) and austenite (Fig. 1d).The combination of 0.3 bar total N 2 gas and 1125 • C temperature offers nitrogen contents of 0.99 wt% and 0.53 wt% at gasferrite interface and gas-austenite interface, respectively.It is noted that the equilibrium N content (0.99 wt%) at the (former) gas-ferrite interface would correspond to the formation of austenite and M 2 N, along with ferrite (cf. the three phase region in Fig. 1c).Under the condition that the metallic composition of austenite is the same throughout the case, for the applied HTSN condition the nitrogen content in γ-phase ranges from ~0.68 wt% at the surface, in equilibrium with the N 2 gas (marked as N s in Fig. 1b), to ~0.55 wt% (N t ) in the sub-surface zone, close to the transition between the austenite case and the dual-phase core.Furthermore, a duplex transition zone with a higher amount of austenite than the core would be present below the monophase austenite.Here the nitrogen content ranges from 0.16 to 0.55 wt% and the ferrite/austenite fraction changes gradually to the level in the core.
To avoid austenite decomposition and the formation of nitrides in the near-surface zone as well as to preserve the structure achieved in the core, the specimens were quenched in 8.0 bar N 2 -gas after HTSN.
For phase identification, X-ray diffractometry was carried out with a Bruker D8 AXS X-ray diffractometer for the scattering angle range 50-160 • 2θ, using Cr-Kα radiation with a step size of 0.03 • 2θ and counting time of 4 s per 2θ step.In the present evaluation, the effect of crystallographic texture on the determined phase fractions was minimized by analyzing a summation over diffractograms measured at psi tilts ranging from 0 to 70 • at 5 • increments.The phase fractions were quantitatively estimated using the direct comparison method [26], according to ASTM standard E975 [27]; the integrated intensities of the 111, 200 and 220 reflections of austenite, and the 111, 220 and 211 reflections of ferrite were considered.For both annealed and HTSN treated conditions, phase analysis was determined on the transverse section.Diffractograms for residual stress analysis with the sin 2 ψ method were recorded in the range 124-132 • 2θ at 0.03 • 2θ increments and a measurement time of 2 s per step, covering the 220 γ peak.Rocking curves were measured in XRD analysis to determine the ψ values used for stress analysis within the range of 0-70 • .The intensity resulting from diffraction of K α2 radiation was stripped in Bruker's DIFFRAC.EVA assuming a 2:1 intensity ratio between K α1 and K α2 .The remaining peaks were fitted with a Pseudo-Voigt function in Origin® software.The slopes obtained from the sin 2 ψ plots were used for the calculation of the stress for the various process conditions; X-ray elastic constants were calculated using single crystal elastic constants of austenitic Fe-12 % Cr-12 % Ni [28].
The annealed and HTSN-treated steel discs were sectioned parallel to the rolling direction, and the surfaces and cross-sections were ground and polished with 1 μm diamond.The annealed steels were etched in HCl/HNO 3 = 3/1 for ~15 s, while the HTSN-treated steels were etched additionally with Kallingʼs reagent to reveal austenite boundaries.Metallographic observations on the annealed and HTSN-treated specimens were performed on a Zeiss Jena Neophot 30 microscope.The ferrite/austenite phase fractions were quantified using electron backscatter diffraction (EBSD) and the elemental partitioning between ferrite and austenite, was determined with energy dispersive X-ray spectroscopy (EDS), both on a Zeiss Supra 35 scanning electron microscope (SEM).The examined objects, surface or cross-section, were electropolished in a "Struers Electrolyte A2" solution at 20 V for ~15 s in a LectroPol-5 electrolytic polishing machine prior to EBSD mapping.EBSD analysis was performed using a step size of 1 μm, with an accelerating voltage of 15 kV and a working distance of 13 mm.Microscope control, pattern acquisition and data solution were carried out with the HKL Channel 5 system.Hardness indentations were performed on a Future-Tech FM700 hardness tester.For all annealed conditions, the bulk hardness was measured at a load 100 g (0.983 N) and a dwell time of 10 s at room temperature, while hardness in ferrite and in austenite was determined using hardness indentation with a 10 g (98.3 mN) load and 10 s duration on the etched surface of the discs in relatively large grains.Hardness depth-profiles of the annealed and HTSN-treated specimens were measured on the transverse sections, in a polished state, applying a 100 g (0.983 N) load and dwell time of 10 s.The hardness values presented are the average of six repetitions.

Microstructural evolution
The microstructural response of 2205 steel to the annealing treatments is revealed by X-ray diffraction analysis and reflected light microscopy.Diffractograms in Fig. 2 and micrographs in Fig. 3 for different annealing temperatures confirm that the typical duplex microstructure, i.e. ferrite and austenite, is present, consistent with the calculated results in Fig. 1a.No indications were found for the development of additional phases.A representative EBSD phase map (inset in Fig. 3b) clearly identifies the duplex microstructure of ferrite and austenite in the annealed 2205 steel.Different from the isotropic structure that was observed in the transverse section (Fig. 3a-c), the grains in the longitudinal section are elongated parallel to the rolling direction, with almost continuous bands of austenite grains (Fig. 3d).The change in annealing temperature leads to a change in the morphology, as particularly observed in the longitudinal section, and changes the fractions and sizes of ferrite/austenite.Annealing at a relatively high temperature, results in more ferrite and an associated breaking up of the continuous austenite bands (Fig. 3f).Moreover, a high annealing temperature leads to ferrite and austenite grain growth as observed in the transverse and longitudinal sections.
The ferrite/austenite phase fractions were quantitatively determined from the X-ray diffractograms, and analyzed with EBSD, as presented in Fig. 4a.Thermo-Calc predictions are incorporated in Fig. 4a for comparison.
All experimental data shows an increase in the volume fraction of ferrite with an increase in annealing temperature, analogous to thermodynamic prediction.Overall, the experimentally obtained ferrite/ austenite phase ratios are higher than the predictions for equilibrium; the agreement appears to improve with annealing temperature.Several reasons can be given for the slight discrepancy.There may be an influence of the starting condition (annealing at 1060 • C), in particular for the lower annealing temperatures, where the deviation from thermodynamic predictions is largest.For the lower annealing temperatures, the annealing time has been too short to reach thermodynamic  equilibrium.For higher annealing temperatures, the faster redistribution of substitutional elements enables approaching equilibrium compositions and thus equilibrium phase fractions of austenite/ferrite.The XRD results may be affected by the presence of crystallographic texture, resulting in either higher or lower ferrite/austenite phase ratios compared to the image analysis result.According to the findings in Ref. [29,30], indeed preferred grain orientations can occur in the annealed states.Quantitative EBSD yielded consistent ferrite/austenite phase ratios for transverse section and longitudinal section.The grain sizes measured with the linear intercept method on EBSD maps are presented in Fig. 4b and reveal progressive grain coarsening in both ferrite and austenite with increasing annealing temperature.Evidently, at all applied annealing temperatures, ferrite grains are coarser than austenite grains, a tendency that is aggravated with annealing temperature and is consistent with previous findings [31].Under the given annealing conditions, DSS 2205 is less sensitive for grain coarsening than austenitic stainless steels, where grain growth is excessive during annealing above 1150 • C [32].

Partitioning
Partitioning of alloying elements among ferrite and austenite depends on the annealing temperature, which is inherent in a two-phase material [13,19,33].The amounts of the main alloying elements as determined by EDS for the applied annealing temperatures are presented in Fig. 5, along with predictions obtained with Thermo-Calc.A fair agreement is obtained between experimental and calculated compositions.Clearly, Cr and Mo are preferentially dissolved in ferrite, while Ni and Mn are preferentially dissolved in austenite, consistent with previous studies [18,26].Carbon and nitrogen contents were not included in the analyses, because their contents are poorly quantifiable with EDS.Thermodynamics predicts that nitrogen is predominantly present in austenite and has very low solubility in ferrite.The precise compositions of the austenite and ferrite phase at the different annealing temperatures determine the nitrogen contents that are dissolvable under

Hardness
The bulk hardness and the hardnesses of the individual phases depend on the annealing temperature (Fig. 6).The bulk hardness, as measured with a load of 100 g (0.983 N), decreases gradually with annealing temperature as a consequence of a gradual increase in grain size.Hardness indentations with a low load of 10 g (98.3 mN) allowed probing the hardness in individual phases.Within experimental accuracy, ferrite and austenite have the same hardness, but a trend is observed that austenite is slightly harder than ferrite for all but the highest annealing temperatures.This is in accordance with previous results [23,34,35], and it was demonstrated that fine-grained austenite in DSS can have a higher hardness than ferrite, associated with the solid solution strengthening provided by interstitial nitrogen [16].Furthermore, the austenite grains are systematically smaller than the ferrite grains (see Fig. 4b).Therefore, it cannot be excluded that neighboring grains affect the hardness measurements, particularly for austenite.The observed trend towards a vanishing difference between hardness values in ferrite and austenite could be associated with grain growth (Fig. 4a), so that the influence of neighboring grains is reduced at the higher annealing temperatures.

Microstructural evolution
X-ray diffractograms of the pre-annealed and subsequently solution nitrided steels are given in Fig. 7 for the five annealing temperatures.In all cases, the diffractograms determined at the surface contain mainly diffraction peaks of austenite.Due to the strong intensity and narrow peaks in the low scattering angle region, all diffractograms contain tiny satellite peaks at smaller diffraction angles relative to the intense 111 γ and 220 γ peaks which arise from diffraction of K β radiation by {111} and {200} families of lattice planes, respectively.Occasionally, very weak iron oxide (Fe 2 O 3 ) peaks are observed; no indications of chromium nitrides (Cr 2 N) were observed.This confirms that for all annealed temperatures, solution nitriding treatments for 1.0 h and 2.5 h, produced an austenite case atop the dual-phase base.Similar to the annealed states, for the solution nitrided steels (Fig. 2), the 111 γ and 200 γ reflections are dominant as compared to the 220 γ reflection; moreover, the 200 γ is more prominent than for the annealed case.This indicates that crystallographic texture changed upon solution nitriding, which could be ascribed to the phase transformation from dual phase to monophase austenite and, possibly, growth selection among the grains in the austenite case.
Cross-sectional micrographs of the solution nitrided specimens are given in Fig. 8. Irrespective of the annealing/nitriding conditions, an austenite case has formed, consistent with the X-ray diffractograms.Grain growth is pronounced in the austenite case, and despite identical solution nitriding parameters, grain growth kinetics is different for different pre-annealing temperatures (Figs.8-9).After 1000 • C annealing, yielding the lowest amount of ferrite, austenite grains grow fastest during solution nitriding (Fig. 9) and develop typical columnar grains bridging the entire case (Fig. 8a and d), similar to the observations in an isobaric treatment in Ref. [11].The trend to develop columnar grains is weaker for pre-annealing at 1100 • C (Fig. 8b and e) and has disappeared upon solution nitriding the 1200 • C pre-annealed condition (Fig. 8c and f), where most ferrite was present prior to solution nitriding and equi-axed austenite grains develop in the case.It is observed that, especially for pre-annealing temperatures up to 1100 • C, the untransformed ferrite in the transition zone limits grain growth, and the progress of grain growth is promoted immediately after the disappearance of ferrite.The grain size reached after 2.5 h solution nitriding depends strongly on the pre-annealing temperature (Fig. 9), and indicates, in comparison to grain size after 1 h HTSN, that the higher the ferrite phase fraction, the slower is grain growth kinetics during HTSN.The depth of the austenite case was taken as the distance from the surface to the depth beyond which continuous ferrite bands are visible.The case depths for all pre-annealing and solution nitriding conditions presented in Fig. 10 are obtained from three measurements.For all applied annealing conditions, prolonged nitriding results in a thicker austenite case, as would be expected.The austenite case depth developing during HTSN shows no obvious dependence on the pre-annealing temperature.Within experimental accuracy, the austenite case depth after HTSN is independent of the pre-annealing temperature up to 1150 • C; HTSN after 1200 • C pre-annealing leads to a slightly thinner austenite case.

Residual stress
Residual stress values determined in the austenite case of the solution-nitrided specimens for the various pre-annealing temperatures are given in Fig. 11.Irrespective of pre-annealing condition, nitriding and cooling to room temperature leads to tensile stresses in the austenite case.Tensile stresses are consistent with previous observations [7].The resulting tension is mainly caused by the difference in thermal contraction between the austenite case and dual-phase core; the latter shrinks less than the case, because of the relatively large thermal expansion coefficient of austenite.Clearly, prolonged nitriding results in a lower tensile stress in the austenite case.This could be an immediate consequence of a deeper case, so that the thermal mismatch between case and core is distributed over a larger volume.

Hardness-depth profiles
Hardness-depth profiles over the austenite case after solution nitriding are presented in Fig. 12.In all cases, the hardness is highest close to the surface and decreases gradually with depth until reaching the core.Prolonged nitriding shows an increase of the hardness in the austenite case for pre-annealing at 1100 and 1200 • C, while the hardness for the condition pre-annealed at 1000 • C slightly decreases.Evidently, the moderate decline in hardness profile for the 1000 • C pre-annealing condition is explained from the substantially coarsened grains in the austenite case after 2.5 h HTSN, as demonstrated in the micrograph in Fig. 8d.For higher pre-annealing temperatures (Fig. 12b and c), the relatively high hardness obtained after longer HTSN treatment indicates that the contribution to hardness increase from nitrogen solid solution strengthening is larger than the hardness decline induced by grain coarsening.Furthermore, in the whole measured depth range, the hardness achieved after 2.5 h HTSN shows slightly higher values than before HTSN, suggesting that nitrogen has penetrated deeper than 500 μm.

Discussion
A high ferrite/austenite ratio obtained by annealing at relatively high temperatures (or high-temperature pre-annealing) shows paramount importance for industrial solution nitriding of DSS, because it effectively limits grain coarsening accompanied with the acceptable nitriding kinetics.The thermodynamics and kinetics of solution nitriding, and the grain growth within the austenite case for the various pre-annealing conditions are discussed below.

Thermodynamics • Equilibria and reactions
During solution nitriding at 1125 • C in 0.3 bar N 2 , the reactions occurring at the gas-solid interface involve the dissociation of molecular nitrogen and the dissolution of atomic nitrogen into the steel, as listed in Table 2. Due to the different compositions of ferrite and austenite for the various pre-annealing temperatures (Fig. 5), the equilibrium nitrogen contents in austenite and ferrite are different under the applied nitriding conditions.Hence, different local reactions occur at the gas-ferrite and gas-austenite interfaces (Fig. 1c and d, and R3 in Table 2).Fig. 13 shows the nitrogen concentration reached for the global steel compositions, as well as the local equilibria at the gas-ferrite and gas-austenite interfaces for the equilibrium metallic compositions after solution annealing at different temperatures, as predicted by Thermo-Calc.Under the applied nitriding conditions no nitride formation was observed (Figs. 7 and 8), so reaction R3 does not appear to have reached equilibrium.Ferrite does transform into austenite under the influence of nitrogen ingress.The high N content predicted for the surface is most likely not reached (T1 in Table 2), because the actual nitrogen content is a result of the competition between the flux nitrogen provided to the surface from the gas and the flux of nitrogen diffusing into the solid.This applies in particular for austenite formed on ferrite, because diffusion of nitrogen into the underlying ferrite proceeds faster than diffusion into austenite.Consequently, the surface nitrogen content changes with time.The change in hardness in the austenite case with HTSN treatment time as observed in Fig. 12c, and to a lesser extent in Fig. 12b, is an indication that this occurs indeed for the specimens pre-annealed at a relatively high temperature.In this respect, it is mentioned that the applied techniques, i.e. light-optical microscopy and X-ray diffraction, are insufficient to confirm the presence/absence of fine precipitation of M 2 N nitrides; the local presence of nitrides at grain boundaries, the most favourable heterogeneous nucleation sites, cannot be excluded.In the lateral direction, i.e. along the surface, there are differences in equilibrium N content

Table 2
Gas-solid reactions relevant to solution nitriding, and phase transformation occurred during solution nitriding. No.
Reaction Site between the austenite and ferrite regions (cf.Fig. 13).These composition differences will only lead to lateral diffusion if they are accompanied by a difference in chemical potential of nitrogen.Considering the dynamic situation at the surface, as caused by the balance of N provision at the surface and N diffusion into the different phases, it can be expected that diffusion of N from austenite to ferrite is more likely than vice versa.Moreover, laterally there will be a difference in chemical potentials of the substitutional elements after transformation of ferrite into austenite, which promotes redistribution of substitutional alloying elements.Recognizing the much faster diffusion of nitrogen as compared to metallic elements, in principle, nitrogen could reach para-equilibrium between parent austenite and transformed ferrite (into austenite) as long as no uniform substitutional composition is reached in the austenite case.
Due to the difference between solution nitriding temperature and the pre-annealing temperatures, temperature-induced transformation (T2 in Table 2) can occur in the entire material during the HTSN treatment.Specifically, in the surface zone of the steels pre-annealed at temperatures below 1125 • C, T1 and T2 occur simultaneously but drive towards different equilibria.T1 promotes austenite formation, while T2 promotes ferrite formation.For steels pre-annealed at temperatures above 1125 • C, both T1 and T2 drive towards an increase in the austenite content.Evidently, according to the obtainable austenite case for all preannealed temperatures, nitrogen-induced phase transformation T1 dominates in the surface-adjacent zone as is driven by interstitial nitrogen diffusion, while T2 is controlled by substitutional diffusion of metal atoms.For the transition zone that has developed in-between the austenite case and the core, ferrite and austenite co-exist and are the net result of T1 and T2.Accordingly, for the pre-annealing temperatures above 1125 • C, a distinct transition zone has developed, while this is less pronounced for specimens pre-annealed at a temperature below 1125 • C (Fig. 12).

Kinetics of solution nitriding
Next, nitrogen-concentration profiles over the nitrided case are simulated with DICTRA (Fig. 14).It is assumed that nitrogen diffusion is independent of the morphology of the phases and that the phase fractions of ferrite and austenite obtained by different pre-annealing temperatures do not change upon solution nitriding.In the simulations, the starting condition for all pre-annealed cases is a shallow austenite case with homogeneous composition at the surface.The nitrogen content in the austenite case at the "interface" with the dual phase transition zone is assumed 0.55 wt% for all conditions.For both nitriding times (1 and 2.5 h), the depth of the austenite case increases with pre-annealing temperature up to 1100 • C. For a higher pre-annealing temperature, the thickness of the austenite case decreases with increasing preannealing temperature (for 1 h only slightly).This trend is in agreement with the experimental findings in Fig. 10, where a reduction in austenite case depth is observed on solution nitriding the specimen that was pre-annealed at 1200 • C. It is noted that a higher pre-annealing temperature leads to faster diffusion in the dual-phase system, leading to a relatively long tail in the nitrogen profile and a thicker transition zone underneath the austenite case.Faster diffusion in the dual-phase system is ascribed to a higher phase fraction of ferrite, wherein nitrogen diffusion proceeds faster than in austenite.The faster diffusion of nitrogen into the dual-phase core implies that the balance of nitrogen fluxes arriving at and leaving from the "interface" between austenite case and dual-phase region leads to less accumulation of nitrogen in austenite, and thus slower growth of the austenite case.The net outcome for the nitrogen-depth profiles is that the deepest nitrided case, i.e. austenite case and transition zone, is obtained for the steel that is preannealed at the highest temperature.The shallowest case-depth is Fig. 13.Nitrogen concentration achieved at the global gas-solid interface (dashed line) and at the local gas-ferrite and gas-austenite interfaces in equilibrium with the applied nitriding condition, in dependence of the preannealing temperature.obtained after pre-annealing at the lowest temperature, for which the highest austenite fraction is obtained.Clearly, the initial phase fractions have a large influence on the achievable case depth, as also reflected by the experimental hardness profiles in Fig. 12. Evidently, relatively simple assumptions in the above diffusion simulations can explain the trends observed in austenite case depth and the width of the transition zone.Further improvement of the diffusion simulation should include a change in the phase fractions of austenite and ferrite, in particular for the higher pre-annealing temperatures.Also, a labyrinth factor could be considered to account for discontinuity in the ferrite phase, particularly for the lower pre-annealing temperatures.These modifications are beyond the scope of the present investigations.

Grain growth in the austenite case
The results clearly demonstrate that the pre-annealing treatment has a significant influence on grain growth in the austenite case during subsequent solution nitriding.Specifically, grain growth in the austenite case is appreciable and non-uniform for a relatively low ferrite/austenite volume ratio.For a high ferrite/austenite volume ratio (Figs. 8 and 10), grain growth is limited and the grain shape remains largely unaltered.The influence of the ferrite/austenite phase ratio on grain growth in the austenite case is discussed as follows.
In the initial stage of solution nitriding, i.e. before the formation of the shallow uniform austenite case, the dual-phase morphology retards grain growth of both phases.This retardation is most effective if equal contents of ferrite and austenite are present.If one of the phases dominates, the majority phase grows fastest [36], as also observed in Fig. 4b for the relatively high pre-annealing temperatures, where ferrite dominates and coarsens faster than austenite.Under the ingress of nitrogen, ferrite transforms into austenite, without changing the grain size of the pre-annealed austenite.The higher ferrite content in specimens that were pre-annealed at a relatively high temperature, could be suggested to provide more new austenite grains than for the lower pre-annealing temperatures.Accordingly, a smaller average grain size of ferrite and austenite would result in the outer part of the nitrided case.However, this reasoning does not provide a consistent explanation for the clear trend that is observed during HTSN: after pre-annealing at temperatures at or below1100 • C, coarsening proceeds (Fig. 9), while the ferrite: austenite ratio changes only slightly (Fig. 4a).
Analogous to the grain growth behavior in austenitic stainless steel [37], in the as-produced austenite case migration of grain boundaries is the main grain coarsening mode, and a strong driving force for coarsening is expected at the applied temperature.It has been recognized that solute atoms often segregate at defects, e.g. grain boundaries, interfaces and dislocations, in crystalline structures and thereby lower the total Gibbs energy [38].During boundary migration, the solute atmosphere diffuses with, but lags behind, the boundary, resulting in a dragging pressure on the boundary, referred to as solute drag [39].In particular Mo appears to retard grain boundary migration in austenitic stainless steel qualities by solute drag [40].The higher ferrite content for the higher pre-annealing temperatures leads to a smaller difference in Mo content between ferrite and austenite (Fig. 5b).It is unclear whether this could lead to more segregation of Mo to grain boundaries during HTSN, which would be necessary if solute drag of Mo is the reason for slow grain growth during HTSN after pre-annealing at a high temperature.Rather, the contrary would be expected.
Another explanation for reduced grain boundary mobility is Zener pinning by the presence of fine particles at the grain boundaries [37].
Here the formation of stable nitrides would be an obvious precipitate for such pinning.If this is the explanation for the observed reduction in grain growth, then most nitrides should have formed during HTSN after pre-annealing at a high temperature.In this respect, the hardness profiles in Fig. 12b/c show that a higher hardness results on prolonged HTSN despite the occurrence of grain growth (Fig. 9).This strongly suggests that the nitrogen content increases gradually and over-compensates the hardness loss owing to grain growth.Thermo-Calc predicted the stability of chromium nitrides under equilibrium conditions (Fig. 13).Even though no X-ray diffraction evidence was obtained for the actual presence of nitrides, small nitride fractions located at grain boundaries can remain undetected with this technique.Perhaps, for the higher pre-annealing temperatures Cr-nitrides developed in ferrite and were retained after ferrite transformed into austenite.

Residual stress
The residual stresses determined with X-ray diffraction in the austenite phase are always tensile, irrespective of the pre-annealing temperature (Fig. 11).The mechanisms for stress development and stress relaxation can be discussed as follows.At the HTSN temperature the main origins for growth stresses are the volume reduction caused by the conversion of ferrite into austenite and the dissolution of nitrogen, which cause tensile and compressive stresses respectively.It is anticipated that the high HTSN temperature leads to relaxation of most of these volume changes by plastic deformation and/or creep at HTSN temperature.Then, the tensile stresses experienced at room temperature (as measured with XRD) in the austenite case can be ascribed to the difference in thermal shrink between the case and the core.The case shrinks most because of the larger thermal expansion coefficient of austenite as compared to ferrite.A reduction in tensile stress on continued HTSN can be caused by a difference in case depth: the thinner the case, the higher are the tensile stresses in the case (and the lower is the average compensating compressive stress in the core).Alternatively, a higher nitrogen content in the case may have caused the reduction in tensile stress.The obtained residual stress in Ref. [7] for the same alloy is reported as 250 MPa, but no details of the specimen geometry were given.The trend observed in Fig. 11 that the tensile stresses are highest for lower pre-annealing temperatures is unexpected.The lower ferrite content in the core was anticipated to lead to a smaller difference in shrink between the core and the case.Hence, lower tensile stresses in austenite case are expected for the lower pre-annealing temperatures.Evidently, higher tensile stresses at higher pre-annealing temperatures are effectively compensated for, for example by (partial) ferrite-toaustenite conversion in the core, a difference in nitrogen content in the case and different temperature dependence of the mechanical properties for austenite and ferrite.

Conclusion
High-temperature solution nitriding of duplex stainless steel leads to grain growth of the nitrogen-enriched austenite case, compromising the mechanical properties.The present work provides a procedure to limit grain growth during solution nitriding by changing the austenite-ferrite ratio.
Annealing treatments of 2205 duplex stainless steel in the temperature range 1000-1200 • C yielded various ferrite/austenite ratios, i.e. from ~0.9 to 1.6, and led to elemental partitioning between ferrite and austenite.The pre-annealed states were subsequently solution nitrided, such that the influence of ferrite-austenite distribution on hightemperature solution nitriding behavior was investigated.In all annealed cases, and under all solution nitriding conditions, the interstitial dissolution of nitrogen into the steel resulted in the formation of an austenite case, as well as a transition zone underneath.The transition zone is dual phase, but the ferrite/austenite ratio evolves gradually towards the level in the core.With an increase of the ferrite/austenite ratio, the thickness of the austenite case is slightly reduced, while the thickness of the transition zone is increased.
Grain over-coarsening or formation of columnar grains in the austenite case, which readily occurs on solution nitriding of duplex stainless steel, is prevented by the relatively high ferrite fraction obtained after high-temperature pre-annealing.
The combined treatment, i.e. high-temperature pre-annealing and solution nitriding, provides a promising prospect for surface nitrogen alloying of duplex stainless steel, because the uniform and controllable grain structure is of key importance for the reliability of the nitrided parts.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.(a) Equilibrium phase fractions in 2205stainless steel as a function of temperature, predicted using Thermo-Calc.(b) Isopleths with N iso-activity lines as adjusted by the indicated total pressures.N s and N t indicate the nitrogen concentration achieved at the surface and at the transition between austenite case and dual phase zone, respectively; N c is the nitrogen content in the core which is equal to the asreceived state.Isopleths corresponding to the equilibrium metallic compositions of ferrite (c) and austenite (d) after annealing at 1100 • C.

Fig. 3 .Fig. 4 .
Fig. 3. Light-optical micrographs of 2205 stainless steels annealed at different temperatures, acquired over (a-c) transverse section and (d-e) longitudinal section.Insert in Fig. 3b is an EBSD phase map showing the typical dual phase structure (red: austenite; blue: ferrite).(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5 .
Fig. 5. Comparison of the contents of the main alloying elements, as calculated with Thermo-Calc, and experimentally determined with EDS.

Fig. 6 .
Fig. 6.Variation of bulk/phase hardness as a function of annealing temperature.

Fig. 7 .
Fig. 7. X-ray diffractograms of the solution-annealed conditions subjected to solution-nitriding at 1125 • C for (a) 1.0 h and (b) 2.5 h as a function of annealing temperature.

Fig. 8 .
Fig. 8. Cross-sectional micrographs of 2205 steels pre-annealed at the indicated temperatures and subjected to solution nitriding at 1125 • C for (a-c) 1.0 h and for (d-f) 2.5 h.

Fig. 9 .
Fig. 9. Mean grain diameter in the austenite case as a function of pre-annealing temperature.

Fig. 10 .
Fig. 10.Thickness of the austenitic cases as a function of pre-annealing temperature.

Fig. 11 .
Fig. 11.Stress values as measured in the austenite case of solution nitrided DSS 2205 pre-annealed at the indicated temperatures.

Table 1
Chemical composition of 2205 DSS used in this study (wt%).
B. Wang et al.