Gradient self-organized dislocation in expanded austenite layer during low-temperature nitriding

A typical nitrogen expanded austenite layer is formed by plasma-based low-energy nitrogen ion implantation (PBLEII) on AISI 304L austenitic stainless steel at a moderate temperature of 380 °C. The dislocation self-organization structure in the nitrogen expanded austenite layer is characterized as an evolution from partial and Lomer-Cottrell dislocations in the inner layer near the interface to multilayer stacking faults in the outer nitrided layer. The self-organized dislocation density and forms are essentially dependent on the plastic deformation, strain-gradient, and nitrogen-related stacking fault energies, respectively, due to the constrained expansion in the nitrided layer. As the nitrogen concentration in the austenitic matrix increases, the stacking fault energy gradually decreases, resulting in the transformation of the defect from Lamer-Cottrell dislocations to multilayer stacking faults. The appropriate stress, which is associated with orderly stress relief during dislocation self-organization, preserves the integrity of the nitrided layer with a combinedly improved in wear and corrosion resistance. Nitriding-induced dislocation self-organization is basically explored as the formation mechanism of the nitrogen expanded austenite layer, contributing to the development of the specific low-temperature nitriding austenitic steel.

and dynamics, a self-organizing system can emerge from the interactions between subsystems. The application of self-organization has been well studied in phase transformation, friction and diffusion systems [6,7]. The selforganized defects not only determine the properties of the nitrided layer, but also reflect the details of the phase formation process during nitriding. The parallelogram and triangle surface relief patterns varied with different oriented grains. Most of the surface relief patterns are parallelograms formed by two groups of parallel relief lines, and even fewer patterns on grains close to 〈111〉 oriented grains are triangles formed by three groups of parallel relief lines [8][9][10]. In the depth direction, these surface relief patterns corresponded to the high-density stacking faults in the nitrided layer. However, the self-organized microstructure of nitriding-induced dislocations in-depth, especially the dislocations in the diffusion front, which may also play an important role in diffusion, phase transformation, and anisotropic property in the nitrided Fe-Cr-Ni austenite stainless steel, has not been clearly characterized and discussed in detail.
In the present work, the defect structure and distribution of the nitrogen expanded austenite layer are investigated by TEM/EDS. A comprehensive description of the origin and evolution of dislocation selforganization is given, taking into account the nitrogen concentration-related plastic deformation, strain gradient, and stacking fault energies.

2. Experimental
Commercial AISI 304L austenitic stainless steel was selected as the test material, consisting of (wt.%) C 0.020, Mn 1.040, Si 0.43, Cr 18.11, Ni 8.14, Mn 1.04, P 0.031, S 0.020 and Fe balanced. The samples were nitrided by plasma-based low-energy nitrogen ion implantation (PBLEII), which combines the features of conventional plasma-based ion implantation and low-energy ion beam implantation. The typical operating parameters used during the nitriding process are listed in table 1. The implantation depth for N 2+ with an ion energy of 2 kV was approximately 2.5 nm. Due to the presence of a temperature field in the sample caused by both the auxiliary heater and the ion bombardment, significant thermal diffusion of the implanted nitrogen took place from the surface inward [11], and the general diffusion coefficient at 400°C could reach to about 1×10 −11 cm 2 s −1 , which varied with N concentration and grain orientation.
The nitrided samples were prepared by the standard metallographic preparation method and the microstructure was revealed by etching in Marble's reagent. The cross-sectional image was observed by OLYMPUS LEXT OLS4000 laser confocal microscope. The depth profile of nitrogen concentration was measured by SHIMADZU EPMA1600. The phase content was analyzed by PANalytical Empyrean diffractometer with CuKα x-ray radiation (λ = 0.154 nm) at 2θ = 20-100°. Transmission electron microscopy (TEM) specimens were prepared according to the standard procedure for cross-sectional TEM specimens procedure, and examined with the FEI Tecnai G2 microscope, and the JEM-F200 microscope in the STEM mode using EDS. The detailed examination procedure is described in [9].

Structure of nitrided layer
The optical micrograph of cross sections over the nitriding layer for AISI 304L austenitic stainless steel is shown in figure 1. The layer depth is about 5-7 μm with a clear anisotropic diffusion in the nitrided layer, which is characteristic of nitrogen expanded austenite layer on stainless steel as a thicker layer in 〈100〉 oriented grain and thinner in 〈111〉 [12]. The nitriding layer appears as a bright and featureless layer compared to the substrate, due to the greatly improved corrosion resistance, especially for pitting corrosion. The depth profile of the nitrogen concentration over the nitrided layer is shown in figure 2. The depth in the concentration profile is about 7 μm, which is similar to the thicker layer zone in figure 1. A high plateau of nitrogen concentration was formed with a steep increase of the diffusion front near the interface, which is a typical N distribution form of the low-temperature nitrided layer on Fe-Cr-Ni austenitic stainless steel. The abnormal N distribution profile to the non-Fick law, especially the steep increase of N at the diffusion front, was caused by the strong Cr-N interaction as trapping of N in the Cr-containing octahedral interstice. The higher the Cr content in an octahedral interstice, the stronger the Cr-N interaction. Combined with the subsequently formed Fe 4 N-like LRO, the supersaturated N concentration could reach more than 30 at.% [9,13]. Figure 3 shows the XRD patterns of the original steel and the steel nitrided by PBLEII. Due to the super-high N concentration, the diffraction peaks were broadened and shifted to lower 2θ. The broadened peaks were essentially formed by two phases of Fe 4 N-like ordered expanded austenite (γ′ N ), and the disordered expanded austenite (γ N ), which have been well confirmed in previous experimental work [9]. The ordered γ′ N phase appeared as nano-sized antiphase domains near the surface at high N concentration [13]. For the nitrided Fe-Cr-Ni alloys with more than about 12 wt% Cr, there is no clear concentration and lattice parameter break between the γ N and γ′ N phases, but only the disordered and ordered transition of the N arrangement. The lattice parameter expanded from 0.358 nm to about 0.394 nm during the nitriding. The expansion of about 10% resulted in high stress and plastic strain in the nitrided layer. This has two main effects on the structures: (1) the  texture of the nitrided layer by grain rotation, represented by the higher intensity of (111) and (200) peaks and weaker of (311) peak, and even the disappearance of (220) peak [14,15]; (2) high density of plastic deformation defects, such as dislocations and stacking faults or twin defects [5]. The high density of stacking faults has been reported as single or multilayer (h.c.p.) structures, which are also marked as the ε N in figure 3 and can be observed in the following TEM results [9]. It should be noted that the ordered γ′ N phase transition has no directional relationship with the plastic deformation defects, which was confirmed by perfect nitrogen expanded austenite obtained from the sample by stress-and strain-free nitriding [4]. The following analyses will focus on exploring the nature and evolution of dislocations in the nitrogen expanded austenite layer, especially those near the interface. Figure 4 shows the (HR)TEM observation with EDS analysis of the nitrided AISI 304L austenitic stainless steel layer, which is consistent with the layer thickness observed in the optical micrograph of figure 1. The crosssection is about 5 μm thick with a clear interface in figure 4(a). The evolution of the defect structure with depth can be observed in the nitrided layer. In the near-surface zone, as shown in figure 4(b), the defects are continuous multilayer stacking faults forming an h.c.p. structure, which is also referred to ε martensite. Figure 4(b-1) shows an HRTEM image of the multilayer stacking structure zone in figure 4(b). The red dashed line is the interface between the h.c.p. structure and f.c.c. structure formed by sliding in (111) plane along 〈112〉 direction, which has also been well studied in previous TEM observation [9]. The stacking faults penetrate most parts of the nitrided layer, while the thickness of the multilayer becomes thinner with the depth. In the front of each stacking fault, there is a partial dislocation at each thinned point of the multilayer stacking faults. In the middle zone of the nitrided layer shown in figure 4(c), it consists of large stacking faults and short dislocations. In the inner layer near the interface, the large stacking faults are completely transformed into short dislocations in the zone near the interface, as shown in figure 4(d). A high density of partial dislocations piled up at the interface is observed, and the dislocations in the two sides of the interface have distinctly different forms.

Distribution of dislocations
The EDS signal of N in the nitrided layer under STEM mode is shown in figure 4(e), corresponding to the white line in figure 4(a), which has been mathematically smoothed to remove the noise in the profile. This is basically similar to the typical N concentration profile of the nitrogen expanded austenite layer in figure 2. Since the relationship between the N concentration and the lattice expansion parameter is basically linear [16], the signal intensity can also represent the lattice expansion. Figure 4(f) shows the expansion gradient in the whole nitrided layer calculated by the first derivative of the N concentration signal intensity. Under the constraint of the austenitic steel substrate, the lattice expansion leads to the formation of high-density defects in the nitrided grains due to significant plastic deformation. Based on the strain and its gradient profiles, the nitrided layer could be roughly divided into two zones: higher strain and lower strain-gradient zone (I), and higher strain and higher strain-gradient zone (II). Zone (I) mainly corresponds to the high-density stacking fault area, where the large lattice expansion is released, and the low expansion gradient is adjusted by gradually thinning the multilayer stacking faults with partial dislocations. Zone (II) corresponds to the high-density dislocations in the nearinterface region, which can adjust the higher strain gradient here.  (1) the distance of the same diffraction spots becomes larger at a deeper position due to the smaller lattice parameter, which has about 10% difference. This is consistent with the N concentration profile; (2) compared with the orange horizon line in each pattern, the diffraction pattern gradually rotated counterclockwise, and completely rotated about 9°from the matrix to the surface. This results from the nitriding-induced plastic strain, which is studied by EBSD and model simulation [14,15]; (3) the diffraction pattern shows a clear evolution process with depth. In figure 4(i), the SAED pattern is composed of three sets of spots, including f.c.c. matrix spots of γ N , superlattice spots of γ′ N , and h.c.p. spots of multilayer stacking faults. At the deeper position of (ii), the superlattice spots disappear, leaving only f.c.c. spots and h.c.p. spots. This evolution was also reported in [9]. Position (iii) is a transition zone from stacking faults to dislocations. No clear h.c.p. spots are observed in the pattern, instead a weak line along the (111) spots. At this position, the multilayer stacking faults are thinned out and mainly transformed into single-layer stacking faults which cannot show complete h.c.p. diffraction. The faint line represents the diffraction of high-density single-layer stacking faults in the (111) plane. Another noticeable phenomenon is that several satellite spots are observed near some diffraction spots, as marked by blue arrows. These satellite spots come from the diffraction of high-density dislocations with partial order. In the deeper position of (iv) near the nitrided layer interface, the single-layer stacking faults disappear completely and only high-density dislocations are present, corresponding to the disappearance of weak lines and the presence of only satellite spots.
Since the number of dislocations is sufficient to compensate for the lattice deformation, the large plastic strain gradient corresponds to a high density of geometrically necessary dislocations. However, the stacking fault energy determines the form and distribution of dislocations, and the different dislocation forms adapt to the same deformation. The dislocation distribution in the nitrided layer is analyzed as follows.

Evolution of dislocations
It has been confirmed in the N distribution and antiphase domain structure that the evolution of the lowtemperature nitrided layer has an equivalence property between time and space, which means that the feature at the diffusion front is equal to the initial state during nitriding [17]. This is mainly because low-temperature nitriding inhibits the movement of metal atoms, and only N atoms diffuse into the deeper layer. All structure evolution is driven by N diffusion and interaction between N and metal atoms. During nitriding, the N atoms are firstly strongly trapped in the octahedral interstice by Cr-N interaction to sharply increase the N concentration, which corresponds to high expansion and strain gradient in the nitrided layer. With the increase in N concentration and lattice expansion, the planar stress in the nitrided layer increases. As a result, the resolved shear stresses in the four (111) planes also increase. When one of the resolved shear stresses reaches a critical value, it forms a partial dislocation and slides deeper on the (111) plane along the direction with the largest Schmid factor. The planar stress is then released, leaving a stacking fault. With further expansion, another partial dislocation forms to accommodate the strain. The new partial dislocation can be formed by two relationships with the previous partial dislocation and stacking fault. One is to form a partial dislocation exactly in the (111) plane of the previous one and enclose the stacking faults, resulting in the perfect crystal being left behind. The other is to form a partial dislocation in another (111) plane parallel to the previous one and form another stacking fault behind it. These two states of the expanded surface have the same plastic strain, strain gradient, and dislocation density, but have different widths in the stacking faults between the two partial dislocations, which is mainly determined by the stacking fault energy of the nitrided layer.
Since the N solution in the Fe-Cr-Ni austenitic stainless steel can further reduce the stacking fault energy, the width of the stacking faults becomes larger with the higher N solution in the austenitic steels [18,19]. The lower stacking fault energy results in two effects: (1) the stacking fault is wider between two partial dislocations [20], and (2) it prefers to form multilayer stacking faults at further lower stacking fault energy, even negative [21]. As nitriding progresses, the formed stacking fault becomes wider than the earlier ones, as shown schematically in figures 5(a) and (b).

4. Discussion
The above dislocation evolution is schematically introduced in a 2D schematic, slipping in only one plane and direction as shown in figure 5. However, in order to relieve the planar stress of the nitrided layer in two principal directions, at least two slip planes must be activated. Therefore, most surface relief patterns are parallelograms formed by two sets of parallel plastic strain lines along two different (111) planes. When the two partial dislocations meet, the Lomer-Cottrell dislocation is formed at the cross-section position. Figures 6(a) and (b) show the general schematic diagram of the Lomer-Cottrell dislocation (denoted as L-C) along the [011] direction on the (100) plane. In the nitrided layer, the two different partial dislocations are generated from the surface and move into the deeper layer in figure 6(c). The frontiers of the partial dislocations are not necessarily straight and are basically parallel to the surface. When the two partial dislocations meet, the formed Lomer-Cottrell dislocation should be a straight line along [011], which gives a set of Lomer-Cottrell dislocations caused by crossing two planes in figure 6(d). In the real state of the nitrided layer, many parallel planes will cross each other in two different (111) plane groups, resulting in two Lomer-Cottrell dislocations formed by three planes in figure 6(e). With further nitriding and dislocation evolution, another partial dislocation completes the stacking fault from the surface and moves deeper, as shown in figure 6(f). This is the whole process of Lamer-Cottrell dislocation formation in the nitride layer, and also the dislocation structure at the diffusion front of the nitride layer, which corresponds to the two-dimensional schematic diagrams in figures 5(a)-(e). Therefore, when the TEM observes the dislocation parallel to the (111) plane of blue, the dislocation will appear as shown in figure 6(g). It is mainly composed of two parts: (1) the curved dislocation lines basically parallel to the surface, which is the partial dislocation of the front or end stacking fault; (2) the straight dislocation lines parallel to the long stacking fault near the nitrided surface, which is the [011] Lomer-Cottrell dislocation. This is in good agreement with the TEM observation at the diffusion front as shown in figure 6(h).
In conventional dislocation evolution during deformation, the Lomer-Cottrell dislocation is considered to be a dislocation lock, which is an important source of strain hardening [22,23]. However, in the nitrogen expanded Fe-Cr-Ni austenite layers, most grains activate only two slip planes to release planar stresses in both principal orientations, except for grains near the 〈100〉 and 〈111〉 orientations, which have highly symmetric orientations, so that all Lomer-Cottrell dislocations in a nitrided grain are of the same form in the same [110] orientation. These self-organized dislocations not only relieve the constrained stress in the layer, but also adjust the lattice difference in depth as a geometrically necessary dislocation. Similar Lomer-Cottrell dislocations have been observed in the interface of epitaxially grown (Ga,In)As/InP layers to compensate for the lattice misfit [24].
In the nitrided layer, as the gradient expansion gradually moves to a deeper layer with nitrogen interstitial diffusion, the Lomer-Cottrell dislocations are not locked but could move deeper due to the advance of lattice expansion. The [110] Lomer-Cottrell dislocations remain in the (100) plane with two forms of movement, such as slipping along the 〈110〉 direction or climbing vertically through different (100) planes. The more [110] Lamer-Cottrell dislocations are parallel to the surface of the nitride layer, the more they tend to move by climbing. The square and parallelogram relief patterns on the nitrided surface caused by crossed multilayer stacking faults in the nitrided layer are essentially similar to the Lomer-Cottrell dislocations, but with multilayer stacking faults typically formed by partial dislocations moving layer by layer. It is noteworthy that the dislocations below the interface of the nitride layer show a different structure, lacking the typical Lomer-Cottrell dislocation features, in particular the absence of [100] dislocation lines parallel to the multilayer stacking faults. These are likely to be low-density partial dislocations created by passive expansion to accommodate the high lattice expansion of the nitride layer, which only activates a (111) plane with the highest Schmidt factor during nitriding.
The driving force for dislocation movement comes from the shear stress resolved by the planar compressive stress. When the shear stress reaches the critical resolved shear stress for dislocation movement, some of the dislocations will generate and move inward to relieve some of the stress in the nitride layer. In general, any obstacle to dislocation movement will lead to an increase in the critical dislocation shear stress, such as precipitates, dislocation traps, and grain boundaries. For nitrogen expanded austenitic layers without precipitates, there are no other barriers to dislocation movement and stress relief because the self-organizing Lomer-Cottrell dislocations also move deeper with the expansion gradient during nitrogen diffusion. The lower nitriding temperature suppresses the diffusion of metal atoms for the nitride layer to relieve stress. Dislocation movement is the only way to relieve stress in the nitrogen expanded austenite layer. The maximum residual stress in the nitrided layer is related to the magnitude of the critical resolved shear stress of the moving dislocations. Self-organized dislocation evolution and stress relief ensure that the expanded austenitic layer is not destroyed by excessive constrained expansion and planar stresses during nitriding. In low-temperature nitriding of martensitic stainless steels, there is no such stress relief path as in austenitic alloys; some surfaceparallel cracks may form in the nitrided martensite layer [25]. Therefore, self-organization is an important factor in the orderly stress relief to maintain the integrity of the low-temperature nitride layer formed on Fe-Cr-Ni austenitic steels. Although these high-density self-organized stacking faults and dislocations relieve stress in an orderly manner and do not hinder each other's movement during nitriding, in the application process they are the main reason for hardening and improving the wear resistance of the nitrided layer by hindering dislocation evolution in other slip planes. The nitriding-induced dislocation self-organization as the formation mechanism of the nitrogen expanded austenite layer is explored to contribute to the development of low-temperature nitrided austenitic steels with better overall performance. By changing the alloy composition, the stacking fault energy and dislocation self-organization behavior during nitriding could be adjusted. This can further influence the properties and behaviors such as diffusion, elasticity, residual stress, hardness, precipitation, and so on [8,[26][27][28]. By obtaining the evolutionary laws and mechanisms of the self-organizing behavior, new alloys and steels suitable for low-temperature nitriding can be designed more purposefully.

5. Conclusions
(1)A typical nitrogen expanded austenite layer of 5-7 μm thickness and supersaturated up to 30 at.% N concentration is formed on AISI 304L stainless steel by PBLEII. The microstructure of the nitrided layer in depth, especially the dislocations near the interface, was characterized by TEM/EDS.
(2)The dislocation self-organization structure is obtained from the constrained expansion in the nitrogen expanded austenite layer with a gradient evolution from partial and Lomer-Cottrell dislocations in the inner nitrided layer to multilayer stacking faults in the outer layer.
(3)The self-organized dislocation density and forms in the nitrogen expanded Fe-Cr-Ni austenite layer are mainly determined by the nitriding plastic deformation. The structure evolution depends on the stacking fault energy, which gradually decreases with the increase of nitrogen concentration.
(4)The integrity of the nitrided Fe-Cr-Ni austenite layer in the highly expanded state is maintained by the orderly stress relief and corresponding stress due to the self-organized evolution of dislocations.