Carbideprecipitationin austenitic stainlesssteel carburized at lowtemperature
F.Ernst * Y. Cao G.M. Michal A.H. Heuer
Deperiment ef Materialb Sciesce oml Engimering. Case Westerw Rerene U/nisersity. Clereland. OfH 44706-7204 US.A
Received 16 August 2006; received in revised form 14 September 2006; acepted 14 September 2006 Available online 12 January 2007
Abstraet
Low-temperature gas-phase carburization can significantly improve the surface mechanical properties and corrosion resistance ofaustenitic stainless steel by generating a single-phase “case”* with concentrations of interstitially dissolved carbon exceeding the equilib- rium solubility limit by orders of magnitude. Upon prolonged treatment however carbides (mostly x M;C) can precipitate and degradethe propertie. High-resolution and spatially resolved analytical transmission eletron microscopy revealed the precise carbideaustenite orientation relationship a highly coherent interface and that precipitation only occurs when (i the carbon-induced lattice expansion ofthe austenite has reached a level that substantially reduces volume-misfit stres and (i) diffusional transport of nickel chromium andiron enhanced by structural defects can locally reduce the nickel concentration to the solubility limit of nickel in z-carbide.
Kryordc Austenitic stainles stee; Fe-CrNi Surface allying: Low-temperatue gas-phase carburization; x-Carbide (MCs Haigg)
1. Introduction
316 the case is ≥25 μm thick and X-ray diffractometry0.12 [4 6]. Recent scanning Auger microprobe (SAM) mea- and X-ray photoelectron spectrometry indicated x|0| surements based on calibration standards even suggestexceeds the equilibrium solubility limit of carbon by a fac- x[0] = 0.14 (Avishai et al. unpublished data). Thistor of ≥700 at the processing temperature and a factor 10² at room temperature. Such a “colossal” supersatura-tion with carbon is possible because Z is high enough forinterstitially dissolved carbon atoms to retain considerable mobility but low enough to limit the mobility of the metalinto the alloy to form a single-phase carbon-rich “case". in metal carbides cannot readily occur [13]. Another atoms to a level at which the usual precipitation of carbonimportant requirement we have identified for obtaining acolossal supersaturation with carbon is the presence of suf- ficient atom fractions of elements with a high affinity forcarbon. In the 316L-type alloys the main element playing this role is Cr (chromium).
In recent work [16] we reported that the surfacemechanical properties and corrosion resistance of 316-typeaustenitic stainless steel (FeCrNi alloy) can be signifi-lok Company. This process after activating the alloy sur- gas-phase carburization process developed by the Swage-face by removing the passivating chromium oxide scalesphere at a processing temperature T = 748 K for a dura- [7] supplies carbon from a conventional carburizing atmo-tion t between 26 and 38 h. As a result carbon difusesWith increasing depth : into the case the concentrationX[] smoothly decreases from its maximum X[0] at the surface to the intrinsic carbon level X? of the non-carbu-rized alloy core. For low-temperature-carburized bulk
As long as carbon stays in solid solution a colossalsupersaturation of the austenite greatly improves the sur- face hardness resistance to fatigue crack nucleation wearresistance and corrosion resistance [1 2 4 6 8]. Precipitation
of carbon-rich second phases in contrast generallycarburization stage of the Swagelok process is designed degrades the properties. Therefore the duration tp of theto maximize the uptake of carbon while entirely retaining it in solid solution. However on prolonged carburizationor multiple application of the Swagelok process whichmay be of considerable interest for increasing the case depth metal carbides can eventually precipitate. They formin a zone below the surface in which the average carbon fraction exceeds the non-equilibrium solubility limit X(≥0.12 for bulk 316L). Carbon uptake in excess of this limit correspondingly increases the volume fraction of thecarbide phase and the depth of the carbide-containing zonebelow the surface.
Previous work [5 6] revealed two different carbide mod-the metal atoms ie. Fe or alloying elements substituting ifications: MC (or ≥ phase) and MC ("M" stands forFe in its lattice sites in appropriate proportions). The(MC). Laterally the particles are not distributed uni- majority of the carbide particles by far are phaseformly within the carbide zone. Rather they occur in groups or "“colonies"” with virtually carbide-free austeniteregions between them. Fig. la presents a typical conven-tional transmission electron microscopy (TEM) bright-field image of a colony [6]. The particles appear with largeaspect ratios typically featuring a short dimension of ≥20 nm and a long dimension up to several micrometers.They contain a high density of planar faults orientedroughly orthogonal to the long axis.
The x particles have the crystal structure of Fe;Cters [5 6] Table 2 piles the lattice parameters of we (*"Hagg carbide') [911] with very similar lattice parame-have determined by X-ray diffractometry (XRD) fromsoftware packages Celref [12 13] and RIETAN [14 15]. low-temperature-carburized 316L powder [6] using theagree very well with those found for Fe;C in two other The lattice parameters obtained from our measurementspublications [16 17] also listed in Table 2. The space groupis C2/c (15). Table 3 lists the atom positions as determined by Retief [16]. The unit cell contains 20 Fe and 8 carbonatoms consistent with the stoichiometry M;C. Fig. 2 pro- jects the structure of FesC in [010] the direction of thesymmetry axis.
Fig. 1. Conventional TEM of z (M;C needles that precipitated in a (austenite) matrix with a colossal supersaturation of carbon (X [0] = 0.12. recorded with a 200 kV Philips CM20 transmission ekeetroncorectly oriented with respet to (a) The patt constitutes a supepo- microscope). (a) Bright-field image. (b) Selected-area difraction pattern jo ued sxeoz[11] pu 7 jo und stxeoz[010] jo uos The white spots superimposed onto the pattem in the lower rightquadrant were obtained by simulating the [010] diffraction pattern of Fe C2.
tation relationship (OR) with the matrix [5]. Based on The x particles grow in a unique crystallographic orien-Fig. Ib originating from a region including several carbide particles and about 20 similar TEM selected-area diffrac-tion (SAD) patterns we have described this OR as [5]
planar faults apparent in the image and identify the fault ing [100]-oriented rows originate from the high density ofplane as (200). As all the carbide particles constituting the colony in Fig. 1a have the same OR with the matrix ed e s red se m se sxe u pueto each other a unique correspondence exists between the (idealized) particle shape and the orientation of the crystallattice therein. According to Fig. la the typical particle extension in [001] is much smaller than in [100] . TEM
(1)
(2)
The direction (2) corresponds to the viewing direction ofan angle of 90° as do [211] and the normal of (111) - Fig. 1. Note that [010] and the normal of (001) make
with respect to Fig. Ia. The streaks between the spots form- The diffraction pattern of Fig. Ib is correctly oriented
Fig. 2. The crystal structure f Fe; C the Hagg” carbide shown in [010]projection. The structure belongs to the monclinic crystal system and has thespace group is C2/c (15). The shaded region corrsponds to one unit cell.
images in [001] [5 6] show that the extension in [010] isthan in [100]. Accordingly. the particle shape resembles parable to that in [oo1l i.e. it is also much smallerthat of needles or laths but not plates or discs.
As carbides generally degrade the surface mechanicalproperties and corrosion resistance it is important to iden-mental understanding of the micromechanism by which tify the parameters that control X? and to gain a funda-the carbide particles nucleate and grow in order to develop strategies for avoiding them. In the work reportedhere we have studied the atomistic structure of the tenite matrix) and the spatial redistribution of atom species interface (the interface between M C particles and the aus-associated with carbide formation under the constraints of low-temperature carburization by advanced methods ofTEM.
2. Experimental methods and procedures
The material we investigated was a low-temperature-car-burized foil of 316L-type austenitic stainless steel (ADT15)with a thickness of 38 μm and annealed at 1338 K for 1.5 h. Table 1 lists the results of a wet-chemical analysis. Carbideprocess corresponding to a total carburization time formation occurred after two applications of the Swagelokfp = 44 h at 7 =748 K.
carburization with a Scintag X-1 X-ray diffractometer uti- X-ray diffractograms were obtained before and afterlizing Cu Kα radiation with the wavelength = 0.154056 nm in the BraggBrentano (020) mode [18]From the diffractograms we determined the peak centerssponding to / = 1 2 3 4 respectively). From these data 0 of the reflections {1111 {200) {220) {311} (corre-
Composition of the AT15 foil (at%) Table 1
C Mn Si P S Cr Ni Mo AI Cu0.16 1.61 0.98 0.05 100 18.76 10.65 1.10 0.210.35N 0.24 0 0.07 0.07 0.190.05 Co B Nb 0.02 Ca 0.140.07 65.29 Se Fe
we obtained best estimates a and β for the austenite latticeparameter and the NelsonRiley coeficient respectively this approach a and β are given by the values of a andthe measured 0 and their theoretical values β that minimize the sum of the square residuals between
where p := of the planes reflecting under 0 The relation (3) follows di- ² and hare the Miler indicesrectly from the Bragg equation [18] and the NelsonRileycorrection [20 18]. For evaluating [p] (an implicit func- tion) and for fitting the parameters a and β we developeda C implementation of the downhill simplex algorithm [21]. To determine standard error limits for the best esti-mate a we generated 10* new data sets {0′) drawing ran-dom values from normal distributions centered on the theoretical [p] obtained from (3) with the best estimatesa and β. The standard deviation of these distributions was chosen to [19]
where P = 4 is the number of peaks and f = 2 the numberof fitting parameters (or degrees of freedom). Fitting (3) to each new data set {0) we obtained histograms of the 104best values af and β’ for the fitting parameters a and β respectively.Sorting these data yielded corresponding cumulative distributions from which we eventually deter-the 68% level of confidence. The symmetric *standard er- mined (asymmetric) confidence intervals for a and β atrors" indicated for the data in the following sectionscorrespond to the largest possible deviation within the asymmetric 68% confidence interval. A corresponding pro-cedure was applied to difusion data from the literature to obtain pre-factors and activation energies.
The particles and the interface were studied byhigh-resolution TEM (HRTEM) scanning TEM (STEM) with a high-angle annular dark-field (HAADF) detector
(3)
(4)
electron-spectroscopic imaging (ESI) and X-ray energy dispersive spectroscopy (XEDS）. The instrument weemployed was a Tecnai F30 S-TWIN (FEI) operating with e dp oe ss tion resolution limit of 0.14 nm.The instrument isequipped with a HAADF detector (Model 3000 Fischione) d 2001" Gatan retrofitted with a 2k x 2k CCD camera) Li detector. Employing these attachments we performed and an X-ray energy-dispersive spectrometer with an Sielemental mapping using the “three-window method" [22] and XEDS line scans. The latter were conducted in STEMmode based on Z-contrast imaging with the aid of theHAADF detector. In these experiments the full width at half intensity maximum of the electron probe wassize # 6 in nanoprobe mode. Specimen drift was pen- ≥0.5 nm corresponding to gun lens setting # 7 and spotsated by the instrument's built-in capability of monitoringthe position of a characteristic image feature.
Fig. 3. NelsonRiley plot of XRD (X-ray diffractometry) data obtainedfrom the 316L-type austenitic stainless steel foil before and after low- temperature carburization. (After [6])
(8)
The increased scatter about the least-square-fit line ob-served for the points representing the carburized material in Fig. 3 does not reflect statistical error but is a systematicand reproducible effect [6] which we attribute to peakshifts introduced by an increased density of stacking faults [3133]. The peak shifts can be modeled [34 35] and thea third fitting parameter in the non-linear regression anal- stacking fault density can be quantified by including it asysis (Ernst unpublished data). However the value for athe result (6) obtained by treating the peak shifts as a sto- obtained in that way is not significantly different fromchastic effect.
Cross-sectional specimens for TEM were preparedfrom the low-temperature-carburized steel foils following the method developed by Strecker et al. [23] using spe-cially prepared specimen holders.Afer mechanicallyreducing the thickness to 80 μm and further reducing the thickness in the center to ≥20 μm by means of a dim-ple grinder (Gatan Inc.) the final thinning to electronprecision ion polishing system (PIPS Gatan Inc.) fromboth sides until perforation occurred. Prior to loading the specimen into the microscope we cleaned the surfacefrom adsorbants and contamination by a means of an Ar* plasma cleaner (Fischione Model 1020). For simulat-for generating ball models of the crystal structures of and we employed the JAVA version of the EMS soft-[9]d r
3.2. Concentional and high-resolution TEM
Fig. 4 presents another conventional TEM bright-fieldthe hole of the TEM specimen. The carbide particles one image of a colony. This image includes the thin edge atof them marked " appear with uniform gray levels indi- cating that they have the same crystallographic orientation.While TEM images recorded in certain crystallographicorthogonal to the long axis of the particles (e.g. Fig. la) directions reveal a high density of planar faults on planesthese defects do not produce much contrast under the imaging conditions of Fig. 4. The regions of carbon-super-saturated matrix (marked *") also have the same crystal-lographic orientation (average gray level) implying that the matrix regions belong to a single grain and the 7ZDifferent from the carbide particles the matrix regions orientation relationship is constant across the entire image.exhibit distinct gray level variations on the subnanometerlength scale indicating a high density of dislocations. While part of these dislocations will have resulted fromfrom cold work (foil rolling) prior to carburization a sig- nificant fraction may have formed by yielding under biaxialpressive stress introduced by the (depth-dependent) lat-tice expansion acmodating the interstitially dissolved carbon [2].
3. Results
3.1. X-ray diffractometry
difractograms of the 316L-type austenitic stainless steel Fig. 3 shows a NelsonRiley plot obtained from X-rayfoil recorded before and after carburization [6]. The best estimates for the lattice parameter a° of the non-carbu-rized austenite and the lattice parameter a*[0] at the sur-face of the low-temperature-carburized material are
(5)
(6)
This corresponds to an expansion factor of 1.024 ± 0.005.Based on the empirical relationship [2730]
(7)
and α = 0.104 nm [30] the value (6) for a[0] corresponds [9] 01
Fig. 4. Conventional TEM bright-field image of x (M C) particles embedded in / (austenite) with a colossal supersaturation of carbonedge and the hole of the TEM specimen. (X[0] 0.084). The upper part of the image features the amorphous
Fig. 5a is an HRTEM image of the interface between a particle (top) and the 7 matrix (bottom). The interfaceresides in the lower third of the image and lies approxi- mately horizontal and parallel to the viewing direction（"edge-on). It can be identified with the transition between the contrast patterns in the upper and lower halfof the image. Within the region included in Fig. 5a theinterface does not exhibit considerable roughness.
Fig. 5b shows the interface at higher magnification. Thecontrast pattern observed inin the lower half featurese te s e pss jo ss om of 71°. Accordingly. the fringes correspond to [111}it exhibits only one set of lattice fringes the viewing direc- and the viewing direction corresponds to [110]. In 7 astion does not coincide with a low-indexed zone axis. Differ- ent from what might have been expected based on (1) and(2) the lattice fringes parallel to the one set of (111}fringes do not originate from (002) planes. This bees obvious by considering the misfit between the correspond-ing plane spacings. According to the lattice parameters ofa = (0.368 ± 0.002） nm of the expanded austenite dis-(0.212 ± 0.001) nm. Taking into account that these planes make an angle of Φ ≥ 20° with the plane of the interface the misfit
Fig. 5. Atomistic structure of the x interface. (a) High-resolution TEMmatrix (bottom). The viewing direction corresponds to [110] in 7 and image of an interface between a needle (top) and the supersaturated 7specimen close to the hole (upper right). The lattice fringes that [124| in x. This image was recorded in an ultra-thin region of the TEM q o pa 1eue ue oms aoq oq o qoe edde coherence of the field-emission electron gun employed for imaging. (b)Enlarged region of the interface in (a).
Table 2 Lattice parameters of x (M/C)
Parameter Dirand and Afqir Retief [16] Cao [6]α (nm) [17] 1.1563 1.1588 ± 0.0002 1.1552 ± 0.0002b (nm) (uu) 3 0.5058 0.4573 0.4579 ± 0.0001 0.5059 0.45638 ± 0.00004 0.50432 ± 0.00006β 90° (invariant fr space group C2/c) 97.7* (97.746 ± 0.002) (97.68 ± 0.01)a
(9)
between the plane spacings would require misfit disloca- tions with an average spacing of
