lable at ScienceDirect

Intermetallics 17 (2009) 886–893

Contents lists avai

Intermetallics

journal homepage: www.elsevier .com/locate/ intermet

Microstructural evolution of spinodally formed Fe35Ni15Mn25Al25

Ian Baker a,*, R.K. Zheng b, David W. Saxey b, Satoko Kuwano b, Markus W. Wittmann a,Johnathan A. Loudis a, K.S. Prasad b, Zongwen Liu b, Ross Marceau b, P.R. Munroe c, Simon P. Ringer b

a Thayer School of Engineering, Dartmouth College, Hanover, NH 03755, USAb Australian Key Centre for Microscopy and Microanalysis, The University of Sydney, Sydney, NSW 2006, Australiac Electron Microscope Unit, University of New South Wales, Sydney, NSW, Australia

a r t i c l e i n f o

Article history:Received 22 September 2008Accepted 16 March 2009Available online 9 May 2009

Keywords:Atom probeDiffraction electron

* Corresponding author.E-mail address: Ian.Baker@Dartmouth.edu (I. Bake

0966-9795/$ – see front matter � 2009 Elsevier Ltd.doi:10.1016/j.intermet.2009.03.016

a b s t r a c t

The microstructural evolution of a b.c.c.-based, spinodally formed alloy Fe35Ni15Mn25Al25 has beenstudied as a function of annealing time at 550 �C using atom probe tomography and transmissionelectron microscopy, including energy-filtered imaging. The sizes, crystal structures, orientation rela-tionships and compositions of the phases present were determined as a function of annealing time. Thehardness showed complicated behavior as a function of annealing time, consisting of initial hardening,followed by softening and finally, by a rapid hardening behavior. The hardness is controlled both by thecoarsening of the spinodally formed phases, and the precipitation and growth of b-Mn structuredparticles.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Starting with the discovery of the spinodally formed alloyFe30Ni20Mn25Al25 (in at. %), several b.c.c.-based spinodally formedalloys have been discovered in the Fe–Ni–Mn–Al system [1,2,17,31–33,37,39]. The parent phase for this spinodal transformation isbelieved to be a B2 (ordered b.c.c.) phase [17]. The alloys typicallyconsist of coherent, alternating rods or platelets of b.c.c. and B2phases aligned along C100D directions [17]. This type of micro-structure is characteristic of a spinodal decomposition in a b.c.c.-alloy in which the phases are coherent. The wavelengths associatedwith the decomposition range from 10 to 120 nm depending on thealloy composition [39]. Interestingly, plastic deformation at roomtemperature, at least in Fe30Ni20Mn25Al25, is accommodated by theglide of paired a/2 C111D dislocations [32].

The strengths of the spinodally formed Fe–Ni–Mn–Al alloys arecomparable to the strongest maraged aircraft steels and hardestbearing steels, but they possess a better strength-to-weight ratio,and the high aluminum content contributes to oxidation resistance.A survey of several as-cast Fe–Ni–Mn–Al spinodally formed alloys[39] showed that Fe35Ni15Mn25Al25 had the highest as-cast hard-ness at 534�12 VPN. This alloy consisted of the usual alternatingordered and disordered b.c.c. rods, but the B2 rods displayedregions of L21 ordering (a further ordering of the B2 structure).

r).

All rights reserved.

Annealing at 625 �C for 115 h appeared to decrease the L21

ordering, but led to both a substantial increase in the spinodalwavelength from w10 nm to 50 nm and to the formation of(Fe,Mn)-rich precipitates that were 100 nm wide and 0.5–1 mmlong [39]. Similar precipitates have been analyzed in some detail inthe related spinodally formed alloy Fe30Ni20Mn25Al25 [33], in whichthe precipitates were shown to have a clear relationship with thematrix.

In an earlier study [17], it was shown that annealingFe30Ni20Mn25Al25 at 550 �C – the temperature at which the greatesthardness increase occurred – led to a substantial (50%) monotonicincrease in hardness for annealing times up to 96 h. In this paper,we examine the effect of annealing at 550 �C for times up to 72 h onthe hardness of the alloy Fe35Ni15Mn25Al25, and investigate theaccompanying changes in microstructure using both atom probetomography (APT) and transmission electron microscopy (TEM),including energy dispersive X-ray spectrometry (EDS) and energy-filtered imaging (EFTEM).

2. Experimental

An ingot of Fe35Ni15Mn25Al25 alloy was prepared by arc-meltingthe constituent elements of >99.9% purity in a water-cooled coppermold under argon. The ingot was melted four times and flippedbetween meltings to ensure adequate mixing. Analysis usinga Cameca SX50 electron probe microanalyzer confirmed that finalcompositions were all within 0.4 at. % of the nominal values.

I. Baker et al. / Intermetallics 17 (2009) 886–893 887

The alloy was annealed at 550 �C in air for up to 72 h, followed byair cooling. As noted above, such annealing has previously been shownto markedly increase both the hardness and yield strength [17].

Hardness measurements were performed at room temperatureusing a Leitz MINIload tester with a Vickers-type indenter usinga 200 g load and a 12 s drop time. Reported values are the averageof at least 5 measurements.

Transmission electron microscopy was used to characterize themicrostructure. Small samples (0.4�1�3 mm) were first cut witha water-cooled diamond saw and mechanically ground with 1 mmdiamond encrusted paper. These ground wedges, were supportedby copper grids and thinned using a Gatan precision ion polishingsystem. The specimens were examined using both a 120 kV PhilipsCM12 TEM and a 300 kV JEOL 3000F TEM equipped with EDS anda Gatan electron energy loss spectroscopy (EELS) system, incorpo-rating a post-column Gatan Image Filter for energy filtered trans-mission electron microscopy.

Atom probe tomography (APT) was used to study the very finescale phase elemental partitioning between the phases. Since theAPT uses a mass spectrometer it is straightforward to distinguishbetween Fe, Ni, Mn and Al, except for the overlap of 58Fe and 58Ni.However, this overlap can usually be accounted for, since thenatural abundance of the isotopes is known. In fact, the naturalabundance of 58Fe is so low that it can be ignored. Anotherimportant aspect of APT analysis to consider is the different evap-oration fields of the elements at the specimen surface: elementswith low evaporation field may evaporate between the evaporationpulses, and thus not be detected, resulting in an underestimatedmeasured concentration. A pulse fraction (pulse voltage overstanding voltage) of 20%, and low specimen temperature of 80 Kwas used to minimize this effect.

Specimens for APT were prepared from square-sectioned blanks(0.4� 0.4�10 mm), cut using a diamond saw, electropolishedusing standard procedures, and examined using an Imago localelectrode atom probe, which provides high-speed data acquisitionover a large spatial and angular field of view.

3. Results

Fig. 1 shows the hardness dependence on annealing time at550 �C. Unlike the related alloy Fe30Ni20Mn25Al25, which showeda monotonic increase in hardness with increasing annealing time at550 �C [17], Fe35Ni15Mn25Al25 shows an initial increase in hardnessfor short annealing times, which then declines slightly for times up

520

540

560

580

600

620

640

660

680

0.1 1 10 100Hours

II

IV

III

I

Fig. 1. Room temperature hardness dependence of Fe–15Ni–25 Mn–25Al on annealingtime at 550 �C. Specimens from the as-cast material (I) and for three differentannealing times (II – 2 h; III – 22 h; IV – 72 h) were examined using TEM and APT.

to 2 h. For a slightly longer annealing time (2 h) the hardness fallsmore rapidly, after which it is independent of time for times up tow40 h. Finally, for longer annealing times (70 h), the hardnessincreases dramatically from 540 VPN to more than 660 VPN.

As noted above, previous TEM studies [39] showed thatFe35Ni15Mn25Al25 has a periodic coherent microstructure, sug-gesting formation by spinodal decomposition. The interconnectednature of the microstructure is clear in the TEM images, seeFig. 2(a–d). The formation of very large lenticular precipitates1–2 mm long, as previously found in Fe30Ni20Mn25Al25 [17], is alsoclearly apparent in the 72 h annealed specimen, see Fig. 2(d).

The very strong elemental partitioning between the phases,previously noted in Fe30Ni20Mn25Al25 [17], is evident in the EFTEMimages. Fig. 3 shows Fe, Ni, Mn, Al maps of the 2 h annealedspecimen filtered using M-edges for Fe, Ni and Mn, and an L-edgefor Al. It is clear that Ni co-segregates with Al and that Fe co-segregates with Mn. It is worth noting the strong orderingtendencies of Ni and Al, and the high solubility of Mn in both Fe andNi. These factors can provide enough flexibility in elementalsegregation to allow for close lattice matching between the twophases. Such segregation during spinodal decomposition involvesa very low coherency strain barrier and, hence, can proceed muchmore rapidly than phase separation by nucleation and growth [7].

The initially insuppressible concentration waves, which definethe beginnings of a regular, periodic, and often interconnectedarray of two coherent phases, is characteristic of the spinodaldecomposition of a solid solution [8,9]. The orientation ofconcentration waves, i.e., the spinodal phases, is determined by theelastic anisotropy of the lattice, so it favors the weak cube direc-tions of low elastic anisotropy [8]. High resolution TEM (HRTEM)was used to determine the spinodal phase orientations in theFe35Ni15Mn25Al25 specimens. Both the spinodal decompositionwaves and crystallographic lattices can be seen simultaneously inthe HRTEM images shown in Figs. 4(a–d). The insets are the cor-responding fast Fourier transform (FFT) patterns of lattices, whichindicate the direction of the lattices. It is clear that the spinodaldecomposition propagates along C100D directions, as previouslyreported [17].

The wavelength (l) and amplitude (A) are the two parametersmost commonly used to describe spinodal alloys [10,12–14,18–20].The average spinodal wavelengths in the current alloy weredetermined by measuring FFT spots, see Figs. 4(a–d). The results aresummarized in Table 1. It is clear from the micrographs and Table 1that the wavelength increases monotonically with increasingageing time.

Figs. 2(e–h and i–l) shows selected area electron diffractionpatterns (SADPs) taken along [001] and [101], respectively, of boththe as-cast alloy and after 2 h, 22 h and 72 h anneals. The BF imagesin Fig. 2, which were taken along C100D, together with the diffractionpatterns again confirm that the spinodally formed microstructure isaligned along C100D. The SADPs from the 2 h annealed specimentaken along both [100] and [101] clearly display the presence of a B2structure, see Figs. 2(f, j) – the b.c.c. diffraction spots are coincidentwith the fundamental reflections of the B2 structure. Additionaldiffraction spots are present in [101] SADPs from both the as-castand 22 h annealed specimens, as shown in Figs. 2(i, k) confirmingthe presence of L21 ordering, as previously observed for this alloy inthe as-cast state [39]. It is not immediately clear why L21 orderingwas found in the as-cast material and after the 22 h anneal, but notafter the 2 h or 72 h anneals. It was previously noted that the L21

ordering decreased after a 112 h anneal at 615 �C [39]. One possibleexplanation is the additional (L21) ordering reflects compositionalheterogeneity. Long streaks through {110} along [110] are alsopresent in the [001] diffraction patterns of the as-cast and 22 hannealed specimens, see Figs. 2(e, g). The origin of this streaking is

Fig. 2. (a–d) Bright field TEM images taken along [001] of Fe35Ni15Mn25Al25 in the as-cast condition and after annealing for 2 h, 22 h and 72 h, respectively; (e–h) The corresponding[001] diffraction patterns; (i–l) [101] diffraction patterns.

I. Baker et al. / Intermetallics 17 (2009) 886–893888

unclear. It does not arise from the fine scale of the spinodalmicrostructure since the streaks would be along [001] and are fartoo long. It is possible that the long streaks arise from very fineregions of L21 phase.

SADPs from the 72 h annealed specimen show diffraction spotsfrom the large precipitates, see Fig. 3(g). The patterns are consistentwith the precipitates having the b-Mn structure. (Precipitate spotsfrom the specific phase are excluded in the SADP of Fig. 3(l),because the complicated [101] SADP from the b-Mn precipitatesoverwhelms that from the B2 matrix.) These precipitates have beenanalyzed in depth for the related spinodally formed alloyFe30Ni20Mn25Al25 [33] where they were shown to have the b-Mnstructure and to have a clear orientation relationship with thematrix, i.e.,

(100)spinodal // (210)b-Mn

ð010Þspinodal==ð120Þb-Mn(001)spinodal // (001)b-Mn

Similar results were obtained for b-Mn precipitates in the b.c.c.-based alloy Fe–9.1Al–29.9Mn–2.9Cr [26–29].

3-dimensional (3D) reconstructions of the spinodal alloysobtained with the topographic atom probe are shown Fig. 5. Eachdot represents a single atom, and different elements are labeled

with different colors. Again, the fine periodic Ni/Al-rich and Fe/Mn-rich regions provide evidence of spinodal decomposition. The 3Dreconstruction provides a good representation of the microstruc-ture morphology. The coarsening of the microstructure withincreased annealing time is clearly evident in the reconstructions.

The 1-dimensional (1D) concentration profiles along the spi-nodal decomposition direction give a possible way to determine thewavelength and amplitude of spinodal decomposition, see Fig. 6.Although the concentration profiles can, in principle, provide thewavelength, it is difficult to align the 1D profile along the spinodalpropagation. Also, the spatial resolution of atom probe is not betterthan TEM, so we adopted the wavelength values given by TEM, seeTable 1.

The solute distributions can be determined using a frequencydistribution, which is a plot of the occurrence frequency of datablocks with a given number of solute atoms against that numberfrom the APT data. Alternatively, the frequency distribution mayalso be plotted in terms of concentration, i.e., the occurrencefrequency of the element at a certain concentration [34]. Thelatter approach was used in the present work and twin Gaussianpeaks fitted to the data based on the Langer et al. [25] treatment ofnon-linear spinodal decomposition. The amplitudes of the spi-nodal decomposition after various times were thus derived andthe results are summarized in Table 1, from which it is evident

Fig. 3. Energy filtered images of Fe35Ni15Mn25Al25, using M-edges for Fe, Ni and Mn and the L-edge for Al, of the 2 h annealed specimens. The bright areas show higherconcentrations of the element used for mapping.

I. Baker et al. / Intermetallics 17 (2009) 886–893 889

that there is no clear correlation between the compositions of thephases and the annealing time.

4. Discussion

Although the yield strengths of spinodal alloys can be quite highon initial formation, their strengths often increase considerablywith ageing as the amplitude of the sinusoidal spatial fluctuation incomposition increases [6,20–22,38,41]. Several models have beendeveloped to explain the yield strength of spinodal alloys. Katoet al. [22] evaluated several of these models using their data forsingle crystals of a Cu–10Ni–6Sn spinodal alloy together with otherdata from the literature.

The first model of strengthening in spinodal alloys was devel-oped by Cahn [10] and based on gliding perfect edge or screwdislocations interacting with internal coherency stresses arisingfrom the periodic shear stress. (Gerold and Haberkorn [13] useda similar approach to Cahn in their model of edge dislocationsinteracting with coherency strain fields in precipitation systems.)The critical resolved shear stress (CRSS) from the movement of anedge dislocation through a compositional fluctuation of amplitude3A was derived to be:

sc ¼A2h2Y2bl

2ffiffiffi

2p

pg

where h¼ 1/a(da/dc), where a is the lattice parameter and c theconcentration of a particular element; b is the magnitude of the

Burgers vector; l is the wavelength or periodicity of the composi-tional fluctuations; g is the line tension; and for compositionalfluctuations in C100D directions in a cubic crystal, Y can be written:Y¼ (C11þ C12)(C11� C12)/C11, where Cij are the elastic stiffnessconstants. Unfortunately, this model, which does not take intoaccount any surface energy effects, not only gives values which aretoo small when compared to experimental results for Cu–Ni–Snand Cu–Ni–Fe alloys [6,21,22] but, more importantly, the relation-ship expressed by the model that sc is proportional to A2l is notusually observed experimentally [6,12,22,36]. On the other hand,Schwartz and Plewes [38] found that the measured yield strengthincrease after ageing Cu–9Ni–6Sn showed a linear relationshipwith both A2 and l, both of which they estimated from X-raydiffraction data rather than measuring directly.

Cahn [10] also considered an Orowan looping-type mechanismand determined that the yield strength would be proportional to(A/l2)1/3 for this case.

Ghista and Nix [14] developed a model based on the interactionbetween gliding dislocations and the periodic shear modulusfluctuations and, hence, obtained for the CRSS:

sc ¼4DGb

l

where DG is the amplitude of the periodic fluctuations in shearmodulus. It is reasonable to assume that to a first approximation,DG is proportional to A [22]. Thus, in this case, sc is proportional toA/l. Interestingly, Ditchek and Schwartz [12] developed a modelbased on mismatch strains caused by motion of dislocations on the

Fig. 4. HRTEM images taken along [100] and corresponding FFT patterns for (a) as-cast, (b) 2 h, (c) 22 h, and (d) 72 h annealed Fe35Ni15Mn25Al25. The white arrows are along [100].

I. Baker et al. / Intermetallics 17 (2009) 886–893890

slip plane and also arrived at sc proportional to A/l. Again, unfor-tunately, this relationship does not appear to be observed experi-mentally [22,36].

A different coherency stress model was proposed by Dahlgren[11] who assumed a lamellae square-wave structure in contrast toCahn’s sinusoidal wave assumption. Unlike Cahn, he did not putconstraints on the dislocation shape, nor did he include line tensionforces. He found that yield stress is proportional to A and inde-pendent of l. While Cahn’s result predicted a yield stress smallerthan the experimentally observed, generally the Dahlgren equationpredicts yield stresses that are too large. Although the spinodaldecomposition of Fe–Ni–Mn–Al alloys has nearly square-waveshape, the spinodal decomposition of Fe35Ni15Mn25Al25 does notappear to follow this model.

Later, Hanai et al. [18] developed a model based on thecreation of new interfaces on the slip plane when a dislocationmoves through a compositionally modulated structure andobtained:

sc ¼ KGA2

l

Table 1The phases, amplitude, wavelength and hardness of as-cast and annealed alloys.

Condition Phases Amplitude (%) Wavelength(nm)

HardnessVPN

Fe Ni Mn Al

As-cast B2, L21 38.3 32 14.4 21 17 5432 h B2 45.5 34 10 22 27 57822 h B2, L21 45.5 35 12 22.5 39 54172 h B2, b-Mn 50 23 9 25.5 50 666

where G is the interfacial energy per unit area of slip plane, and K is4O2p/9 for a screw dislocation and O6p/3 for an edge dislocation.This model produces both a different magnitude for sc anda different functional dependence than Cahn’s model, i.e., sc isproportional to A2/l rather than A2l, but has the same general lackof agreement with experimental data [22].

Kato et al. [20] developed a model similar to Cahn’s based on theinteraction of a gliding dislocation with the periodic fluctuation ininternal coherency stresses. However, their model assumeda gliding 60� dislocation as opposed to the screw or edge disloca-tions considered in Cahn’s analysis. Thus, they obtained thefollowing expression for the CRSS:

sc ¼AhYffiffiffi

6p

Unlike any of the earlier models, this model does not depend on l. Itappears to work well for the Cu–Ni–Sn alloys that they studied atsmall values of A, but does not work well at larger values [22]. Katoet al. [22] suggested that this breakdown in their model happenswhen the compositional fluctuations deviate from a sine wave andbecome more like a square wave. An additional factor is thatprecipitation may have occurred in the alloy at the same time as thespinodal decomposition or coarsening was occurring. For example,Ditchek and Schwartz (1980) suggested that precipitation of a D022-structured (CuxNi1�x)3Sn compound occurred in tin-rich regions ofa similar Cu–Ni–Sn alloy as early as after 600 s at 350 �C. Suchprecipitation was also noted by Kratochvil et al. [23] in a similar Cu–Ni–Sn alloy, but at longer times, and is observed in the commercialToughMet Cu–Ni–Sn spinodal alloys [5]. Similar precipitation

Fig. 5. 3-Dimensional reconstructions of Fe–15Ni–25Mn–25Al in (a) the as-cast condition and after annealing for (b) 2 h, (c) 22 h and (d) 72 h. Scales are in nm.

I. Baker et al. / Intermetallics 17 (2009) 886–893 891

behavior has been observed in spinodally decomposed Cu–3.2Ni–0.75Si, in which D022-structured (Cu,Ni)3Si precipitates formduring ageing [40], and spinodal decomposition precedes Cu4Tiprecipitation in Cu-0.9–1.1 at.% Ti [15,24]. Clearly, such precipitation

0

10

20

30

40

50

60

Al Fe Mn Ni

Com

po

sitio

n (at.%

)

Distance (nm)

0 5 10 15 20 25 30 35

0

10

20

30

40

50

60

70 Al Fe Mn Ni

Co

mp

ositio

n (%

)

Distance (nm)

0 20 40 60 80

a

c

Fig. 6. One-dimensional composition profiles from atom probe reconstructions of Fe35Ni15MNote the dramatic increase in phase width (wavelength) in IV.

complicates evaluation of the strengthening from the spinodaldecomposition, as is the case with the alloy studied here.

Like the observations of Kato et al. [22] on Cu–Ni–Sn alloys, Satoet al. [36] noted that in a Fe–30Mn–9Al–0.9C (in wt.%) alloy

10 20 30 40 50 600

10

20

30

40

50

60 Al Fe Mn Ni

Co

mp

ositio

n (%

)

Distance (nm)

-5 0 5 10 15 20 25 30 35 400

10

20

30

40

50

60

70 Al Fe Mn Ni

Co

mp

ositio

n (%

)

Distance (nm)

b

d

n25Al25 in (a) the as-cast condition and after annealing for (b) 2 h, (c) 22 h and (d) 72 h.

Table 2The dependency of the yield stress on A and l predicted by various models forspinodally formed alloys.

Author A, l dependency Comments

Cahn [10] A2l f.c.c. shearingCahn [10] (A/l2)1/3 f.c.c. loopingGhista and Nix [14] A/l shearingDahlgren [11] A shearingHanai et al. [18] A2/l surface energyKato et al. [20] A f.c.c. shearingKato [19] C1Aþ C2A/l b.c.c. shearing

I. Baker et al. / Intermetallics 17 (2009) 886–893892

a substantial increase in yield strength occurred during the earlystages of ageing without any significant changes in l. Similarly, inboth Cu–Ni–Fe [30] and Cu–Ni–Cr alloys [41,42], l was found to beessentially constant during the early stages of aging during whichthe strength increases substantially. Interestingly, at longer ageingtimes, the strength in the Cu–Ni–Fe studied by Livak and Thomas[30] reached a plateau, the beginning of which corresponded towhen l started to increase. Sato, Tagawa and Inoue [36] also notedthat the increase in l at longer ageing times was not accompaniedby any significant changes in the yield strength. Both, Livak andThomas [30] and Sato et al. [36] concluded that the strength waslinearly related to A. (As in most studies, Sato et al. [36] did notmeasure A directly, but instead estimated it from the strainamplitude obtained by deconvolution of X-ray diffractionmeasurements and using (C11þ2C12)/C11¼1.) The data cited heresuggest that the yield strength does not depend on l, in directcontradiction to the models of Cahn [10], Ghista and Nix [14] andHanai et al. [18], but does depend on A.

Butler and Thomas [6] studied the microstructure of a spino-dally decomposed Cu–Ni–Fe alloy with a composition at the centerof the miscibility gap and found that initially it consisted of rod-likeblocks with diffuse interfaces, which changed into distinct {100}planar coherent interfaces as ageing proceeded and l (measureddirectly with a TEM) increased. At longer ageing times, coherencywas lost and interfacial dislocations appeared. Deformation in allcases appeared to involve shearing of the phases. In contrast to thestudies cited above, Butler and Thomas found that during ageingthe yield strength initially increased with increasing l, but even-tually became roughly constant even though l continued toincrease with increasing ageing time. They did not directly measureA, but inferred the change in its value from magnetic measure-ments, and concluded that the yield strength was linearly related toA. More recently, Gudladt et al. [16] also found that the yieldstrength increased at the same time as l (measured using a TEM)increased during ageing of a spinodally decomposed Cu–Ni–Fealloy (in line with Cahn’s model). Although they did not measurethe chemistry, they concluded that the hardening must be domi-nated by the increase in A with increased ageing time.

Kato [19] developed a similar model to that produced by Katoet al. [20], but specifically focused on strengthening in spinodallydecomposed b.c.c.-alloys. In the model, the strengthening arisesfrom the internal coherency stresses due to the misfit effects arisingfrom composition differences as well as from the spatial variationin elastic modulus, i.e.,

Dsc ¼AhY

2þ 0:65DGb

l

In one of the few studies of strengthening in b.c.c. spinodal alloys,Park et al. [35] ascribed the yield strength increases in Fe–30 wt.%Cr during ageing to the coherent internal stress field. This materialis different from many of the other spinodal alloys discussed abovein that the phases do not show any crystallographic alignment, butinstead exhibit a vein-like mottled structure, due to the very lowelastic strains [4]. Interestingly, in the same Fe–30 wt.% Cr alloy,Brenner et al. [4] found that the hardness increased as both A and lincreased during ageing.

The dependencies on A and l of the yield stress for variousmodels are summarized in Table 2.

As noted earlier, the related spinodally formed alloyFe30Ni20Mn25Al25 showed a monotonic increase in hardness afterannealing for various times at 550 �C. The final hardness increase inthat alloy coincided with the nucleation and growth of large b-Mnprecipitates. In the present Fe35Ni15Mn25Al25 alloy the hardnessshowed much more complex behavior, i.e., upon annealing the

alloy showed an initial hardening followed by softening and finallya dramatic increase in hardness for anneals up to 72 h at 550 �C. Inan earlier study on Fe35Ni15Mn25Al25 that had been directionallysolidified, the initial hardness was lower (437–462 VPN) than in thepresent study and did not change for annealing times up to 5 h at550 �C, but the hardness after annealing for 72 h at 550 �C wasw670 VPN [3], i.e., comparable to the alloy studied here. Again,large b-Mn precipitates were present, suggesting that at longerannealing times the strength depends strongly on these particles.

Clearly, the hardness of the as-cast and subsequently annealedFe35Ni15Mn25Al25 specimens studied here likely arise from otherfactors besides spinodal decomposition. The strength at longannealing times probably is controlled by the large b-Mn precipi-tates, and the material should be modeled as a composite. The B2/b.c.c.-modulated structure seems to lead to high hardness since 2 hannealed specimens display higher hardness than the as-cast and22 h annealed specimens, when the more highly-ordered L21 phaseis present. It is clear that there is no simple relationship between A,l and the yield strength even before the b-Mn precipitates areimportant. One difficulty with applying current models for spinodalstrengthening to the B2/b.c.c. spinodal studied here is that whilea/2 C111D dislocations can glide in the b.c.c. phase, these have to bepaired up to glide in the B2 phase to avoid formation of large areasof antiphase boundary. Hence a model similar to that for g/g0

superalloys may be more appropriate.

5. Conclusions

A microstructure formed by spinodal decomposition has beenobserved in Fe35Ni15Mn25Al25 using both TEM and APT. Chemicalanalysis revealed that the elements partitioned into a Fe/Mn-richb.c.c. phase and a Ni/Al-rich B2 phase with the phases aligned alongC100D. Evidence of some L21 ordering was found in both the as-castalloy and in the specimen annealed for 22 h. Upon annealing, thealloy showed an initial hardening, followed by softening and finallya dramatic increase in hardness for anneals up to 72 h at 550 �C. Thesoftening at intermediate times is probably due to coarsening ofthe spinodally formed matrix, whereas the formation and growthof b-Mn precipitates is probably responsible for the very high finalhardness after long annealing times.

Acknowledgements

This research was supported by Award #DE-FG02-07ER46392from the Division of Materials Sciences, U.S. Department of Energywith Dartmouth College The authors in Sydney also acknowledgethe facilities as well as scientific and technical assistance from staffin the Australian Microcopy & Microanalysis Research Facility(AMMRF) at the Australian Key Centre for Microscopy and Micro-analysis at the University of Sydney. Any opinions, findings, andconclusions or recommendations expressed in this material arethose of the author(s) and do not necessarily reflect the views of theU.S. Government.

I. Baker et al. / Intermetallics 17 (2009) 886–893 893

References

[1] Baker I, Hanna JA, Wittmann MW and Munroe PR, The mechanical propertiesof Fe30Ni20Mn25Al25. In: Processing and fabrication of advanced materials XIVwith frontiers in materials science. 2005-I. p. 237–247.

[2] Baker I, Hanna JA, Wittmann MW, Munroe PR. Microstructure of a spinodalFe–Ni–Mn–Al alloy. In: Proceedings Microscopy & Microanalysis. 2005-II. p.1864–1865.

[3] Baker I, Liao Y, and Wu X. Microstructure and mechanical properties of novelFe–Ni–Mn–Al alloys. Transactions of Nonferrous Metals Society of China, inpress.

[4] Brenner SS, Miller MK, Soffa WA. Spinodal decomposition of Iron-32 at. %Chromium at 470 �C. Scr Metall 1982;16:831–6.

[5] Brush Wellman, Inc.. TOUGHMET specifications, www.brushwellman.com.[6] Butler EP, Thomas G. Structure and properties of spinodally decomposed Cu–

Ni–Fe alloys. Acta Metall 1970;18:347–65.[7] Cahn JW. On spinodal decomposition. Acta Metall 1961;9:795–801.[8] Cahn JW. Spinodal decomposition in cubic crystals. Acta Metall 1962;10:

179–83.[9] Cahn JW. Phase separation by spinodal decomposition in isotropic systems. J

Chem Phys 1965;42:93–9.[10] Cahn JW. Hardening by spinodal decomposition. Acta Metall 1963;11:1275–85.[11] Dahlgren SD. Correlation of yield strength with internal coherency strains for

age-hardened Cu–Ni–Fe alloys. Metall Trans A 1977;8:347–51.[12] Ditchek B and Schwartz LH. On the strength of spinodal alloys. In: Proceedings

of the 4th international conference strength of metals and alloys (ICSMA4).1976. 3: p. 1319.

[13] Gerold V, Haberkorn H. On the critical resolved shear stress of solid solutionscontaining coherent precipitates. Phys Stat Sol 1966;16:675–84.

[14] Ghista DN, Nix WD. A dislocation model pertaining to the strength of elasti-cally inhomogeneous materials. Mater Sci Eng 1969;3:293–8.

[15] Greggi J, Soffa WA. Anomalous age hardening effects in Copper–Titanium alloysstrengthened by modulated microstructures. Scr Metall 1980;14:649–52.

[16] Gudladt HJ, Wunderlich W, Costlas E. The influence of microstructure on themechanical properties of a spinodally decomposed Cu–Ni–Fe alloy. Z Metal-lkunde 1997;88:642–7.

[17] Hanna JA, Baker I, Wittmann MW, Munroe PR. A new high-strength spinodalalloy. J Mater Res 2005;20:791–5.

[18] Hanai Y, Miyazaki T, Mori H. Theoretical estimation of the effect of interfacialenergy on the mechanical strength of spinodally decomposed alloys. J Mat Sci1979;14:599–606.

[19] Kato M. Hardening by spinodally modulated structure in BCC alloys. ActaMetall Mater 1981;29:79–87.

[20] Kato M, Mori T, Schwartz LH. Hardening by spinodal modulated structure.Acta Metall Mater 1980;28:285–90.

[21] Kato M, Schwartz LH. The temperature dependence of yield stress and workhardening in spinodally decomposed Cu–10Ni–6Sn. Mat Sci Eng 1979;41:137–42.

[22] Kato M, Katsuta S, Okamine S, Sato A. Deformation behavior and micro-structure of Cu–10Ni–6Sn spinodal alloy single crystals. Mat Sci Eng1986;77:95–102.

[23] Kratochvil P, Mencl J, Pesicka J, Komnik SN. The structure and low temperaturestrength of the age hardened Cu–Ni–Sn alloys. Acta Metall 1984;32:1493.

[24] Kratochvil P, Haasen P. A model for an anomaly in the age hardening of Cu–Tisingle crystals. Scr Metall 1982;16:197–200.

[25] Langer JS, Bar-on M, Miller HD. New computational method in theory ofspinodal decomposition. Phys Rev A 1975;11:1417–29.

[26] Liu TF, Tasy JC. Morphology of A12-Alpha-Mn structure. Scr Metall1987;21:1213–8.

[27] Liu TF, Wan CM. Alpha-Mn structure in a Fe–Al–Mn–Cr alloy. Scr Metall1985;19:727–32.

[28] Liu TF, Wu CC. Beta-Mn structure in an Fe–Al–Mn–Cr alloy. Scr Metall1989;23:1087–92.

[29] Liu TF, Wu CC. The orientation relationship between A13 Beta-Mn and ferrite.Scr Metall 1989;23:1243–8.

[30] Livak RJ, Thomas G. Spinodally decomposed Cu–Ni–Fe alloys of asymmetricalcompositions. Acta Metall 1971;19:497–505.

[31] Loudis JA, Boyd TC, Coen D, Baker I. Microstructure and mechanical propertiesof an extruded Fe30Ni20Mn25Al25 alloyAdvanced intermetallic-based alloys.Proc Mat Res Soc 2007;980. 0980-II01-02980.

[32] Loudis JA, Baker I. Dislocation identification and in situ straining in the spi-nodal Fe30Ni20Mn25Al25 alloy. Micros Res Tech 2008;71:489–96.

[33] Loudis JA, Baker I. a- and b-Mn precipitates in the spinodal Fe30Ni20Mn25Al25alloy. Philos Mag 2007;87:5639–56.

[34] Miller MK. Atom probe tomography: analysis at the atomic level. New York:Kluwer Academic/Plenum Publishers; 2000. p. xiv, 239–506.

[35] Park KH, Lasalle JC, Schwartz LH, Kato M. Mechanical properties of spinodallydecomposed Fe–30 wt%Cr alloys: yield strength and aging embrittlement.Acta Metall 1986;34:1853–65.

[36] Sato K, Tagawa K, Inoue Y. Spinodal decomposition and mechanical propertiesof an austenitic Fe–30 wt%Mn–9 wt%Al–0.9 wt%C alloy. Mater Sci Eng1989;A111:45–50.

[37] Saxey DW, Hanna J, Zheng R, Marceau RKW, Baker I and Ringer SP. Nano-structural analysis of advanced alloys in a local electrode atom probe. In:Proceedings of Microscopy & Microanalysis. 2005. p. 872–873.

[38] Schwartz LH, Plewes JT. Spinodal decomposition in Cu-9 wt percent Ni-6 wtpercent Sn.2. Critical examination of mechanical strength of spinodal alloys.Acta Metall Mater 1974:911.

[39] Wittmann M, Baker I, Hanna J, Munroe PR. Microstructure and mechanicalproperties of Fe–Ni–Mn–Al alloys. Proc Mat Res Soc 2005;842. S5.17.1–6.

[40] Zhao DM, Dong QM, Liu P, Kang BX, Huang JL, Jin ZH. Structure and strength ofthe age hardened Cu–Ni–Si alloy. Mater Chem Phys 2003;79:81–6.

[41] Rao RP, Agrawal BK, Rao AM. Studies of spinodal decomposition in Cu-27Ni-2Cr alloy. J Mater Sci 1986;21:3759.

[42] Rao RP, Agrawal BK, Rao AM. Hardening mechanism in spinodal Cu-Ni-Cralloys. Mater Sci Eng 1987;92:199.