O

Cfc

DCa

b

B

a

A

R

A

A

K

F

S

m

X

M

1

Fe

h2a

j m a t e r r e s t e c h n o l . 2 0 1 9;8(1):923–934

www.jmrt .com.br

Available online at www.sciencedirect.com

riginal Article

haracterization of an ASTM A335 P91erritic-martensitic steel after continuous coolingycles at moderate rates

enise Alejandra Carrizoa, Jorge Ignacio Besokyb, María Luppoa, Claudio Danona,inthia Paula Ramosb,∗

Gerencia Materiales, Comisión Nacional de Energía Atómica, Av. Gral. Paz 1499, B1650KNA San Martín, Buenos Aires, ArgentinaGerencia Investigación y Aplicaciones, Comisión Nacional de Energía Atómica, CONICET, Av. Gral. Paz 1499, B1650KNA San Martín,uenos Aires, Argentina

r t i c l e i n f o

rticle history:

eceived 19 January 2018

ccepted 9 July 2018

vailable online 16 August 2018

eywords:

erritic-martensitic steels

canning and transmission electron

icroscopy

-ray diffraction

össbauer spectroscopy

a b s t r a c t

In this contribution some aspects of the behavior of the ASTM A335 P91 (9Cr1MoVNbN)

steel subjected to continuous cooling cycles under fixed austenitization conditions are

studied. Representative samples of the structures obtained after cooling at moderate rates

(i.e. pure martensitic and mixed martensitic-ferritic) were analyzed in order to incorpo-

rate information to the Continuous Cooling Transformation (CCT) diagram of this material.

The characterization was carried out by means of scanning and transmission electron

microscopy (TEM), X-ray diffraction and Mössbauer spectroscopy. For the working conditions

here employed, the results showed the existence of retained austenite within both, the pure

martensitic and the mixed martensitic-ferritic domains of cooling rates, adding important

information to the CCT diagram since retained austenite presence could be detrimental to

the mechanical properties of the steel. Second phase precipitates, being relevant to fix the

mechanical behavior of the material, were identified by means of TEM on carbon replicas.

Mössbauer spectroscopy was very helpful to detect a low volume fraction of cementite-type

phase in all of the samples, considering that it could not be conclusively determined by XRD

due to orientation effects.

© 2018 Brazilian Metallurgical, Materials and Mining Association. Published by Elsevier

Editora Ltda. This is an open access article under the CC BY-NC-ND license (http://

. Introduction

erritic-martensitic steels of the 9%Cr1%Mo type have beenxtensively used in conventional power plant components,

∗ Corresponding author.E-mail: ciramos@cnea.gov.ar (C.P. Ramos).

ttps://doi.org/10.1016/j.jmrt.2018.07.004238-7854/© 2018 Brazilian Metallurgical, Materials and Mining Assocrticle under the CC BY-NC-ND license (http://creativecommons.org/lic

creativecommons.org/licenses/by-nc-nd/4.0/).

heat exchangers, piping and tubing, etc., due to an excellentcombination of properties such as creep resistance, tough-ness and corrosion resistance at high temperatures. In the fourlast decades, 9%Cr1%Mo-based modified alloys have also beendesigned and studied for applications as structural materials

in the new generation of nuclear fission reactors, expectedto work at high to very high temperatures (Generation IVreactors) and in fusion reactors as well. These new envis-aged applications have raised very exigent requirements, as

iation. Published by Elsevier Editora Ltda. This is an open accessenses/by-nc-nd/4.0/).

n o l . 2 0 1 9;8(1):923–934

Table 1 – Chemical composition of the studied ASTMA335 grade P91 steel.

C Mn Si Cr Mo V Cu Sn

0.100 0.360 0.240 8.370 0.880 0.211 0.160 0.008

924 j m a t e r r e s t e c h

elevated temperatures and aggressive environments in stan-dard service conditions, which must be coupled with theexistence of strong irradiation fields [1–3].

Grade P91/T91, P denoting pipes and T tubes, is a ferritic-martensitic 9%Cr1%Mo steel microalloyed with V and Nb. Itsenhanced properties enabled steam-plant components to bemanufactured with thinner walls, thus minimizing thermallyinduced stresses. Nb, V and N addition also favored the creep-rupture strength.

The standard manufacturing steps of P91/T91 steelsinclude normalizing at 1050–1060 ◦C and tempering at750–760 ◦C. The resulting microstructure is a lath martensite,and the main type of precipitates are M23C6 carbides (M = Cr,Fe) and MX carbonitrides (M = Nb, V; X = C, N). Although thestandard heat treatment ensures the basic sought-after prop-erties of the alloy, the procedure has recently been subjectedto review. Some variations have been proposed in order toattain a better performance in the mechanical properties ofthe material. For instance, two-step tempering treatments(low and high temperature stages, respectively) have beendeveloped to combine a narrow average lath width along withadequate precipitation strengthening [4].

The Continuous Cooling Transformation (CCT) diagrams ofthese alloys are relatively simple and display basically threephase fields; that is, austenite, ferrite and martensite, withthe additional presence of second phase precipitates in all ofthem [5–7]. The key features of the CCT diagram of P91/T91steels include two critical cooling rate values: c˛, which rep-resents the maximum cooling rate to achieve a fully ferriticmicrostructure (on increasing cooling rates), and cm, whichamounts the minimum cooling rate to have a fully marten-sitic microstructure (on decreasing cooling rates). For coolingrates between these two values, the final microstructure is amix of martensite and ferrite (with precipitates).

Despite a good knowledge of the CCT diagram of thesealloys, some issues still deserve a renewed attention. Theexisting literature on the characterization of the metallurgi-cal state of 9%Cr steels after continuous cooling at moderaterates is scarce because, traditionally, cooling behavior studieshave been directed toward welding or normalizing conditions,however, some differences in the measured transformationtemperatures have been found [8]. In this way, novel ideas tendto the design of new heat treatments for these steels, whichinclude the introduction of a non-vanishing ferrite fraction inthe final product via isothermal steps or cooling at moderaterates [9,10].

The relevance of microstructural studies on moderate con-tinuous cooling also comes from manufacture processes ofthick-walled components. As the wall becomes thicker, thecooling rate of the central section slows down and, as a result,martensite domains or ferrite grain size (if any) may coarseand the precipitation conditions may change [11].

On the other hand, there is also ample evidence of thepresence of retained austenite (RA), not reported in the CCTdiagrams, in the microstructure of these steels under differentexperimental or processing conditions [12–16]. In particular,

when following the conventional normalization and temper-ing treatment and depending on the tempering temperature,RA could promote the occurrence of non-tempered or inho-mogenously tempered martensite in steel components or

Al Nb N Ti P S Co0.009 0.075 0.061 0.004 0.014 0.001 0.010

parts, producing a negative influence in the mechanical prop-erties of the material during its service life [17]. Even a lowfraction of film-like, interlath RA could strongly affect thetoughness of some 9Cr steels if a lower-temperature temper-ing –such as the one used in manufacturing steam turbinerotors [18]- is expected. Hence the importance of assessingthe RA content after continuous cooling cycles.

In the present contribution, the microstructure of P91steel samples tested under continuous cooling conditions atmoderate rates is studied taking advantage of the diverseinformation brought by different characterization techniques.

2. Methods

The starting material is an ASTM A335 grade P91 steel whosechemical composition is shown in Table 1.

Specimens – 25 mm × 5 mm × 5 mm rectangular prisms –were machined from a seamless pipe 219 mm in externaldiameter and 32 mm in wall thickness following the rollingdirection.

Heat treatments were done in a high temperature furnaceAdamel DHT 60 under vacuum better than 10−5 Torr. Thermalcycles were chosen – on the grounds of previous information-in order to ensure a fully martensitic or a mixed (marten-site + ferrite) microstructure: heating at 300 ◦C/h up to 1050 ◦C(that is, austenitization), holding at 1050 ◦C during 30 min andthen cooling at 200, 150, 125 and 100 ◦C/h. Austenitization tem-perature was chosen following the standard manufacturingconditions for P91/T91 steels, and holding time was fixed asspecified in the reported CCT diagram [5].

Samples were characterized by Field Emission Gun Scan-ning Electron Microscopy (FEG-SEM), transmission electronmicroscopy (TEM), X-ray diffraction (XRD) and Mössbauerspectroscopy.

2.1. FEG-SEM

The surface of each sample was prepared for examination bythe standard metallographic procedures. Grinding with SiCpapers (220, 320, 400 and 600 grit) and a final polishing with a1 �m diamond cloth were carried out. Chemical etching wasaccomplished with the Villela or Nital reagents, depending onthe microstructure expected to be revealed. Observations werecarried out in a Zeiss Supra 40 microscope operated at 5 keVaccelerating voltage and using a working distance of ∼2–3mm.

2.2. TEM

Replicas were obtained by carbon evaporation in order to iden-tify the various precipitated second phases. The analysis was

o l . 2

csd

2

MigKcwTwsIwdirybwt

2

PMtss

Fo

Fpm

j m a t e r r e s t e c h n

arried out in a Phillips CM200 transmission electron micro-cope (TEM) equipped with an EDAX-DX4 system for energyispersive X-ray spectroscopy (EDS).

.3. XRD

easurements were done on bulk samples using a Panalyt-cal Empyrean diffractometer with the �–2� Bragg–Brentanoeometry, a flat graphite monochromator and Ni-filtered Cu-� radiation. The mid-plane of each sample was exposed byutting and the resulting surface was prepared in the sameay as described above, adding an electrolytic polishing step.he diffracted radiation was collected between 35 and 120◦ (2�)ith an ultra-fast PixCel area detector. The step scan angle and

tep scan time were settled in 0.026◦ and 939 s respectively.n order to deal with orientation effects, XRD measurementsere performed in two configurations, i.e. with the mainimension of the sample parallel and perpendicular to the

ncident beam. The spectra were processed by the Rietveldefinement method with the software MAUD (materials anal-sis using diffraction) [19]. Peak broadening was consideredy modeling microstrains and crystallite sizes. Texture effectsere taken into account considering previous work in litera-

ure [20].

.4. Mössbauer spectroscopy

owders of the samples were prepared with a diamond file.

easurements were performed at room temperature (RT) in

he transmission geometry with a conventional Mössbauerpectrometer set in constant acceleration mode with a 57Coource in a Rh matrix. Spectra were recorded at 11 mm/sec

a

300 nm1 µm

b

ig. 1 – FEG-SEM micrographs of the sample cooled at 200 ◦C/h. Lf needle-like precipitates (c).

ba

200 nm1 µm

ig. 2 – FEG-SEM micrographs of the sample cooled at 150 ◦C/h. Lrior austenite grain boundaries (a). Precipitates in the ferrite noartensite (c).

0 1 9;8(1):923–934 925

and fitted by using the Normos program developed by Brand[21], according to two different procedures; i.e., by decompo-sition in broadened discrete sextets with a least-squares fitto Lorentzian lines and then by using hyperfine field distribu-tions. Isomer shift (IS) values were referred to that of �-Fe atRT.

3. Results

3.1. FEG-SEM observations

Figs. 1–4 representative FEG-SEM micrographs of the mainmicrostructures found as a result of continuous cooling atmoderate rates, i.e., pure martensitic and mixed martensitic-ferritic.

Fig. 1 corresponds to the sample cooled at 200 ◦C/h. Fig. 1aand 1b evidence a typical lath martensite structure, similarto that obtained after air cooling [13,22], with the presence ofsecond-phase precipitates. Fig. 1c shows details of the pre-cipitates with two different morphologies, some of them lookcoarser and the others are oriented, needle-shaped ones.

A different pattern for the sample cooled at 150 ◦C/h canbe seen in Fig. 2, where an incipient new constituent is form-ing at prior austenite grain boundaries. Fig. 2a displays thelath martensitic structure but this time small nodules of fer-rite begin to appear. This ferrite includes carbides grownduring cooling before the martensitic transformation. Themorphology of these second-phase precipitates within the fer-

rite nodules (Fig. 2b) is different from the morphology of thoseobserved within martensite laths (Fig. 2c).

Micrographs corresponding to the samples cooled at 125and 100 ◦C/h are displayed in Figs. 3 and 4 respectively, where

200 nm

c

ath martensite microstructure (a, b). Preferred orientation

c

100 nm

ath martensite microstructure –including ferrite nodules atdules (b). Oriented needle-like precipitates within

926 j m a t e r r e s t e c h n o l . 2 0 1 9;8(1):923–934

a

200 nm2 µm 1 µm

cb

Fig. 3 – FEG-SEM micrographs of the sample cooled at 125 ◦C/h. Ferrite nodules in martensite (a) Precipitates in ferrite (b).Precipitates in martensite (c).

a cb

200 nm2 µm10 µm

Fig. 4 – FEG-SEM micrographs of the sample cooled at 100 ◦C/h. Ferrite nodules in martensite (a). High magnification of oneof the ferrite nodules displayed in (a). Precipitates in ferrite (c).

a mixed martensitic-ferritic microstructure is seen with anincreasing ferrite fraction in lowering the cooling rate. Withinthe ferritic phase a great quantity of precipitates can be seenwith different sizes and morphologies (fibers and faceted)while in the martensitic structure the same type of precipi-tates described in Fig. 1c are again distinguished.

The size difference in precipitates may be due to nucle-ation by different mechanisms and in different temperatureranges. Those with larger size were likely nucleated at highertemperatures, having in consequence more time to grow.

3.2. TEM observations

In Figs. 5 and 6 TEM micrographs of the sample cooled at150 ◦C/h, chosen as representative to identify the different sec-ond phase precipitates mentioned above, are displayed.

In the matrix zone of the sample (Fig. 5a) precipitates withdifferent sizes and morphologies are observed. Needle-shapedparticles show an irregular morphology at higher magnifica-tions and were identified as M3C (Fe ∼ 76 wt%, Cr ∼ 24 wt%;only metallic elements) on the basis of selected area electrondiffraction (SAED) patterns (Fig. 5b) and EDS spectra (Fig. 5d).Needles, with an average length of ∼100 nm and an averagewidth of ∼ 10 nm (Fig. 5a), are oriented in preferential direc-tions and these orientations change in different austenitegrains and even in different regions within a grain. On theother hand the precipitates with a coarser aspect were iden-

tified as V-rich MX (Fig. 5a). Fig. 5c shows the SAED pattern ofthe V-rich MX precipitate arrowed in white in Fig. 5a. Fig. 5edisplays the EDS analysis performed on the V-rich MX precip-itates, plotted as a Cr-V-Nb ternary composition diagram.

In the ferrite nodules two kinds of precipitates were dis-tinguished, M23C6 (Fe ∼ 40 wt%, Cr ∼ 53 wt%, Mo ∼ 7 wt%) andhexagonal carbonitrides M2X (V ∼ 4 wt%, Cr ∼ 76 wt%, Mo ∼15 wt%, Fe ∼ 5 wt%) (Fig. 6).

3.3. XRD analysis

3.3.1. Peak broadening effectsThe major matrix constituents were considered to be bctmartensite (space group: I4 mmm) and bcc ferrite (spacegroup: Im-3m). This assumption was made in each case onthe basis of the FEG-SEM observations introduced in Section3.1.

The XRD doublets characteristic of the bct martensitestructure failed to be resolved in the sample belonging to thefully martensitic domain because they overlapped, yieldinga single peak in angular positions very close to those of bccferrite. This is an expected consequence of the low carboncontent of the steel and of the peak broadening due to theintrinsically deformed character of martensite [23].

Furthermore, martensitic transformation produces astrained matrix which exhibits diffraction domains muchsmaller than the ferrite ones. Thus, in line with the broad-ening effects, the width of the matrix peaks diminishes withthe cooling rate for the martensitic-ferritic mixed domainsamples. This is consistent with the increment of the ferritecontent produced by lowering the cooling rate, clearly shownin FEG-SEM micrographs.

As an example, in Fig. 7 the effect of the cooling rate on thewidth of the (1 1 0) peak of the matrix in the transition fromthe fully martensitic sample (cooled at 200 ◦C/h) to the mixeddomain sample (cooled at 100 ◦C/h) is shown.

j m a t e r r e s t e c h n o l . 2 0 1 9;8(1):923–934 927

cb

d e

a

M3C

M3C

z = [120] z = [101]

210

002

111

020

FcKa

00

25

25

25

50

50

N = 27

V (wt%)

Cr (w

t%)N

b (w

t%)

50

75

75

75

100

100

100

0

CrKaCuLa

CrLa

CuKa

MX V-rich

MX V-rich

200 nm

MX V-rich

MX V-rich

M3C

M3C

M3C

Fig. 5 – (a) TEM micrograph in the matrix zone of the sample cooled at 150 ◦C/h where M3C and V-rich MX precipitates canbe distinguished. (b) SAED pattern of the M3C carbide arrowed in white in (a), zone axis z = [1 2 0]. (c) SAED pattern of theV-rich MX precipitate arrowed in white in (a), zone axis z = [1 0 1]. (d) EDS spectrum of the M3C precipitates. (e) Ternarycomposition diagram of V-rich MX precipitates.

ba

M23C6

M2X

111133

2 µm 500 nm

Fig. 6 – (a) TEM micrograph in a ferrite nodule of the sample cooled at 150 ◦C/h where M23C6 precipitates can bedistinguished and the corresponding SAED pattern of the carbide arrowed in white, zone axis z = [3 1 2]. (b) Zoom of thecircled zone in (a) where M2X precipitates are present with an insert in which the corresponding ring diffraction pattern isd

3Fr2fii

isplayed.

.3.2. Phase identificationig. 8 shows, as an example, the XRD pattern and the cor-

esponding Rietveld refinement fit for the samples cooled at00 ◦C/h (Fig. 8a) and 100 ◦C/h (Fig. 8b) along with a magni-ed view of the 37–47◦ and 47–120◦ angular regions in order to

dentify the matrix phases in those samples.

Martensite was identified as the major phase in all of thesamples. According to FEG-SEM observations, ferrite was not

included in the Rietveld refinement of the sample cooled at200 ◦C/h and was refined as a strain-free, minor fraction phasein the sample cooled at 150 ◦C/h; the ferrite content increasedon decreasing the cooling rate as expected.

928 j m a t e r r e s t e c h n o l . 2 0 1 9;8(1):923–934

0

Inte

nsity

2θ46454443

10000

20000

200 °C/h100 °C/h

30000

40000

50000

Fig. 7 – Effect of the cooling rate on the width of the (1 1 0)peak of the matrix for the samples cooled at 100 and200 ◦C/h.

Table 2 – Lattice parameters for the matrix phases. M:martensite, F: ferrite, A: austenite.

(◦C/h) M F A

a [Å] c [Å] a [Å] a [Å]

200 2.8747 2.8771 – 3.5996150 2.8859 2.8874 2.9143 3.6142

a

2010000

1208040

4540

120 1002θ (°)

8060

0

0

40

40

A (111)

A (310)A (220)A (200)

Inte

nsity

M (101) y (110)

M (202) (220)

M (002) (200)

M (112) (211)

20000

20000

4

8

1000

3000

2000

20

Fig. 8 – Diffractograms for the samples cooled at 200 ◦C/h (a) andcorresponding Rietveld refinement fit. M: martensite, F: ferrite, Ashown in (b) due to their extremely low intensity.

125 2.8744 2.8798 2.8749 3.6005100 2.8760 2.8800 2.8768 3.6074

Besides the main peaks of the martensite, or overlappedmartensite-ferrite structures, retained austenite presence wasdetermined in all of the samples. The lattice parameters cal-culated from Rietveld refinement for the matrix phases in thefully martensitic sample (cooled at 200 ◦C/h) and the mixeddomain samples (cooled at 150, 125 and 100 ◦C/h) are displayedin Table 2.

At the same time, a few peaks of low to moderate inten-sity were assigned, tentatively, to cementite – type precipitates(M3C). No information about other type of precipitates could be

b

000

1208040

40 45

1201002θ (°)

8060

0

0

000

000

F (310)

F (211)

F (200)

F (220)

A (111)

F (110)

000

000

000

100 ◦C/h (b) with their respective magnified views and their: austenite. The A peaks for the 50–120◦ region are not

j m a t e r r e s t e c h n o l . 2 0 1 9;8(1):923–934 929

a25000

55 5550 5045 4540 4035 35

30000 b

2θ (°) 2θ (°)

M3C

M3C

(210)

Inte

nsity

(102)

(210)

Fig. 9 – Diffractograms of the sample cooled at 125 ◦C/h taken with the sample perpendicular to the incident X-ray beam (a)a

dtoaa

tdwottt

b(dTFhf

ptTcMtp

3

3Mpmiitsatf

nd with the sample parallel to the incident X-ray beam (b).

etermined from this analysis; however, it must be mentionedhat the peaks identified as the (2 1 0) and (1 0 2) reflectionsf the M3C phase could include a contribution of the (1 1 1)nd (2 0 0) peaks of the VN phase, respectively, in view of theirlmost complete overlapping in each case.

The intensity of these cementite-assigned peaks showedo be strongly dependent on the direction of beam incidence,enoting a pronounced preferred orientation effect, whichould be consistent with the one already noticed by FEG-SEMbservations (Section 3.1). To illustrate this point two diffrac-ograms of the sample cooled at 125 ◦C/h and obtained withhe main sample dimension kept perpendicular and parallelo the incident beam are displayed in Fig. 9.

The configuration with the sample parallel to the incidenteam gives a noticeably reinforced intensity for the (2 1 0) and

1 0 2) M3C-assigned peaks. Even so, these peaks were hardlyistinguishable in the samples cooled at 200 ◦C/h and 100 ◦C/h.his fact becomes evident from the comparison betweenigs. 9 and 8; in the former, the possible cementite peakeights reach a significant fraction of the main martensite-

errite peak, a feature absent in Fig. 8.The knowledge of the behavior of precipitated second

hases and their control throughout the thermal history ofhe material is relevant to fix its final mechanical properties.aking into account the difficulties described above for a con-lusive determination of cementite, a subsequent analysis byössbauer Spectroscopy was necessary to univocally confirm

hat the cementite-type phase was present in all of the sam-les.

.4. Mössbauer spectroscopy

.4.1. General aspects of Mössbauer measurementsössbauer spectroscopy was performed on powdered sam-

les. Even if this technique is suitable to distinguish theatrix component phases (martensite, ferrite) from austen-

te – because the former are ferromagnetic, while the latters paramagnetic at RT – powdered samples (as an alterna-ive to thin foil samples) could be detrimental in this type of

tudy. Powdering could promote martensitic transformationnd, provided the low fraction of retained austenite, informa-ion about this phase could be lost. This effect was observedor example in a duplex steel UR45N, in which the difference

in the relative abundance of the austenite between bulk (i.e.,thin foils) and powder samples was about 25% on the average[24]. A similar transformation due to mechanical stresses wasalso observed by Skrzypek et al. [25] in hard ball bearing steels.In the same way a reduction in the austenite content was alsofound due to mechanical polishing with different grain size inan austempered ductil iron by Mercader et al. [26].

However, taking into account that retained austenite phasewas already determined from XRD analysis in this work, Möss-bauer spectroscopy was carried out on powdered samplesresigning eventual information about austenite presence butgaining additional data about cementite-type precipitates (notconclusively identified by XRD in some samples). Powderedsamples are easier and less time consuming to prepare thanthin foils and give a better final signal to noise ratio in the spec-tra, key fact when a so minor phase as M3C has to be detectedbetween the major matrix components. An ulterior, detailedstudy of thin foils of this steel was also done, determining theretained austenite presence and its content in each samplewith more accuracy than XRD; those results will be publishedelsewhere.

Unfortunately, because of the low carbon content in thesteel, a distinction between ferrite and martensite by Möss-bauer Spectroscopy is not possible in this case, due to theirsimilar Fe environments. That is why reference in the fol-lowing text is made to matrix phases (i.e. martensite and/orferrite) in general.

3.4.2. Sites fitting of Mössbauer spectraThe measured Mössbauer spectra are typical of low car-bon steels, showing the presence of several superimposed,magnetically split subspectra. Most of them with IS andquadrupole shift (2�Q) near 0.0 mm/s (these values are zerofor metallic iron �Fe), denoting the matrix elements. In afirst attempt spectra were fitted to six sextets with differenthyperfine parameters reflecting the influence of the alloyingelements. As all the spectra look similar, an example is dis-played in Fig. 10 for the sample cooled at 100 ◦C/h. Fittingswere made based on our previous work in a T91 steel sub-

mitted to a tempering process at 780 ◦C during different timeintervals [27], in which five sextets were employed to representthe matrix phases. The existence of different neighbors of theFe atom governed by the concentration of alloying elements

930 j m a t e r r e s t e c h n o l

matrix sextets

-5 10

vel [mm/s]

rel.

abso

rptio

n

50-100.93

0.94

0.95

0.96

0.97

0.98

0.99

1.00(Fe, Cr)3C precipitates

Fig. 10 – Sites-fitted Mössbauer spectrum corresponding to

the sample cooled at 100 ◦C/h.

was analyzed; in particular, the behavior of Cr was studied, asthe main substitutional atom.

Thus, the sextet having the highest Bhf (∼34 T) representsFe atoms in a pure Fe matrix, i.e. those not surrounded byalloying-element atoms. The successive four sextets, withBhf ∼ 31, 29, 27 and 24 T, are associated with Fe atoms sur-rounded by an increasing number of Cr atoms. A sixth sextetwas also added to account for a small contribution at low fieldsthat can be perceived in the recorded spectra. This last sextetwith average Bhf ∼ 15–18 T would represent some iron-bearingprecipitates with alloying elements.

It is also worth to mention that the presence of a minutecontribution (<1%) seems to arise from at least one incipientdoublet in the central part of all of the Mössbauer spectra.This contribution could be related to the formation of para-magnetic cementite [28] and/or other types of precipitatediron-containing carbides [29,30] and/or to the presence ofretained austenite [31,32]. Due to the scarce content and tothe overlap of signals around the weak fields, it is impossibleto univocally identify this/these Fe-bearing phase/phases.

3.4.3. Further analysis in terms of distributions ofhyperfine magnetic fieldBeing the main purpose of the Mössbauer spectra analysisthe differentiation between matrix and precipitated secondphases, a detailed study was then performed to deal with thesuperposition of many magnetically split components of thespectra. Fitting was performed in consequence by consideringdistributions of hyperfine magnetic fields caused by the mul-tiple local configurations of the alloying elements around theFe atom.

Two independent hyperfine field distributions wereemployed. For the sake of simplicity and completeness theyare drawn one after the other in Fig. 11, delimited by a dot-ted line. One of them, ranging from 24 to 38 T, represents thematrix elements and shows a dominant maximum fluctuat-

ing around 31 T. The corresponding IS and 2�Q values are closeto zero. This distribution accounts for the different numberof Fe nearest neighbors of the resonant atom. The relativelylarge alloy-element concentration (especially Cr) found for this

. 2 0 1 9;8(1):923–934

steel explains why the maximum Bhf value deviates from theone found for �Fe (33 T). The broad distribution represents theinfluence of the main alloying element (Cr), which has a highprobability of substituting Fe atoms in the crystal lattice (Crand Fe atomic radii are similar [33]). It is expected that sub-stitutional non-magnetic atoms, such as Cr, will diminish thehyperfine magnetic field: the higher the number of Cr neigh-boring atoms, the lower the hyperfine magnetic field [34].

The other hyperfine field distribution instead (ranging from10 to 23 T), has an average Bhf between 17.8 and 19.1 T and an ISvalue different from zero, suggesting the presence of carbidesof the cementite type. The hyperfine fields are slightly lowerthan those reported for Fe3C (Bhf ∼ 20.8 T) [28,35]; once againthis is due to the presence of alloying elements. As mentionedabove for the first hyperfine field distribution, the main alloy-ing element (Cr) would be the responsible for the change inthe average Bhf value. Analyzing the distribution, two or threemain maxima can be distinguished (near 12, 15 and 20 T), cor-responding to precipitates of the cementite type – (Fe1–xCrx)3C– with different Cr content.

4. Discussion

4.1. Matrix phases and retained austenite

The ferritic and martensitic domain boundaries of the CCTdiagram are determined by the critical cooling rates, c˛ and cm,which in turn depend both on the chemical composition of thematrix and the austenitic grain size through the austenitiza-tion temperature. In the experimental conditions employed inthe present work, cm is near 150 ◦C/h. This value arises fromFEG-SEM observations, considering that tiny nodules of fer-rite + carbides begin to come out within the martensitic matrixafter cooling at that rate. As the cooling rate lowers from150 ◦C/h the ferrite fraction gradually increases; XRD analysissupports this observation as well.

XRD results also showed systematically the presence ofretained austenite in low volume fraction in all the sam-ples. This fact suggests that there is a broad range of coolingrates not allowing the complete suppression of austenite;in particular, RA is detected in samples cooled within themartensite-ferrite mixed domain, thus incorporating impor-tant information to the CCT diagram of this steel. These dataare relevant to identify a potential degradation source of thematerial properties in certain manufacture processes. A smallfraction of retained austenite in a component normalizedor subjected to a welding process could imply microstruc-tural heterogeneities at the subsequent tempering stage.Depending on tempering temperature, retained austenitecould decompose to non-tempered martensite (below 680 ◦C),or right to ferrite (above 680 ◦C) [36]. This means that atempered martensite matrix could include non-tempered orinhomogeneously tempered isolated regions, which in turncould alter the strength properties of the material. As men-tioned before work is in progress with a detailed study of

retained austenite in these samples and will be publishedelsewhere.

Regarding the lattice parameters of the matrix phasesobtained by Rietveld analysis it is worth to mention that the

j m a t e r r e s t e c h n o l . 2 0 1 9;8(1):923–934 931

150 °C/h200 °C/h

100 °C/h125 °C/h

P(B

hf)

Bhf [T]35 3530 3025 2520 2015 1510 10

Fig. 11 – Distributions of hyperfine fields (Bhf) employed in the alternative fitting of Mössbauer spectra.

hcftw

4

Tsmvthttr

cctatpa

lo

igh value determined for the lattice parameter of ferrite afterooling at 150 ◦C/h, close to the critical rate, could accountor carbon supersaturation not released by precipitation. Forhe other cooling rates, the determined lattice parameters areithin the expected values.

.2. Precipitated phases

he expected precipitation sequence in high Cr steels dependstrongly on the chemical composition and the applied thermo-echanical treatment. Thus, Kipelova et al. [37] have reviewed

arious isothermal precipitation sequences as a function ofhe precise chemical composition and steel generation withinigh-Cr alloys. For instance, during high (≥750 ◦C) temperature

empering of 9Cr1Mo steels after quenching or normalizing,he commonly reported sequence for the precipitation of Cr-ich carbides – as a function of tempering time – is [17,22]:

Matrix – M3C – M7C3 – M23C6, where M: (Fe, Cr) (1)In grade 91 steels, sequences involving carbides, nitrides or

arbonitrides of the MX type (M = Cr, Nb, V; X = C, N) should beonsidered as well. Furthermore, depending on the temperingemperature, hexagonal carbonitrides of the M2X type havelso been observed [38]. For the 9Cr steel studied here, Hur-ado Norena et al. [13] provided a thorough discussion of therecipitation sequence as a function of the tempering temper-

ture.

The sequence (1) derives from an isothermal path to equi-ibrium, instead, a variety of precipitates with different typesf structures, morphologies and sizes are expected to be

formed during continuous cooling, in different temperatureintervals and according to the transformation behavior of thematrix. This assumption is based on previous, detailed studieson the isothermal austenite-ferrite transformation in differenttemperature ranges and on the anisothermal transformationbehavior of 9Cr steels with various compositions [39–41].

In the present work, FEG-SEM and TEM observationsintroduced in Sections 3.1 and 3.2 allowed to describe the mor-phology and to estimate the size of the precipitates. Withinthe martensitic structure mainly two different morphologiesof precipitates were distinguished, that is, coarse and needle-shaped. On the other hand, in the ferritic phase of the mixedstructure samples, another two different shapes of precipi-tates were visualized (fibers and faceted).

4.2.1. Cementite and V-rich MX precipitation in themartensitic matrixThe features of needle-shaped precipitates observed inmartensite by FEG-SEM agree with those reported by HurtadoNorena et al. [13] and Jones et al. [22], strongly suggestingthat they can be identified as the (Fe,Cr)3C phase. On theother hand, XRD analysis would also indicate the formationof cementite; however, due to the low volume fraction of theprecipitates – consequence of the low C content in the steel-

and to orientation effects, its identification was not conclu-sive in all of the examined cases. On the opposite, Mössbauerspectroscopy results unambiguously established the presenceof M3C-type precipitates in all of the samples.

n o l

932 j m a t e r r e s t e c h

In order to confirm these findings TEM analysis was per-formed in a representative sample, the one cooled at 150 ◦C/h,where both matrix constituents (martensite and ferrite) werepresent with their corresponding precipitates. In fact, theneedle-shaped second phase was indexed as (Fe,Cr)3C by SAEDpatterns (Fig. 5b).

Regarding the origin of the cementite particles foundwithin the martensitic regions of the matrix, the mostaccepted explanation is that they come from an auto-tempering process [22,39]. However, and on the grounds ofdata obtained by Differential Scanning Calorimetry (DSC), Liuet al. [42] claimed that precipitation of (Fe,Cr)3C particles in T91steel samples cooled under argon atmosphere after austeni-tization starts above 750 ◦C. In other words, they suggest thatcementite does not come from an auto-tempering process butthat it is completely formed in the � (austenite) phase beforethe start of the transformation to martensite. Kirana et al. [43]have in turn criticized this hypothesis, discarding cementiteprecipitation before reaching the Ms temperature. The inho-mogeneous feature of cementite precipitation observed in thepresent work, which seems to follow the lath subdivision ofthe martensitic matrix (i.e. densely populated laths vs. hardlypopulated laths of cementite particles, Fig. 3c), could be an ele-ment to support the hypothesis of an auto-tempering process.

On the other hand, for the case of cooling rates resultingin a mixed martensite-ferrite structure, Saroja et al. [41] pro-posed that the supersaturation of carbon in ferrite is relievedby precipitation within the (ferrite) phase, subsequent to theformation of � grains from � and not by partitioning of carbonacross the �/� interface. In their work, for a non-stabilized,plain 9Cr1Mo steel, the M3C carbide was proposed as thephase observed to precipitate within the ferrite nodules nucle-ated and grown during continuous cooling. In summary, inabsence of the stabilizing elements Nb and V, previous reportsaccount for cementite precipitation within martensite [39]and –probably- ferrite [41] for cooling rates in the mixedmartensite-ferrite domain.

As for the V-rich MX particles, it is worth to mention thatprevious studies [13,22] report not having found this phaseafter normalizing (i.e., air cooling) steels of similar composi-tion at 1045–1050 ◦C. Instead, clusters of fine, elongated V-richMX particles were detected after 16 min tempering at 760 ◦C[22] and not detected after 1 h tempering at temperatures up to600 ◦C [13]. Another study [44] draws a similar conclusion, i.e.,MX particles are found after normalizing a 9Cr1MoVNb steelfor 15 min at temperatures ranging from 1050 to 1200 ◦C, butexperimental evidence allows to make sure that they do notprecipitate on cooling. The comparison of these observationswith the results in the present work suggests that a coolingrate of 150 ◦C/h is enough to promote V-rich, MX particle pre-cipitation within martensite on cooling.

4.2.2. M23C6 and M2X precipitation in ferritic nodulesThe precipitates present in the ferrite phase nodules wereidentified by SAED and EDS as M23C6 and M2X particles. It isnatural to assume that the low solubility of C and N in ferrite

promotes the precipitation of these or other carbide, nitride orcarbonitride phases. Thus, new thermomechanical processesenvisaging the introduction of a ferrite phase fraction in thestandard metallurgical state of the material – that is, prior to

. 2 0 1 9;8(1):923–934

entering in service – should take into account that a precipi-tate fraction will be introduced along with it. In this way, ferritecould act as a “carbon sink” via second phase precipitation init, leaving less carbon available to form the fine MX precipitatedispersion that enhances the creep strength of the martensiticmatrix on further cooling and eventual tempering.

The finding of the M23C6 and M2X phases within the fer-ritic nodules in the anisothermal process in this work is fullyconsistent with the already reported observations on isother-mal and anisothermal austenite-ferrite transformation in 9Crsteels with varying composition [39,40]. In the isothermal case,these phases were observed to form early in the transforma-tion, the C/N ratio being the main factor in promoting a higherfraction of one or another [39].

5. Conclusions

The transformation behavior of ASTM A335 P91 steel sam-ples under continuous cooling conditions has been studiedin the range of moderate cooling rates. A thorough char-acterization of the resulting samples showed that retainedaustenite was systematically identified for all cooling rates,i.e., within the martensitic and part of the mixed martensitic-ferritic domains of the CCT diagram. This fact adds importantinformation to that diagram and opens to debate about whatthe conditions are for a four-phase metastable state to existinvolving austenite, ferrite, martensite and carbides in 9Crsteels.

Among the precipitated second phases observed by FEG-SEM and TEM, (Fe, Cr)3C and V-rich MX were identified inmartensite and M2X and M23C6 in the ferrite phase. A cool-ing rate of 150 ◦C/h was enough to promote V-rich MX particleprecipitation on cooling.

Mössbauer spectroscopy showed to be a good tool to differ-entiate the minor second phases with magnetic character as(Fe, Cr)3C which due to the scarce content and to orientationeffects were not conclusively determined by XRD. Mössbaueranalysis revealed the presence of ferromagnetic cementite-type precipitates in all of the examined samples. The questionabout the precise temperature range for cementite nucleationduring cooling remains opened, but this work suggests that forthe mixed-domain cooling rate range, cementite could nucle-ate and grow within martensite by an auto-tempering process.

Conflicts of interest

The authors declare no conflicts of interest.

Acknowledgements

Mrs. Alicia Petragalli is kindly acknowledged for her alwayswilling support in XRD determinations.

This work has been supported by the National AtomicEnergy Commission and partly funded by the Generation IV

Public Fund Project N◦ 46813 of the Planning, Public Fund andServices Ministry of Argentina and the PICT 2014 No. 2170contract with the National Science and Technology Agency ofArgentina.

o l . 2

r

j m a t e r r e s t e c h n

e f e r e n c e s

[1] Luzginova NV, Rensman JW, Jong M, Pierick P, Bakker T,Nolles H. Overview of 10 years of irradiation projects onEurofer97 steel at High Flux Reactor in Petten. J Nucl Mater2014;455(1–3):21.

[2] Anderoglu O, Byun T-S, Toloczko M, Maloy SA. Mechanicalperformance of ferritic martensitic steels for high doseapplications in advanced nuclear reactors. Metall MaterTrans 2013;44A suppl. 1:S70.

[3] Yvon P, editor. Structural materials for generation IV nuclearreactors. 1st ed. Woodhead Publishing; 2016.

[4] Liu CH, Yan ZS, Dong ZZ, Liu YC, Ning BQ. Effects of two-steptempering treatment on the microstructural formation ofT91 ferritic steels. Solid State Phenom 2011;172:875–80.

[5] Haarmann K, Vaillant JC. The T91/P91 Book. 2nd ed.Düsseldorf: Vallourec & Mannesmann Tubes; 2002.

[6] Sasaki T, Kobayashi K, Yamura T, Kasuya T, Masuda T.Production and properties of seamless modified 9Cr-1Mosteel boiler tubes. Kawasaki Steel Techn Rep 1991;25(9):78–87.

[7] Béres L, Balogh A, Irmer W. Welding of martensitic creepresistant steels. Weld Res Suppl 2001, 191s-5s.

[8] Nitsche A, Mayr P. Round Robin test on measurement ofphase transformation temperatures in 9Cr1Mo steel. WeldWorld 2017;61(1):81–90.

[9] Tamura M, Kumagai T, Sakai K, Shinozuka K, Esaka H. A newapproach to improve creep resistance of high Cr martensiticsteel. J Nucl Mater 2011;417(1–3):29–32.

[10] Plesiutschnig E, Beal C, Sommitsch C, Paul S, Zeiler G. Ferriticphase transformation to improve creep properties ofmartensitic high Cr steels. Scr Mater 2016;122(9):98–101.

[11] Ohgami M, Naoi H, Kinbara S, Mimura H, Ikemoto T, Fujita T.Development of 9CrW tube, pipe and forging for ultrasupercritical power plant boilers. Nippon Steel Techn Rep1997;72(January):59–64.

[12] Tsuchida Y, Tokuno K, Hashimoto K. Development and BOFmanufacture of modified 9Cr-1Mo steel plates with excellentstrength and toughness. Nippon Steel Techn Rep1993;58(July):27–35.

[13] Hurtado-Norena C, Danón CA, Luppo MI, Bruzzoni P.Evolution of minor phases in a 9PctCr steel: effect oftempering temperature and relation with hydrogen trapping.Metall Mater Trans 2015;46A(9):3972–88.

[14] Shiue RK, Lan KC, Chen C. Toughness and austenite stabilityof modified 9Cr-1Mo welds after tempering. Mater Sci Eng2000;A287(1):10–6.

[15] Mayr P, Palmer TA, Elmer JW, Specht ED. Direct observation ofphase transformations in the simulated heat-affected zoneof a 9Cr martensitic steel. Int J Mater Res 2008;99(4):1–6.

[16] Ning B, Liu Y, Shi Q, Gao Z, Yang L. Effects of austenitestabilization on the onset of martensite transformation inT91 steel. J Mater Sci Technol 2008;24(2):202–6.

[17] Bhadeshia HKDHR. Honeycombe. Steels: microstructure andproperties. 4th ed. Oxford: Butterworth-Heinemann; 2017.

[18] Mito Y, Miki K, Azuma T, Ishiguro T, Tamura O, Murata Y,Morinaga M. Effects of Cr and W content in high Cr ferriticheat-resistant steels on long-term creep rupture strength. In:Advances in materials technology for fossil power plants.Proceedings of the 7th international conference. 2013. p.627–8.

[19] Lutterotti L. Total pattern fitting for the combinedsize-strain-stress-texture determination in thin filmdiffraction. Nucl Inst Methods Phys Res B

2010;268(3–4):334–40.

[20] Eng-Poh NG, Chateigner D, Bein T, Valtchev V, Mintova S.Capturing ultrasmall EMT zeolite from template-freesystems. Science 2012;335(6064):70–3.

0 1 9;8(1):923–934 933

[21] Brand RA. WinNormos for Igor; 2010.http://www.wissel-gmbh.de.

[22] Jones WB, Hills CR, Polonis DH. Microstructural evolution ofmodified 9Cr-1Mo steel. Metall Trans 1991;22A(5):1049–58.

[23] Sprael JM, Michaud H. Contribution to X-ray analysis ofcarbo-nitrided steel layers. J Appl Crystallogr2001;34(5):549–57.

[24] Blachowski A, Ruebenbauer K, Jura J, Bonarsky JT, Baudin T,Penelle R. Mössbauer study of deformation inducedmartensitic phase transformation in duplex steel.Nukleonika 2003;48 suppl. 1:S9–12.

[25] Skrzypek S, Kolawa E, Sawicki JA, Tyliszczac T. A study of theretained austenite phase transformation in low alloy steelusing conversion electron Mössbauer spectroscopy andX-ray diffraction. Mater Sci Eng 1984;66(2):145–9.

[26] Mercader RC, Desimoni J. Mössbauer studies of phasetransformations in iron alloys. Hyperfine Interact1997;110:101–9.

[27] Ramos CP, Sztrajman A, Bianchi R, Danón CA, Saragovi C.Mössbauer spectroscopy analysis on a tempered martensitic9%Cr steel. Hyperfine Interact 2010;195(1):257–63.

[28] Schaaf P, Wiesen S, Gonser U. Mössbauer study of ironcarbides: Cementite (Fe, M)3C (M = Cr Mn) with variousmanganese and chromium contents. Acta Metall Mater1992;40(2):373–9.

[29] Principi G, Frattini R, Magrini M. Mössbauer analysis ofcarbides extracted from heat-treated alloy-steels. GazzettaChim Ital 1983;113:281–4.

[30] Liu XW, Zhao S, Meng Y, Peng Q, Dearden AK, Huo CF, Yang Y,Li YW, Wen XD. Mössbauer spectroscopy of iron carbides:From prediction to experimental confirmation. Sci Rep2016;6:26184.

[31] Tenuta Azevedo AL, Galvão da Silva E. Mössbauer study ofretained austenite in a low C low alloy steel. Scr Metall1978;12(2):113–7.

[32] Pierce DT, Coughlin DR, Williamson DL, Kähkönen J, ClarkeAJ, Clarke KD, Speer JG, De Moor E:. Scr Mater 2016;121:5–9.

[33] Krauss G. Steels: processing, structure and performance. 2nded. ASM International; 2015.

[34] Wertheim GK, Jaccarino V, Wernick JH, Buchanan DNE.Range of the exchange interaction in iron alloys. Phys RevLett 1964;12(1):24–7.

[35] Ron M, Schechter H, Hirsh AA, Niedzwiedz S. On theMössbauer study of cementite. Phys Lett 1966;20(5):481–3.

[36] Shiue RK, Lan KC, Chen C. Toughness and austenite stabilityof modified 9Cr-1Mo welds after tempering. Mater Sci Eng A2000;287(1):10–6.

[37] Kipelova AY, Belyakov AN, Skorobogatykh VN, ShchenkovaIA, Kaibyshev RO. Tempering-induced structural changes insteel 10Kh9K3V1M1FBR and their effect on the mechanicalproperties. Metal Sci Heat Treat 2010;52(3–4):100–10.

[38] Saroja S, Parameswaran P, Vijayalakshmi M, RaghunathanVS. Prediction of microstructural states in Cr-Mo steels usingphase evolution diagram. Acta Metall Mater1995;43(8):2985–3000.

[39] Brachet JC. Alliages martensitiques 9Cr1Mo: influence del’addition de l’azote, du niobium et du vanadium sur lamicrostructure, les transformations de phase et lespropriétés mécaniques [PhD dissertation]. Orsay Centre:Paris Sud University; 1991.

[40] Brachet JC. Décomposition isotherme et anisotherme del’austénite en ferrite dans des aciers de base 9Cr1Mo:aspects cinétiques et microstructuraux. J Phys IV 1994;4(C3):83–91.

[41] Saroja S, Vijayalakshmi M, Raghunathan VS. Influence ofcooling rates on the transformation behavior of9Cr-1Mo-0.07C steels. J Mater Sci 1992;27(9):2389–96.

n o l

934 j m a t e r r e s t e c h

[42] Liu C, Zhang D, Liu Y, Wang Q, Yan Z. Investigation on the

precipitation behavior of M3C phase in T91 ferritic steels.Nucl Eng Design 2011;241(7):2411–5.

[43] Kirana R, Raju S, Mythili R, Saibaba S, Jayakumar T, KumarER. High-Temperature phase stability of 9Cr–W–Ta–V–C

. 2 0 1 9;8(1):923–934

based reduced activation ferritic–martensitic (RAFM) steels:

effect of W and Ta additions. Steel Res Int 2015;86(7):825–40.

[44] Yoshino M, Mishima Y, Toda Y, Kushima H, Sawada K, KimuraK. Phase equilibrium between austenite and MX carbonitridein a 9Cr–1Mo–V–Nb Steel. ISIJ Int 2005;45(1):107–15.