published flash process paper

Development of rapid heating and cooling(flash processing) process to produceadvanced high strength steel microstructures

T. Lolla*1, G. Cola2, B. Narayanan3, B. Alexandrov1 and S. S. Babu1

Flash processing of an AISI8620 steel sheet, which involves rapid heating and cooling with an

overall process duration of ,10 s, produced a steel microstructure with a high tensile strength

and good ductility similar to that of advanced high strength steels. Flash processed steel [ultimate

tensile strength (UTS): 1694 MPa, elongation: 7?1%], showed at least 7% higher UTS and 30%

greater elongation than published results on martensitic advanced high strength steel (UTS:

1585 MPa, elongation: 5?1%). The underlying microstructure was characterised with optical,

scanning electron, transmission electron microscopy as well as hardness mapping. A complex

distribution of bainitic and martensite microstructures with carbides was observed. A mechanism

for the above microstructure evolution is proposed.

Keywords: Flash processing, Advanced high strength steels, Heat treatment, Bainite, Phase transformation, Microstructure characterisation

IntroductionThere is a need for weight reduction of automotivestructures in order to achieve improved fuel efficiency,while not compromising the safety of passengers. Tomeet these two demands, a new class of steels known asadvanced high strength steels (AHSS) was developed bythe ultra light steel automotive body advanced vehicleconcept consortium.1,2 This family of steels shows goodformability while maintaining very high strength values[engineering yield strength (YS).300 MPa and ultimatetensile strength (UTS).700 MPa].3 These AHSS gradesinclude dual phase, transformation induced plasticity,complex phase and partially martensitic steels. All thesesteels achieve their mechanical properties by engineeringthe fractions of ferrite, bainite and martensite micro-structure as well as austenite phase. This is achievedeither through energy intensive thermomechanical pro-cessing steps and/or by expensive alloying additions (e.g.manganese).3

The need to focus on the area of advanced steeldevelopment is imminent. This is demonstrated byresearch directions of many researchers throughout theworld.4–7 In addition to the production of these steels,there exists a critical need to develop reliable and robustprocesses for forming8 and welding9 of these steels too.The above examples are not a comprehensive represen-tation of the published literature. However, it proves

that there is an impetus to develop processing techniquesto produce AHSS.

The present paper pertains to an innovative heattreatment procedure that shows a potential as analternative route for the current production of AHSSsteels.10 This process has been termed as ‘flash proces-sing’ for the incredibly short time (,10 s) for heatingand cooling of the steel sheets. Throughout the presentpaper this term or an acronym ‘FP’ will be used todenote steels that are processed using this method.Preliminary work has shown that FP of plain carbonsteels (.0?15C) may lead to high YS (.1200 MPa),tensile strength (.1500 MPa) and appreciable ductility(.7%). As a result, FP processed steels could beclassified under the category of AHSS.10

Cursory evaluation of the above claim leads toskepticism, since it contradicts the well establishedtheories of phase transformation and microstructureevolution in steels.11 Heating of steels to the homo-genous austenite phase field and rapid quenching willlead to the formation of martensite. During tensiletesting of this martensitic microstructure, no appreciableplastic strain is expected. Therefore, the objective of thepresent paper is to relate the observed thermal cycle,microstructure and the observed properties by develop-ing a mechanistic understanding of the microstructureevolution.

Experimental

Steel composition and initial conditionIn earlier work, Cola has investigated many steelsincluding AISI8620 steels by FP.10 In the current paper,to limit the effect of the steel compositions on theinterpretations of the microstructure and propertycorrelations, only AISI8620 grade steel sheets were

1Industrial Systems and Welding Engineering, Ohio State University,Columbus, Ohio 43221, USA2SFP Works LLC, Washington, MI, USA3Materials Science and Engineering, Ohio State University, Columbus,Ohio 43221, USA

*Corresponding author, email [email protected]

� 2009 Institute of Materials, Minerals and MiningPublished by Maney on behalf of the InstituteReceived 1 February 2009; accepted 9 March 2009DOI 10.1179/174328409X433813 Materials Science and Technology 2009 VOL 000 NO 000 1

investigated. The measured composition of the steel isFe–021C–0?27Si–0?002S–0?009P–0?73Mn–0?48Cr–0?48Ni–0?007Co–0?156Mo–0?178Cu (wt-%). The original steelwas under the annealed condition and the microstruc-ture constitution was ferrite and uniform dispersion ofcarbides.

Flash processing set-up and heat treatmentsA schematic of the process set-up is shown in Fig. 1. Theassembly consists of a pair of rollers that transfers thesteel sheets through a heating and cooling stage. At acontrolled distance from the top pair of rollers, theheating stage is placed. The heating stage is based oneither oxy propane flame or electrical induction heating.The flame heating consists of 17 flame nozzles spacedevenly to spread the heat over the steel sheets. The aboveset-up has also been equipped with infrared pyrometers,instrumentation to drive the feed rollers, mechanicalfixtures for positioning heating and cooling units, andheating intensity controls. Spot temperature check usingan Ircon infrared pyrometer showed a small variation oftemperature of approximately ¡10uC from the middleto the edges of the sample. Immediately below theheating port, a cooling trough is placed. To avoidinefficient heating due to the upward flow of steam, agraphite separator film is used to separate the heatingand cooling ports. In this experiment, the steel sheetswere fed through the heating and cooling system at arate of 28 cm min21 while maintaining the peaktemperature at 1100uC. In order to compare the flashprocessed samples with a reference sample, anotherAISI8620 unprocessed steel sheet was subjected to aquenched and tempered (QT) treatment (1000uC/2 minRquenchedRtempered at 200uC/4 min).

Thermal cycle measurements and analysesCola10 calculated thermal cycles in different regionsusing infrared pyrometers and imposed feedrates of steelsheets. However, the detailed temporal variations oftemperature were not measured using contact thermo-couples. Such measurements are necessary to develop amechanistic understanding of microstructure evolutionduring FP. However, the oxy propane flame heatingmethod does not lend itself for the direct measure-ment of temperature due to possible deterioration ofthermocouples by the high intensity flame. Therefore,

the induction heating method was used during tempera-ture measurement experiments. In order to impose thesimilar heating conditions as of that of the flameheating, feedrates, temperature gradients and the peaktemperature were controlled to be identical. Type K(Chromel–Alumel) thermocouples were attached to themiddle of the steel sheets. The temperature of the steelwas measured, as the sheet traverses through the heatingand cooling ports. These measurements are made with ahigh speed data acquisition system capable of recordingtemperature at a sampling rate of up to 5 KHz. Themeasured thermal cycles were analysed to determine theinstantaneous heating and cooling rates throughoutthe FP. In addition, ferrite (a) to austenite (c) trans-formation was evaluated by analysing thermal cycle datausing single sensor differential thermal analysis(SSDTA) technique.12 The sensitivity and accuracy ofSSDTA technique, in measuring a to c transformationduring high heating rates (y500 K s21), has beendocumented earlier13 by direct comparison with dilato-metric techniques.

Mechanical property measurementsIn order to evaluate the spatial variations of steelmicrostructures, two-dimensional hardness distributionswere measured in the cross-section of samples. For eachsample, .2000 indents were made on QT and FPsamples using an AMH43 automatic hardness testingsystem with a load of 300 g. The measured hardnessdistributions were analysed in a map format and usingfrequency distribution curves to provide quantitativemeasure of mechanical heterogeneity of the samples.

Tensile samples were extracted from the edge regionsand from the middle regions. Tensile tests wereperformed along the longitudinal direction of the coils.The gauge length of the test samples was 25 mm. Thewidth of the samples at the gauge length was 6?3 mm.The tests were performed at room temperature with acrosshead displacement rate of 1?27 mm min21. All thetensile testing were performed using testing machinesavailable at Edison Welding Institute. Similar tests wererepeated in QT samples also.

Microstructure characterisationStandard optical microscopy was used to characterisemicrostructure from the FP and QT samples. In the nextstep, the samples were characterised using a Quanta200scanning electron microscope (SEM), equipped withsecondary electron, backscattered and X-ray energydispersive spectroscopy (EDS) detectors. The micro-scope was operated at an accelerating voltage of 25 kV.

Transmission electron microscopy (TEM) sampleswere prepared by electrolytic thinning of 3 mm discs of90 mm thick samples. A chemical solution of 33% Nitricacid, 67% Methanol was used as the electrolyte. Thepolishing was carried out at 25 V at a current of 10–15 mA and the temperature of the electrolyte wasmaintained at 225uC. The samples were imaged usingtwo types of transmission electron microscopes. Generalmicrostructure and electron diffraction analyses wereperformed in a Phillips CM12 transmission electronmicroscope operating at 120 kV. In addition, high angleannular dark field (HAADF) and bright field imagingwas performed using STEM technique in a Tecnai F-20FEG/TEM operating at 200 kV.

1 Schematic illustration of experimental set-up of FP

technique

Lolla et al. Development of rapid heating and cooling – Flash processing

2 Materials Science and Technology 2009 VOL 000 NO 000

Computational modellingThermoCalc14 software with TCFE5 database was usedto estimate the thermodynamic quantities and equili-brium phase transformation temperatures. In addition,the time temperature transformation (TTT) diagramsfor the initiation of reconstructive and displacive trans-formations were predicted using the methodologydeveloped by Bhadeshia.15,16 The TTT data were thenconverted into a continuous cooling transformation(CCT) diagram using the methodology developed byBabu and Bhadeshia.17 The source code for these calcu-lations can be downloaded from an internet location18

and example calculations can be made using an onlinecalculator.19

Results and discussion

Calculated phase transformation characteristicsIn order to analyse the results from the FP experimentswith reference to expected transformation behaviours,the transformation temperatures were calculated usingthe methodologies described before14–19 for theAISI8620 steel composition. The equilibrium A1 andA3 temperatures of this steel were calculated to be 702and 806uC respectively. In addition, the calculatedequilibrium volume fractions of cementite (M3C) andalloy carbides (M23C6 and MC) that can form in thesesteels as a function of temperature are shown in Fig. 2a.This result shows that as the steel heats above 810uC, itshould transform to 100% austenite.

The bainitic and martensitic start temperatures15,16

were calculated to be 553 and 426uC respectively.Calculated TTT and CCT diagrams for the initiationof paraequilibrium (transformation controlled by car-bon partitioning and the ratio of substitutional to ironatoms are configurationally frozen in place on either sideof austenite and ferrite transformation interface) recon-structive mode of austenite to allotriomorphic ferritetransformation and displacive mode of austenite toWidmanstatten and bainitic transformations are shownin Fig. 2b. Since the FP involves rapid cooling process,the assumption of paraequilibrium is indeed valid. Basedon the above results, one can expect complete transfor-mation of austenite to 100% martensite in AISI8620steels at cooling rates faster than 70 K s21. Thesepredictive methodologies have been developed using awide range of steels and their TTT data presented in theliterature.16,20 Therefore, the predictions are expected tobe fairly accurate for low alloy steels including AISI8620grades. For example, these calculations have accuratelypredicted the transformation start during continuouscooling of low alloy steels.21

Thermal cycles during FPA typical heating and cooling thermal cycle during FP isshown in Fig. 3. The plot also shows the heating andcooling rates as a function of time. These thermalprofiles can be divided into four regions based on thetemperature range. In region I, the steel temperatureincreases gradually due to heat conduction from the hotstage. In region II, the steel temperature increasesrapidly due to its proximity to heat source. In thisregion, maximum heating rate of 410 K s21 wasobserved at 780uC. This temperature is between thecalculated A1 and A3 transformation temperatures for

AISI8620 steel. Above 780uC, the heating rate startsreducing indicating that there is a dynamic equilibriumbetween the heat flux, the heat conduction, andendothermic/exothermic effects due to phase transfor-mations.12 In region III, the heating rate reaches zerowhen the sample reaches a peak temperature of 1100uC.As per the equilibrium thermodynamic calculations, atthis temperature, the sample should be 100% austenite.After reaching the peak temperature, the sample startscooling down gradually. The total dwell time for thisslow cooling near 1100uC is measured to be 2 s. Inregion IV, within 1?2 s of reaching peak temperature,the sample reaches the water bath and starts coolingrapidly. The maximum cooling rate of 3150 K s21 isachieved when the sample reaches a temperature of393uC. This temperature is much below the calculated Bs

(553uC) and Ms (426uC) temperatures.15 As per themeasured thermal cycles and the calculated CCTdiagram (see Fig. 2b), the austenite phase from hightemperature should transform to 100% martensiteduring FP. Figure 4 shows the complete thermal profileof two separate FP treatment runs recorded fromthermocouples attached at the centre of the sheet. Thisshows the thermal cycle repeatability that can beachieved in FP.

Alexandrov and Lippold12 have developed theSSDTA software to detect transformation temperaturesfrom measured thermal cycles, in a well described(Newtonian, Gaussian or Rosenthal type) heat transfer

2 a calculated variation of carbide volume fraction as

function of temperature obtained using ThermoCalc14

and TCFE5 database and b TTT and CCT diagrams

showing reconstructive transformation and displacive

transformation regions


Materials Science and Technology 2009 VOL 000 NO 000 3

condition. The transformation temperatures aredetected by looking for a departure of measuredtemperature from a reference heating (or cooling) curvedue to endothermic (or exothermic) effects during phasechanges. Figure 5 shows the results of the SSDTAanalyses of thermal profiles , for the heating regime only.In these analyses, the reference curves were described by

a polynomial function that describes a well controlledheat transfer conditions.

Both the results show a small endothermic phenom-enon around the calculated14 Curie temperature(y750uC) of bcc ferrite and another large endothermicpeak ,900uC. These large endothermic peaks areinterpreted as Ac1 (926 and 930uC) and Ac3 (1046 and1052uC) temperatures by SSDTA software. The repro-ducibility of these measurements is within the 0?4%accuracy of the type K thermocouple. The accuracy ofthese analyses is also affected by the electromagneticnoise generated by the induction heating. It is note-worthy that SSDTA technique estimates a higher andnon-equilibrium Ac1 and Ac3 temperatures, in compar-ison to equilibrium A1 (702uC) and A3 (806uC) tem-peratures. Possible mechanisms for these are discussedlater.

Hardness and tensile propertiesMeasured engineering tensile properties are presented inTable 1. Table 1 also shows the tensile properties of theFP samples tested by Cola10 and the current work. Thedata show that the flash processed AISI8620 steels dohave yield and tensile strengths greater than 1300 and1500 MPa respectively. In addition, the elongation andreduction in area indicate appreciable ductility in these

3 a heating/cooling rate and b temperature variation versus time in typical FP

4 Thermal profile of two samples showing repeatable

thermal cycles



samples. The ductility of 8?8–9?9% is indeed greater thanthe levels reported for martensitic AHS steels.22 TheUTS of the FP tensile specimen is comparable to that ofthe QT sample; however, the FP samples, consistently,show higher elongation values.

In order to evaluate the formability of FP steels, theratio of YS and UTS and true stress true strain

characteristics were calculated (see Fig. 6). The strainhardening exponent n and pre-exponent strength coeffi-cient K for uniform plastic strain were estimated byfitting the following equation to the plastic regime of thetrue stress–strain curve.

s~Ken (1)

The calculated K and n values (see Table 2) are similarvalues to the published grades of AHSS steels for dualphase and complex phase steels.22 The measured UTSand YS from the current work are overlaid in a map ofcurrent AHSS steels. This comparison confirms that theflash processed steels (see Fig. 7) are indeed a part ofAHSS steel family. It is interesting to note thatboth UTS and the elongation are higher than thereported values for AHSS based on a martensiticmicrostructure.

Table 1 Measured engineering tensile properties of FP (from previous and current work) and QT samples (current work).The tests were performed according to ASTM E8 specifications*

Sample IDUltimate tensilestrength, MPa

0.2% yieldstrength, MPa

Reduction ofarea, % Elongation, % Ratio of YS to UTS Source

Ref-FP1 1685.77 1241.75 … 6.3 0.737 Ref. 10Ref-FP2 1694.04 1314.83 … 7.3 0.776 Ref. 10Ref-FP3 1676.12 1275.53 … 7.1 0.761 Ref. 10Ref-FP4 1669.91 1292.77 … 6.9 0.774 Ref. 10FP no. 1 1664.1 1442.8 39.4 8.8 0.867 Present paperFP no. 2 1619.3 1386.9 38.0 9.9 0.856 Present paperFP no. 3 1520.7 1300.0 38.3 9.9 0.855 Present paperQT no. 1 1607.6 1333.8 37.4 6.8 0.829 Present paperQT no. 2 1657.9 1464.1 49.8 10.0 0.883 Present paperQT no. 3 1642.8 1487.6 14.8 4.3 0.905 Present paper

*Sample dimensions used in Ref. 10: width 13 mm; thickness 1?5 mm. Sample dimensions in the present work: width 6?4 mm;thickness 1?23 mm.

6 Comparison of true stress versus true strain graphs of

QT and flash processed specimen

5 Single sensor differential thermal analysis of heating

region of a sample 1 and b sample 2 showing two

endothermic thermal events

Table 2 Calculated pre-exponent K and strain hardening exponent n derived by fitting equation (1) to plastic regime oftrue stress–strain curve

FP no. 1 FP no. 2 FP no. 3 QT no. 1 QT no. 2 QT no. 3

K, MPa 569.6 622.1 613.7 1305.6 536.3 543.9n 0.1542 0.1568 0.1348 0.1950 0.1342 0.1471



Microstructure of heat treated samplesIn this section, the underlying microstructure in theoriginal steel sheets before processing and after the FPand QT heat treatments are presented.

Initial microstructure of annealed AISI8620 steel sheets

Optical microscopy of original steel sheets showed thepresence of ferrite grains with carbides (Fig 8a). Most ofthe carbides were ,1 mm and some of them were.2 mm, suggesting a bimodal nature of carbide dis-tribution. These carbides also appear to be aligned inbetween pancaked ferrite structure suggesting that thesecould have precipitated during legacy thermomechanicalprocessing of these sheets. Scanning electron microscopy(see Fig. 8b) with backscattered imaging showed thatmost of these carbides were associated with ferrite grainboundaries. Energy dispersive X-ray spectroscopy ana-lyses (see Fig. 8c) of these carbides show a higher ratioof Cr/Fe peak intensities compared to that of ferrite

7 Overview of published tensile strength and ductility

that can be achieved for a wide range of advanced

high strength steels22 and measured tensile strength

and ductility of flash processed steels

8 a optical image microstructure of AISI8620 unprocessed sample showing extensive carbide distribution, especially

along grain boundaries, in ferrite matrix, b SEM backscattered image showing presence of carbides along grain bound-

aries in microstructure of AISI8620 unprocessed sample and c comparison of EDS signals obtained from matrix and

carbide particle observed in AISI8620 unprocessed sample (signals from particle show higher Cr/Fe ratio indicating

that particle could be chromium carbide)



matrix. Based on the above analyses and thermody-namic calculations, the carbides are assumed to be Crrich cementite (M3C).

Microstructure of QT AISI8620 steel sheets

Since the QT and FP microstructures showed similarproperties, the microstructure of the QT samples wascharacterised with optical microscopy and scanningelectron microscopy. Owing to the fine nature of themicrostructure, optical microscopy (see Fig. 9a) couldnot discern the martensitic and bainitic microstructures.The scanning electron microscopy (see Fig. 9b) withbackscattered electron imaging revealed the martensiticlath structure with brightly imaging carbides. By tracingthe lath packet boundary, typical prior austenite grainsize was estimated to be around 20–30 mm. Owing to theshort duration of the austenitising time (2 min at1000uC), extensive austenite grain growth was notobserved. A quantitative distribution of the austenitegrain size in the overall sample was not performed dueto the difficulty of identifying all the prior austenitegrain boundaries by this technique. A backscatteredelectron diffraction technique with crystallographicorientation analyses is required23 and is the focus ofthe ongoing work.

The volume fraction and size of carbides in the QTsamples is smaller than that of the untreated samples(see Fig. 8b). It is possible that the microstructure maycontain nanometer sized carbides that formed during thetempering (220uC for 4 min) stage. Usually the sizes ofthese carbides that form during early stages of temperingare below the resolution of scanning electron micro-scopy.24–27 However, it is fair to conclude that initialcarbides that are present in the untreated samples havedissolved during austenitising at 1000uC for 2 min.

Microstructure of flash processed AISI8620 steel sheets

Since the focus of this paper is related to FP, extensivecharacterisation was performed on FP samples. Opticalmicroscopy (see Fig. 10a) of the FP samples againshowed features that were difficult to distinguish.However, scanning electron microscopy with backscat-tered electron imaging revealed interesting features (seeFig. 10b). Fine lath structures with brightly imaging

particles could be observed. By tracing the lath packetboundary (see Fig. 10b), typical prior austenite grainsize was estimated. This qualitatively shows that theprior austenite grain size of FP samples ((10 mm) ismuch finer than that of the (20–30 mm) QT samples.This could be attributed to the rapid heating and coolingabove the Ac3 temperature. In the next step, thecomposition of carbides was evaluated with EDSanalyses. The measured spectrum and the calculatedcomposition of analysed regions are presented inFig. 10c. Surprisingly, similar to the original steelsamples before processing, these carbides were foundto be rich in Chromium. In addition coarser and finercarbides were also observed which is similar to untreatedsamples. Since the FP process involves rapid heating(.400 K s21) and cooling (.3000 K s21) rates, thesecarbides are interpreted as the undissolved carbides thatwere present in the original base material.

With cursory observation of optical and scanningelectron micrographs, one may conclude that the FPsamples are predominantly martensitic. However, tosubstantiate this conclusion, detailed transmission elec-tron microscopy was performed. A low magnificationTEM image (see Fig. 11a) shows a prior austenite/austenite (c/c) grain boundary with presence of fine scalemartensitic laths with high dislocation density andcoarser bainitic ferrite with slightly reduced dislocationdensity. The identification of bainitic ferrite was madebased on the size of these plates and the presence ofsheave-like morphology with subunits as shown byBhadeshia.26

In another region of the same TEM sample (seeFig. 11b), three bainitic (marked as 1, 2 and 3 in theimages) sheaves were observed. It is interesting to notethat each and every individual sheave is made up ofmany ferrite subunits with similar orientation in space.Electron diffraction analyses failed to identify anyretained austenite film between these subunits.27

Crystallographic orientation relationships between thesebainitic sheaves were calculated and the summary of theresults is presented in inset diffraction patterns inFig. 11b. The sheaves nos. 1 and 2 are indexed to bein the [111] bcc zone axis with a relative misorientation

9 a optical image QT AISI8620 sample showing fine microstructure consisting of martensitic and bainitic structures and

b SEM backscattered image of AISI8620 QT sample showing martensitic lath structures with prior austenite grain

boundary



of only 10 degrees. Assuming that these bainitic sheaveshave a Kurdjumov–Sachs/Nishiyama–Wassermann(KS/NW) orientation relationship,28 with the parentaustenite (fcc) phase, one can conclude that the bainiticsheaves nos. 1 and 2 could have formed a singleaustenite cA grain, within the regions that are separatedby small angle boundaries. In contrast, the sheave no. 3was indexed to be close to the [011] zone axis. Using theaxis angle pair analyses,29 the relative orientationbetween no. 3 and no. 1 or between no. 3 and no. 2was found to be of high angle type. Using the sameassumption of KS/NW orientation relationship ofbainitic sheaves with austenite and symmetry relations,28

one can also conclude that the no. 3 sheave formedin adjacent austenite grain cB and the orientation

relationship between cA and cB should be of high angletype. Copious presence of such austenite grains withlarge angle boundaries will promote bainitic and/ormartensitic plates with wide varying orientations andmay improve the properties.30 It is stressed that theabove results are typical and one cannot concludesimilar crystallographic conditions throughout the FPsample. To evaluate this throughout the sample, we needto do extensive backscattered electron diffractionanalyses and correlate the orientation distributionfunctions to the properties, which is the focus of theongoing work.

In another FP sample, HAADF STEM image showedextensive carbide distributions. Some of them wereassociated with the ferrite subunit grain boundaries (see

10 a optical image of flash processed AISI8620 sample showing very fine, indiscernible microstructure, b SEM backscat-

tered image of AISI8620 flash processed sample showing fine lath structures with brightly imaging particles and typi-

cal prior austenite grain boundary and c comparison of EDS signals obtained from matrix and carbide particle

observed in AISI8620 FP sample (signals from particle show higher Cr/Fe ratio indicating that particle could be chro-

mium carbide)



Fig. 12a). Size of these carbides was ,100 nm.Elongated carbides between the lath boundaries areprobably cementite. The precipitates are predominantlydarker than the matrix suggesting that their averageatomic number is lower than the iron matrix. However,the precipitate thickness effect is convoluted with theatomic number contrast and further work is required toestablish the chemical composition of these precipitates.Some of the large carbides (marked by arrow) show nocorrelation to the ferrite boundaries (Fig. 12b). TheEDS analyses show that these carbides were rich insilicon and chromium (Fig 12c). The detailed analyses ofelemental distribution within these carbides can only beattained by atom probe field ion microscopy.24,25 Owingto the rapid rate of heating and cooling, the diffusioncontrolled dissolution and/or growth is expected to besluggish. Therefore, these large carbides are interpretedas the undissolved carbides from the original steel sheetsthat existed before processing. From the above TEManalyses one can conclude that the FP microstructurescontain bainitic ferrite, martensitic laths as well asuniform distribution of large and small carbides.

Evaluation of mechanical heterogeneity in heattreated samplesAlthough, the FP samples showed this uniquemixture of bainite, martensite and carbide microstruc-ture, it is necessary to evaluate the heterogeneity ofthese microstructures within the overall sample. Since

it is practically impossible to evaluate the microstruc-tural heterogeneity through the series of optical,SEM and TEM techniques throughout the sample,hardness mapping was adopted as a measure of theheterogeneity.17 With this method, series of hardnessindents were made on the samples and the mechanicalheterogeneity31 is estimated by analysing the hardness

11 Images (TEM) showing a c/c grain boundary with par-

allel laths growing out of it and b three bainitic

sheaves (1, 2, 3) in FP sample

12 a HAADF STEM image showing extensive carbides

with different size range, b HAADF STEM image show-

ing fine carbides in between bainitic subunits and c

measured energy dispersive X-ray spectrum obtained

from matrix and carbides showing higher Cr/Fe ratio

similar to that of carbides from unprocessed samples



distribution. The average size of the indents in the FPsample was 42 mm which is bigger than the prioraustenite grain size and also the bainitic and martensiticpacket sizes. It is highly possible that the differentmicrostructure distribution may provide similar hard-ness values. Therefore, the above measure has to be usedonly as a way to evaluate the possible scatter in finalmechanical properties.32 The mechanical heterogeneityresults from QT and FP samples are discussed below(see Fig. 13–16).

The hardness maps from the cross-section of QT(Fig. 13) sample showed a large soft region (blue colour)in the edge 1 region. In addition, the hardness variationsare larger in the QT sample. This is attributed to theinefficient quenching during manual insertion of the

austenitised sample into the water. This is also reflectedin the results of the tensile tests, where QT samplesshowed low YS (see Fig. 6a and Table 1) and high strainhardening exponent. The hardness distribution curvefrom the QT sample is shown in Fig. 14. The above datawere interpreted by fitting multiple peaks to thedistribution. The fitted peaks show that the QT samplecan be divided into two regions, i.e. large (vQT1588%)region 1 with an average hardness of 475¡47 HV withlarge scatter and a small (vQT2512%) region 2 with anaverage hardness of 483¡5 HV with small scatter.

In contrast, the mechanical heterogeneity of the flashprocessed samples was less than QT samples (seeFigs. 15 and 16). The image also shows that a slightly

13 Hardness contour map of cross-section of QT sample cut into three pieces showing distribution of hardness as mea-

sured from over 2000 micro-indent points over cross-section area

14 Histogram plotted from hardness values observed in

cross-section of QT sample, shows large scatter peak

(vQT1588%) and small scatter peak (vQT2512%)

15 Hardness contour map of cross-section of FP sample cut into three pieces showing distribution of hardness as mea-

sured from over 2000 micro-indent points over cross-section area

16 Histogram plotted from hardness values observed in

cross-section of FP sample shows three distinct

regions of different hardness range; 498¡7, 512¡11

and 534¡9 HV



harder region (yellow colour) is in the middle of the FPsteel sheet compared to the edge of the steel sheet. Thereader may recall the description of the FP process set-up (see Fig. 1) in which the middle of the steel sheet getsexposed to the highest heat density, as a result of highestheating rate and high peak temperature. Similar to QTsample analyses, the calculated hardness distributionwas analysed by fitting multiple peaks to the distribu-tion. The fits show that the QT sample can be dividedinto three distinct regions, i.e. region 1 (vFP1 14%) withan average hardness of 498¡7 HV, region 2 (vFP25

74%) with an average hardness of 512¡11 HV andregion 3 (vFP3512%) with an average hardness of534¡9 HV. This presence of distinct and homogeneousmixing of soft and hard regions in the FP sample istentatively attributed to the improved mechanicalproperties of the flash processed steels.

Mechanism of microstructure evolution duringFPBased on the above results, one can conclude that the FPleads to a complex microstructure that contains mar-tensite, bainite and carbides. In addition the abovecomplex microstructure also leads to a preferredmechanical heterogeneity that is better than the refer-ence QT sample. In addition, the FP steels can beclassified under the category of AHSS based on thestrength observed. Furthermore, the flash processedsteels show at least 7% higher UTS and 30% higherelongation than published martensitic advanced highstrength steels22 (see Fig. 7). However, these results haveto be discussed in context of the mechanism for themicrostructure evolution during FP.

On heating carbide dissolution

As per equilibrium thermodynamic calculations, the8620 steels should be 100% austenite above 800uC. Fromthe measured thermal cycle during the flash process, thepeak temperature was 1100uC. Therefore, the sampleshould have completely transformed into austenite atthis temperature and all the carbides should havedissolved. However, electron microscopy showed thatthe carbides have not dissolved completely. Thisdiscrepancy is attributed to a very short dwell time inthe austenitising region which limits the dissolution ofcarbides. This hypothesis is supported by the observa-tion of fewer carbides in the QT samples, which wereaustenitised for 2 min at 1000uC before quenching. It iswell known that, extended time in the austenite phasefield will promote the complete dissolution of carbides33

by enhancing the substitutional element diffusion.

On heating ferrite to austenite transformation

Another possibility of such unique microstructureevolution can be attributed to an increase in Ac1 andAc3 temperatures due to rapid heating rates(.400 K s21) during FP. This hypothesis is in agree-ment with neural network analysis34 and Gaussianprocess modelling.35 The level of superheating requiredto nucleate and grow the austenite phase depends on thealloy composition, initial microstructure and the heatingrate. This increase in Ac1 temperature also beenrationalised with the need to nucleate austenite phasefrom a ferritezcarbide microstructure.36 On the otherhand, if the initial microstructure contains retainedaustenite, there is no need for nucleation barrier.37 Since

the initial microstructure of AISI8620 steel sheetscontained only ferrite and carbide, there is a need tonucleate austenite. Therefore, the rapid heating rateshould lead to an increase in Ac1 temperature aboveequilibrium value. Similar arguments can be made for anincrease in Ac3 temperature due to reduced dwell timefor diffusion controlled growth of austenite into a ferritematrix. It noteworthy that the published research34,35

did not consider very high (.100 K s21) heating rates.Therefore, we need to consider the validity of thecontinued increase in Ac1 and Ac3 temperatures farabove the equilibrium Ac3 temperature.

Recently, Elmer et al.38 have developed an overalltransformation kinetic model for ferrite to austenitetransformation formation in 1005 steel based onsynchrotron diffraction measurements during a heatingcycle of a weld. The model was calibrated using the datafrom low heating rates ,100 K s21. However, themodel calculations were extrapolated to higher heatingrates. The calculations suggest a superheating of 100uCtemperature or more above the equilibrium Ac3 tem-perature for the completion of austenite formation atheating rates .300 K s21. This result supports thenotion of non-equilibrium Ac1 and Ac3 temperaturesduring FP (heating rate .410 K s21). Moreover, theSSDTA analyses of the heating curve (Fig. 5a and b)indicate that the Ac1 temperature is definitely above900uC; 925?7uC (Fig. 5a) and 929?6uC (Fig. 5b).Similarly the Ac3 temperature is calculated to be above1000uC; 1045uC (Fig. 5a) and 1052uC (Fig. 5b), which is200uC higher than the equilibrium Ac3 temperature(806uC). Such a large increase in Ac1 and Ac3

temperatures will also affect the carbide dissolution,since the presence of austenite is necessary for theinitiation of dissolution. In addition, the time taken foraustenite grain growth will also be reduced. This issupported by the small prior austenite grain size(,20 mm) measured in the flash processed samples.

As a result of small dwell time above Ac3 temperature,the carbon diffusion and redistribution within theaustenite grain is expected to be incomplete.39 Hencethe steel that is being cooled from peak temperature willhave inhomogeneous carbon distribution within theaustenite matrix, which could trigger complex decom-position of austenite into different ferrite morphologiesand martensite.

On cooling austenite decomposition

Based on the calculated CCT diagram and cooling rate(.3000 K s21) during FP, the steel should havetransformed to 100% martensite. To evaluate the effectof carbon concentration gradients within the austeniteon the final microstructure, the CCT diagrams forAISI8620 steels with different carbon concentrationsranging from 0?03 to 0?21 wt-%, with the samesubstitutional alloying element concentrations, werecalculated using the methodology described earlier.16,17

These series of CCT diagrams are also overlaid with themeasured cooling curve (see Fig. 17). The comparisonshows that even with very low carbon content in theaustenite the microstructure should be 100% martensite.However, the microstructure contains bainitic ferrite,martensite and carbides.

This discrepancy is attributed to the inadequacies oftransformation kinetic modelling methodologies. Thesemodels were developed using the TTT data generated by



traditional heat treatment conditions16,20 involving slowheating and isothermal hold at austenitising temperatureexceeding 5 min and homogenous austenite. The exten-sion of these models to austenite with inhomogeneouscarbon concentration may not be reliable. Thesesituations are similar to the rapid heating and coolingconditions experienced by heat affected zone of welds.

Based on the above results and discussions, ahypothetical microstructure evolution during FP isproposed (see Fig. 18). At room temperature, the initialmicrostructure contains ferrite and carbides (Fig. 18a).On rapid heating, the austenite nucleation is expected tooccur at the carbide–ferrite boundaries36 and theaustenite is expected to grow at the expense of ferrite

and carbide dissolution. Owing to rapid heating rates,the temperature at which the austenite nucleation andgrowth occurs may be elevated to higher temperatures(.900uC) (Fig. 18b). But before the complete dissolu-tion of cementite, the cooling cycle stars. At thisjuncture, the austenite is expected to have inhomoge-neous distribution of carbon content (Fig. 18c). Withthe on-set of rapid cooling, the carbon enriched regionsmay transform to martensite and carbon depletedregions may transform to bainite or low carbonmartensite. This leads to a final microstructure withcomplex distribution of bainite, martensite and carbides(Fig. 18d). This hypothetical mechanism suggests thatthe initial microstructure will have a strong effect on thefinal microstructure and properties achieved by FP.Ongoing experiments are focusing on evaluating theseeffects and methods to track the phase transformationsduring heating and cooling for a wide range of steels toprove or disprove this hypothesis.

ConclusionsA rapid heating (.400 K s21) and cooling(.3000 K s21) flash process has been developed thatproduces a microstructure with good combination ofyield (1280 MPa), UTS (1600 MPa) and appreciableductility of up to 9?9%. The above strength and ductilitylevels are significantly better than martensitic basedAHSS (1400–1500 MPa UTS). Based on the above data,the flash processed steels can be classified as part of theadvanced high strength steel family. As the overallprocessing time (,10 s) is very short, this process couldbe an alternative route for producing AHSS sheets.

The underlying microstructures in flash processedsamples were characterised with optical and analyticalelectron microscopy. The characterisation resultsshowed that the steel contains bainite, martensite and

17 Calculated CCT curves for AISI8620 steel austenite

with same substitutional concentration (Fe–021C–0?27Si–

0?002S–0?009P–0?73Mn–0?48Cr–0?48Ni–0?007Co–0?156Mo–

0?178Cu, wt-%) but with different carbon concentra-

tions (diagram is overlaid with measured cooling

curve during FP)

a unprocessed ferrite with carbides; b growth of austenite and start of carbide dissolution; c inhomogeneous carbondistribution in austenite with reduction in carbides; d martensite with bainitic plates formed from inhomogeneous carbondistribution in austenite

18 Schematic illustration of microstructure evolution



carbides. The hardness map analyses showed that theflash processed samples contain a minor fraction ofsoft microstructure interspersed within a harder micro-structure. This unique distribution of soft and hardmicrostructure is correlated with the improved YS, UTSand ductility.

Traditional phase transformation models indicatedthat, for the cooling rates measured in the flash process,the microstructure should be 100% martensite afterprocessing. This discrepancy is addressed with atheoritical mechanism that involves an increase in Ac1

and Ac3 temperatures due to rapid heating rate,incomplete dissolution of carbides and decompositionof austenite with non-uniform carbon concentration.

References1. ‘Body structure materials’, Technical transfer dispatch no. 6,

ULSAB-AVC, May 2001.

2. Porsche Engineering Services, Inc., ULSAB-AVC engineering

report, October 2001

3. R. Kuziak: ‘Advanced high strength steels for automotive industry:

a review’, Arch. Civil Mech. Eng., 2008, 8, (2), 102–177.

4. D. V. Edmonds, K. He, F. C. Rizzo, B. C. de Cooman, D. K.

Matlock and J. G. Speer: Mater. Sci. Eng., 2006, 438–440, 25–34.

5. F. G. Caballero, H. K. D. H. Bhadeshia, K. J. A. Mawella, D. G.

Jones and P. Brown: Mater. Sci. Technol., 2002, 18, 279–284.

6. T. Inoue, S. Torizuka and K. Nagai: Mater. Sci. Technol., 2001, 17,

1329–1338.

7. T. Ohmura and K. Tsuzaki: ‘A new aspect of the strengthening

factors f Fe-C martensite through characterization of nano-

indentation-induced deformation behavior’, Proc. 1st Int. Symp.

on ‘Steel science’, (ed. T. Furahara and K. Tsukzai), 35–46; 2007,

Tokyo, The Iron and Steel Institute of Japan.

8. W. Gan, S. S. Babu, N. Kapustka and R. H. Wagoner: Metall.

Mater. Trans. A, 2006, 37A, 3221–3231.

9. N. Kapustka, C. Conrardy, S. Babu and C. Albright: Weld. J. Res.

Suppl, 2008, 87, 135s–148s.

10. G. M. Cola, Jr: ‘Properties of bainite nucleated by water quenching

in 80 ms’, Proc. 1st Int. Symp. on ‘Steel science’, (ed. T. Furahara

and K. Tsukzai), 187–190; 2007, Tokyo, The Iron and Steel

Institute of Japan.

11. H. K. D. H. Bhadeshia and R. Honeycombe: ‘Steels:

Microstructure and Properties’, 3rd edn; 2006, Oxford,

Butterworth-Heinemann.

12. B. T. Alexandrov and J. C. Lippold: ‘Method and device for

investigation of phase transformations in metals and alloys’, US

patent no. 7473028, 6 January 2009.

13. B. T. Alexandrov and J. C. Lippold: Weld. World, 2007, 51, 48–59.

14. B. Sundman, B. Jansson and J. O. Anderson: Calphad, 1985, 9,

153–190.

15. H. K. D. H. Bhadeshia: Acta Metall., 1981, 9, 1117–

1130.

16. H. K. D. H. Bhadeshia: Met. Sci., 1982, 16, 159–165.

17. S. S. Babu and H. K. D. H. Bhadeshia: Mater. Sci. Technol., 1990,

10, 1005–1020.

18. Materials Algorithm Project, available at: http://www.msm.cam.

ac.uk/map/mapmain.html

19. http://calculations.ewi.org/VJP/MaterialsModeling.html

20. BISRA: ‘Atlas of isothermal transformation diagrams of B. S. EN

steels’, Special report no. 56, 2nd edn, The Iron and Steel Institute,

London, 1956.

21. S. S. Babu, G. M. Goodwin, R.J. Rohde and B. Sielen: Weld. J.

Res. Suppl., 1998, 77, 249s–253s.

22. J. Zrnik, I. Mamuzic and S. V. Dobatkin: Metallugija, 2006, 4,

323–331.

23. B. Narayanan: Ohio State University, Columbus, OH, to be

published.

24. F. Caballero, M. Miller, C. Garcia-Mateo, C. Capdevila and S.

Babu: Acta Mater., 2008, 56, 188–189.

25. S. S. Babu, K. Hono and T. Sakurai: Metall. Trans. A., 1994, 25A,

499–508.

26. H. K. D. H. Bhadeshia: ‘Bainite in Steels’, 2nd edn; 2001, London,

Institute of Materials.

27. G. I. Rees and H. K. D. H. Bhadeshia: Mater. Sci. Technol., 1992,

8, 994–996.

28. S. S. Babu and H. K. D. H. Bhadeshia: Mater. Sci. Eng. A, 1991,

A142, 209–220.

29. H. K. D. H. Bhadeshia: ‘Worked Examples in the Geometry of

Crystals’, 1st edn; 1987, London, Institute of Metals.

30. T. Furuhara, S. Morito, T. Maki: ‘On the controlling factors of

grain sizes in martensite and bainite’, Proc. 1st Int. Symp. on ‘Steel

science’, (ed. T. Furahara and K. Tsukzai), 51–56; 2007, Tokyo,

The Iron and Steel Institute of Japan.

31. S. S. Babu, S. A. David, J. M. Vitek and R. W. Reed: Sci. Technol.

Weld. Join., 2001, 6, 31–40.

32. A. A. B. Sugden and H. K. D. H. Bhadeshia: ‘Microstructure

entropy and the scatter in toughness’, Proc. 2nd Int. Conf. on

‘Trends in welding research’, Gatlinburg, TN, USA, May 1989,

ASM International, 745–748.

33. Z. K. Liu, L. Hoglund, B. Jonsson and J. Agren: Metall. Mater.

Trans. A., 1991, 22A, 1745–1752.

34. L. Gavard, H. K. D. H. Bhadeshia, D. J. C. MacKay and S. Suzuki:

Mater. Sci. Technol., 1996, 12, 453–463.

35. C. A. L. Bailer-Jones, H. K. D. H. Bhadeshia and D. J. C.

MacKay: Mater. Sci. Technol., 1999, 15, 287–294

36. R. C. Reed, T. Akbay, Z, Shen, J. M. Robinson, J.H. Root: Mater.

Sci. Eng., 1998, 256, 152–165.

37. J. R. Yang and H. K. D. H. Bhadeshia: Mater. Sci. Eng. A, 1989,

A118, 155–170.

38. J. W. Elmer, T. A. Palmer, W. Zhang and T. DebRoy: Sci. Technol.

Weld. Join., 2008, 13, 265–277.

39. J. W. Elmer, T. A. Palmer, W. Zhang, S. S. Babu: Weld. J. Res.

Suppl., 2004, 83, 244s–253s.



published flash process paper

Documents