A

otusTmwb©

K

1

beIaahtscsvbmTtm

0d

Materials Science and Engineering A 464 (2007) 157–169

Mitigation of sensitisation effects in unstabilised 12%Crferritic stainless steel welds

Martin van Warmelo, David Nolan ∗, John NorrishFaculty of Engineering, University of Wollongong, NSW 2522, Australia

Received 16 November 2006; received in revised form 25 January 2007; accepted 25 February 2007

bstract

Sensitisation in the heat-affected zones of ferritic stainless steel welds is typically prevented by stabilisation of the parent material with titaniumr niobium, and suitable design of the overall composition to produce a suitably high ferrite factor. However, such alloy modification has proveno be economically unviable for thick gauge (>10 mm) plate products and therefore unstabilised 12%Cr (3CR12) material is still currently beingsed for heavy gauge structural applications in many parts of the world. The aim of the current work was to review the mechanisms responsible forensitisation in these unstabilised ferritic stainless steels, and to characterise the sensitisation effects arising from multipass welding procedures.he objective was to determine the influence of welding parameters, and thereby to recommend mitigating strategies. Two particular sensitisation

odes were found to occur in the current work, although only one was predominant and considered problematic from a practical perspective. Itas found that with proper positioning of weld capping runs and control of weld overlap, it is possible to ensure that sensitising isotherms remainuried beneath the parent surface, and so reduce harmful corrosion effects. 2007 Elsevier B.V. All rights reserved.

mwndbiptiAcw

c4t

eywords: Sensitisation; 12%Cr steel; Welding; Ferritic stainless steel

. Introduction

In recent years, sensitisation in 12%Cr (3CR12) steels haseen the subject of intensive investigation in Australia after sev-ral corrosion failures were reported in welds on coal wagons.t was subsequently determined that sensitisation could occurs a result of a number of different mechanisms, manifestingfter different heat treatments and at different positions in theeat-affected zone (HAZ). The main conclusion of the work washat sensitisation could only be totally eliminated by effectivetabilisation by titanium or niobium additions, and by suitableontrol of the ferrite factor. While simple in theory, this presentsome problems for the steel manufacturer. Since titanium is aery strong ferrite stabiliser, the presence of titanium needs toe offset by appropriate amounts of � stabilisers, but the twoost potent candidates (carbon and nitrogen) are undesirable.

his leaves manganese and nickel as suitable candidates and

hese elements are typically added in concentrations of approxi-ately 2 and 1%, respectively. With these concentrations a fully

∗ Corresponding author. Tel.: +61 2 42215549; fax: +61 2 4221 3112.E-mail address: dnolan@uow.edu.au (D. Nolan).

imfiwrmc

921-5093/$ – see front matter © 2007 Elsevier B.V. All rights reserved.oi:10.1016/j.msea.2007.02.113

artensitic high-temperature HAZ (HTHAZ) can be achievedhich has the additional advantage of having increased tough-ess, since the presence of coarse � ferrite is eliminated. Thisoes, however, present a major challenge to the steel makerecause both these elements depress the Ac1 temperature andncrease the tempering resistance of the steel [1]. High Ac1 tem-eratures are advantageous because they allow higher annealingemperatures, which decreases the required holding time andncreases throughput, resulting in lower cost. Steels with lowc1 temperatures require batch annealing which is a slower pro-

ess compared to continuous line annealing, which is possibleith higher Ac1 temperatures.These measures, however, do not provide a solution in all cir-

umstances. In comparison to fully ferritic grades like 409 and30, 3CR12 is considered to have good weldability and HAZoughness in thick as well as thinner gauges, and it is suppliedn thicknesses up to 30 mm [2]. Material 10 mm and greater is

ost commonly supplied as shear plate and not as coil, that is, thenal flat strip is cut directly to the desired length after hot rolling

ithout being coiled. With the relatively high cooling rates expe-

ienced during the cutting process, complete transformation toartensite will take place, irrespective of the Ac1 and chemical

omposition. Annealing is therefore required, but an equivalent

1 ce and Engineering A 464 (2007) 157–169

oaai

palmoitosaeiti

dwicp

2

ipTcrtpch

tgFfa3ffd

cwtcef(H

F[

Aomcamtratfi

tcchc

2

This mode of sensitisation is linked to the presence of untem-pered martensite in the steel before it is exposed to the sensitisingtemperature. This means that as a result of a single weld pass,

58 M. van Warmelo et al. / Materials Scien

f batch annealing for stacks of cut plate becomes impracticalnd expensive. As a result, titanium stabilised material in gaugesbove 8 mm is generally not available, and unstabilised materials still being imported into Australia.

Unstabilised thick gauge material is theoretically highlyrone to sensitisation. The work by Williams and Barbaro [3]nd Matthews et al. [4,5] showed that intersecting isotherms canead to sensitisation, and thick gauge material, which requires

ultiple passes for the simplest joints, will naturally have severalverlapping weld beads with intersecting isotherms. However,t is not a foregone conclusion that critical sensitisation willake place, since the position of the beads, and the degree ofverlap, will determine whether any particular HAZ is sen-itised by subsequent weld beads. The work by Matthews etl. [4] also showed to what extent the degree of overlap influ-nced the presence of sensitisation and IGC. A significant factors also whether the sensitised region reaches the surface ofhe plate or whether it remains within the interior where it isnnocuous.

The current investigation was initiated in an attempt toetermine the extent to which thick gauge 3CR12 materialas susceptible to sensitisation, to characterise its occurrence

n multiple pass welds, and to determine to what extent itould be minimised by appropriate control of the weldingarameters.

. Background

Sensitisation in stainless steel can be defined as susceptibil-ty to intergranular corrosion (IGC), which occurs due to theresence of chromium-depleted zones at the grain boundaries.his depletion of chromium is generally associated with the pre-ipitation of chromium carbides on the grain boundaries, whichemove chromium from the matrix and thereby locally reducehe corrosion resistance along the grain boundaries. In their sim-lest form, the carbides would be Cr23C6 or Cr7C3, and theyonsume a large quantity of chromium on formation due to theigh stoichiometric ratio.

Predicting the sensitisation behaviour of 3CR12 during weldhermal cycles is complicated by the presence of the so-calledamma loop in the typical Fe–Cr phase diagram, illustrated inig. 1. In high purity Fe–Cr systems, the gamma loop extends asar as about 13.5%Cr [1], after which the structure is fully ferritict all temperatures. Due to its otherwise low alloying content,CR12 lies in the dual phase region, and the structure will there-ore consist of a mixture of delta ferrite (untransformed), alphaerrite (transformed from austenite on cooling) and martensite,epending on the cooling rate.

The phase diagram naturally does not provide a good indi-ation of the structure which will be found in the HAZ of aeld, but in conjunction with the typical continuous cooling

ransformation (CCT) diagram shown in Fig. 2 [6], the typi-al features of the HAZ expected in 3CR12 type steels can be

xplained. Unlike the HAZ for plain carbon steels, the HAZ zoneor 3CR12 has two visually distinct zones; the high-temperatureor coarse grained) HAZ (HTHAZ) and the low temperatureAZ (LTHAZ). Material heated close to the liquidus (above the

Ft(

ig. 1. Typical Fe–Cr phase diagram illustrating the gamma loop (after Demo1]).

5) transforms completely to � ferrite and rapid grain growthccurs. On cooling, the amount of reversion to � will be deter-ined by the CCT diagram and therefore the HTHAZ frequently

onsists of coarse-grained � ferrite, with islands of martensitet the grain boundaries. The CCT diagram indicates that if theaterial temperature reaches 1050 ◦C within 1–2 s, no reversion

o � will occur and the � ferrite structure will be maintained tooom temperature. However, material which has been heatedbove the Ac1 but below the Ac5 will contain significant frac-ions of � which will transform to martensite, resulting in a toughne-grained structure.

Extensive work by Williams and Barbaro [3] has shown thathe origin of sensitisation in 3CR12, that is, the creation ofhromium-depleted zones, can be ascribed to four different pro-esses or modes. These modes distinguish between where andow the chromium-depleted zone will be formed and the thermalonditions required to create the chromium depletion.

.1. Mode 1 sensitisation

ig. 2. Proposed continuous cooling transformation diagram for the transforma-ion of �-ferrite to � in the high-temperature HAZ during weld thermal cyclesadapted from Pistorius and van Rooyen [6]).

M. van Warmelo et al. / Materials Science

Fs

smtiaaiseharemta

2

mipHcmiwpabmpI

ebtwa

2

gHfoaoAs

mawd1tascpruaadw�whrcsb

omhwbetids

ig. 3. Macrosections illustrating the influence of weld geometry and weldingequence on mode 2 sensitisation (after Williams and Barbaro [3]).

ensitisation can occur parallel to the weld bead wherever theaterial reached the critical sensitising temperature. In practice,

he presence of substantial amounts of untempered martensiten 3CR12 sheet will only occur if the material was incorrectlynnealed (that is the Ac1 temperature was exceeded duringnnealing or any other form of heat treatment before process-ng). Mode 1 is potentially the most severe manifestation ofensitisation in these steels, principally because it is likely toxtend over a large area. If a plate or edge of a coil is over-eated during the final annealing stage, it renders the entirerea susceptible to sensitisation if it is welded. The sensitisedegion can therefore be very widespread and extend along thentire length of a weld bead. However, commercially availableaterial should not contain any untempered martensite since

he material is not deliberately heated above the Ac1 duringnnealing.

.2. Mode 2 sensitisation

Fundamentally, the mechanism for mode 2 is identical toode 1, but the difference lies in how the untempered martensite

n the material is created. Mode 2 assumes at least two weldingasses where the first pass created untempered martensite in theAZ and the critical sensitising isotherm from the second pass

auses carbide precipitation in the first HAZ. Essentially, bothodes require two exposures to high temperature but the man-

festation will be different. In mode 1, the intergranular attackill be associated directly with the weld bead that caused therecipitation, while in mode 2, corrosive attack will be associ-ted with weld 1 while the precipitation was effectively caused

y weld 2. Depending on the weld geometry and dimensions,ode 2 sensitisation can manifest itself on the opposite side of a

late from where the weld was positioned, as illustrated in Fig. 3.t has been shown by Williams and Barbaro [3] and Matthews

2

a

and Engineering A 464 (2007) 157–169 159

t al. [5] that the joint configuration and positioning of the weldeads can significantly affect whether sensitisation occurs. Ifhe sensitising isotherm from the second weld only intersectsith the filler metal or unaffected base metal, sensitisation is

voided.

.3. Mode 3 sensitisation

Mode 3 sensitisation occurs in the HTHAZ (i.e. the coarse-rained region adjacent to the fusion line) in material where theTHAZ is predominantly ferritic. Since the material close to the

usion zone is heated well above the Ac5 temperature, this modef sensitisation is independent of any previous heat treatmentnd material condition. Unlike mode 2 sensitisation, mode 3ccurs after a single exposure to high temperatures above thec5 temperature and has been shown to occur even in titanium-

tabilised steels with higher ferrite factors.Although not well understood, recent work has shown that

ode 3 is caused by extremely rapid cooling rates generallyssociated with very low heat input welds [7] and by shalloweld toe cusps and arc strikes [3]. As can be seen from the CCTiagram for 12%Cr steels [6], very rapid cooling from above350 ◦C will result in a fully � ferrite structure. At these elevatedemperatures, even TiC and TiN can dissolve to release carbonnd nitrogen back into the � matrix. If the � transformation isuppressed during cooling, there will be a strong tendency forarbides to precipitate at the grain boundaries as the materialasses through the critical temperature range. It appears thateformation of Ti(C, N) precipitates is kinetically unfavourablender these conditions, and preventing sensitisation is thereforechieved by the presence of sufficient � during cooling whichbsorbs and traps the carbon and nitrogen rejected by the � ferriteuring cooling. In the recent work by Greef and Du Toit [7], itas shown that for equal heat input levels, material with a higherpotential had a lower propensity for sensitisation than materialith lower � potential. It was also shown that increasing theeat input resulted in reduced sensitisation due to lower coolingates, which in turn resulted in more � reverting to �. This workonfirms previous reports by Gooch et al. [8] that the degree ofensitisation in single pass welds would depend on the phasealance of the material.

It is therefore desirable to increase the level of austenite ref-rmation during cooling in order to maximise the volume ofartensite in the HAZ. Gooch and Ginn [9] suggested that pre-

eating in the region of 300 ◦C would result in slower coolingith a subsequent increase in the volume of � formed on cooling,ut also commented that very little success had been achievedxperimentally. Even if moderate success had been observed,his would not necessarily represent a practical solution becauset negates the benefits that the utility ferritics were designed toeliver, specifically ease of fabrication, similar to that of carbonteels.

.4. Mode 4 sensitisation

Mode 4 sensitisation is only associated with steels which haverelatively high Ac1 temperature, where the Ac1 for common

160 M. van Warmelo et al. / Materials Science and Engineering A 464 (2007) 157–169

Table 1Chemical analysis of 3CR12 materials (all values in weight percent)

Material ID C S P Mn Si Ni Cr Mo Ti N Cu Co

A1 0.016 0.0005 0.023 0.957 0.247 0.416 12.45 – 0.003 0.0080 – –A2 0.012 0.0016 0.027 0.910 0.280 0.588 12.20 – 0.004 0.0073 – –C 0

1fbtpscaofvbtuWdtaol[oAfrdwtvbfr

dcawsm

3

3

eimiTitvsaTmKBaoeistucadceaTam

TM

M

AAC

1 0.030 0.0040 0.028 0.410 0.640

2%Cr steels ranges between 760 and 840 ◦C. The mechanismor mode 4 has not been confirmed but it is believed that car-ides start to dissolve in the region just below the Ac1 andhen precipitate as chromium carbides on cooling. The modelroposed by Williams and Barbaro [3] suggests that elements,uch as boron and vanadium are involved. It is suggested thathromium borides and vanadium carbides dissolve at temper-tures around 800 ◦C and then chromium carbides precipitaten cooling. In their review on stabilisation and potential carbideorming elements, Gordon and van Bennekom [10] report thatanadium is unsuitable because of sluggish precipitation, andecause VC dissolves close to 800 ◦C, which correlates withhe hypothesis by Williams. Vanadium was present as a resid-al in all the materials showing mode 4 in the work done byilliams and Barbaro [3] and, if the resulting sensitisation is

ue to VC dissolution, this would explain the observation thathe severity of mode 4 sensitisation is generally low. Williamslso suggests that some dissolution of carbo-nitrides can startccurring from temperatures above 790 ◦C since the solubilityimit of carbon at this temperature can be as high as 0.03%11]. In steels with low Ac1 temperatures, any dissolved carbonr nitrogen will be absorbed by the austenite formed when thec1 temperature is exceeded. The severity of mode 4 is there-

ore strongly dependent on the amount of carbon and nitrogeneleased into solution and the temperature range between theissolution temperature and the Ac1. The severity of mode 4ould be expected to increase with higher heat input and resul-

ant increased time at the dissolution temperature. Conversely,ery low heat input and fast travel speeds have been shown toe instrumental in causing mode 3 sensitisation [7], and there-ore the heat input needs to be controlled within an optimalange.

Mode 4 can occur after a single heat cycle and is not depen-ent on martensite being present or created by previous heatycles. A third zone has therefore been identified in the HAZ

nd is generally referred to as the subcritical HAZ [5]. This areaas previously considered as being unaffected by welding cycles

ince no phase transformations occur, but with the isolation ofode 4 sensitisation, this region becomes highly significant.

emw

able 2aterial characteristics of the 3CR12 materials (temperatures in degrees celsius)

aterial ID KFF [13] FFBSL [3] Ac1 (Folkhard) [11] Ac1 (SmTarboton

1 9.42 8.70 824.5 765.02 8.97 8.16 819.9 751.91 11.40 9.21 797.7 826.4

.380 11.32 0.02 0.033 0.0140 0.12 0.02

. Experimental methods

.1. Materials

The main aim of the work carried out was to determine thextent to which multiple weld passes and varying heat inputnfluences the degree of sensitisation in thick gauge 3CR12

aterial. The experimental work was carried out in two parts,nvolving bead on plate (BOP) welds and straight butt welds.he material for the investigation was supplied by Atlas Special-

ty Metals, who import the steel from various suppliers aroundhe world. Material from two different manufacturers was pro-ided and plates from two different batches were provided byupplier A. The nominal chemical analysis for each material,s shown on the material test certificates, is given in Table 1.hese analysis values have been used to calculate the variousaterial characteristics, as given in Table 2. As expected, thealtenhauser Ferrite Factor (KFF) value [13] is higher than theluescope Steel Ferrite factor (FFBSL) for each material and,ccording to the criteria given by Williams and Barbaro [3],nly material A2 should be immune from mode 3 sensitisation,ven though the difference between the analyses of A1 and A2s typical of the variation between production casts from theame manufacturer. While the prediction for the martensite startransformation temperature is fairly consistent for the formulassed by Smith and Tarboton [12] and Gooch et al. [8], the sameannot be said for the Ac1 temperatures. Given that all the steelsre unstabilised, the Ac1 temperature is an important factor inetermining susceptibility to mode 4 sensitisation. The valuesalculated by Smith and Tarboton [12] are determined from anmpirical model based on dilatometry results of several hundrednalyses and was specifically developed for 12–14%Cr steels.he Folkhard equation [11] is based on more limited data and,s such, the results given by Smith et al. are more likely to beore accurate.

A set of preliminary trials were required in order to optimise

xperimental welding conditions and ensure weld quality. Asaterial availability from supplier C was limited, the initial trialork was undertaken using material A1. Once the experimental

ith and) [12]

HAZ M% (Matthewset al.) [5]

Ms (Smith andTarboton) [12]

Ms (Goochet al.) [8]

77.8 385.7 376.580.8 404.3 375.277.5 431.7 385.5

cience

dwvpmte

3

wlTwctoduiobotaerbvibargtlol

gus

1Nt1poc

3

Pfs0spw

aioprditdaashvi

wfvu

M. van Warmelo et al. / Materials S

ifficulties had been overcome, material from both suppliers waselded and tested to obtain comparative results between the twoariants. Since the available A1 material was consumed for thereliminary experimentation, direct comparisons have only beenade between material A2 and C1. In any case, the variation in

he analyses of A1 and A2 is typical of the level of variationxpected in commercial production.

.2. Welding procedures

Butt welds were produced between two 10 mm thick flat bars,hich had a double sided 60◦ V weld preparation with a 1.5 mm

anding (as per the manufacturer’s recommended practice [2]).he samples were tack welded at either end so that a 2 mm gapas left at the bottom of the V-notch. The tacked sample was

lamped to prevent bowing but a 10 mm gap was left betweenhe sample and the base plate. As per the manufacturer’s rec-mmended welding practice, the filler metal used was 0.9 mmiameter solid austenitic wire (309 L), and the shielding gassed was StainShield Lite (argon, 1%O2). Shielding gas contain-ng carbon dioxide is not recommended, due to the possibilityf carburisation. The interpass temperature was controlled toelow 60 ◦C and the plate was allowed to cool naturally with-ut forced air or water quenching between passes. The weldingorch was clamped to a Bug-O® system trolley which ran along

rail mounted parallel to the base plate. For the preliminaryxperiments on material A1, no backing bar was used for theoot pass and the back side of the root pass was ground outefore the backing pass was laid down. Root penetration was notery good on these initial weld runs and thorough back grind-ng was done before the back pass was laid down. A ceramicacking strip was used on the later welds using material A2nd C1. Better control of the root pass could be achieved andoot penetration was significantly better on these welds. Backrinding was still done on all samples, but it was less extensivehan for material A1. The welds were cleaned with a stain-ess wire brush between passes but grinding was only donen the root run. The back pass was always done as the veryast pass.

Transverse samples were cut across the welds for metallo-raphical analysis and sensitisation tests. The samples were cutsing a band saw so as to prevent any local overheating of theurface and then ground to 1200 grit finish and polished to a

wvac

Fig. 4. Cross-section showing more copper deposition on (a) the horizontally

and Engineering A 464 (2007) 157–169 161

�m finish. Etching with acidified FeCl3 solution or Kallingso. 2 revealed the coarse primary ferrite in the CGHAZ and

he martensite/ferrite HAZ and matrix. Electrolytic etching in0% oxalic acid clearly revealed the regions of fresh untem-ered martensite where effectively no carbide precipitation hadccurred, and severely attacked the grain boundaries where pre-ipitation had occurred.

.3. Modified Strauss test

The modified Strauss test is a variation on ASTM A 763-93ractice Z [14] for detecting susceptibility to sensitisation inerritic stainless steels. Samples are placed in a copper sulphateolution (60 g CuSO4 in 1l of water) which is acidified by adding.5% sulphuric acid (3 mL H2SO4 in 1l water). A layer of copperhot is placed on the bottom of the vessel and the samples arelaced in the solution in such a way as to prevent direct contactith the welded region. The solution is then boiled for 20 h.During the test, chromium-depleted regions will undergo

nodic dissolution and copper will be deposited on the corrod-ng regions. This generally highlights the areas where IGC isccurring but sometimes this can be masked by other corrosionroducts. Copper will deposit anywhere that corrosion is occur-ing and if any pitting or crevice corrosion takes place, coppereposition will be seen. It should also be noted that position-ng of the samples can also influence the test results, in thatest surfaces positioned horizontally tend to show more coppereposition than those positioned vertically. This may be due tosettling effect, and subsequent enhancement of corrosion viacrevice corrosion mechanism. An example of this effect is

hown in Fig. 4, where the sample in Fig. 4(a) was positionedorizontally while the sample shown in Fig. 4(b) was positionedertically during testing, and the decrease in copper depositionn the case of vertical positioning is clear.

The samples which had been analysed metallographicallyere tested by this method, as well as two additional samples

rom every weld. The polished specimens were placed in theessel so that the polished surface was horizontal and facingp while the other specimens were placed so that the weld bead

as parallel with the base of the vessel, i.e. the cut surfaces wereertical. After the samples were removed from the test solution,ll surfaces were scrubbed with a soft brush to remove any excessopper and other surface accumulation.

positioned sample, compared to (b) the vertically positioned sample.

162 M. van Warmelo et al. / Materials Science and Engineering A 464 (2007) 157–169

Table 3Welding parameters used for preliminary butt welds on material A1

Sample ID Heat input [kJ/mm] Current [A] Potential [V] Travel speed [mm/min] Wire feed rate [m/min]

BWA 01 0.63 110 25 200 6.1BWA 02 0.48 107 25 250 6.0BWA 03 0.55 104 25 220 6.2BWA 04 0.40 114 20 250 6.5BWA 05 0.37 112 22 300 6.5BWA 06 0.61 109 25 200 6.5BWA 07 0.68 117 26 200 7.2BWA 08 0.39 105 24 300 6.0BWA 09** 0.38 107 22 280 6.2BWA 10 0.45 109 25 270 6.5

Constant parameters for all samples

GC

ht

4

4

murmiwuir

wbtm

momiw

obsftwstsrs

TW

S

BBBBBBBB

C

GCC

as flow rate: 15 L/minontact tip to work distance: 15 mm

If metallographical analysis was required after the samplesad been tested, the samples were cleaned in nitric acid to removehe copper and then polished lightly.

. Results

.1. Preliminary welding trials: material A1

A satisfactory root run could be obtained when a dip transferode was set on the welding machine and so this mode was

sed for the root run and the two capping runs on all welds. Theemainder of the passes were generally done using spray transferode settings. The heat input and welding parameters are given

n Table 3. From literature [15], the arc efficiency for GMAelding is in the region of 60–75% and a factor of 75% wassed in the calculation for heat input. The heat input value givenn the table represents the average of all the passes excluding theoot passes done using the dip transfer parameters.

The heat input was varied by adjusting the travel speed and the

ire feed rate, and the arc current and potential were determinedy the welding machine. The values given in Table 4 for thesewo parameters are the final values displayed by the welding

achine after welding had been completed.

eomc

able 4elding parameters for comparative trials on materials A2 and C1

ample ID Heat input [kJ/mm] Current [A] Poten

WA 201 0.59 113 25WA 202 0.40 115 25WA 203 0.77 132 26WA 204 0.47 113 22WC 101 0.57 111 25WC 102 0.39 112 25WC 103 0.48 113 22WC 104 0.79 137 25

onstant parameters for all samples

as flow rate: 15 L/minontact tip to work distance: 15 mmeramic backing strip on root pass

Filler material: 0.9 mm 309 LInterpass temperature <50 ◦C

Given that the recommended heat input range for 3CR12 typeaterials is 0.4–1.0 kJ/mm, these values are all on the lower side

f the range. Nevertheless, the material deposition rate provedore than adequate to fill the V-notch, and at the higher heat

nput values, eight passes were not required to complete theeld.Consistency in the weld bead positioning created the majority

f the variation between the welds. In many cases, the weldead tended to ride up on one side of the V, leaving a veryharp notch at the base on the opposite side. Fusion defects wererequently noted at this position. The weld bead also frequentlyended to wander and this created very uneven and lopsidedelds, specifically with the last cover passes. This also created a

ituation where subsequent beads overlapped a great deal morehan would be expected in an expertly welded joint. As washown in literature [4], the degree of overlap plays a significantole in whether sensitisation will occur. Examples of this will behown in later sections.

Samples for metallographical examination were cut from

ach test piece, which were then polished and etched inxalic acid. It was apparent that the precipitate-free untemperedartensite was essentially unattacked. However, in some cases it

ould clearly be seen where intersecting isotherms had reached

tial [V] Travel speed [mm/min] Wire feed rate [m/min]

220 6.5320 6.5200 7.5240 6.5220 6.5320 6.5240 6.5200 7.5

Filler material: 0.9 mm 309 LInterpass temperature <50 ◦CRoot gap: 2 mm

M. van Warmelo et al. / Materials Science and Engineering A 464 (2007) 157–169 163

F hin th

totstwrstFi

smmoaaeiabne

fi

iFara

wmHmscbpssc

Fc

ig. 5. Macrographs showing the intersecting HAZ isotherms (a) contained wit

he surface, indicating areas where IGC might be expected toccur. Two contrasting examples are shown in Fig. 5. In Fig. 5(a),he HAZ is relatively small and the intersecting areas are sub-tantially below the plate surface. In Fig. 5(b), it can be seenhat the root run generated a very large HAZ on the left side,hile the HAZ from the back pass is significantly smaller. As a

esult, a narrow sensitised strip has been created on the bottomurface of the plate. This was confirmed by copper deposition onhe underside of the sample during the Strauss test, as shown inig. 6. In general, copper deposition could be seen on all samples

n the region of the root passes.When the samples were examined under an optical micro-

cope, it could clearly be seen that extensive corrosion of theartensite phase had occurred. In the most severe cases, theartensite regions were effectively completely corroded and

nly a network of � ferrite grains remained. Theoretically, thebsence of martensite grains could be due to grain droppingfter the grain boundaries had been totally dissolved. How-ver, given that the material is predominantly martensite withslands of ferrite and the fact that grain boundary precipitation

nd chromium depletion is much more likely to occur on �–�oundaries, grain dropping should involve the ferrite grains andot the martensite. In addition, samples cut perpendicular to thexposed surface showed dissolved martensite grains while the

i

cm

ig. 6. The same macrosections as shown in Fig. 5, illustrating the deposition of coppontained within the material in (a), but it is reaching the back surface in (b).

e materials, and (b) reaching the surface on the backside (oxalic acid etch).

errite grains above them were still in place. This is clearly seenn Fig. 7.

The HTHAZ was similarly affected when the sensitisingsotherms intersected through these regions. As can be seen inig. 8, the martensite grains beading the coarse �-ferrite grainslong the fusion line are essentially removed. However, the fer-ite grains also appear to be marginally affected by corrosivettack along what appear to be sub-grain boundaries.

This level of corrosive attack might suggest that the materialas not sensitised, but rather that preferential corrosion of theartensite was occurring due to the reduced chromium content.owever, if no elemental partitioning occurs during welding, asaintained by Gooch and Ginn [9], the martensite on the whole

ample should suffer similar corrosive attack, and not just spe-ific regions within the HAZ. In areas where the martensite hadeen less severely corroded, it could be seen that the martensitehase was not being uniformly corroded but rather that corro-ion was taking place on a network of sub-grain boundaries, ashown in Fig. 9(a). In some instances a web of copper depositionould clearly be seen on the martensite grains as in Fig. 9(b),

ndicating that corrosion was not uniform over the whole phase.

Samples were also examined under the SEM and typi-al results are shown in Fig. 10 Extensive pitting of theartensite can be seen on the transverse section (Fig. 10(a))

er on overlapping HAZ isotherms during Strauss test. Note the sensitisation is

164 M. van Warmelo et al. / Materials Science and Engineering A 464 (2007) 157–169

Fig. 7. Micrographs showing extensive corrosion of martensitic regions resulting from Strauss test: (a) attack on exposed surface, and (b) attack perpendicular toexposed surface. Micron bar represents 50 �m.

Fig. 8. Two micrographs showing examples of corrosion of martensite regions surrounding �-ferrite in HTHAZ (micron bar represents 38 �m).

Fig. 9. Micrographs showing (a) preferential corrosion of martensite regions along sub-grain boundaries (micron bar represents 20 �m at 500×), and (b) depositionof copper along sub-grains within regions of martensite (1000×).

Fig. 10. SEM images showing corrosion of martensitic regions in (a) surface directly exposed during Strauss test, and (b) surface perpendicular to exposed surface.

M. van Warmelo et al. / Materials Science and Engineering A 464 (2007) 157–169 165

on in

wdc

SaoipFacss

gpctwb

cmstSai

na

4

t(fctw

ptwTrp

wfwtw

Fig. 11. Two SEM images showing typical variati

hile the intergranular nature of the attack is clearly evi-ent in Fig. 10(b), which was taken perpendicular to the weldross-section.

The extent of the copper deposition on the samples during thetrauss test (as shown in Fig. 6) indicates that reasonably largereas are sensitised. However, within these regions, the extentf corrosion could be extremely uneven with large crevices orrregular pits forming within regions of general attack. The SEMhotographs in Fig. 11 represent the section from the sample inig. 6(b) on the bottom right hand side of the weld bead, justbove the HAZ from the back pass. In Fig. 11(a), extensiveorrosion is visible in the bottom left hand corner while fairlyevere but isolated crevices have formed at the extremity of theensitised region.

The uneven nature of the martensite corrosion and the inter-ranular corrosion evident in Fig. 10(b) indicate strongly thatreferential corrosion of the martensite due to lower chromiumontent cannot be the only factor involved. If any element par-itioning had taken place during cooling, the HAZ of all theeld beads should be affected and not just the intersecting areasetween beads.

Any doubt as to whether corrosion was due to sensitisationould be settled by applying a healing heat treatment to theaterial. To confirm this, pieces were cut from several weld

amples and annealed at 700 ◦C for 10 min. The samples were

hen ground to remove the annealing scale and subjected to thetrauss test. As expected, none of the annealed samples showedny copper deposition and no IGC could be detected after pol-shing and examination under a microscope. Fig. 12 shows that

nTtc

Fig. 12. Micrographs showing CGHAZ after annealing, which is free of corrosi

the extent of corrosion within sensitised regions.

o IGC occurred in both the high and low temperature HAZfter the samples were annealed.

.2. Comparative welding trials: material A2 And C1

For the comparative work, the welding procedures used werehe same as those outlined for the previous section, that is, pulseddip transfer) mode was used for the root passes and spray trans-er weld mode was used for the remaining passes. However, aeramic backing strip was used on all the welds and hence bet-er penetration was obtained on most welds. Superficial grindingas done on the root pass before the back pass was laid down.Despite this, the weld bead still tended to wander and bead

ositioning was not good, giving rise to a high level of inconsis-ency between the various runs. The aim was to produce similarelds in the two different materials from two manufacturers.he welding parameters used are given in Table 4 and compa-

able heat input values were obtained, but the effect of thesearameters is masked by inconsistencies in the welds.

As for the previous procedures, samples were cut from all theelds for metallographic analysis and Strauss testing. Samples

rom the two materials with comparable heat input in the weldsere tested simultaneously. From purely visual observation of

he samples when they were removed from the test solution, itas clearly visible that the samples from material C had sig-

ificantly greater areas of copper deposition than material A.he samples were also visually different, in that for material A,

he polished surfaces were dull grey while for material C, theorroded surfaces were almost black.

on effects (micron bars represent: (a) 75 �m and (b) 50 �m, respectively).

166 M. van Warmelo et al. / Materials Science and Engineering A 464 (2007) 157–169

Fig. 13. Macrosections illustrating differences in copper deposition for different materials A and C subjected to modified Strauss test.

g and

sstigitcgisw

chetssscm

e

sitiweld. If the bead had been positioned correctly, the sensitisingisotherm would have intersected with the lower weld bead andnot reached the surface.

Fig. 14. Mode 2 sensitisation at weld toe due to poor positionin

From the analysis of the materials given in Table 1, it can beeen that material A contains significantly more chromium andlightly higher nickel, which would have a significant effect onhe overall corrosion resistance of material A. The differencen material discolouration is most likely due to this differingeneral corrosion resistance. It is also likely that material Cs inherently more susceptible to IGC since Demo [16] main-ains that higher levels of carbon can be tolerated with higherhromium content, but material C has a higher carbon and nitro-en content as well as having less chromium. The differencen the level of copper deposition is illustrated in Fig. 13. Theamples shown are BWA 202 and BWC 102, which were bothelded with a heat input of roughly 0.4 kJ/mm.Within the range of heat input used, very little difference

an be detected between the level of copper deposition at lowereat inputs compared to higher heat input. What was far morevident was that bead positioning and the degree of overlap ofhe beads is extremely important in order to prevent mode 2ensitisation on the surface of the material. In this set of trials,everal samples showed copper deposition on the upper or lowerurface of the plate, parallel to the weld beads, but these defects

ould be attributed to poor welding rather than to an inherentaterial defect or the variation in the welding parameters.As was discussed previously, it has been shown [4] that the

xtent of overlap is a significant contributing factor in causingFp

excessive overlap: (a) weld macrosection and (b) weld surface.

ensitisation. This can be seen in Fig. 14 where poor position-ng of the final pass resulted in a high degree of overlap onhe weld bead below and subsequently the sensitising isothermntersected with the surface of the plate close to the toe of the

ig. 15. Macrosection showing mode 2 sensitisation occurring as a result ofoor weld positioning.

M. van Warmelo et al. / Materials Science and Engineering A 464 (2007) 157–169 167

ensive

FoppcweoF

5

siisp

shevtpab

pdotftatatws

cw

dsossofot

snaf

csfmaiwhmtogt

c2HTt

Fig. 16. (a) Optical image and (b) SEM image showing ext

The effect of bad positioning of the back pass is shown inig. 15 below. The base of the root pass was not properly groundut and a serious fusion defect can be seen on the left of the rootass. The back pass position was badly off centre of the rootass and subsequently the HAZ from the back pass does notover the HAZ from the root pass, as was the case in weldsith reasonable positioning of the back pass. As a result, a very

xtensive sensitised region has been created adjacent to the toef the back pass and severe corrosion was evident, as shown inig. 16.

. Discussion

The two most significant results from the work on both sets ofamples were confirmation of the concept that high heat inputs detrimental to the corrosion resistance and that poor weld-ng, that is, excessive overlap or weaving, can result in severeensitisation of the previous weld beads at the surface of thelate.

Quite apart from increasing the time that the material willpend in the sensitising temperature range, unnecessarily higheat input will also result in a very large HAZ. While not nec-ssarily detrimental on its own, a large HAZ with obviouslyery susceptible martensite increases the potential for sensitisa-ion from subsequent parallel or intersecting weld beads. Thisroblem has been clearly illustrated in Fig. 6. The problemsssociated with excessive overlap and bad positioning have alsoeen illustrated in Figs. 14 and 15.

As noted previously, despite the relatively low heat inputarameters selected, in many instances the volume of metaleposited appeared to be excessive for the recommended numberf passes. However, any attempt at reducing the metal deposi-ion rate by increasing the travel speed or reducing the wireeed rate resulted in poor fusion and very bad welds. A compe-ent welder would be in a position to overcome these problemsnd produce consistent welds with smaller, more regularly posi-ioned beads which should produce significantly smaller HAZs

nd subsequently less sensitisation. As a result, the only sensi-ised regions within a professionally completed weld would beell below the surface. While this may not represent an ideal

ituation, the use of unstabilised material could still represent a

gtos

corrosion at weld toe resulting from mode 2 sensitisation.

ost effective solution over the additional problems associatedith producing stabilised thick gauge plate.For all the immersion tests, it could be seen that more copper

eposited on the upper surface of the sample than on the sameurface if the sample was positioned vertically. While the extentf copper deposition is not necessarily indicative of the extent ofensitisation, it does provide a measure of the amount of corro-ion taking place. The actual mechanism by which corrosion isccurring is not particularly relevant but it does indicate that anyor of fouling on the weld beads is likely to increase the chancesf corrosion and therefore cleaning, pickling and passivation ofhe beads after welding is very important.

With respect to the various modes of sensitisation, mode 1ensitisation was not observed, nor expected, since it shouldot occur if the parent material has not been heated above theustenitising temperature before fabrication and the material isree from untempered martensite.

Mode 3 has been shown [3,7] to be attributable to very rapidooling rates resulting from low heat input and weld cusps andignificantly affects steels with high ferrite factor. As such, theerrite factor data given in Table 2 suggests that all the testedaterials should be susceptible to mode 3 sensitisation under

ppropriate conditions. However, mode 3 sensitisation was notdentified in any of the experimental work described herein. Thisas to be expected since the range in the heat input would notave resulted in the extremely rapid cooling rates required forode 3 sensitisation to occur. Therefore the austenite potential of

he material is adequate to ensure some austenite reversion willccur and prevent sensitisation from occurring on the � ferriterain boundaries. This does however not rule out the possibilityhat mode 3 sensitisation will occur at weld cusps or at arc strikes.

Modes 2 and 4 sensitisation were seen separately, and inombination, but will be addressed as separate issues. Modesensitisation was seen on all the welded samples where theAZ from one bead overlapped the HAZ from a previous bead.he oxalic acid etch technique clearly showed where significant

empering of the HAZ martensite had occurred, and therefore

ave a strong indication of where corrosion would be expectedo occur. Even though the ASTM standards [14,17] define anxalic etching test, it is not considered suitable for low chromiumteels and therefore did not conclusively indicate the presence

1 ce and Engineering A 464 (2007) 157–169

ogwmettsbpetscaagtrTthr

ihalwitooelcw

mraAsdshwneCmcTtmot

Fig. 17. Composite micrograph illustrating mode 4 sensitisation at the surfaceof sample BWA 09.

Fm

sasc

6

68 M. van Warmelo et al. / Materials Scien

f sensitisation on these materials. Furthermore, it is not sug-ested as an appropriate etching technique for 12%Cr steels,ith Kallings No. 2 [18] and Picric acid [19] being the com-only recommended solutions. The oxalic acid etch, which

nables a much shorter etching time, was tried in an attempto get a better picture of the grain boundary precipitation andhe results were unexpected but highly informative. Mode 2ensitisation was seen on all the areas which were indicatedy the oxalic acid etch and the correlation between the etchattern and the copper deposition during the Strauss test wasxtremely good. While some IGC was expected, the extent ofhe observed corrosion was surprising. As has been shown ineveral figures, overall corrosion of the martensite was the mostommonly observed corrosion, rather than specific grain bound-ry attack. As is shown in Fig. 7, sensitisation is not occurringt the �-ferrite grain boundaries but rather within the martensiterains. It is likely that chromium-based carbides are forming athe martensite lath boundaries during the thermal cycle, and thisesults in a network of chromium-depleted sub-grain boundaries.he fact that a chromium depletion mechanism is responsible for

he observed sensitisation is confirmed by the fact that a healingeat treatment at 700 ◦C restores the corrosion resistance andemoves any sign of intergranular attack in the Strauss test.

Even though sensitisation could not be eliminated by reduc-ng the heat input, it could be seen that reducing the size of theeat-affected areas dramatically reduced the sensitised regions,s shown in Fig. 6. This indicates that any welding techniquesike weaving, and high deposition consumables like flux coredire, will more likely result in inferior properties. The worst

nstances of sensitisation and corrosion could clearly be linkedo poor welding, specifically bad bead positioning and excessiveverlap on subsequent beads. As shown in Fig. 15, positioningf the back pass is critical, so that the HAZ from the root pass isngulfed by the back pass. This should however not be a prob-em for a competent welder. Similarly, proper positioning of theover passes will ensure that the sensitising isotherm intersectsith the filler metal of the previous pass and not with the HAZ.There were a number of samples that showed indications of

ode 4 sensitisation, as shown in Fig. 17. However, no obviouselationship between the occurrence of mode 4 sensitisationnd composition or welding conditions could be determined.s with the mode 2 sensitisation, the mode 4 attack does not

how typical intergranular corrosion but manifests as generalissolution of the martensite grains. In this instance it can beeen as a fine row of pits along the edge of the LTHAZ. Higheat input, together with a high Ac1 temperature are the factorshich are believed to influence the creation of mode 4. However,either of these should be a factor on sample BWA 09. A similarxample of mode 4 sensitisation that occurred on material

is shown in Fig. 18. This is particularly interesting in thatode 4 was clearly seen in two steels of significantly different

omposition and with substantially differing Ac1 temperatures.his would appear to indicate that some mechanism other than

he dissolution of less stable carbides (specifically of vanadium)ust be resulting in carbon being available for precipitation

f chromium carbides on cooling. This mechanism also haso be independent of material condition and heat treatment,

1ii

ig. 18. Micrograph showing modes 2 and 4 sensitisation and IGC onaterial C.

ince mode 4 sensitisation appears after a single weld pass andpparently over a wide range of heat input. In theory, stabilisedteels should be immune to this form of sensitisation, since allarbon should be tied up with titanium or niobium.

. Conclusions

In conclusion, it can be said that unstabilised thick gauge2%Cr steels are susceptible to mode 2 sensitisation. However,n the majority of cases, this should not present a problemn service since the sensitised regions remain buried below

cience

ttaiowwrc

aittli

A

LwLtw

R

[

[

[

[[

[[[

shohocken, PA, 1998, pp. 50–65.[18] A.M. Meyer, M. Du Toit, Weld. Res. Suppl. 80 (12) (2001) S275–

M. van Warmelo et al. / Materials S

he surface of the plate. Conditions which would exacerbatehe problem need to be carefully controlled (or even avoided)nd these include high heat input, techniques which resultn very high metal deposition rates, inappropriate weld beadverlap and weaving or stop/start operation. It was found thatith proper positioning of weld capping runs and control ofeld overlap, it is possible to ensure that sensitising isotherms

emain buried beneath the parent surface, and so reduce harmfulorrosion effects.

Mode 4 sensitisation was detected irrespective of heat inputnd material transformation temperature. However, low heatnputs and fast cooling rates will reduce the level of sensitisa-ion. In addition, the level of mode 4 sensitisation is much lowerhan that observed for mode 2. As such, mode 4 is only likely toead to service problems in severe corrosion environments, andn conjunction with high stress levels.

cknowledgements

The authors would like to thank Atlas Steels (Australia) Ptytd. for financial and material contributions in support of thisork, and Jim Williams and Frank Barbaro from Bluescope Steeltd. for valuable discussions and background data regarding

heir experience of sensitisation in 12%Cr steels used for railagon manufacture in Australia.

eferences

[1] J.J. Demo, in: D. Peckner, I.M. Bernstein (Eds.), Handbook of StainlessSteels, McGraw-Hill Book Company, 1977, pp. 5.1–5.40.

[

and Engineering A 464 (2007) 157–169 169

[2] Columbus Stainless-3CR12 Technical Manual, April 2001.[3] J.G. Williams, F.J. Barbaro, Proceedings of AISTech 2005, Charlotte, North

Carolina, USA, 2005.[4] L.M. Matthews, et al., Proceedings of the14th International Corrosion

Congress, Cape Town, South Africa, Corrosion Institute of Southern Africa,1999.

[5] L.M. Matthews, et al., Proceedings of the IIW Asian Pacific WeldingCongress, Melbourne, Australia, 2000.

[6] P.G.H. Pistorius, G.T. van Rooyen, Weld. World 36 (6) (1995)65–72.

[7] M.L. Greef, M. Du Toit, The sensitisation of two 11–12% chromium typeEN 1.4003 ferritic stainless steels during continuous cooling after welding,IIW Document No. IX-2182-05, 2005, pp. 1–11.

[8] T.G. Gooch, P. Woollin, A.G. Haynes, Microstructural development onwelding low carbon 13% Cr martensitic steels, IIW Document No. IX-H-449-99, 1999, pp. 1–15.

[9] T.G. Gooch, B.J. Ginn, Weld. Res. Suppl. 11 (1990) S431–S440.10] W. Gordon, A. van Bennekom, Mater. Scie. Technol. (1996) 126–

131.11] H. Folkhard, Welding Metallurgy of Stainless Steels, Springer Verlag,

Vienna, Austria, 1988.12] D.A.A. Smith, Tarboton J.N., Personal Communication, Columbus Stain-

less Pty Ltd., 2005.13] R.H. Kaltenhauser, Met. Eng. Q. 11 (2) (1971) 41–47.14] ASTM A 763-93, Annual Book of ASTM Standards, ASTM, West Con-

shohocken, PA, 1993, pp. 421–431.15] R.G. Campbell, Eng. Mater. 69/70 (1992) 167–216.16] J.J. Demo, Corrosion 27 (12) (1971) 531–544.17] ASTM A 262-98, Annual Book of ASTM Standards, ASTM, West Con-

S280.19] P.C. Pistorius, M. Coetzee, J. South Afr. Inst. Min. Metall. 96 (3) (1996)

119–125.