nhydstremiuermbtait thsistmpe

rotecten incrn occuositionen thaed byrengthompon

centration of hydrogen produced at the surface of the exposed steelsubstrate, with more active metals leading to higher hydrogen con-centrations. The structure of the coating is also a factor as a denselayer acts as an effective barrier, whereas a cracked or porous oneprovides a short circuit path to the steel surface for hydrogen up-take to occur. The effects of composition and structure of different

2.1. Materials

The rst steel was 300 M (UNS K44220) produced by vacuumarc melting. It was tested in the hardened condition by austenitiz-ing at 870 C and oil quenching to room temperature, followed by adouble tempering at 315 C to give a tensile strength of 2250 MPa.The second steel was AerMet100 (UNS K92580), which is a high al-loy steel containing nickel, cobalt and chromium. This was heattreated by austenitizing at 843 C, oil quenching to room tempera-ture followed by 1 h at 73 C to transform retained austenite andnally tempered at 485 C for 5 h. The resulting tensile strength

* Corresponding author. Tel.: +44 01234 758 066; fax: +44 01234 750 875.E-mail address: m.robinson@craneld.ac.uk (M.J. Robinson).

Corrosion Science 52 (2010) 15931602

Contents lists availab

Corrosion

.e l1 Present address: Corus Strip Products UK, South Wales, UK.200 C to remove the hydrogen [1,2]. Failure resulting fromhydrogen uptake during electroplating is referred to as directembrittlement.

Second, hydrogen can also be absorbed when a sacricial coat-ing corrodes in service. In this case, exposed areas of the steel sub-strate act as cathodic sites where hydrogen is generated by thereduction of water. If this hydrogen is absorbed in sufcientquantities it can also lead to failure of high strength steel and thisprocess is termed hydrogen re-embrittlement [3,4].

The composition of the sacricial coating determines the con-

sile tests were performed to measure the extent of embrittlementin each experimental condition and the results are discussed interms of the hydrogen permeation rates in the two materials,determined by two electrochemical techniques. Finally, mathe-matical modelling has been used to relate the uptake of hydrogenin the two steels to the fractography observed in the tensile tests.

2. Methods1. Introduction

When high strength steels are psacricial metallic coating there is ahydrogen embrittlement and this cathe coating is applied by electrodepthe plating current produces hydrogin the coating and then slowly absorbcan lead to delayed failure of high stis the usual practice to bake such c0010-938X/$ - see front matter 2010 Elsevier Ltd. Adoi:10.1016/j.corsci.2010.01.001d from corrosion by aeased risk of failure byr in two ways. First, ifa small proportion oft is reversibly trappedthe steel. This hydrogensteel under load and itents for up to 24 h at

sacricial coatings on direct embrittlement and re-embrittlementof high strength steels have been discussed in a previous paper [4].

The extent of hydrogen embrittlement generally increases withthe tensile strength of the steel [5] but the grain size and thephases present in the microstructure also have an important inu-ence by affecting hydrogen trapping and thereby controlling therate of hydrogen uptake [6]. This paper compares the hydrogenembrittlement and re-embrittlement of two ultra high strengthsteels coated with electroplated cadmium and a non-electrolyticaluminium-based SermeTel 1140/962 coating. Slow strain rate ten-Hydrogen transport and embrittlement istrength steels

D. Figueroa 1, M.J. Robinson *

School of Applied Sciences, Craneld University, Bedford, UK

a r t i c l e i n f o

Article history:Received 24 April 2009Accepted 11 January 2010Available online 20 January 2010

Keywords:A. SteelB. Hydrogen permeationB. GalvanostaticC. Hydrogen embrittlement

a b s t r a c t

This paper describes howand AerMet100 ultra highwith electrodeposited cadmeasured using two-cell p1.0 109 cm2 s1 were omodel was used to predicsuperior embrittlement remartensite laths during te

journal homepage: wwwll rights reserved.rogen transport affects the severity of hydrogen embrittlement in 300 Mngth steels. Slow strain rate tests were carried out on specimens coatedm and aluminium-based SermeTel 1140/962. Hydrogen diffusivities wereeation and galvanostatic charging methods and values of 8.0 108 andned for 300 M and AerMet100, respectively. A two-dimensional diffusione hydrogen distributions in the SSR specimens at the time of failure. Theance of AerMet100 was attributed to reverted austenite forming aroundring.

2010 Elsevier Ltd. All rights reserved.300 M and AerMet100 ultra high

le at ScienceDirect

Science

sevier .com/ locate /corsc i

was 2100 MPa. The chemical compositions of both alloys areshown in Table 1.

The microstructures of the 300 M and AerMet100 steels con-

1594 D. Figueroa, M.J. Robinson / Corrosionsisted of lath/plate martensite with M7C3, M23C6 and e-carbidesdistributed within the matrix and at the boundaries [79]. In300 M, examination in a SEM showed that some retained austenitewas present at prior austenite grain boundaries. Retained austenitewas also visible in AerMet100 and, in addition, a thin lm of re-verted austenite had formed at lath/plate boundaries during the -nal tempering treatment. In each case, the quantity of austenitewas too low to be detected by X-ray diffraction. The same micro-structure has been reported in the literature for AerMet100, fol-lowing similar heat treatments [10,11]. The carbides, and theretained and reverted austenite are all known to act as stronghydrogen traps [8,12] and their signicance in the microstructurewill be discussed in a later section.

2.2. Hydrogen diffusivities

Two methods were used to measure the hydrogen diffusioncoefcient, D, of the two steels. The hydrogen permeation tech-nique was suitable for 300 M steel but was insufciently sensitiveto measure the lower diffusion coefcient of AerMet100. Instead, agalvanostatic charging technique was adopted, which was satisfac-tory for both materials.

2.2.1. Hydrogen permeation techniqueThe measurements were made using the double cell permeation

technique of Devanathan and Stachurski [13]. A membrane of thesteel, 1 mm thick was polished to a 1 lm diamond nish and heldbetween two electrochemical cells with an exposed surface area of50.3 cm2 on each side. Both cells contained an electrolyte of 0.2 MNaOH solution. Prior to hydrogen charging, the membrane was de-pleted of residual hydrogen by holding the exit surface at a poten-tial of 0.094 V (SCE) until the current density was lowered to0.01 lA cm2. Charging was then carried out on one side only bymaintaining the entry surface at a constant potential of 1.250 V(SCE) to promote hydrogen generation and uptake. The hydrogenthat diffused through the membrane was oxidised and detectedon the exit side. After an interval known as the breakthrough time,tb, the oxidation current displayed a rising transient and eventuallyreached a steady state value, J1.

J1 DCoL

1

Where the D is the hydrogen diffusion coefcient, Co is the con-stant hydrogen concentration at the charging side of the mem-brane, and L is the membrane thickness. The aim of theexperiment was to measure the hydrogen diffusion coefcient ofthe steel, which was calculated from the breakthrough time usingEq. (2).

tb L2

15:3D2

In this case, the Co value in the charging side of the membranewas not of interest as it was not required to determine D. Using apotential of 1.250 volts (SCE) and 0.2 M NaOH electrolyte gave ahigh rate of hydrogen charging and avoided corrosion on the entryside of the steel membrane.

Table 1Compositions (wt.%) of 300 M and AerMet100 high strength steels.

Alloy Fe C Ni Cr Si Mo V Nb Co Mn300 M bal 0.4 1.8 0.8 1.6 0.25 0.7AerMet100 bal 0.23 11.5 3.1 13.5 2.2.2. Galvanostatic methodThe electrochemical principle of the method has been clearly

reviewed by Wen et al. [14] and its application is reported in sev-eral other studies [1517]. The technique involves using a currentpulse to cathodically charge a thin sample of the material fromboth sides in a two-electrode cell and record the cell potential dif-ference, E, as a function of time. The charging current produces aconstant ux of hydrogen at the electrolyte/metal interface, x = 0,and the increasing cell voltage that is recorded provides a directmeasure of the hydrogen activity in the steel at the interface. Un-der these experimental conditions the initial and boundary condi-tions for a sample of thickness 2L, charged from both sides, are:

C C0 0 6 x 6 L; t 0 3

DdCdx I

zFSx 0; t > 0 4

dCdx 0 x L; t P 0 5

where C, t, D and x are the molar hydrogen concentration, diffusiontime, hydrogen diffusion coefcient and distance from the metalsurface, respectively, and I, S, z and F are the charging current, sam-ple area, charge number of the ion reduced in the electrolyte (i.e.H+) and Faradays constant.

The solution of Ficks second law for one dimensional diffusioncan be expressed as either error function or trigonometric series. Itcan be shown that for very short times, (t L2/D), the concentra-tion of hydrogen at the interface (as measured by the cell potentialdifference) can be approximated by the rst term of the error func-tion series. In this condition, the potential difference, E, is linearlyrelated to t1/2, as shown in Eq. (6). The gradient dE/dt1/2 can be ob-tained from the gradient of the graph of E vs. t1/2.

dEdt1=2

2IVmzFSpD1=2

dEdd

if t L

2

D6

where Vm is the molar volume of the steel sample and dE/dd is thechange in cell voltage with change in hydrogen concentration.

At longer times, (t > L2/D), the surface hydrogen concentration isapproximated by the rst two terms of the innite trigonometricseries and dE/dt is linearly related to the diffusion time, as givenin Eq. (7).

dEdt

IVm

FSLdEdd

if t >

L2

D7

Eqs. (6) and (7) can be combined to eliminate dE/dd and the dif-fusion coefcient is then calculated from the gradients dE/dt1/2 anddE/dt at short and long times, respectively.

D 2LdEdt

p1=2 dEdt1=2

!28

In the practical measurements, the 300 M steel sample was2 mm thick (i.e. 2L = 2 mm) with a charging area of 1.44 cm2 andthe AerMet100 sample was 1 mm thick (2L = 1 mm) with an areaof 2.64 cm2. Both were polished to a 1 lm nish with diamondpaste and the edges were protected with Lacomit. In each case, aplatinum counter electrode was used and the current densitywas 1 mA cm2. The electrolyte was 0.2 M NaOH, with the additionof 1 g l1 of thiourea as a recombination poison to ensure thathydrogen generated at the surface entered the metal rather thanescaping as bubbles.

2.3. Hydrogen embrittlement tests

Science 52 (2010) 15931602The two high strength steels were machined in the form of ten-sile specimens, with a gauge length of 25 mm and diameter of

NaCl solution to simulate service conditions and establish represen-

specimen not failing within time, t, is given by;

s s i

In this case, Eq. (10) can be modied to include ti.

Ps 1 Pf extti 11

3.2.2. Student t-testIn addition to the Weibull model, the Student t-test [21] for

small sample sizes was used to determine whether a set of failuretime data, from the four replicate specimens, was signicantly dif-ferent from other sets of data measured under different experi-mental conditions, or whether they should be considered to bepart of the same overall population.

The test compared the means of the two data sets,m1 andm2, tocalculate the test statistic, t, as follows;

t m1 m2s p 1=n1 1=n2 12

where n1 and n2 were the number of failure times in each data set(in this case n1 = n2 = 4) and s was found from the standard devia-tions for the two sets.

Fig. 1. Hydrogen permeation transient for 300 M steel charged at 1.25 V (SCE) in

osiontative surface hydrogen concentrations. In these tests the gaugelength was scribed to remove the coating and expose the substrateover a width of 1 mm. Similar tests were performed on specimenscoated with SermeTel 1140/962 to measure the re-embrittlementcaused by corrosion of this aluminium-based coating.

2.3.1. Embrittlement indexIn each case, the time to failure of a coated tensile specimen (tc)

was recorded and compared with the mean failure time (ta) of un-coated specimens tested in air. The embrittlement index (EI) wascalculated as follows;

EI 1 tc=ta 9

3. Results

3.1. Hydrogen diffusivities

A typical transient for 300 M from the two-cell permeationtechnique is shown in Fig. 1. The mean breakthrough time for threereplicate experiments was 1.9 0.06 h and this corresponded to ahydrogen diffusion coefcient of 9.6 0.3 108 cm2 s1 from Eq.(2). It was evident that AerMet100 had a much lower diffusivityas the current transient was very small and the breakthrough timeand steady state current were difcult to detect (Eq. (1)).

The results of the galvanostatic method for both 300 M and Aer-Met100 are shown in Fig. 2. The plot of cell potential differenceagainst time displays a parabolic shape for short charging times,t L2/D, followed by a linear region at longer times when t > L2/D. The corresponding graphs of E vs. t1/2 and E vs. t are shown inFigs. 3 and 4, respectively and their gradients were used to calcu-late the hydrogen diffusion coefcients from Eq. (8).

The hydrogen diffusion coefcients for 300 M and AerMet100steels were 8.0 108 and 1.0 109 cm2 s1 respectively andthe results for the two methods are summarised in Table 2.

3.2. Hydrogen embrittlement testing2.75 mm. Some of the tensile specimens were coated with 8 lm ofelectrodeposited cadmium and others were coated with SermeTel1140/962 by Sermatech International Inc in Lincoln, UK. This coat-ing consisted of densely packed aluminium particles in a chro-mate/phosphate inorganic binder and was applied as two spraycoats, each being cured at 315 C to give a total thickness of70 lm. As the process was non-electrolytic it was not expectedthat these coatings would cause direct embrittlement of the steelsubstrates. A modied polyurethane top-coat was applied to en-hance the corrosion resistance and barrier properties of thecoating.

The extent of direct hydrogen embrittlement and re-embrittle-ment in the coated specimens was assessed by performing slowstrain rate (SSR) tests. Initial tests were carried out in air on un-coated specimens to provide a baseline for comparison. The strainrate was 9.7 107 s1 and four specimens were tested for eachexperimental condition. Some of the cadmium plated specimenswere tested in the as-plated condition. (They were stored in liquidnitrogen within 8 h of plating to immobilize hydrogen in the cad-mium and the steel until the test could be carried out). Other platedspecimens were tested after being given a de-embrittlement treat-ment by baking for 24 h at 200 C. Further tests investigated the ef-fects of corrosion of the cadmium coating by surrounding the gaugelength of plated and baked specimens with a cell containing 3.5%

D. Figueroa, M.J. Robinson / CorrHydrogen embrittlement tests on high strength steel often dis-play a range of times to failure when carried out under apparentlyPs 1 Pf ext 10Where Pf is the probability of failure and x is a shape parameter

termed the Weibull slope [1820]. In the present context, x repre-sents the probability per unit time that during time, t, a crack willform in the specimen of sufcient size to cause failure. The value ofx can be obtained from the negative gradient of a graph of ln Psagainst t. In practice, there is a minimum time, ti, for a crack to ini-tiate and the graph is displaced along the time axis by this amount.(When P = 1, then ln P = 0 and t = t ).identical experimental conditions. This is due to differences in thenumber, size and distribution of microstructural aws within thespecimen at which hydrogen accumulates and leads to fracture. Inaddition, aws and lack of uniformity in the sacricial coating willlead to differences in the rate of hydrogen uptake by the steel andthis too will affect the failure time. In this work, Weibull statisticswas used to display the probability of each of the four replicatespecimens in each group surviving for a specic time before thenfailing by hydrogen embrittlement or re-embrittlement.

3.2.1. Weibull model of failure timesIt can be shown that for a Weibull model the probability, Ps, of a

0.2 M NaOH solution.0.3

0

0.1

0.2

0 8 16 24 32 40Time [h]

I [uA/

cm2 ]

300M steel

Science 52 (2010) 15931602 1595s2 s21 s222

13

300M

AerMet100

sion Science 52 (2010) 15931602-1

0

renc

e [V

]

1596 D. Figueroa, M.J. Robinson / CorroThe test statistic was then compared with tabulated values of at-distribution for a 95% condence level with 6 degrees of freedom(n1 + n2 2). If the test static was less than the t-distribution thenit could be concluded with this condence level that the two datasets were part of the same overall population (the null hypothesis).

3.2.3. Slow strain rate testsThe slow strain rate test results for 300 M and AerMet100 spec-

imens, tested in air and when cadmium coated and exposed to 3.5%

Time [s]

-4

-3

-2

0 5 x 103 1.0 x 104 2.0 x 104 2.5 x 104 3.0 x 104

Cell P

oten

tial D

iffe

1.5 x 104

Fig. 2. Graphs of cell potential difference vs. time obtained

EA100= -0.0042(t1/2) - 3.103

E 300M= -0.0008(t1/2)- 1.099

-4

-3

-2

-1

0

0 10 20 30 40 50 60

Time1/2 [s1/2]

Cell

Pote

ntia

l Diff

eren

ce [V]

300M A100

Fig. 3. Linear plots of cell potential difference against (time)1/2 for 300 M andAerMet100 over short charging times (t L2/D).

E300M= -1x10-6 t - 1.134

EA100 = -8x10-7t - 3.397

-4

-3

-2

-1

0

0 2.0 x 104 4.0 x 104 6.0 x 104 8.0 x 104

Time [s]

Cell P

oten

tial D

iffer

ence

[V]

300MA100

Fig. 4. Linear plots of cell potential difference against time for 300 M andAerMet100 at long charging times (t > L2/D).NaCl solution, are shown in Figs. 5 and 6. The mean failure timesfor the four replicate specimens in each condition and the corre-sponding embrittlement indices are compared in Table 3. 300 M

for 300 M and AerMet100 in the galvanostatic method.

Table 2Comparison of hydrogen diffusion coefcients for 300 M and AerMet100 steelsmeasured by the permeation and galvanostatic methods.

Steel Diffusion coefcient (cm2 s1)

Permeation method Galvanostatic method

300 M 9.6 108 8.0 108AerMet100 1.0 109displayed much more direct embrittlement than AerMet100 andwhen electroplated with cadmium its embrittlement index, EI,was 0.54, compared to only 0.05 for AerMet100. Post-plating bak-ing the cadmium plated specimens at 200 C was effective inremoving hydrogen from the microstructure and the mechanicalproperties were almost completely recovered to give an EI for300 M of 0.04. Corrosion of the cadmium coating in 3.5% NaCl solu-tion caused signicant re-embrittlement of 300 M and an EI of0.16. However, AerMet100 displayed no re-embrittlement in 3.5%NaCl solution. Its lower susceptibility was attributed to the distri-bution of phases in its microstructure, as discussed in a latersection.

-2

-1.5

-1

-0.5

0

0 5 10 15 20 25 30 35

Time to Failure [h]

Ln P

s

Uncoated

Cadmium Plated

Cadmium Baked

Cadmium Baked/Corroded

Plated/

SermeTel Corroded

Fig. 5. Weibull plots for 300 M showing the effects of direct embrittlement fromcadmium plating and re-embrittlement from corrosion of cadmium and SermeTel1140/962 coatings.

concentration of absorbed hydrogen at the specimen surface must

-1

-0.5

0

Ln P

s

Uncoated

Cadmium Plated

D. Figueroa, M.J. Robinson / CorrosionThe results for the re-embrittlement tests on the aluminium-based SermeTel 1140/962 coated specimens are also shown in Figs.5 and 6. The open circuit potential of SermeTel 1140/962 in 3.5%NaCl solution (0.76 V (SCE)) was very close to that of electrode-posited cadmium (0.75 V (SCE)) yet it led to greater re-embrittle-ment of 300 M, with an EI of 0.27. The structure of the coating isbelieved to be an important factor affecting hydrogen uptake, inaddition to its electrochemical potential [4]. In contrast, Aer-Met100 specimens showed no reduction in failure times whentested in 3.5% NaCl solution. In fact, there was a small increase inthe mean failure time, as shown in Table 3, which resulted in anegative value for the calculated embrittlement index. It is possiblethat the baking treatment at 315 C to cure the coating may haveslightly affected the mechanical properties of the steel. However,this temperature was well below the tempering temperature ofthe material (485 C for 5 h) and the Student-t test showed, witha 95% condence level, that the coated and uncoated AerMet100specimens could not be considered to belong to different statisticalpopulations.

3.3. Appearance of the fracture surfaces

-2

-1.5

0 5 10 15 20 25 30 35 40Time to Failure [h]

Cadmium Baked

Cadmium Plated/Baked/CorrodedSermeTel Corroded

Fig. 6. Weibull plots for AerMet100 showing low levels of direct embrittlementfrom cadmium plating and re-embrittlement from corrosion of cadmium andSermeTel 1140/962 coatings.After the SSR tests, the fracture surfaces were examined in ascanning electron microscope. The different failure modes werecarefully mapped across the surfaces and are shown schematicallyfor each material in Figs. 7 and 8. The 300 M specimens had verylittle reduction in area due to their high level of embrittlement,whereas the AerMet100 fractures were more ductile and neckinghad occurred. In 300 M, a brittle fracture had initiated at a singlepoint at the edge of each specimen and propagated inwards untilthe stress on the remaining section exceeded the UTS, whereuponit then failed in a ductile manner by microvoid coalescence. The

Table 3Mean failure times and embrittlement indices (EI) in the slow strain rate test for300 M and AerMet100 steels in the uncoated condition and after Cd plating, bakingand corroding and after coating with SermeTel 1140/962 and corroding.

Mean failure times (h) & embrittlement indices

Uncoated Cdplated

Cdbaked

Cdcorroded

Alcorroded

300 M (EI) 31.6 14.5 30.2 26.6 23.2 (0.54) (0.04) (0.16) (0.27)

AerMet100(EI)

32.9 31.1 32.0 33.8 33.9 (0.05) (0.03) (0.03) (0.03)both exceed threshold values for a brittle crack to initiate [23,24].While the stress increased slowly during the SSR test, hydrogenwas diffusing into the specimen from the reservoir of hydrogenin the unbaked cadmium coating [25]. Eventually the stress ex-ceeded the threshold value and a brittle crack initiated. The crackthen propagated rapidly towards the centre of the specimen, asthe material in this region contained more than the thresholdhydrogen concentration. When the stress on the remaining cross-section reached the UTS, a ductile overload failure occurred.

Clearly, the depth of the brittle region observed on the fracturesurface is a measure of the depth to which hydrogen had diffusedin that specimen. Therefore, by determining this distance it is fea-sible to model hydrogen diffusion across the section and to calcu-late the corresponding diffusion coefcients [26] which can thenbe compared with those obtained from the electrochemical mea-surements. In practice, some hydrogen would have diffused furtherthan the embrittled region, as the concentration at the crack tip isrequired to reach the threshold value. Therefore, the diffusion coef-cients assessed from the fracture surfaces should be regarded asminimum values. This model does not preclude any specic cracktip mechanism for hydrogen embrittlement [27]. For example, theslip softening mechanism [28] can explain how, in certain circum-stances, crack growth can be accompanied by the enhanced gener-ation and mobility of dislocations and an observed lowering of thestress required for plastic deformation. However, such crack prop-agation would still require absorbed hydrogen to have diffused tosufcient depth below the surface and for a threshold concentra-tion to have been exceeded.

4.2. Modelling hydrogen uptake by the SSR specimens

The case of two-dimensional, radial diffusion in a cylinder hasbeen considered. The rate of hydrogen uptake at the surface, F, isdependent on the hydrogen diffusion coefcient of the steel, D,and the concentration gradient, dC/dx, according to Ficks rstlaw of diffusion [29].

F DdC=dx 14Before the start of the test the residual hydrogen concentration

in the specimen, C1, was assumed to be zero and as soon ashydrogen charging began the surface concentration took on a va-lue, C0, and remained constant. These boundary conditions, whichapply for the case of potentiostatic charging [13] are dened be-low, where r is the distance from the centreline of a specimen ofradius a.mean depth of brittle fracture was 1.0 mm. The initial brittle frac-ture was intergranular, suggesting that the applied stress had beenrelatively low [22], and this changed to a transgranular cleavagemode as the crack grew in length and the stress increased(Fig. 9). In contrast, all the regions of brittle fracture in AerMet100specimens were transgranular cleavage. They had initiated frommany sites around the circumference but had propagated only ashort distance (mean depth of 0.13 mm) before the nal ductilefailure occurred.

4. Discussion

4.1. Conditions for embrittlement in SSR tests

For hydrogen embrittlement failures to occur in SSR tests, twoprincipal conditions need to be met. The applied stress and the

Science 52 (2010) 15931602 1597C C1 0; 0 < r < a; t 0 15C C0; r a; t P 0 16

sion1598 D. Figueroa, M.J. Robinson / CorroAs the test proceeded, the hydrogen content of the steel in-creased such that the concentration gradient (C0C) was progres-sively reduced and the rate of hydrogen uptake decreased also.

Fig. 7. Schematic illustration of different failure modes on the f

Fig. 8. Schematic illustration of different failure modes on the fraScience 52 (2010) 15931602The solution of the diffusion equation for the case of hydrogentransport into a solid cylindrical specimen [29,30], of innitelength, is given in Eq. (17).

racture surfaces of four replicate 300 M tensile specimens.

cture surfaces of four replicate AerMet100 tensile specimens.

osion Science 52 (2010) 15931602 1599D. Figueroa, M.J. Robinson / CorrC C1C0 C1 1 2

X1n1

expa2nDt=a2J0ran=aanJ1an

17

where an, n = 1, 2, . . .. . . are roots of J0(a) = 0, the Bessel function ofzero order and J1(a) is the Bessel function of the rst order. The ser-ies is convergent and was evaluated for the rst 150 Bessel roots.For the initial condition, C1 = 0, the equation simply gives the con-centration C/C0.

4.3. Minimum rates of hydrogen uptake

Fig. 10 shows the least depth to which hydrogen could havepenetrated into the cadmium plated, but unbaked, 300 M SSR spec-imens in the mean failure time of 14.5 h, to explain the observedfailures. The graph represents the hydrogen concentration, C, as aproportion of the surface concentration, Co, across the 2.75 mmdiameter of the specimen, when charged from both sides. In eval-uating the series in Eq. (17), the hydrogen diffusion coefcient hasbeen adjusted so that hydrogen just reached a depth of 1.0 mm,which was the mean depth of the embrittled region on the fourreplicate 300 M specimens. For this condition to be satised, thediffusion coefcient would need to be 1 108 cm2 s1, comparedto 8.0 108 cm2 s1 measured using the galvanostatic method.

Similar calculations were carried out for AerMet100 such thathydrogen just penetrated 0.13 mm from the surface; the meandepth of the embrittled region in cadmium plated specimens of

Fig. 9. SEM Micrographs showing fracture surfaces of cadmium plated slow strain raboundaries, (B) transgranular fracture of 300 M, (C) transgranular fracture of AerMet100this alloy, as shown in Fig. 11. In this case, D needed to be8 1011 cm2 s1, whereas the measured value was 1 109 cm2 s1. As the measured diffusion coefcients were the high-er values in both steels, it appears that at the time of failure hydro-

te specimens. (A) brittle intergranular fracture of 300 M at prior austenite grain.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.138-0.138 0Distance from centreline (cm)

C/C o

52200Time (s)

-0.038

1.0 mm

Fig. 10. Hydrogen distribution in 300 M at the failure time of 14.5 h for a diffusivity,D, of 1 108 cm2 s1.

hydrogen had just penetrated from each side of the specimen toreach the centre (the breakthrough time tb = 0.04L2/D). At failure

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

831.00831.0-

Distance from centreline (cm)

C/C o

111960

Time (s)

-0.125

0.13 mm 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

831.00831.0-

Distance from centreline (cm)

C/C o

111960

Time (s)

Cth

-0.125

Fig. 13. Predicted hydrogen distribution in AerMet100 at the failure time of 31.1 hfor the measured diffusivity, D, of 1 109 cm2 s1.

1600 D. Figueroa, M.J. Robinson / Corrosion Science 52 (2010) 15931602(52200s), the hydrogen concentration, C/Co, at a depth of 1.0 mmwas predicted to be 0.6. That is to say, the concentration was 0.6of that at the surface due to hydrogen charging from the cadmiumcoating. This value represents the threshold hydrogen concentra-gen had diffused further than the embrittled region, indicating thata signicant hydrogen concentration was required at the crack tipfor the brittle fracture to propagate.

4.4. Threshold hydrogen concentrations

Figs. 12 and 13 show the hydrogen distribution for each alloythat would be predicted from its failure time and measured diffu-sion coefcient. For 300 M (Fig. 12) distributions are shown for arange of times up until failure occurred. For example, at 10440s

Fig. 11. Hydrogen distribution in AerMet100 at the failure time of 31.1 h for adiffusivity, D, of 8 1011 cm2 s1.tion for embrittlement in 300 M.The corresponding hydrogen distribution for AerMet100 at the

failure time of 111960s is shown in Fig. 13. At the mean embrittle-

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0831.0-

Distance from cen

C/C o

Cth

-0.038

Fig. 12. Predicted hydrogen distribution in 300 M at the failure timment depth of 0.13 mm, the hydrogen concentration is predicted tobe 0.4 C/Co.

4.5. Comparison of the embrittlement susceptibilities

The SSRT results indicate that 300 M was the more susceptibleof the two alloys to direct hydrogen embrittlement from cadmiumplating and re-embrittlement from hydrogen generated when thecadmium and aluminium-based sacricial coatings were corroded.The hydrogen diffusion coefcient was shown to be approximately80 times higher in 300 M than in AerMet100 and it appears thatthe faster rate of hydrogen uptake in 300 M was a major contribu-tory factor in its greater susceptibility in the SSR tests. WhetherAerMet100 would display greater embrittlement if more timewas available for hydrogen uptake is an interesting point to con-sider and this could be investigated by performing tests at a slowerstrain rate to allow more time for diffusion.

A comparison of the threshold hydrogen concentrations, Cth, re-quired to propagate a brittle crack (Figs. 12 and 13) might suggestthat AerMet100 was slightly more susceptible than 300 M (C/Co of0.4 for AerMet100 and 0.6 for 300 M). However, this difference isnot thought to be particularly signicant. Both steels have tem-pered martensite in the microstructure and embrittlement of thisphase is expected to be similar in each case. In addition, due to831.0

treline (cm)

52226101044020880313204176052200

Time (s)

e of 14.5 h for the measured diffusivity, D, of 8 108 cm2 s1.

received equivalent heat treatments to that used in the presentstudy [8,12]. Ritchie et al. [31] showed that the presence of re-

of the SermeTel 1140/962 resulted in more hydrogen penetrating

bou

osiontained austenite was benecial in controlling hydrogen embrittle-ment and attributed this to a reduction in the diffusivity ofhydrogen due to its high solubility in austenite. In the case of Aer-Met100, the austenite was thought not to have been retained dur-ing quenching but to have formed by partial reversion ofmartensite during tempering for 5 h at 485 C [32,33]. Austenite,is known to have a high hydrogen solubility and low hydrogen dif-the steepness of the hydrogen distribution prole in Fig. 13, a smalldifference in the measured depth of embrittlement in AerMet100would result in a large difference in the predicted C/Co value. How-ever, there are features of the microstructure of AerMet100 thatexplain both its low embrittlement susceptibility and the lowhydrogen diffusion coefcient.

4.6. The inuence of microstructure

Both steels consisted of tempered martensite and the markeddifference in the two hydrogen diffusivities is thought to resultfrom the concentration and distribution of trap sites in their micro-structures. In particular, the presence of retained and reverted aus-tenite is an important factor [4]. Fig. 14 shows that in 300 Mretained austenite was present at prior austenite grain boundaries,whereas in AerMet100 the austenite was distributed more widely,on both prior austenite boundaries and martensite lath boundaries.Previous work reported similar microstructures in 300 M that had

Fig. 14. SEM micrographs showing (A) retained austenite at prior austenite grainaustenite boundaries (3) in AerMet 100.

D. Figueroa, M.J. Robinson / Corrfusivity [34] and therefore acts as an effective trap when present inan essentially martensitic microstructure. The hydrogen diffusioncoefcient of austenitic stainless steel is reported [35] to be inthe range 1.8 8 1012 cm2 s1 and, although these alloys con-tain chromium and nickel, this indicates the low rate of diffusionthat can be expected in high strength steels when austenite is pres-ent. Similarly, it has been reported that the apparent diffusivity ofduplex stainless steel containing 44% austenite is reduced byapproximately 400 times in comparison with that of a fully ferriticsteel, due to hydrogen trapping in the austenite phase [35]. In300 M, austenite retained at prior austenite grain boundarieswould have trapped some hydrogen but this would have providedlimited benet as the martensite laths within the grains wouldhave remained vulnerable to embrittlement. In AerMet100, how-ever, the presence of a thin (3 nm) but essentially continuous enve-lope of reverted austenite around the martensite laths has beenreported [36] and this lowered the risk of embrittlement in thepresent work by acting as a very effective barrier to restrict hydro-gen transport and prevent a critical hydrogen concentration beingreached in the more susceptible martensite phase. The factorsthis coating than passed through the cadmium. This led to a higherhydrogen concentration at the SermeTel/steel interface, with aaffecting hydrogen uptake and embrittlement have been discussedin more detail in a previous paper [4].

4.7. Effect of coating type on hydrogen re-embrittlement

It was shown in Table 3 that less re-embrittlement occurredfrom corrosion of the cadmium coating (EI 0.16) than from thatof the SermeTel 1140/962 coating (EI 0.27), although they had verysimilar open circuit potentials in 3.5% NaCl solution. Generally, amore active coating would be expected to cause more hydrogenuptake by the steel and a higher level of re-embrittlement. Theinuence of coating type on hydrogen re-embrittlement has beendiscussed in detail in a previous paper [4] and the structure ofthe coating has been shown to have an important effect, in additionto its electrochemical potential. Etching the coating surfaces in afocused beam of gallium ions (FIB) and examining the resultingtrough in a scanning electron microscope revealed the cadmiumto have a dense structure, giving better barrier properties. In con-trast, the SermeTel 1140/962 coating contained extensive nepores that are thought to have formed during drying of the inor-ganic binder in the curing treatment. The pores were mostly closedand, while undesirable, they were less damaging than if they hadbeen through-thickness aws, which would have left areas of thesteel substrate exposed to act as sites where cathodic hydrogencharging would have occurred more readily. The greater porosity

ndaries in 300 M and (B) reverted austenite at lath boundaries (1, 2) and at prior

Science 52 (2010) 15931602 1601higher rate of hydrogen uptake by the specimen and more re-embrittlement. In general terms, the effect of the coating is todetermine the hydrogen concentration at the steel surface andthe extent of hydrogen uptake and diffusion in the substrate isthen the consequence of this concentration.

5. Conclusions

[1] 300 M displayed signicant direct embrittlement by hydro-gen uptake from electrodeposited cadmium coatings andre-embrittlement from the corrosion of both cadmium andaluminium-based SermeTel 1140/962 coatings. In contrast,AerMet100 showed very little embrittlement in any of theseconditions.

[2] The relative embrittlement susceptibilities of the two steelscan be explained by their different hydrogen diffusivities,which were measured by the galvanic charging method tobe 8.0 108 and 1.0 109cm2 s1 for 300 M and Aer-Met100, respectively.

[3] The threshold hydrogen concentrations, Cth, for propagationof brittle cracks in the two steels were estimated to be in therange 0.40.6 of the surface concentration produced byunbaked electrodeposited cadmium.

[4] The greater embrittlement resistance of AerMet100 is attrib-uted to the low rate of hydrogen transport in the revertedaustenite phase that forms around martensite laths in itsmicrostructure during tempering. The austenite effectivelytraps absorbed hydrogen, preventing it being taken up bythe more susceptible martensite phase.

Acknowledgement

The research described in this paper was supported by AirbusUK. The authors acknowledge the assistance of Dr. A. Skordos withthe modelling of hydrogen diffusion.

[15] P.A. Sundaram, D.K. Marble, Hydrogen diffusivity in AerMet100 at roomtemperature under galvanostatic charging conditions, Journal of Alloys andCompounds 360 (1-2) (2003) 9097.

[16] P.A. Sundaram, E. Wessel, H. Clemens, H. Kestler, P.J. Ennis, W.J. Quadakkers, L.Singheiser, Determination of the diffusion coefcient of hydrogen in gammatitanium aluminides during electrolytic charging, Acta Materialia 48 (2000)10051019.

[17] L.B. Anderson, D.J. Macero, Generalised chronopotentiometric potential-timeequations and their application to quasi-reversible processes, AnalyticalChemistry 37 (3) (1965) 322326.

[18] M.J. Robinson, R.M. Sharp, The effect of post-exposure heat treatment on thehydrogen embrittlement of high strength steel, Corrosion 41 (10) (1985) 582586.

[19] T. Yokobori, in: P. Noordhoff (Ed.), Strength Fracture and Fatigue of Materials,Groningen, 1965, pp. 2228.

[20] E.M.K. Hiller, M.J. Robinson, Hydrogen embrittlement of high strength steelelectroplated with zinccobalt alloys, Corrosion Science 46 (2003) 715727.

[21] C. Chateld, Statistics for technology a course in applied statistics, third ed.,Chapman & Hall, 1983.

[22] C.D. Beachem, A new model for hydrogen assisted cracking (hydrogen

1602 D. Figueroa, M.J. Robinson / Corrosion Science 52 (2010) 15931602References

[1] Defence Standard 03-4/2, The pre-treatment and protection of steel parts ofspecied maximum tensile strength exceeding 1450 N mm2, 1977.

[2] E. Kossoya, Y. Khoptiara, C. Cytermannb, G. Shemesha, H. Katza, H. Sheinkopfa,I. Cohena, N. Eliaz, The use of SIMS in quality control and failure analysis ofelectrodeposited items inspected for hydrogen effects, Corrosion Science 50(2008) 14811491.

[3] G. Chalaftris, M.J. Robinson, Hydrogen re-embrittlement of high strength steelby corrosion of cadmium and aluminium based sacricial coatings, CorrosionEngineering Science and Technology 40 (1) (2005) 2832.

[4] D. Figueroa, M.J. Robinson, The effects of sacricial coatings on hydrogenembrittlement and re-embrittlement of ultra high strength steels, CorrosionScience 50 (2008) 10661079.

[5] D. Hardie, E.A. Charles, A.H. Lopez, Hydrogen embrittlement of high strengthpipeline steels, Corrosion Science 48 (2006) 43784385.

[6] L.W. Tsay, H.L. Lu, C. Chen, The effect of grain size and aging on hydrogenembrittlement of a maraging steel, Corrosion Science 50 (2008) 25062511.

[7] J.L. Youngblood, M. Raghavan, Correlation of microstructure with mechanicalproperties of 300 M steel, Metallurgical Transactions A 8A (1977) 14391448.

[8] R. Padmanabhan, W.E. Wood, Microstructural analysis of a multistage heat-treated ultra-high strength low alloy steel, Materials Science and Engineering66 (1984) 125143.

[9] B.G. Pound, Hydrogen trapping in high strength steels, Acta Materials 46 (16)(1998) 57335743.

[10] C.H. Yoo, H.M. Lee, J.W. Chan, J.W. Morris, M2C precipitates in isothermaltempering of high CoNi secondary hardening steel, Metallurgical &Metallurgical Transactions A 27A (1996) 34663472.

[11] R. Ayer, P.M. Machmeier, Transmission electron microscopy examination ofhardening and toughening phenomena in AerMet100, MetallurgicalTransactions A 24A (1993) 19431955.

[12] P. Padmanabhan, W.E. Wood, Stress corrosion cracking behaviour of 300Msteel under different heat treatment conditions, Corrosion 41 (12) (1985) 688699.

[13] M.A.V. Devanathan, Z. Stachurski, The absorption and diffusion of electrolytichydrogen in palladium, Proceeding of Royal Society A270 (1962) 90102.

[14] C.J. Wen, C. Ho, B.A. Boukamp, I.D. Raistrick, W. Weppner, R.A. Huggins, Use ofelectrochemical methods to determine chemical-diffusion coefcients inalloys: application to LiAl, International Metals Reviews 5 (1981) 253268.embrittlement), Metallurgical Transactions 3 (1972) 437450.[23] M.J. Robinson, P.J. Kilgallon, Hydrogen embrittlement of cathodically protected

high strength, low alloy steels exposed to sulphate-reducing bacteria,Corrosion 50 (8) (1994) 626635.

[24] R.P. Gangloff, A review and analysis of the threshold for hydrogen environmentembrittlement of steel, in Corrosion Prevention and Control, in: M. Levy, S.Isserow (Eds.), US Army Sagamore Materials Research Conference, vol. 33,Watertown, MA, US Laboratory Command, 1986, p. 64.

[25 ] H.E. Townsend, A study of the entry and removal of hydrogen during coatingand thermal treatment of steel, Corrosion 37 (2) (1981) 115119.

[26] A. Duval, M.J. Robinson, Measurement and prediction of hydrogenembrittlement in high strength carbon steel, Corrosion Engineering, Scienceand Technology 44 (5) (2009) 340346.

[27] A.W. Thompson, I.M. Bernstein, The role of metallurgical variables inhydrogen-assisted environmental fracture, in: M.G. Fontana, R.W. Staehle(Eds.), Advances in Corrosion Science and Technology, Vol. 7, Plenum Press,1980, pp. 53175.

[28] C.D. Beachem, A new model for hydrogen-assisted cracking (hydrogenembrittlement), Metallurgical Transactions 3 (February) (1972) 437450.

[29] J. Crank, The Mathematics of Diffusion, Second ed., Oxford University Press,1975. p73.

[30] H.S. Carslaw, J.C. Jaeger, Conduction of Heat in Solids, Second ed., OxfordUniversity Press, 1959. p199.

[31] R.O. Ritchie, M.O. Castro Cedeno, V.F. Zackay, E.R. Parker, The effect of siliconadditions and retained austenite on stress corrosion cracking in ultra highstrength steels, Metallurgical Transactions A 9A (1978) 3540.

[32] R.L.S. Thomas, D. Li, et al., Trap-governed hydrogen diffusivity and uptakecapacity in ultra high-strength AerMet

100 steel, Metallurgical and Materials

Transactions 33A (2002) 19912004.[33] R.L.S. Thomas, J.R. Scully, et al., Internal hydrogen embrittlement of ultrahigh-

strength AerMet100 steel, Metallurgical and Materials Transactions A 34A (2)

(2003) 327344.[34] M.J. Robinson, P.J. Kilgallon, M. Newman, Hydrogen embrittlement of

cathodically protected duplex stainless steels in seawater containingsulphides, in: Proceedings 15th International Conference on OffshoreMechanics & Arctic Engineering, Florence, Italy, June 1620, 1996.

[35] E. Owczarek, T. Zakroczymski, Hydrogen transport in a duplex stainless steel,Acta Materialia 48 (2000) 30593070.

[36] Raghavan Ayar, P.M. Machmeier, Transmission electron microscopyexamination of hardening and toughening phenomena in AerMet100,Metallurgical Transactions A 24A (1993) 19441955.

Hydrogen transport and embrittlement in 300M and AerMet100 ultra high strength steelsIntroductionMethodsMaterialsHydrogen diffusivitiesHydrogen permeation techniqueGalvanostatic method

Hydrogen embrittlement testsEmbrittlement index

ResultsHydrogen diffusivitiesHydrogen embrittlement testingWeibull model of failure timesStudent t-testSlow strain rate tests

Appearance of the fracture surfaces

DiscussionConditions for embrittlement in SSR testsModelling hydrogen uptake by the SSR specimensMinimum rates of hydrogen uptakeThreshold hydrogen concentrationsComparison of the embrittlement susceptibilitiesThe influence of microstructureEffect of coating type on hydrogen re-embrittlement

ConclusionsAcknowledgementReferences