2012 jackson lake paper gangloff

8/11/2019 2012 Jackson Lake Paper Gangloff

http://slidepdf.com/reader/full/2012-jackson-lake-paper-gangloff 1/15

1

Measurement and Modeling of Temperature Dependent

Internal Hydrogen Assisted Cracking in Cr-Mo Steel

ABDULLAH AL-RUMAIH

Refinery Operations

Saudi Aramco Oil Company

Dhahran, Saudi Arabia

RICHARD P. GANGLOFF

Department of Materials Science

University of Virginia

Charlottesville, VA 22901 (US

ABSTRACT

The temperature dependence of internal hydrogen assisted cracking(IHAC) of 2¼Cr-1Mo steel is measured with several fracture mechanicsapproaches, including a new-slotted specimen. The slow-rising displacement

threshold stress intensity (K IH) for IHAC increases with temperature from -50

to 125oC. The hydrogen concentration proximate to the crack tip (CTσ), stressenhanced and microstructure trapped, provides a similitude parameter; equal

K IH is produced for equal CTσ. CTσ from diffusion analysis decreases with

decreasing-bulk H concentration and decreasing-exponential of the H-trap binding energy (along the bainite interface crack path) to temperature ratio.

IHAC is eliminated below a critical CTσ, and thus above a critical temperaturewhich promotes H detrapping. The relationship between K IH and CTσ is

predicted from a decohesion-based micromechanics model. Crack tipsimilitude provides the fundamental basis to predict minimum pressurizationtemperature for pressure vessel operation without IHAC.

INTRODUCTION

Interaction of internal hydrogen assisted cracking (IHAC) and temperembrittlement can adversely impact the fitness-for-service of thick-wall 2¼Cr-1Mo steel in long duration service in high pressure H2 [1-7]. Between 1 and 6wppm of atomic hydrogen (H) dissolves in the steel lattice during elevatedtemperature (~400oC) operation in pressure H2 (10-15 MPa), and becomessupersaturated in the lattice and at microstructural trap sites upon shutdown to

ambient temperature and low H2 pressure. Slow cooling during shut down doesnot sufficiently remove H and such operation is costly. Subcritical IHAC couldoccur in the Cr-Mo steel during shutdown and subsequent start up. Moreover,temper embrittlement occurs in this class of steels during elevated temperatureservice, adding the concern that impurity segregation will exacerbatesusceptibility to IHAC [8]. Hydrogen embrittlement of steel is typicallymaximized near ambient temperature and eliminated above a sufficiently hightemperature which is often in the vicinity of 100 oC depending on steel strengthand H concentration [9,10]. Knowledge of this temperature dependence iscritical to safe operation of H2 pressure vessels.

A joint industry program was conducted to measure the effect oftemperature, H concentration, and temper embrittlement on IHAC in ten heats of



3

approach was developed to further prove that increasing temperature eliminatesintrinsic IHAC in 2¼Cr-1Mo steel. The results are reported here.

The objective of this research is to: (a) characterize the effects oftemperature and H concentration on K IH for IHAC in 2¼Cr-1Mo steel using anew specimen approach to retain crack tip H during slow-rate loading, (b)establish the ability of a crack tip hydrogen concentration similitude parameterto uniquely describe the H concentration and temperature dependence of K IH,and (c) predict the critical temperature above which subcritical IHAC is not

likely during startup of a thick wall pressure vessel fabricated from Cr-Mo steel.

EXPERIMENTAL PROCEDURE

A CT specimen with 10% sidegrooves was designed to electrochemically produce H on slot surfaces (Fig. 3), throughout slow-rising CMOD loading anddelivered to the crack tip [13,14]. The goal is to sustain a steady state level ofcrack tip fracture process zone (FPZ) H at any temperature and prolongedloading time. Fracture experiments focused on weld metal produced by KobeSteel Co. Thecomposition was Fe-2.33Cr-1.06Mo-0.61Mn-0.28Si-0.12Cu-0.09C-0.016P-0.015S

(wt pct). The multi- pass welded block was post-weld heat treated(PWHT for7 h at690oC) and laboratorystep cooled (SC) tosimulate substantialtemper embrittlement.Properties are: (σys =508 MPa, σuts = 639MPa, FATT = 43

oC,

K IC = 103-133 MPa√mat 25oC) [13].

Three-dimensional finite element analyses produced elastic K and elastic- plastic J-integral solutions which account for the effect of the two-symmetricslots on crack tip stress and displacement fields [13]. Electrochemical studiesdefined the concentration of H produced on a 2¼Cr-1Mo slot surface for a givenelectrolyte (e.g., 0.5 M H2SO4 + 10-3M K 2S2O4 polarized at -5.0 mA/cm2, andother solutions) and applied-cathodic polarization [14]. Two- and three-dimensional finite element analyses established both transient and steady state Hdiffusion profiles between the electrochemically reacting stress-free slot surfaceand stressed crack tip [13]. Crack length during loading was measured by thedirect current potential difference method [13].

Figure 3 Slotted compact tension specimen used to

characterize the intrinsic effect of temperature on IHAC ofsteel.



4

RESULTS

The slotted specimen method produced significant IHAC in temperembrittled 2¼Cr-1Mo weld metal, as shown in Fig. 4. Each of three replicatespecimens loaded under slow rising CMOD (filled points) at 25oC yieldedreproducible K IH in the range of 35 to 41 MPa√m, evidencing subcritical crackgrowth since K IC is 103-133 MPa√m at 25oC. Continued loading produced aStage II type K-independent crack growth rate for each specimen. Crack growthrate declined when CMOD was held constant for a single specimen (•), then

resumed at the Stage II level when rising K was restarted. A specimen loaded atfixed CMOD and initial K of 90 MPa√m exhibited subcritical IHAC (o), but theK TH at crack arrest was higher than K IH and growth rates were lower comparedto rising CMOD loading, consistent with the rise-hold-rise (•) experiment, asseen in Fig. 4. Other studies showed that K IH for the onset of IHAC under slow-rising CMOD issubstantially less thanthe threshold (K TH)characteristic of crackarrest under falling K.[10-12,15]. The data inFig. 4 confirm thiseffect. Slow-rising Kloading is used in the

present study to producea conservative lower- bound threshold forIHAC. Application ofthis parameter in

pressure vessel fitness-for-service assessmentmay need to considerthis complicating effectof loading rate.

The K IH values for the 25oC experiments represented in Fig. 4 are plotted in

Fig. 2, along with the results of slotted CT experiments with this weld metalstressed at temperatures up to 60oC. These results were obtained for severalslot-surface H concentrations (CH-Total-Slot of 3.0 wppm from 0.5 M H2SO4 + 10-

3M K 2SO4 at -5.0 mA/cm2, 1.8 wppm from 0.1 M NaOH at -15 mA/cm2, and 1.1wppm from 0.5 M H2SO4 at -10 mA/cm2) [13]. (The CH-Total-Slot listed as 2.6wppm (♦) in Fig. 2 should be 1.1 wppm.) The number of replicate experimentswhich yielded essentially the same K IH is represented by (2) or (4) in Fig. 2. The

beneficial effect of temperature in reducing IHAC is apparent based on theslotted specimen data in Fig. 2; K IH rises with increasing temperature.Moreover, K IH rises with decreasing CH-Total produced on the slot surface.

ANALYSIS

It is hypothesized that the hydrogen concentration proximate to the cracktip, CTσ, uniquely describes the temperature and H concentration dependencies

Figure 4 da/dt vs. elastic K for replicate specimens of step

cooled 2¼Cr-1Mo weld metal stressed under either slow-rising

K (dK/dt = 0.004 MPa√m/s prior to crack growth , , ) or

fixed CMOD (o) at 25oC and with slot surface total Hconcentration (CH-Total-Slot) = 3.0 wppm.



5

of K IH presented in Figs. 1 and 2. A similitude hypothesis follows; K IH forIHAC is equal for equal CTσ, independent of bulk-H concentration, temperature,and cracked body geometry. The importance of H concentration is wellrecognized, but the challenge is to determine CTσ, which will be enhanced overCH-Total-Bulk or CH-Total-Slot due to H trapping at microstructural features such asdislocations, grain boundaries, bainite interface boundaries and carbides[9,16,17], and also increased by lattice dilation due to crack tip hydrostaticstresses [9,18]. Hydrogen trapping and stress state interaction complicate

diffusion analysis of CTσ; however, a solution to this problem was reported forstandard H precharged and slotted electrochemically charged CT specimens of2¼Cr-1Mo steel [13,19]. This analysis is summarized here, leading to a uniquecorrelation of K IH vs CTσ, which is then mechanistically affirmed and furtherdeveloped for use in pressure vessel fitness-for-service prognosis.

The CTσ is calculated at an FPZ reference location, δ FPZ , set by themaximum in the hydrostatic stress distribution typically located at twice the

blunted crack tip opening displacement [20]. For the Cr-Mo weld metal, δ FPZ varies from 5 to 85 μm ahead of the crack tip, depending on K. Given that K IH is in the range of 15-40 MPa√m for IHAC, it is assumed that δ FPZ = 9 μm(specific to K of 33 MPa√m) for all K IH. CTσ is given by [9]:

Eq. 1

where CL-FPZ is the concentration of H present in lattice sites at δ FPZ defined bythe highest hydrostatic tension (σH), VH is the partial molar volume of H in iron,R and T have the normal meaning, and EB is the binding energy of H to trapsites that constitute the IHAC path. For Cr-Mo steel, the H crack follows bainitelath and packet boundaries plus prior austenite grain boundaries [12,13]. Thereare three uncertainties associated with Eq. 1; specifically, the correct value ofCL-FPZ for a given-bulk level of dissolved H, as well as values of σH and EB for ahighly strained crack tip and complex-steel microstructure; each has been theobject of substantial literature debate [9]. Reasonable values of σH (2.5 σys,confirmed by crack tip elastic plastic FEA for the slotted specimen) and E B forH trapping at bainite lath interfaces (38 kJ/mol) are used [13].

The CL-FPZ at δ FPZ was determined for the standard and slotted CTspecimens and appropriate boundary conditions using a 3-dimensional FEAanalysis[13,19]. The effect of the crack tip stress gradient on H chemical

potential was not included and diffusion occurred at constant temperature.Diffusion analysis was based on an effective H diffusivity (DEff ) whichaccounted for the effect of microstructural trapping on H mobility and solubility.The steady state H concentration profile was used for the slotted specimen,while a transient solution was used for the H2 precharged standard CT specimen,consistent with the loading and H boundary conditions for each experiment.The DEff was estimated as a function of temperature, diffusible H concentration(for any location, CH-Diff = CH-Total – CL), and crack tip plastic strain (fromelastic-plastic FEA). DEff increases with increasing CH-Diff and temperature, anddecreases with increasing plastic strain following H-trapping literature[17,21,22], and estimated algorithms [13]. Exemplar values of DEFF (cm2/s) aregiven here for temperature (oC), CH,Diff (wppm), and plastic strain (%): (a)



6

3.7x10-7 at 25, 0, 0; 1.3x10-6 at 25, 1, 0; 1.3x10-8 at 25, 0, 50; 4.6x10-8 at 25,1, 50; 1.2x10-6 at 100, 0, 0; 4.2x10-6 at 100, 1, 0; 4.2x10-8 at 100, 0, 50;1.5x10-7 at 100, 1, 50). The ratio of H diffusivity in the Fe lattice to DEff established that CL ~ 0.1 CH-Diff and literature data plus high temperature Hextraction measurements show that CH-Diff = 0.65 CH-Total [13]. The crack tip andwake H egress flux was finite, as dictated by CL = 0 on this free surface and DEff was slowed by dislocation trapping in the crack wake [17]. To reduce thenumber of FEA models, CH-Diff on the slot, or the initial CH-Diff throughout the H2

charged CT specimen, was equated to 1.0 wppm. The FEA-predicted values ofCH-Diff at δ FPZ = 9 μm were linearly scaled by the ratio of actual slot surface orH2 charged starting CH-Diff to 1.0 wppm. This scaling is accurate provided that Hdiffusivity is reasonably independent of concentration. Changes in the crack tipCH-Diff profile during slow-rising K, governed by bold-surface H loss andhydrostatic stress and plastic strain enhancement, were approximated byestimating changes in CH-Diff during rising K with a series of steady statesolutions. Some of these solutions utilized a fine crack tip mesh in a 2-dimensional model for sufficiently high fidelity given the strong gradients instress and strain [13,20]. All H concentration profiles that yielded CTσ weretaken at the mid-thickness plane of the CT specimens where IHAC ismaximized. These assumptions and precise values of the input parameters arenot of primary importance provided that the CTσ similitude parameter is basedon the same estimates and assumptions when a variety of laboratory data are

correlated and used to predict pressure vessel IHAC [13,19].Measured K IH decreases as a unique function of rising CTσ independent of H

concentration, temperature and CT specimen geometry-boundary conditions; asshown in Fig. 5. Slotted specimen data (, and ◊) from the current study aretaken from Fig. 2. Standard H2 charged CT specimen data for a similar weldmetal (step cooled, FATT = 28oC) are given by Δ and for test temperatures

between 0oC and 100oC and initial CH-Total-Bulk of 3.0 and 5.0 wppm [12].Standard H2 chargedCT specimen data for asimilar lot of temperembrittled base plate(FATT = 6oC) are given

by , Δ (dotted center),and for test

temperatures between -47oC and 100oC andinitial CH-Total-Bulk of 1.5,3.0 and 5.0 wppm [12].The agreement betweenthese sets of temperembrittled 2¼Cr-1Mosteel data is excellent,

particularly given thatcomplex experimentswere conducted at 3different laboratories.

Figure 5 Correlation between measured K IH from slotted and

H2 precharged standard CT specimens vs CTσ for temper

embrittled 2¼Cr-1Mo weld metal (28 < FATT < 43oC) and

base plate (FATT = 6oC) tested under slow rising CMOD at

various temperatures from -47 to 100oC [13]. Vertical arrowsindicate that crack growth was not resolved.



7

All data in Fig. 5 are well described within the plotted band. At a finer scale, thesolid line in the middle of the data band describes the correlation for the slottedspecimens of weld metal tested with several levels of CH-Total-Slot andtemperatures from 25 to 60oC (Fig. 2). The shaded Δ and show that the H2 charged CT specimens from a similar lot of weld metal are well described bythis single trend line for two levels of CH-Total-Bulk and temperatures from 0 to100oC [12]. The proposed CTσ similitude parameter is validated.

The proposed CTσ similitude parameter was affirmed by independent

experiments with large (90 mm thick) H2 precharged CT specimens, which weredesigned to reduce H loss compared to the standard 25.4 mm thick CT specimen[23,24]. For each K IH measured with these specimens at various temperaturesand initial CH-Total-Bulk , CTσ was computed using the H diffusion model and Eq. 1.A 2-dimensional FEA model was used to approximate nil through-thickness Hloss in these thick specimens [19]. All other elements of the model parallelthose used for 3-D analyses of the small and slotted CT specimens in Fig. 5.The results of this thick-specimen analysis confirm the master correlation

between K IH and CTσ determined for the smallCT specimens (Fig. 5),as demonstrated in Fig.6. The thick specimenresults for the same heat

of temper embrittled2¼Cr-1Mo weld metalrepresented in Figs. 2, 4and 5 are given by ♣ [24]. The agreementwith the trend line forthis weld metal (Fig. 5)is excellent. The resultsfor a separate lot of2¼Cr-1Mo base plate(SC, FATT = 22

oC) are

given by () [23], andare also in goodagreement with the band

redrawn from Fig. 5 toencompass all smallspecimen data.

The master correlation between K IH and CTσ (Figs. 5 and 6) suggests thatIHAC is eliminated as K IH approaches K IC, below a critical value of this cracktip FPZ concentration parameter. The vertical lines in Fig. 6 show conservativelower bound values of CTσ-Crit = 15,000 wppm for the two moderate temperembrittled weld metals (red solid line) and CTσ-Crit = 24,000 wppm for a singlemoderate temper embrittled base plate (blue dashed line) of 2¼Cr-1Mo steel.Each of these values of CTσ-Crit is unique to the assumed values of EB = 38kJ/mol and σH = 2.5σYS. Changes in these parameters, and other assumptions ofthe diffusion model, will alter these absolute values, but not degrade the master

Figure 6 Correlation between measured K IH from slotted

and H2 precharged standard CT specimens vs CTσ for temper

embrittled 2¼Cr-1Mo weld metal (28 < FATT < 43oC) and

base plate (FATT = 6oC) tested under slow rising CMOD atvarious temperatures from -47 to 100oC [13]. Data noted for

90 mm thick H2 precharged CT specimens are indicated by ♣ for weld metal from the lot represented in Figs. 2, 4 and 5 [24]and for step cooled base plate (FATT = 22 oC) [23].

Vertical arrows indicate that crack growth was not resolved.



8

correlation between K IH and CTσ. In this regard values of CTσ in Fig. 5 are high,over 95 atomic pct H, and justified by the strong levels of H trapping predictedthrough Eq. 1 with these values of EB and σH. The master correlation shown inFig. 6 is likely valid for any 2¼Cr-1Mo base plate or weld metal which ismoderately temper embrittled, characterized by an FATT near 25oC [12].

Temper embrittlement intensifies hydrogen embrittlement [8], as affirmedin the JIP study with H2 precharged small CT specimens (see Fig. 1) of high

purity modern base plate and weld metal with FATT values below -28oC [12].

Thus, the master correlation in Figs. 5 and 6 is likely to depend on Cr-Mo steel purity and temper embrittlement exposure. The H diffusion analysis wasapplied to the low-temper embrittlement data set contained in Fig. 1 [12]. Theresults presented inFig. 7 demonstrate thatK IH is effectivelycorrelated by a uniquefunction of CTσ. (AllK IH values in thisfigure were measuredwith variable CH-Total-

Bulk , but only at 25oC.The effect oftemperature on CTσ is

not proven.) A CTσ-Crit of 180,000 wppm H issuggested by thiscorrelation. Thereduced role of temperembrittlement ismanifest by this higherCTσ-Crit compared to thevalues of 15,000 and24,000 wppm for themoderate purity steelsin Figs. 5 and 6.

DISCUSSION

The strong correlation between K IH and crack tip stress-microstructuretrapped H concentration, as well as the fundamental basis for CTσ provided by Htrapping theory, validate the proposed similitude hypothesis and raise twoimportant questions. First, how is this similitude concept and correlation used tosupport safe operation of H2 pressurized Cr-Mo vessels in petrochemicalservice? Second, is this correlation predictable through mechanism-basedmodeling of IHAC, including interaction of temper embrittlement?

Appl ication to Pressure Vessel Fitness-for -Service

The correlations in Fig. 6 and 7 can provide a specific value of K IH formoderate purity (temper embrittled) or high purity (low temper embrittlement)of 2¼Cr-1Mo steel weld metal or base plate. K IH is used in a fracture mechanics

Figure 7 Correlation between measured K IH from H2

precharged standard CT specimens vs CTσ for modern purity

2¼Cr-1Mo weld metal (FATT < -57oC) and base plate (-90oC <FATT < -28oC) tested under slow rising CMOD at 25oC [12].

Vertical arrows indicate that crack growth was not resolved.

The two trend lines represent separate least squares analyses of

the weld metal and base plate subsets of data. Specific steelsare identified in [19].



9

analysis of a cracked pressure vessel with known crack size/shape, appliedstress, and H concentration from a diffusion model for a specific crack tipreference location, δ FPZ . From the temperature perspective, Eq. 1 is rewritten as:

TC .σVEBR ln Cσ C LFPZ

Eq. 2

where TCrit is the temperature above which K IH approaches K IC for elimination ofIHAC and the remaining terms were previously defined. CTσ-Crit is determined

from the K IH correlation for a given level of temper embrittlement and selected

δ FPZ (Figs. 6 and 7). Equation 2 captures the physical interpretation that IHAC

is eliminated with increasing temperature due to reduced H-trapping in the FPZand along the crack path.

A diffusion model of the pressure vessel is needed to compute CL-FPZ at the crack

tip reference δ FPZ which is used in the master correlation for use in Eq. 2. A

small value of δ FPZ (e.g., 9 μm in Figs. 6 and 7) complicates this modeling.

However, a strong corrrelation between measured K IH and diffusion-model-

calculated CTσ was demonstrated for δ FPZ = 470 μm (which is the outer boundary of the plane strain plastic zone for this class of Cr-Mo steel at K of 33

MPa√m) [19]. The major difference is that the absolute value of CTσ-Crit

increases as δ FPZ increases from 9 to 470 μm; however, there is an analogousrise in CH-Diff such that the effect of crack tip reference location cancels in the

calculation of TCrit through Eq. 2. The similitude concept, and Eq. 2, apply for

any value of δ FPZ , which is reasonably close to the crack tip, provided that the

same location (and same values of EB, α, and σH/σys) is used to develop the

master K IH vs CTσ correlation and estimate the value of CH-Diff used in Eq. 2.

Equation 2 and the K IH vs CTσ correlation provide the foundation to predictthe temperature above which IHAC is eliminated for a specific level range oftemper embrittlement, as a function of the concentration of H present in the bulkof the pressure vessel. An example is shown in Fig. 8 for moderate purity2¼Cr-1Mo steel base plate or weld metal. The steady-state total concentrationof H, dissolved in the bulk of the pressure vessel and remote from the crack tip,

is plotted on the abscissa. This CH-Total-Bulk must be related by H diffusionanalysis to the value of CL-FPZ at the selected crack tip δ FPZ . Time-temperaturehistory and cracked vessel geometry will affect the relationship between CH-Total-

Bulk at the start of loading and CL-FPZ, and thus the dependence of TCrit on theeasily accessible bulk-total H concentration. The predictions in Fig. 8 wereestablished using Eq. 2 and the CTσ-Crit = 117,000 wppm for moderate puritytemper embrittled 2¼Cr-1Mo steel. This critical H concentration was obtainedfor δ FPZ = 470 μm, using the correlation shown in Fig. 6, with all Hconcentrations increased by a factor of 4.4 to scale for the increase in δ FPZ from9 to 470 μm [19].

To use Fig. 8, it is necessary to estimate the concentration of diffusible Hdissolved at a location 0.5 mm ahead of the crack tip. For a thick wall



10

hydroprocessingreactor, this Hconcentration shouldreflect: (a) the steadystate total Hconcentration fromelevatedtemperature-high

pressure H2 exposureof the ID surfaceduring service, (b) Hloss and partition totrap sites during thetime-temperaturecool-down history,(c) the high-Hsolubility austenticstainless steelcladding on the IDsurface, and (d)crack depth ahead ofthe clad-alloy steel

interface wherestresses are highest. This assessment can be based on several levels of Hdiffusion modeling, varying from an FEA analysis including the complexitiesoutlined in a previous section [12,19], or an approximate engineering solution to1-dimensional H diffusion [2]. The ratio of CH-Diff-470 μm/CH-Diff-bulk which isdetermined from this diffusion analsyis establishes the governing trend line inFig. 8, which in turn gives TCrit for any starting value of bulk CH-Total. This tem

perature is an important element of next generation specification of the so-calledminimum pressurization temperature (MPT) [25]. As an example, a TCrit of65oC is indicated for a stainless steel clad pressure vessel which: (a) containedCH-Total-Bulk of 4.0 wppm at operating temperature and H2 pressure, and (b) wherea 100 h slow cool to 25oC in the absence of H2 produced a CH-Diff at δ FPZ-470 μ m =25% of the original level of CH-Diff-Bulk , as reported to be typical by a 1-dimensional H diffusion analysis of a stainless steel clad Cr-Mo pressure vessel

[2]. If it is conservatively assumed that CH-Diff at δ FPZ-470 μ m equals the totalconcentration of H dissolved at operating temperature (CH-Diff 470 μm / CH-Total-bulk = 1.54 in Fig. 8), then the MPT is predicted to be 114oC. This prediction can beintergrated with an MPT value determined based on avoidance of H-freeunstable cleavage fracture to determine which damage mechanism limits

pressure vessel start-up [25].

A similar set of curves was developed for modern high purity 2¼Cr-1Mosteel base plate or weld metal [19]. This analysis demonstrated that TCrit at agiven CH-Total-Bulk was typically reduced by about a factor of 2 due to the higherresistance of this modern steel to IHAC, compared to legacy temper embrittled2¼Cr-1Mo steel; compare Figs. 6 to 7. Pressure vessel start-up operation isfacilitated by reduced MPT.

Figure 8 The effect of crack tip diffusible H concentration (CH-Diff

470 μm), localized at δFPZ = 470 μm ahead of the crack tip, on the

predicted TCrit for elimination of IHAC in a cracked sectionfabricated from moderate FATT 2¼Cr-1Mo steel, as a function of

total-dissolved H concentration and based on the short term

laboratory value of CTσ-Crit = 117,000 wppm. In this

representation, CH-Diff470 m = 010 CL-FPZ at this reference location.



11

Mechanistic Interpretation An approach by Gerberich a coworkers provides a theoretical basis to

predict the crack tip H-concentration dependence of the threshold stress intensityfor IHAC and HEAC [26]:

′ exp σ ′′ Eq. 3

where β’ (1/MPa√m) and α’’ (MPa.m) are constants in a dislocation basedmodel of crack tip stresses, CHσ (atom fraction H) is the crack tip diffusible and

stress enhanced H concentration adjacent to the H-assisted crack path in theFPZ, and α (MPa√m(atom fraction H)-1) is an adjustable weighting factor whichdefines the lowering of the Griffith toughness (k IG, MPa√m) per unit Hconcentration. This lowering of fracture resistance follows from H-enhanceddecohesion of various interfaces in the microstructure [9]. From Eq. 1, CHσ istypically less than CTσ because the H-trap binding energy that governs FPZ Hmobility and concentration (11.5 kJ/mol for carbides in 2¼Cr-1Mo steel) is lessthat the H-trap binding energy for crack path trap features (38 kJ/mol for bainiticinterfaces) [16,21]. Equation 3 was validated for both H environment andinternal H assisted cracking of ferritic and tempered martensitic steels over awide range of tensile yield strength [10]. With fixed α” (0.00021 MPa.m) andβ’ ((0.20 MPa√m)-1), fits of K IH vs CHσ data for each steel yielded reasonable butdifferent sets of k IG, α, and the σH/σys ratio used to compute CHσ through Eq. 1.The average parameters for all steel data sets examined are: k IG = 0.67 MPa√m,

α = 1.11 MPa√m(at frac H)-1, and σH/σys = 18.5 [10]. Equation 3 does notconsider the effect of rising K vs falling K loading on the threshold for IHAC.

No current model is sufficient to couple this effect with a H concentrationdependence [15]. Equation 3 is used as the best available predictive tool.

For moderate purity 2¼Cr-1Mo base plate and weld metal, Eq. 3 with theliterature average values of the model parameters reasonably predicts therelationship between K IH and CH-Total for both H2 charged standard and slottedCT specimens, when the values of CTσ in the correlation and CHσ in Eq. 3 werereferenced to δ FPZ = 470 μm [19]. In this analysis CH-Total was related to cracktip CH,Diff at a reference δ FPZ through the diffusion model summarized in a

previous section [13]. However, small changes in the combined-parameter setnecessary for K IH prediction through Eq. 3 result in significant changes in model

predictions. As such, a modest level of experimental calibration is required for

modeling of a specific steel such as 2¼Cr-1Mo. This is illustrated in Fig. 9where the prediction for the moderate purity Cr-Mo steel is referenced to δ FPZ =9 μm. This model used the results of the 3-dimensional FEA diffusion analysis,which suggested a factor of 4.4-times reduction in CH-Diff as δ FPZ decreased from470 to 9 μm. The tri-linear trend line is the lower bound of a large number ofK IH vs CH-Total-Bulk data measured for moderate purity 2¼Cr-1Mo base plate andweld metal [12]. The dotted line is the prediction of Eq. 3 with the followinginput parameters necessary to achieve this fit: k IG = 0.67 MPa√m, α = 1.11MPa√m/at frac H, σH/σYS = 25.0, σYS = 500 MPa, α” = 0.00030 MPa m, β’= 0.1(MPa√m)-1, and T = 298 K. Only k IG and α values equal the steel averages fromindependent experiments, while the other parameters were adjusted to achievethe good fit in Fig. 9. Since these necessary values are within the range of those



12

parameters defined fora wide range of steels[10], Eq. 3 supportsthe proposedcorrelation betweenK IH and CTσ. Thisfundamental basisstrengthens the H

concentrationsimilitude concept.

The modelrepresented by Eq. 3can reflect the benefitof increased steel

purity, or decreasedtemper embrittlement,on resistance toIHAC. McMahon andcoworkers argue thattemper embrittlement reduces the intrinsic k IG due to lowering of interfacecohesive strength by independent amounts associated with the concentrations ofsegregated impurities and trapped H [8]. To assess this explanation, the

approach represented in Fig. 9 for moderate purity steel was applied to themeasured K IH vs CH-Total-Bulk data shown in Fig. 1 for high purity 2¼Cr-1Moweld metal and base plate. The results are shown in Fig. 10, where k IG wasincreased from 0.67 MPa√m for the moderate purity case (Fig. 9) to 0.78MPa√m (16% increase) to achieve a best fit with the data for the higher puritysteel. All other model

parameters wereidentical.

The model ofIHAC threshold (Eq.3) can be used toestimate the value ofCTσ-Crit for prediction

of an MPT.Specifically, assume avalue of K IH which issufficiently high torepresent little chanceof IHAC in a pressurevessel; for example,assume K IH is 50% ofthe upper shelf planestrain fracturetoughness when CTσ =CTσ-Crit. Equation 3 is:

Figure 9 Comparison of measured K IH ( ) for moderate purity

Cr-Mo steel with Eq. 3 predictions (……

) for CL-FPZ at δFPZ = 9 μmfrom H diffusion analysis applied to a given initial level of CH-

Total-Bulk which is plotted on the abscissa. The K IH predictions

used average values of k IG and α from the literature [10], best fit

values of α’’, β’, and σH/σYS, and constant temperature of 298 K.

0

20

40

60

80

100

120

140

160

0.0 1.0 2.0 3.0 4.0 5.0 6.0

P r e d i c t e d T h r e s h o l d S t r e s s I n t e n s i t y

K T H

( M P a

m )

Initial‐Bulk Total H Concentration CH‐Total

(wppm)

2.25Cr‐1MoSteel

High Purity

Best Fit Model Parameters

9 μm Crack Tip Reference

Figure 10 Comparison of measured K IH ( from Fig. 1) for high

purity Cr-Mo steel with Eq. 3 predictions (……) for CL-FPZ at δFPZ =

9 μm from H diffusion analysis applied to a given initial CH-Total-

Bulk plotted on the abscissa. The K IH predictions used values of α,

α’’, β’, and σH/σYS, and constant temperature of 298 K, identical

to the model in Fig. 9. K IG was increased to 0.78 MPa√m

consistent with reduced temper embrittlement.



13

KIH 0.5 K IC β ′ expG αCσ

α ′′ σYS Eq. 4

Using the Eq. 3 model parameters associated with the best-fit to the data formoderate (Fig. 6) and high purity (Fig. 7) 2¼Cr-1Mo steel plates, with H-freeK IC ~ 300 MPa√m, CHσ-Crit is predicted to equal 0.029 atom fraction H and 0.130atom fraction H, respectively. These values are equivalent to CTσ-Crit of 28,000wppm and 138,000 wppm for the moderate and high purity steels, with theseinput parameters and a temperature of 25oC. The measurement-inferred values

of CTσ-Crit are similar; 24,000 wppm for moderate purity (Fig. 6) and 180,000wppm for high purity (Fig. 7) 2¼Cr-1Mo plates. This agreement is good;however, the predicted value of CTσ-Crit depends sensitively on the value selectedfor K IC as well as the temperature used to convert CHσ-Crit to CTσ-Crit.

These analyses establish that a cutting edge theoretical model of thethreshold stress intensity for IHAC supports the empirical correlationsdeveloped to describe the effect of H concentration, temperature, and temperembrittlement on K IH. This strengthens the proposed engineering approach to

predict MPT, and provides a method to leverage complex and costlyexperiments. However, absolute predictions of K IH from the model dependsensitively on multiple-interacting input parameters. Parameters are producedfor two purities of 2¼Cr-1Mo steel, which are consistent with literature averagesfor a wide range of steels. Moreover, such modeling can leverage IHACexperiments with necessarily limited Cr-Mo steel specimens taken from archivaltest blocks which were temper embrittled by long term reactor service. Thisapproach would produce confirmed or refined model parameters, particularly α and k IG, which capture metallurgical effects on IHAC. These parameters canthen be used to estimate K IH vs CTσ and CTσ-Crit for fitness-for-service modelingand prediction of MPT. This threshold approach is conservative since rising-displacement K IH is used and subcritical crack growth is not considered.

CONCLUSIONS

1. Intrinsic resistance to internal hydrogen assisted cracking (IHAC) of2¼Cr-1Mo steel plate and weld metal increases with increasingtemperature in the range from 60 to 125oC, established by diverse fracturemechanics experiments including a new slotted compact tension specimen.

2. The hydrogen concentration proximate to the crack tip fracture processzone (CTσ), stress enhanced and microstructure trapped, provides anenvironmental similitude parameter; equal threshold stress intensity (K IH)for IHAC is produced for equal CTσ.

3. Crack tip CTσ from diffusion analysis decreases proportionate todecreasing-bulk H concentration and decreasing-exponential of the H-trap

binding energy (along the bainite interface crack path) to temperatureratio. IHAC is eliminated below a critical CTσ as K TH approaches steelfracture toughness. Hydrogen detrapping explains the beneficial effect ofincreasing temperature on IHAC.

4. Resistance to IHAC increases with increasing steel purity, manifest in a



14

systematic change in the correlation between K IH and CTσ and CTσ-Crit.

5. The unique relationship between increasing K IH and decreasing CTσ, aswell as the effect of steel purity, are predicted from a decohesion-basedmicromechanics model.

6. The crack tip similitude function yields a geometry specific criticaltemperature (TCrit), above which IHAC is eliminated, and provides a basisto predict the minimum pressurization temperature for a thick-wall

pressure vessel in H2 service.

ACKNOWLEDGEMENTS

This research was supported by two Joint Industry Programs on AgingHydroprocessing Reactors, led by Ara Bagdasarian and David Cooke (Chevron-Texaco), and more recently by James McLaughlin and Vance McCray (Exxon-Mobil) under American Petroleum Institute sponsorship. Saudi-Aramcosupported the PhD dissertation research of AA-R. Standard compact tensionexperiments were conducted by Japan Steel Works, Kobelco, and Acelor-Mittal[12]. Ted Anderson and David Cooke provided important technical insights.

REFERENCES

(1) A.J. Bagdasarian, E.L. Bereczky, T. Ishiguro, K. Kimura and T. Tahara, in 2nd InternationalConference on Interaction of Steels with Hydrogen in Petroleum Industry Pressure Vessel and

Pipeline Service, M. Prager, ed., Materials Property Council, New York, NY, pp. 19-21 (1994).

(2) T. Sakai, T. Takahashi, M. Yamada, S. Nose, and M. Katsumata, in High Pressure Technology,PVP-VOl. 344, ASME, New York, NY, pp. 79-94 (1997).

(3) T. Iwadate, T. Nomura and J. Watanabe, Corrosion, vol. 44, pp. 103-112 (1988).

(4) Y. Wada, T. Hasegawa, R. Kayano and H. Inoue, “Method of Judging Hydrogen

Embrittlement Cracking of Material Used in High-Temperature High-Pressure HydrogenEnvironment”, United States Patent No. US 7,035,746 B2, April 25, 2006.

(5) R.C. Brouwer, in Hydrogen Transport and Cracking in Metals, A. Turnbull, ed., Institute ofMaterials, London, UK, pp. 62-76 (1995).

(6) K. Smit and P.F. Ivens in International Conference on Interaction of Steels with Hydrogen in

Petroleum Industry Pressure Vessel and Pipeline Service, M. Prager, ed., Materials PropertyCouncil, New York, NY, pp. 305-310 (1993).

(7) S. Pillot and L. Coudreuse, in Gaseous Hydrogen Embrittlement of Materials in EnergyTechnologies: The Problem, Its Characterization, and Effects on Particular Alloy Classes,Volume 1, eds., R.P. Gangloff and B.P. Somerday, Woodhead Publishing Limited,

Cambridge, UK, pp. 51-93 (2012).

(8) C.J. McMahon, Jr., Engr. Frac. Mech., vol. 68, pp. 773-788 (2001).

(9) R.P. Gangloff, in Comprehensive Structural Integrity, I. Milne, R.O. Ritchie and B. Karihaloo,

Editors-in-Chief, J. Petit and P. Scott, Volume Editors, Vol. 6, Elsevier Science, New York, NY, pp. 31-101 (2003).

(10) Richard P. Gangloff, in Hydrogen Effects on Materials, B.P. Somerday, P. Sofronis, and R.H.Jones, eds., ASM International, Materials Park, OH, pp. 1-21 (2009).

(11) R.P. Gangloff, R.P. Gangloff, “Fracture Mechanics Characterization of Hydrogen

Embrittlement in Cr-Mo Steel”, in Present Situation on Steels for Hydrogen Pressure Vessels ,Creusot-Loire Industrie, Le Creusot, France (1998).



15

(12) R.P. Gangloff, Laboratory Studies of Hydrogen Embrittlement in Aging Cr-Mo Steel for Thick-

Wall Reactors, Final Report for Phase I to the JIP on Aging Hydroprocessing Reactors, AraBagdasarian, Chairman, (2000).

(13) Abdullah M. Al-Rumaih, Measurement and Modeling of Temperature Dependent Hydrogen

Embrittlement of Cr-Mo Steel to Enable Fitness-for-Service Modeling, PhD Dissertation,University of Virginia, Charlottesville, VA (2004).

(14) A.M. Al-Rumaih and R.P. Gangloff, “New Fracture Mechanics Approach to Characterize

Internal Hydrogen Embrittlement of Steel for Fitness-for-Service Modeling”, in Proceedings,

11th International Conference on Fracture, Paper 5044, A. Carpinteri et al., eds., Politecnico

de Torino, Turin, Italy (2005).(15) K.A. Nibur, B.P. Somerday, C. San Marchi, J.W. Folk, III, M. Dadfarnia, P. Sofronis, and

G.A. Hayden, “Measurement and Interpretation of Threshold Stress Intensity Factors for Steels

in High Pressure Hydrogen Gas”, Report No. SAND2010-4633, Sandia National Laboratories,Livermore, CA (2010).

(16) F.G. Wei and K. Tsuzaki, in Gaseous Hydrogen Embrittlement of Materials in Energy

Technologies: The Problem, Its Characterization, and Effects on Particular Alloy Classes,

Volume 1, eds., R.P. Gangloff and B.P. Somerday, Woodhead Publishing Limited,Cambridge, UK, pp. 493-525 (2012).

(17) A. Taha and P. Sofronis, Engr. Frac. Mech., vol. 68, pp. 803-837 (2001).

(18) Tong-Yi Zhang and J.E. Hack, Metall. Mater. Trans. A, vol 30A, pp. 155-159 (1999).

(19) R.P. Gangloff, “Impact of Hydrogen Environment Embrittlement on Minimum Pressurization

Temperature for Thick-wall Cr-Mo Steel Reactors in High Pressure H2 Service”, Final Reportto API Subcommittee 934-F on Heavy Wall Reactor Operations, Vance McCray, Chairman,

(2011).

(20) T. L. Anderson: Fracture Mechanics: Fundamentals and Applications, CRC Press, Taylor and

Francis Group, Boca Raton, FL (2005).

(21) J.P. Hirth, Metall. Trans. A, vol. 11A, pp. 861-890 (1980).

(22) A.J. Kumnick and H.H. Johnson, Acta Metall., vol. 28, pp. 33-39 (1980).

(23) Y. Wada, R. Kayano, T. Hasegawa and H. Inoue, in 10th International Conference on Pressure

Vessel Technology, J.L. Zeman, ed., vol. 1, pp. 543-552 (2004).

(24) Additional Experiments on Mid X B Weld Metal Tested by Hydrogen Charged CT Specimens,Japan Steel Works, Phase II Report to the JIP on Aging Hydroprocessing Reactors, D.L.

Cooke, Chairman, (2005).

(25) Recommended Practice 934-F, Guidance on Heavy Wall Reactor Startup and Shutdown, Task

Group API-934 on Cr-Mo Steels, Subcommittee on Corrosion and Materials, AmericanPetroleum Institute, Washington, DC.

(26) Katz, N. Tymiak and W.W. Gerberich, Engr. Fract. Mech., vol. 68, pp. 619-646 (2001).

2012 jackson lake paper gangloff

Documents