Enhancing Reliability of Hydrogen-Assisted Cracking Properties Measured in H2 Gas
Brian SomerdaySandia National Laboratories, Livermore CA, USAInternational Institute for Carbon-Neutral Energy
Research (I2CNER), Fukuoka, Japan
Hydrogen embrittlement - Multi-scale modelling and measurement:What is the impact?
National Physical Laboratory, Teddington, UKOctober 7, 2014
Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy under contract DE-AC04-94AL85000
SAND2014-18591 C
Steel pipelines and pressure vessels are central components for delivery and storage of H2 fuel
• ASME developed Article KD-10 in Section VIII-3 of Boiler and Pressure Vessel code for high-pressure H2 vessel design– “Special Requirements for Vessels in High Pressure Gaseous
Hydrogen Service”– Mandatory for seamless vessels with H2 pressure > 41 MPa and
welded vessels with H2 pressure > 17 MPa (upper limit 100 MPa)– Design method also considered for H2 pipelines (ASME B31.12)
Codes require fracture mechanics measurements on containment materials in high-pressure H2 gas
Design life analysis in ASME Article KD-10 employs damage-tolerance principles
Need reliable measurements of subcritical cracking thresholds and fatigue crack growth relationships in H2 gas
measured in laboratory H2 gas
p
RiRo
at
H2
H2
H2H2
K = p[f(a, t ,Ro, Ri)]
“Reliable”: measurement integrity and transferability
• Are the measurements accurate?– Best practices for materials testing in H2 gas: ANSI/CSA CHMC1-2014, “Test methods for evaluating material compatibility
in compressed hydrogen applications – Metals” Advancing Materials Testing in Hydrogen Gas workshop, SNL, April 2013
(www.sandia.gov/matlsTechRef/advmat.html)
• Does the measurement represent non-conservative behavior?– Example: effect of load-cycle frequency on H2-assisted fatigue
A3750 steel4 MPa H2
A.H. Priest, British Steel, EHC-(1)42-012-81UK(H), 1983
4 MPa H2
H2-assisted cracking thresholds depend on test method
Key difference in test methods: propagating crack vs. stationary crack
H2H2
H2H2
H2
load
HH
H2H2
H2H2
H2
load
HH
H2H2H2
H2
loadHH
H2H2H2
H2
loadHH
constant displacement method:“arrest” threshold, KTHa
rising displacement method:“initiation” threshold, KTHi
K.A. Nibur et al., Metall Mater Trans A, 2013
KTHi
time
load
KTHa
crack extension
Kapp
H2-assisted cracking thresholds measured for martensitic pressure vessel steels
• Material– Cr-Mo(-Ni) pressure vessel steels– Yield strengths = 600 to 1050 MPa
• Mechanical loading– Constant displacement– Rising displacement = 0.05 mm/min
(~3 MPa m1/2/min)
• Environment– Supply gases: 99.9999% H2– Pressure = 103 MPa– Room temperature
H2H2
H2H2
H2
load
HH
H2H2
H2H2
H2
load
HH
H2H2H2
H2
loadHH
H2H2H2
H2
loadHH
K.A. Nibur et al., Metall Mater Trans A, 2013
Crack arrest thresholds (KTHa) higher than crack initiation thresholds (KTHi) at lower YS
Can observed trends be rationalized based on crack-tip mechanics?
• Divergent thresholds not attributed to difference in fracture mechanism
• Assume fracture mechanism is strain-controlled, i.e., governed by local– Consistent with plasticity-related
hydrogen-induced cracking (PRHIC)Takeda and McMahon, Metall Trans A, 1981
KTHa
KTHi
pressure vessel steels 4130X pressure vessel steelKTHa KTHi
K.A. Nibur et al., Metall Mater Trans A, 2013
Mechanics of propagating vs. stationary cracks illuminated by crack growth resistance curves
Crack-tip strain field weaker for propagating crack vs. stationary crack
steady state propagation
blunting
rKss
local
2
ln~
rK
local
2
~
a
Hutchinson; Rice and Rosengren, J Mech Phys Solids, 1968
Rice et al., ASTM STP 700, 1980
Weaker crack-tip strain field for propagating crack can account for higher crack-arrest thresholds
Strain-controlled fracture criterion (local) coupled with strain field amplitudes leads to KTHa > KTHi
rK
local
2
~
rK
local
2
ln~
r*
*local
local = *localK = KTH
H2H2
H2H2
H2
load
HH
H2H2
H2H2
H2
load
HH KTHi
H2H2H2
H2
loadHH
H2H2H2
H2
loadHH
KTHa
Crack arrest thresholds (KTHa) depend on initial applied stress-intensity factor (Kapp)
Can observed trends be attributed to intrinsic behavior of propagating cracks?
K.A. Nibur et al., Sandia Report SAND2010-4633, 2010
C110 steelYS = 793 MPaDCB specimens
S.K. Saha and X. Long, SteelyHydrogen2014 Conference Proceedings, 2014
Crack growth resistance curves for lower-strength martensitic steels suggest KTH depends on a
• Crack growth resistance curve represents locus of KTH points– Bounding KTH values (KTH
stationary and KTH
SS) reflect limits of crack-tip strain fields
– For strain-controlled fracture, KTHdepends on crack extension for either test method
• For Kapp > KTH, crack at “instability” condition relative to KTH locus– As crack extends, K declines
following method-dependent K vs. a trajectory
– Crack arrests at KTH locus (e.g., KTHa), arrest point depends on strain field evolution as crack grows
Crack extension-dependent KTH may account for observed KTHa vs. Kapp behavior
H2H2H2
H2
loadHH
H2H2H2
H2
loadHHKTHaKapp
KTHa
Kapp
KTHstationary
KTHSS
rK
local
2
ln~
rK
local
2
~H2
H2
H2H2
H2
load
HH
H2H2
H2H2
H2
load
HH
What is the impact?
• Crack-arrest thresholds measured under constant displacement can be non-conservative
– Micromechanics interpretation illustrates that trend may be universal for strain-controlled, H2-assisted cracking
– Standardized test methods may be non-conservative ASTM E1681 NACE TM0177 - Method D
– Reliable KTHa measurements expected for stress-controlled, H2-assisted cracking
H2-accelerated fatigue crack growth rates may unexpectedly decline at lower load-cycle frequency
A.T. Stewart, Mechanisms of Environment Sensitive Cracking of Materials, 1977
A. Macadre et al., Engineering Fracture Mechanics, 2011
• Do oxygen impurities in H2 gas lead to inhibition at low frequency?
• Does intrinsic H-material interaction account for reduced da/dN at low frequency?
Systematically measured fatigue crack growth rates in high-pressure H2 with controlled O2 levels
• Material– X52 ERW pipeline steel
• Instrumentation– Internal load cell in feedback loop– Crack-opening displacement measured
internally using LVDT– Crack length calculated from compliance
• Mechanical loading– Triangular load-cycle waveform– Constant load amplitude or constant K
• Environment– Supply gases: 99.9999% H2
H2 with 10-1000 vppm O2– Pressure = 21 MPa– Room temperature
B.P. Somerday et al., Acta Mater, 2013
Fatigue crack growth measurements performed on X52 line pipe steel
• API 5L X52 steel composition
• Tensile properties– Yield strength: 428 MPa– Ultimate tensile strength: 483 MPa
C Mn P S Si Cu Ni Cr V Nb Al CE
0.06 0.87 0.011 0.006 0.12 0.03 0.02 0.03 0.002 0.03 0.034 0.11
ERW seam
324 mm OD x 12.7 mm WT
ferrite + 8 vol% pearlite
B.P. Somerday et al., Acta Mater, 2013
Measurements demonstrate pronounced inhibition at lower frequency and higher O2 content
Can interplay between mechanical and environmental variables be quantified through physics-based model?
B.P. Somerday et al., Acta Mater, 2013
• Initial inert-environment crack growth modeled by blunting-resharpening
• Oxygen out-competes hydrogen for adsorption sites on freshly exposed crack-tip surface
• Extent of oxygen adsorption depends on crack-tip area, proportional to crack-growth increment (da)– when da < dacrit, crack tip fully
passivated by oxygen– when da > dacrit, crack tip not fully
passivated H uptake
• Develop model that quantitatively relates adsorbed oxygen (H uptake) to mechanical and environmental variables
O2
O2
K = Kmin
H2
a
H2O2H2
H2
H2 H2H2OO
O
a
H2
H2 H2O2
O2
O2
H2 H2H2
H2
H2
O2 H2
O2
O2
O
O
O
O
O O2
H2
O2
H2
H2
O2
H2
O
O
O
H
H
HHH
H
H2
H2
H2H2
O2
H2H2H2H2
H2H2 H2
H2H2
H2
H2
da > dacrit
K = Kmax2 > Kmax1
a
O2
H2
H2H2O2
H2
O2
H2H2
H2
H2
H2
O2 H2
H2
H2
O O
O
OOO
O
O2H2
O2
H2
H2
H2
H2
da < dacrit
K = Kmax1
Central physical premise: oxygen adsorption at crack tip can inhibit H uptake
B.P. Somerday et al., Acta Mater, 2013
Model developed based on idealized crack geometry and diffusion-limited oxygen adsorption
a
O2
O2 O2
O2
O2O2
O2h ~ Kmax
O2HH
Ha*
OO
O
a*distance from crack tip, x
ptot
pO2
• Goal: quantify amount of adsorbed oxygen (n) during load-cycle time (t)
• Key assumption: adsorption rate-limited by O2 diffusion in crack channel– constant crack-channel height (h)
during diffusion– steady state pO2 profile
• Model foundation: oxygen delivered to crack tip (Jht) = oxygen adsorbed on crack tip (Sa)
*flux
TaRpDJg
tot
2
0
02
)1(160height channel
RσK
Eσ)ν(.h
coverageoxygen
ft /1
2
0
2
)1(*)1(3.0)(
Ra
KTERS
Dpfagcrit
totcrit
H uptake and accelerated crack growth when = crit
density site surfaceS
B.P. Somerday et al., Acta Mater, 2013
Model formulation quantifies interplay between f, O2 concentration, da/dN, and R ratio on onset of H2-accelerated crack growth
Model predictions consistent with da/dN vs. frequency data measured in H2+O2 gas
2
0
2
)1(*)/()1(3.0
Ra
KTERSdNda
Dpfgcrit
totcrit
Scrit from measured da/dN vs. K data in H2+100 vppm O2
B.P. Somerday et al., Acta Mater, 2013
Enhanced inhibition at higher R ratio associated with effect of crack channel height on O2 transport
Model identifies variables that can distinguish effects of O2 inhibition → example: R ratio
2
0
2
)1(*)1(3.0
Ra
KTERSf
DpdNda
gcrit
tot
crit
What is the impact?
• Measured fatigue crack growth rates in H2 gas with O2impurities may be non-conservative
– Measurements and modeling demonstrate that O2 inhibition depends on interplay between da/dN, load-cycle frequency, R ratio, and O2 concentration
– Inhibition can be pronounced in H2 gas with ppm-level O2concentrations, particularly at low frequency
– Physics-based model of O2 inhibition can aid in distinguishing mechanism for reduced H2-assisted crack growth rates at low frequency
Enhancing Reliability of Hydrogen-Assisted Cracking Properties Measured in H2 Gas
• Comprehensive experiments coupled with mechanistic interpretations enhanced reliability of subcritical cracking threshold and fatigue crack growth rate measurements– Impacts life prediction of pipelines and pressure vessels
for H2 gas delivery and storage
Acknowledgments
• Sandia National Laboratories– Kevin Nibur (currently at Hy-Performance Materials Testing)– Chris San Marchi– Jay Foulk– Joe Ronevich– Zach Harris– Ken Lee– Jeff Campbell– Mark Zimmerman
• International Institute for Carbon Neutral Energy Research– Prof. Petros Sofronis (University of Illinois)– Prof. Reiner Kirchheim (University of Göttingen)– Prof. Alex Staykov (Kyushu University)– Dr. Mohsen Dadfarnia (University of Illinois)
Back-up Slides
SNL core capability in hydrogen embrittlementfeatures Hydrogen Effects on Materials Lab
• Static-loading crack-growth system– Wedge opening load (WOL) and double
cantilever beam (DCB) specimens– H2 pressure up to 200 MPa– Temperature -70 to 170 oC
• Dynamic-loading crack-growth system– Compact tension (CT) and single edge
notch (SEN) specimens– H2 pressure up to 138 MPa– New pressure vessel design with
target temperatures -100 to 200 oC
Materials testing in H2 supports technology development in several mission areas
Yield strength and alloy composition of martensitic pressure vessel steels
KTH vs. yield strength trends can be rationalized based on crack-tip mechanics framework
Assumption that local decreases as yield strength increases can explain converging KTHa and KTHi
*local
*local
local = *localK = KTHlow-strength steel
local = *localK = KTHhigh-strength steel
K.A. Nibur et al., Metall Mater Trans A, 2013
KTHi
KTHa
decreasing *local
KTHi reaches lower bound as loading rate decreases
KTHa
KTHi
Clark and Landes, ASTM STP 610, 1976 Nibur et al., SAND2010-4633, 2010
Higher KTHa compared to KTHi not attributed to crack arrest near back face in conventional WOL
• Chevron-notched WOL specimen– Reduced compliance leads to lower initial loads– Lower initial loads allow shorter precracks, which
limit final crack length at KTHa
Conventional WOL specimen
Chevron-notched WOL specimen
Plasticity-related hydrogen-induced cracking assumed for martensitic pressure vessel steels
4130X100 MPa H2 gas
slip lines at notch root plasticity-related hydrogen-induced cracking (PRHIC): glide plane decohesion
HY 130-type0.2 MPa H2 gas
Takeda and McMahon, Metall Trans A, 1981
Elastic unloading behind crack tip leads to distinct strain field for propagating crack vs. stationary crack
Plane strain slip-line representation of crack tip stress states of Prandtl field for stationary crack and propagating crack
Ritchie and Thompson, Metall Trans A, 1985
unloading
stationary crack
propagating crack
Stress fields near crack tip similar for stationary and propagating cracks
Local stresses ahead of crack tip as a function of angle for stationary crack and propagating crack
Ritchie and Thompson, Metall Trans A, 1985
Gas constituents such as O2 can inhibit H2-accelerated fatigue cracking in steels
R. Wei and R. Gangloff, ASTM STP 1020, 1989
H. Johnson, Stress Corrosion Cracking and Hydrogen Embrittlement of Iron Based Alloys, 1977
4340 steel
How is O2 inhibition affected by variables such as K, frequency, R ratio, and O2 content?
Density functional theory (DFT) simulations suggest mechanism for O2 inhibition of H uptake
Preadsorbed oxygen on iron surface raises activation energy for H2 dissociation
Ene
rgy
(eV
)
Distance from surface (Å)
transition stateEA = 0.6 eV
Potential energy surface scan for H2 approaching Fe(100) surface
H2 molecule approaches directly on top Fe atom
Distance from surface (Å)
Ene
rgy
(eV
)
dissociationtransition stateEA = 0.1 eV
dissociation
Potential energy surface scan for H2 approaching Fe(100) surface with preadsorbed O atoms
H2 molecule approaches directly on top Fe atom
Staykov et al., Int J Quantum Chemistry, 2014
Electron density difference method provides insight into dissociation inhibition mechanism
• Oxygen atoms on surface localize e- density, reducing ability of neighboring Fe atoms to transfer e- to H2
• Less e- density available for H2 activation: dissociation hindered
H2 approaching Fe(100) surface H2 approaching Fe(100) surface with preadsorbed O atoms
FeO
H2
e- e-
e- e-
enhanced e- density
depleted e- density
DFT simulations provide rationale for preferential adsorption of O2 in mixed H2+O2 gas
Strong attractive force and absence of activation barrier allow O2 to out-compete H2
F dEdR
Potential energy surface scan for H2 approaching Fe(100) surface
• H2 is detected at 2.6 Å• Weak attractive force• Activation barrier: not all H2 molecules dissociate
Potential energy surface scan for O2 approaching Fe(100) surface
• O2 is detected at 3.6 Å• Strong attractive force• No activation barrier: all O2 molecules dissociate
EA = 0.1 eV
Ene
rgy
(eV
)
Distance from surface, R (Å) Distance from surface, R (Å)
Ene
rgy
(eV
)
2.6 Å 3.6 ÅF dE
dR
DFT simulations define potential scenario for O2 inhibition of H uptake
Key elements in O2inhibition scenario:
• O2 detected deeper in gas volume compared to H2
• Force on O2 >> force on H2
– O2 can out-compete H2for adsorption sites
•Adsorbed O leads to repulsive force on H2
Density Functional Theory (DFT): Methods of Calculation
Theory: plane wave DFT
Program: VASPFunctional: optB86b-vdW(includes long-range interactions)Results verified with GGA-PBE
Spin polarized calculations(accurate spin guess applied)
Energy cut-off: 400 eVK-points: 8x8x8 bulk; 3x3x1 surface
Transition state search:climbing nudged elastic band method8 images per band
Analysis:Bader population analysisElectron density difference method
BCC Fe latticeFerromagnetic coupling
2x2 surface supercell6 layers Fe atoms12 A Vacuum slab
Top 3 layers are relaxedBottom 3 layers are fixed
H2 starting geometries on Fe surface
On topside-on
On topend-on
Bridge Bridgecross
H2 dissociation on Fe surface
On topside-on
chemisorptionstate
transitionstate
trajectory
Single H2 moleculeapproaches the surfacein parallel orientation
Bridge cross
Distance from surface (Å)
trajectory
Single H2 moleculeapproaches the surfacein parallel orientationacross a bridge site
Ene
rgy
(eV
)
H2-accelerated fatigue crack growth not absolutely inhibited by O2: depends on da/dN and R ratio
• At lower K, crack growth rates in H2 environments same as rates in air
• At R=0.1, hydrogen-accelerated crack growth observed at higher K– da/dN at onset of hydrogen-
accelerated crack growth depends on O2 concentration
• At R=0.5, hydrogen-accelerated crack growth not observed
• Why is da/dN at onset of H2-accelerated cracking a function of O2 concentration?
• Why does higher R promote inhibition?
B.P. Somerday et al., Acta Mater, 2013
Favorable correlation between measurements and model gives credence to physical assumptions
Model predictions consistent with da/dN vs. K data measured in H2+O2 gas at R = 0.1 and 0.5
2
0
2
)1(*)1(3.0
Ra
KTERSf
Dpagcrit
totcrit
Scrit calculated from measured da/dN at onset of accelerated crack growth for H2+100 vppm O2
assume a ~ da/dN
B.P. Somerday et al., Acta Mater, 2013
Why H2-induced intergranular cracking under cycling loading for lower-strength ferritic steels?
X52 steel21 MPa H2
Rationale: onset of accelerated crack growth dictated by threshold values of both da/dN and Kmax
Model does not predict onset of accelerated crack growth for high-purity H2 case
Scrit from measured da/dN vs. K data in H2+100 vppm O2
2
0
2
)1(*)1(3.0
Ra
KTERSf
DpdNda
gcrit
tot
crit
Suresh and Ritchie, Metal Science, 1982
B.P. Somerday et al., Acta Mater, 2013
• Question: inhibition governed by O2 concentration or absolute O2partial pressure?
• Results at constant system pressure inconclusive both concentration and absolute partial pressure vary
• Can isolate partial pressure by varying system pressure at constant O2 concentration– Results show no effect on critical
da/dN for accelerated crack growth– Conclusion: inhibition dictated by
O2 concentration
• Model predicts governing role of O2 concentration– Dptot term is constant
Model critically tested by performing fatigue crack growth experiments at lower pressure
2
0
2
)1(*)1(3.0
Ra
KTERSf
DpdNda
gcrit
tot
crit
Establishing physical description allows rationalization of O2 concentration and R ratio effects
2
0
2
)1(*)1(3.0
Ra
KTERSf
Dpagcrit
totcrit
• O2 concentration () → affects gradient in crack channel• R ratio → affects height of crack channel
More advanced model accounts for varying O2 profile in “breathing” crack
haTERfdNda
DpSg
tot
*)/()1(3.0
0
2
Model based on “breathing” crack retains dependence on O2 concentration () and frequency (f)
crack length, a*crack length, a*
gray-scale shading → oxygen concentration, cO
B.P. Somerday et al., Acta Mater, 2013
Does load-cycle waveform have intrinsic effect on H2-accelerated fatigue crack growth rates?
Does load-cycle waveform affect H-material interaction during fatigue crack growth?
f = 1 Hz
A.T. Stewart, Mechanisms of Environment Sensitive Cracking of Materials, 1977
B.P. Somerday and M. Barney, J Pressure Vessel Technology, 2014
triangular wave
trapezoidal wave