Metallurgical and Materials Transactions A

, Volume 41, Issue 11, pp 2816–2828

Hydride-Phase Formation and its Influence on Fatigue Crack Propagation Behavior in a Zircaloy-4 Alloy


    • Department of Materials Science and EngineeringThe University of Tennessee
    • Applied Technologies DivisionY-12 National Security Complex
  • Hahn Choo
    • Department of Materials Science and EngineeringThe University of Tennessee
    • Neutron Scattering Science DivisionOak Ridge National Laboratory
  • Gongyao Y. Wang
    • Department of Materials Science and EngineeringThe University of Tennessee
  • Peter K. Liaw
    • Department of Materials Science and EngineeringThe University of Tennessee
  • Bjørn Clausen
    • LANSCE-LCLos Alamos National Laboratory
  • Donald W. Brown
    • MST-8Los Alamos National Laboratory
  • Jungwon Park
    • Department of Materials Science and EngineeringThe University of Tennessee
    • dpiX, LLC
  • Philip D. Rack
    • Department of Materials Science and EngineeringThe University of Tennessee
  • Edward A. Kenik
    • Material Science and Technology DivisionOak Ridge National Laboratory

DOI: 10.1007/s11661-010-0342-z

Cite this article as:
Garlea, E., Choo, H., Wang, G.Y. et al. Metall and Mat Trans A (2010) 41: 2816. doi:10.1007/s11661-010-0342-z


The hydride-phase formation and its influence on the fatigue behavior of a Zircaloy-4 alloy charged with hydrogen gas are investigated. First, the microstructure and fatigue crack propagation rate of the alloy in the as-received condition are studied. Second, the formation and homogeneous distribution of the delta zirconium hydride in the bulk and its effect on the fatigue crack propagation rate are presented. The results show that in the presence of hydrides, the zirconium alloy exhibits reduced toughness and enhanced crack growth rates. Finally, the influence of a preexisting fatigue crack in the specimen and the subsequent hydride formation are examined. The residual lattice strain profile around the fatigue crack tip is measured using neutron diffraction. It is observed that the combined effects of residual strains and hydride precipitation on the fatigue behavior are more severe leading to propagation of the crack under near threshold loading.

1 Introduction

Zirconium alloys are used extensively in the nuclear industry as the nuclear fuel sheathing material and fuel pressure tubes because of the alloys’ high performance under severe pressure and temperature conditions coupled with high transparency to thermal neutrons.[13] During service, these alloys undergo a series of simultaneous processes, including redistribution of stresses, irradiation, changes in temperature, and hydrogen diffusion. The combined effects of these processes can lead to the precipitation of hydrides and thus to hydride-induced embrittlement.[46] Hydrogen embrittlement is a complex mechanism, not fully understood, which causes changes in the bulk physical and mechanical properties, such as diminution of ductility, decrease of fracture toughness, and shortening of fatigue life.[46] Throughout the years, several theories have been developed in regard to the mechanisms that govern the formation of metal hydrides and their effects on alloy performance. In particular, the diffusion of hydrogen to high-stress regions, the nucleation and growth of hydrides in such regions followed by the reorientation of hydrides under external stresses, and the subsequent fracture have been studied.[57]

A specific issue of hydrogen embrittlement that occurs mostly during reactor shutdowns is the crack propagation.[3] For example, during the Pickering fuel pressure tubes incident,[3] it was observed that cracking was induced by the high residual stresses in the tube’s inner wall in combination with the hydrogen (typically about 10 ppm of hydrogen present) in the pressure tube that embrittled the alloy.[3] Subsequently, when the stress intensity factor (K) became high enough at some defect sites, the crack initiation occurred. The crack propagation was accelerated by fracture of the brittle hydride precipitates aligned perpendicular to the tensile stress. The stress intensity threshold for crack initiation in the presence of hydrides was about 5 to 6 MPa\( \sqrt {\text{m}} \)[3,8] compared with \( 1 3\,{\text{MPa}}\sqrt {\text{m}} \) for a Zr-2.5 pct Nb alloy in the as-received condition.[8]

The influence of hydrides on the crack propagation was investigated under cyclic loading conditions.[7] Dutton et al.[7] reported that in the presence of zirconium hydrides in Zr-2.5 pct Nb, stage II of fatigue crack growth (FCG) exhibits negative stress dependency, which also was observed in steel.[9] However, a different result was reported previously on Zircaloy-4,[10] suggesting that during stage II, the crack propagation rate (da/dN) increases with stress intensity factor (K), and the trend is enhanced by the increase in temperature. An increase in the hydride concentration shortens stage II and decreases the critical stress intensity factor.[11]

The crack propagation behavior also is influenced by residual stress near the crack and its interaction with hydrogen. Residual stresses develop in a material after the load is released in the case of fatigue cyclically loaded specimens.[12,13] In general, tensile stresses enhance the crack propagation and reduce the fracture toughness, whereas compressive stresses have an opposite effect.[14] The knowledge of the nature of preexisting residual stresses near the crack tip and their influence on hydrogen diffusion, hydride formation, and distribution can help predict crack growth behavior in hydrogen-rich environments. The residual strains in the crack-tip regions can be investigated directly by neutron diffraction. The high penetration of thermal neutrons through engineering materials, such as steel, aluminum, or zirconium, makes neutron diffraction an excellent nondestructive technique to measure the elastic lattice strain response to external factors (e.g., stresses and hydrogen-rich environments).[1216]

In this article, hydride phase formation and its influence on FCG behavior in a Zircaloy-4 alloy will be presented and discussed for three different cases, representing three different specimen histories. These cases correspond to the zircaloy in the as-received condition (case 1), to the homogenous distribution of hydrides in the bulk (case 2), and finally, to the inhomogeneous distribution of hydrides as a result of the existence of a crack prior to hydrogen charging (case 3). This approach mimics a situation of a zircaloy in service in which a fuel pressure tube that contains internal cracks is exposed to a hydrogen source (e.g., reactor coolant) and to an internal heavy water coolant pressure, as well as to bending loads from the weight of the fuel and coolant (simulated here by fatigue loading).[3]

2 Experimental Details

2.1 Material and Specimens

A commercial zirconium alloy, Zircaloy-4 (Zr-4) obtained from ATI Wah Chang (Albany, OR)[17] was used for this study. The nominal chemical composition (wt pct) of the starting Zr-4 is 1.5 Sn, 0.22 Fe, 0.12 Cr, 0.13 O, and balance Zr.[17]

Three different cases (compact tension [CT] 1, 2, and 3) were studied using CT specimens. Specimen drawing and dimensions are shown in Figure 1(a), and the specimen preparation and fatigue testing steps are illustrated in Figures 1(b) through (d). The CT specimens were machined in accordance with the American Society for Testing and Materials (ASTM) Standard E647–86,[18] with a width (W) of 50.80 mm, length of 63.50 mm, and a thickness (B) of 6.35 mm. For CT 1, an as-received specimen was fatigued to failure following the steps depicted in Figure 1(b), such as precracking to the crack length a of 11.43 mm and FCG to failure. This specimen is considered the reference case. The CT 2 specimen was charged with hydrogen prior to fatigue testing to investigate the effect of homogenously distributed hydrides on the FCG (Figure 1(c)). However, for CT 3 condition, the specimen was fatigued to a crack length of 22.86 mm and then charged with hydrogen, as shown in Figure 1(d). After the charging, the specimen was fatigued to failure in an attempt to investigate the impact of the preexisting crack (and the residual stresses associated with the crack tip) on the hydride formation and, in turn, its subsequent influence on the crack propagation rate.
Fig. 1

CT specimen dimensions and preparation steps. (a) Schematic of CT specimen (dimensions are in millimeters). The area coated with nickel for hydrogen charging also is indicated. The fatigue precracking, crack growth, and hydrogen charging steps for the three different cases studied are shown in the following: (b) CT 1: fatigue of the as-received specimen; (c) CT 2: hydrogen charged and fatigued; and (d) CT 3: fatigued (FCG1) to a = 22.86 mm, Pmax = 4000 N, hydrogen charged, and continued to be fatigued (FCG2) with Pmax = 1300 N. “A” denotes when the neutron diffraction residual strain measurements were performed before H charging at a = 22.86 mm

2.2 Hydrogen Charging

CT 2 and CT 3 specimens were charged with hydrogen at an elevated temperature to form zirconium hydrides. To promote the hydrogenation process, an area of 25 × 30 mm2 around the crack on both sides of the specimen, indicated in Figure 1(a) by the shaded area, was coated with a nickel (Ni) layer of 200 nm thickness. Nickel dissociates the molecular H2 to atomic H by chemisorption, leading to the diffusion of H in the alloy’s bulk.[19] The Ni thin film was deposited on the zirconium CT specimens using an AJA 2000RF (AJA International, Inc., North Scituate, MA) magnetron-sputtering system. The deposition was performed at room temperature in argon with a total pressure of 5 mTorr at a radio frequency of 13.56 MHz and power of 200 watts.

Prior to the application of the Ni coating, each specimen was polished mechanically to 800 grit size and then polished chemically by immersion in a solution of 45 ml hydrogen peroxide (H2O2), 45 ml nitric acid (HNO3), and 10 ml hydrofluoric acid (HF).[20] The chemical polishing was employed to ensure the removal of the protective oxide layer from the specimen surface that could interfere with the hydrogen diffusion. The CT 2 and CT 3 specimens with the Ni film applied were hydrogen charged using a 12.5 vol pct hydrogen in argon mixture in a tube furnace at a temperature of 687.15 K (414 °C), under a pressure of 3.8 kPa, for 50 minutes. Subsequently, the specimens were annealed at 658.15 K (385 °C) for 5 hours in the same hydrogen-argon mixture. After annealing, the samples were furnace cooled to room temperature at a rate of 3 to 5 deg/min.[21]

2.3 FCG Testing

For this study, the macroscopic mode I of fracture was chosen, also known as the tensile opening mode, in which the crack faces separate in a direction normal to the plane of the crack. The CT specimens were precracked and fatigued using a servo-hydraulic material test system (MTS) testing machine, interfaced with a Teststar IIs controller. The details of precracking, fatigue cracking (fatigue crack growth – FCG), and hydrogen charging steps are illustrated in Figures 1(b) through (d). Each specimen was precracked under a constant stress-intensity-factor range (ΔK) to the crack length a of 11.43 mm. The FCG tests were performed under an increasing stress-intensity-factor range (ΔK) control mode (a constant load) at a frequency of 10 Hz and a constant load ratio, R = Pmin/Pmax = 0.1 at ambient temperature. The maximum stress intensity Kmax can be calculated as follows[22]:
$$ K_{\max } = {\frac{{P_{\max } (2 + \alpha )}}{{B\sqrt W (1 - \alpha )^{3/2} }}}\left( {0.886 + 4.64\alpha - 13.32\alpha^{2} + 14.72\alpha^{3} - 5.6\alpha^{4} } \right) $$
where Pmax is the maximum load applied, and α is the ratio between the crack length and the specimen width α = a/W. Crack length was measured by the crack-opening-displacement gauge using the unloading-compliance method.[23,24] In a FCG test, data is recorded using a computer data-acquisition system and then plotted as crack length avs the total number of elapsed cycles N. The crack growth rate da/dN is taken as the slope at some point of the curve, and it is a function of the stress level, crack size, and material properties. Mathematically, this crack propagation rate may be expressed in terms of the stress intensity factor, K[11] as follows:
$$ \frac{da}{dN} = C\left( {\Updelta K} \right)^{m} $$
where C and m are scaling constants and depend on material microstructure, fatigue frequency, load ratio, environment, and test temperature. In our case (tensile fatigue), the ΔK is the mode I stress intensity factor range during the stress cycle at the crack tip, and it is defined as follows:
$$ \Updelta K = K_{\max } - K_{\min } $$
Although the precracking part is performed with variable loads (constant ΔK), the Pmax value must be chosen carefully for the FCG tests such that the fatigue tests could be continued after the hydrogen charging, which was expected to embrittle the specimen significantly. The Pmax employed in this study was 4000 N, and the FCG tests were conducted to the failure of specimens. However, for CT 3, the specimen first was fatigued with the same Pmax to the crack length of 22.86 mm, marked with FCG 1 in Figure 1(d). At that point, the specimen was removed from the MTS machine and charged with hydrogen for 50 minutes, as described in the previous section. After hydrogen charging, the fatigue test on a similar specimen was resumed, using Pmax = 4000 N, which resulted in an abrupt failure during the first cycle. Therefore, Pmax = 1300 N was used after the hydrogen charging for the FCG test of CT 3, as indicated in Figure 1(d) by FCG 2. The load ratio R was kept constant at 0.1. Because fatigue crack growth rates generally are controlled by ΔK, it is reasonable to compare the crack-propagation rates at a given ΔK level under different loading and hydrogen-charging conditions regardless of the applied loads and a/W values.

2.4 Neutron Diffraction Residual Strain Measurements Around the Fatigue Crack

Spatially resolved strain measurements were performed on the CT specimen, CT 3, prior to hydrogen charging, using the Spectrometer for Materials Research at Temperature and Stress (SMARTS) instrument at the Los Alamos National Laboratory (Los Alamos, NM).[25] The fatigue test, using Pmax = 4000 N, was stopped at a = 22.86 mm, and the area of interest in the crack-tip region is marked by “A” in Figure 1(d). The residual strain distribution in this crack tip region was studied by measuring the elastic lattice strains using neutron diffraction. Strain scanning was conducted along the crack, starting from 4 mm behind the crack tip to 20 mm in front of the crack tip, as indicated in Figure 2(a). Figure 2(b) shows a schematic representation of the neutron experiment setup. The two detector banks at the SMARTS instrument allow simultaneous measurements of two strain components, such as in-plane (IP or y-direction, parallel to the fatigue loading direction) and through-thickness (TT or x-direction, perpendicular to the fatigue loading direction) strain components, respectively. However, for brevity, in this article, only the IP residual strain component is presented. In more detail, a cubic diffraction gauge volume was defined at the center of the specimen using 2-mm radial collimators to delineate the gauge volume along the 2-mm-high and 2-mm-wide incident beam. Note that within ±2 mm from the crack tip, the incident beam height was reduced to 1 mm to achieve a finer spatial resolution along the crack length. The diffraction data were measured at a total of 21 positions along the crack as shown in Figure 2(a). The lattice strains εa and εc along the a and c axes of the hexagonal-close-packed (hcp) crystal structure of the Zr-4 were obtained by Rietveld analyses[26] of the measured diffraction patterns, using the general structure analysis system[27] and Eq. [4], which is expressed as follows:
$$ \varepsilon_{a} = {\frac{{a_{i} - a_{0} }}{{a_{0} }}}\;{\text{ and }}\;\varepsilon_{c} = {\frac{{c_{i} - c_{0} }}{{c_{0} }}} $$
where ai, ci, and a0, c0 are the lattice parameters of the strained and unstrained conditions, respectively. The strained lattice parameters were measured from the area around the fatigue crack tip. The strain-free lattice parameters, a0 and c0, were measured from a small Zr-4 piece of 25 × 30 × 6 mm3 cut from the same ingot.
Fig. 2

CT specimen with (a) dimensions including the fatigue crack length of a = 22.86 mm and the crack tip position also indicated by “A” (Figure 1(d)). All dimensions are in millimeters. The magnified scale indicates the positions where neutron diffraction data were collected. Spatially resolved residual strain measurements were conducted from 4 mm behind the crack tip to 20 mm in front of the crack tip. (b) Top view of the neutron diffraction measurement setup. The detector banks 1 and 2 record the signal from the families of grains with their normals perpendicular (through thickness, TT or x-direction) and parallel (IP or y-direction) to the loading direction. The gage volumes were defined within the center of the thickness of the specimen by a 2-mm incident beam height (along the crack length) and 2-mm diffracted beam width. Note that the beam height was reduced to 1 mm within ±2 mm from the crack tip

2.5 Metallography and X-Ray Diffraction

The microstructure of the as-received zircaloy was investigated by optical microscopy under polarized light and by electron backscattered diffraction (EBSD). The EBSD experiment was conducted on a mechanically polished mirror finished surface (final polishing with 50-nm colloidal silica). The EBSD experiments were performed at 15 kV in a Philips XL30-FEG scanning electron microscope (SEM) equipped with a EBSD/OIM (orientation image mapping) system. Though it had been intended to analyze both the as-received and the hydrided materials, the higher strain levels associated with hydride formation resulted in poor or nonexistent EBSD patterns for the hydrided material. The specimens (in the as-received and the hydrogen-charged conditions) for the optical microscopy investigation were polished mechanically to 800 grit size. To reveal the zirconium hydrides, each specimen was etched by immersion in a solution of 46 ml nitric acid (HNO3), 4 ml hydrofluoric acid (HF), and 50 ml lactic acid (C3H6O3).[20] Using an image plot profile software, Image Processing and Analysis in Java (ImageJ),[28] the hydrides distribution was visualized via the contrast variation in each micrograph. Using the same software and a point counting method, the volume fraction of hydrides was determined.[28,29] X-ray diffraction was used to identify the phases in the near surface region of the hydrogen-charged samples. The scans were performed using Cu Kα radiation (λ = 1.540 Å), 2 × 2 mm2 incident beam, 0.01 deg step size, and 5 s/step scan rate.

Each specimen that was fatigued to failure was analyzed using an SEM. The fracture surfaces of the as-received and the hydrogen-charged conditions were investigated, using a Leo 1526 SEM (Electron Microscopy. Ltd., Cambridge, UK). Two different regions, the FCG area and the fast fracture area (after failure), were investigated. Images were taken on each sample at specific stress intensity factor (K) values to correlate the fatigue and fracture behaviors.

3 Results

3.1 Microstructure

The microstructure of the as-received specimen CT 1 is presented in Figure 3. The as-received Zircaloy-4 alloy was purchased from Wah Chang[17] as an ingot, which exhibited the α-Zr Widmanstätten basketweave morphology.[1,30,31] This morphology forms during the transformation from β body-centered cubic (bcc) to α (hcp) phase, by cooling from 1271.15 K (998 °C) to room temperature at moderate cooling rates.[1,30,31] The prior β–Zr grains have a mean diameter of approximately 700 μm (Figure 3(a)) and each grain contains multiple α-Zr plates (Figure 3(b)). Also in Figure 3(b), the presence of second-phase precipitates can be observed that appear as white spheroids under polarized light. In this case, the second-phase precipitates are Fe-Cr with an initial reported content of 0.33 wt pct Fe-Cr.[17] The distribution of the crystallographic orientation of α-Zr plates (variants) in a single, original β grain was investigated by EBSD, and an inverse pole figure map is shown in Figure 3(c). Each color in the EBSD map represents a different orientation based on the color-coded stereographic triangle in Figure 3(d). The dark regions are areas in which no indexable EBSD pattern was obtained because of the highly distorted lattice. The schematic hexagonal unit cells in Figure 3(c) show the local crystallographic orientation of several variants out of the 12 possible variants.[15] Two pole figures obtained from the EBSD maps are presented in Figure 3(e). The black circle in the basal 0001 pole figure is associated with two α variants in the material (the pink variant that runs from lower left to upper right and the orange variant that runs from upper left to lower right in Figure 3(c)). Three sets of \( 10\bar{1}0 \) poles (appropriately colored circles) for each variant fall on a great circle defined by the circled direction in the 0001 pole figure. The large misorientations between adjacent variants indicate that these interfaces are high-energy boundaries where hydrides can grow.
Fig. 3

Polarized light optical microstructure of mechanically polished and etched Zircaloy-4; (a) showing the Widmanstätten basketweave morphology including the former β-Zr grains with a mean diameter of approximately 700 μm, (b) magnified view of the basketweave morphology showing the α-Zr plates and the second-phase precipitates as the white spheroids. (c) EBSD inverse pole figure map of α plates within a former β grain. The schematic unit cells indicate the local crystallographic orientation of several variants. The dark regions are a result of distorted lattices. (d) Color-coded stereographic triangle in which each color represents a different orientation in the EBSD map. (e) Prism and basal pole figures obtained from the EBSD analyses. Points with confidence index <0.1 in indexing were excluded from these pole figures. See text for explanation of the pole figures relative to the spatial distribution and misorientation of α variants

The phases formed from hydrogen charging were studied by X-ray diffraction on the surface of each CT specimens. To ensure that this formation is not only a surface effect, the X-ray pattern was collected after approximately 1.4 mm from the initial thickness were removed by mechanical polishing. A typical X-ray diffraction pattern, intensity vs 2-theta, is presented in Figure 4. Three reflections corresponding to the hcp α-Zr phase, \( {\text{Zr}}\left( {10\bar{1}0} \right), \) Zr(0002), and \( {\text{Zr}}\left( {10\bar{1}1} \right) \) as well as the two peaks ZrH2(111) and ZrH2(200) corresponding to the face-center-cubic zirconium hydride, δ-ZrH2, are identified by peak indexing. It should be pointed out that the room-temperature hydride phase for zircaloy alloys is usually the δ-ZrH2 phase.[7,16,32,33]
Fig. 4

Typical X-ray diffraction pattern from the surface of a hydrogen-charged specimen. Three zirconium reflections Zr \( \left( {10\bar{1}0} \right), \) Zr (0002), and Zr \( \left( {10\bar{1}1} \right) \) and two zirconium hydride peaks ZrH2 (111) and ZrH2 (200) were identified. The hydride peaks correspond to δ-ZrH2

Figure 5 displays the morphology and distribution of zirconium hydrides in the CT specimens. A schematic of the upper half of the CT specimen is presented in Figure 5(a), where the green square at the crack tip indicates the location of the low- and high-magnification micrographs shown in Figures 5(b), (d), and (f) and Figures 5(c), (e), and (g), respectively. These micrographs represent the microstructure of the zircaloy before and after the formation of hydrides. The images were taken after approximately 1.4 mm from the initial thickness was removed (the original specimen thickness was ~6 mm). The microstructure for the Zircaloy-4 in as-received condition (CT 1 specimen) is presented in Figures 5(b) and (c) for comparison, whereas the microstructure for the CT 2 (H-charged and fatigued) specimen is shown in Figures 5(d) and (e) at lower and higher magnifications, respectively. The presence of a hydride phase can be observed in the zircaloy microstructure at the boundaries of α-Zr plates, as suggested by the EBSD investigation. During the 50-minute hydrogen charging at 687.15 K (414 °C), H2 dissociates on the area coated with Ni, and in turn, atomic H forms a thin layer of zirconium hydrides at the surface of the specimen. This process is followed, during the 5-hour annealing, by the diffusion of hydrogen into the specimen’s bulk and, subsequently, by the precipitation of δ-ZrH2 plates during cooling. The microstructure at the crack tip for the CT 3 (fatigued and H-charged) specimen is shown in Figures 5(f) and (g) exhibiting a higher concentration of hydrides than CT 2. Also shown in Figures 5(h) through (j) are the distributions of hydrides from the gray shaded area of 30.4 × 14.8 mm2 marked in Figure 5(a) for each of the three specimens in which the horizontal red dotted line denotes the outer edge of the Ni-coated area. The CT 2 specimen (Figure 5(i)) exhibits a reasonably uniform but darker image than CT 1, which is the as-received condition (Figure 5(h)). The distribution of hydrides through the width of the CT 3 sample is presented in Figure 5(j), which reveals a significant darker localization near the crack. Furthermore, the contrast given by hydrides in these images is quantified and graphically represented in Figure 5(k) by plotting the color contrast integrated across the entire area imaged, 30.4 × 14.8 mm2. The high values at the beginning and the end of each graph represent the edges of the specimens (the black areas in Figures 5(h) through (j)). Although CT 1 is the base line, the hydrides in CT 2 specimen are distributed rather uniformly through its width, with a slightly higher concentration on the Ni-coated area. For CT 3, hydrides are more localized around the crack, and interestingly, further away from the crack, the contrast becomes similar to the as-received condition. The vertical black dash–dot lines along the width of the half-specimen indicate the position of the crack tip for CT 3 during hydrogen charging. They also represent scan lines along which the hydride volume fractions were obtained by a point counting method for both specimens CT 2 and CT 3. Thus, the variation of hydride volume fractions on a specimen’s surface at 1.4 mm from the initial specimen thickness is displayed in Figure 5(l) by symbols and lines and exhibit consistent trends with those observed in Figure 5(k). In addition, the hydride volume fraction evolution at the middle of the specimen’s thickness was plotted by lines in Figure 5(l). Both specimens have comparable contents of hydrides at the middle of their thickness, which clearly indicates that hydrogen’s diffusion and hydride precipitation occurred throughout the specimen’s thickness.
Fig. 5

Optical microstructures of the following three cases studied: CT 1, CT 2, and CT 3. (a) Schematic of upper half of the CT specimen. The green square at the crack tip indicates the location for the low magnification micrographs in (b), (d), and (f) and at higher magnification in (c), (e), and (g) to show the zircaloy microstructure and the hydride plate formation. The gray shaded area of the three specimens is shown in (h) through (j) to illustrate the hydride distributions in each case. The red dotted area denotes the outer edge of a nickel-coated area. (k) Quantification of the obtained image contrast in (h) through (j) representing the distribution of the hydrides. (l) Variation in hydride volume fraction at 1.4 mm from the specimen surface (symbols and lines) and at the middle of the specimen thickness (lines) obtained along the vertical dash–dot lines (line scan)

3.2 Fatigue Crack Propagation Behavior

The FCG behavior plotted as the fatigue crack propagation rate (da/dN) vs stress intensity factor range (ΔK) for the three CT specimens is shown in Figure 6. The as-received specimen, CT 1 (Pmax = 4000 N) exhibits only two stages (I and II) of the crack growth. During stage I, the crack propagates relatively rapidly from 3.5 × 10−5 mm/cycle to 1.6 × 10−4 mm/cycle with increasing ΔK across a short range from \( 1 1. 5\,{\text{MPa}}\sqrt {\text{m}} \) to \( 1 3\,{\text{MPa}}\sqrt {\text{m}} . \) At ΔK values greater than \( 1 3\,{\text{MPa}}\sqrt {\text{m}} , \) the crack enters stage II (the plateau), also known as the Paris regime in which a power law dependence prevails, leading to a linear relation between log(da/dN) and log(ΔK). The crack propagates steadily until failure of the specimen occurs at \( \Updelta K_{\text{f}} = 3 7. 5\,{\text{MPa}}\sqrt {\text{m}} . \) The slope of regime II represents m in Eq. [2] and was approximately 2.8, a value typical for ductile materials.[11] The C parameter in Eq. [2], obtained by extrapolation, is \( 1. 5 4\times 10^{ - 5} {\frac{\text{mm/cycle}}{{\left( {{\text{MPa}}\sqrt {\text{m}} } \right)^{m} }}}. \)
Fig. 6

Fatigue behavior (continuous lines) for the three cases CT 1 through CT 3 plotted as crack propagation rate (da/dN) vs stress intensity factor range (∆K). Note that for the CT 3 specimen, it is shown that the FCG 2 (Figure 1(d)) obtained using Pmax = 1300 N is after the specimen was charged with hydrogen. The crosses indicate the failure of the specimen. The fatigue striation spacing values obtained from SEM investigation and represented by symbols also are shown

The presence of zirconium hydrides influences the fatigue behavior for CT 2, which is shown in Figure 6. Unlike CT 1, the CT 2 specimen exhibits the typical three stages for the FCG behavior. During stage I, the crack growth extends at slightly larger ΔK values and slower propagation rates. For example, at \( \Updelta K = 1 3\,{\text{MPa}}\sqrt {\text{m}} , \)da/dN for CT 2 is 1.1 × 10−4 mm/cycle, whereas for CT 1, it is 1.6 × 10−4 mm/cycle. CT 2 has a shorter stage II \( \left( {\Updelta K_{\text{II}} = 1 4\,{\text{to}}\, 1 7. 5\,{\text{MPa}}\sqrt {\text{m}} } \right), \) and crack propagation is increased by the end of this stage. The failure occurs after the crack became unstable during stage III (stage III greater than \( \Updelta K = 1 7. 5\,{\text{MPa}}\sqrt {\text{m}} \)). The stress intensity at failure is \( \Updelta K_{\text{f}} = 2 1. 9\,{\text{MPa}}\sqrt {\text{m}} , \) indicating that CT 2 exhibits a reduction in toughness. The slope of the Paris regime gives m = 4.6 and \( {C} = 4. 2 4\times 10^{ - 6} {\frac{\text{mm/cycle}}{{\left( {{\text{MPa}}\sqrt {\text{m}} } \right)^{m} }}}. \) A summary of the fracture toughness, m, and C values is shown in Table I.
Table I

Summary of Fatigue Behavior: Stress Intensity Factor at Failure (ΔKf), and Scaling Constants, m and C, In Eq. [2] for the Three Cases Studied

Specimen/Fatigue Behavior

CT 1

CT 2

CT 3

ΔKf\( \left[ {{\text{MPa}}\,\sqrt {\text{m}} } \right] \)








C\( \left[ {{\frac{\text{mm/cycle}}{{\left( {{\text{MPa}}\sqrt {\text{m}} } \right)^{m} }}}} \right] \times 10^{ - 6} \)




The FCG behavior for CT 3 after hydrogen charging is presented in Figure 6 as well. This specimen has a similar behavior with the as-received condition until the crack length reaches 22.86 mm, prior to H charging. After charging, the fatigue test resumed with a load of 1300 N, as mentioned in Section II–C. The fatigue profile of the CT 3 specimen, shown in Figure 6, displays only two stages, II \( \left( {\Updelta K_{\text{II}} = 7\,{\text{to}}\, 1 3\,{\text{MPa}}\sqrt {\text{m}} } \right) \) and III (greater than \( 13\,{\text{MPa}}\sqrt {\text{m}} \)). It is worth noting the small value for ΔK, about \( 7\,{\text{MPa}}\sqrt {\text{m}} \) at the beginning of the Paris regime, which is comparable with the stress intensity factor threshold value of \( 6\,{\text{MPa}}\sqrt {\text{m}} \) reported in Reference 3. The specimen fails at \( \Updelta K_{\text{f}} = 1 7. 5\,{\text{MPa}}\sqrt {\text{m}} , \) which indicates that CT 3 undergoes a considerable decrease in toughness. The observed fatigue profile for the CT 3 specimen after H charging (Figure 6) combined with the metallographic evaluation (Figure 5) suggests that hydrides play a significant role in the mechanical behavior of this specimen.

On the other hand, a possible impact to the fatigue profile of CT 3 could be an overloading effect caused by the reduction in applied load from 4000 N to 1300 N. However, calculation of ΔK using Eq. [1], a = 22.86 mm (crack length when the H charging was employed) and Pmax = 1300 N, yields a value of \( \Updelta K = 7. 8\,{\text{MPa}}\sqrt {\text{m}} . \) Previous tests on Zircaloy-4 specimens, similar to CT 1, have showed that the near-threshold ΔK is about \( 10. 7\,{\text{MPa}}\sqrt {\text{m}} . \) Therefore, under the current fatigue conditions, the crack would not have propagated at such a low ΔK in the absence of zirconium hydrides. The values for m and C are 5.9 and \( 2. 9\times 10^{{ - 6 { }}} {\frac{\text{mm/cycle}}{{\left( {{\text{MPa}}\sqrt {\text{m}} } \right)^{m} }}}, \) respectively. The different m and C values obtained for the three cases, Table I, are from the reduction of toughness with the hydride content and the increase in the crack propagation rates. Thus, the slope for the Paris regime becomes steeper from CT 1 to CT 2 and to CT 3.

The crack propagation rates for all three cases are compared and presented in Table II. First, at \( \Updelta K = 1 3\,{\text{MPa}}\sqrt {\text{m}} , \) as discussed for CT 1 and CT 2, da/dN increases from 1.1 × 10−4 mm/cycle to 1.6 × 10−4 mm/cycle, whereas da/dN for CT 3 is about 2.9 × 10−4 mm/cycle. At \( \Updelta K = 1 6. 5\,{\text{MPa}}\sqrt {\text{m}} , \) which is the end of Paris regime for CT 2, the crack growth rates increase from 3.2 × 10−4 mm/cycle for CT 1 to 5.6 × 10−4 mm/cycle for CT 2 and to 2.9 × 10−3 mm/cycle for CT 3. Thus, one can observe that at both positions, CT 3 exhibits significant faster crack propagation rates.
Table II

Summary of Fatigue Behavior: Crack Propagation Rate (da/dN) at Two Different ΔK Values for the Three Cases Studied

Specimen/Fatigue Behavior

CT 1

CT 2

CT 3

da/dN × 10−4 at \( \Updelta {\text{K}} = 1 3\,{\text{MPa}}\sqrt {\text{m}} \)




da/dN × 10−4 at \( \Updelta {\text{K}} = 1 6. 5\,{\text{MPa}}\sqrt {\text{m}} \)




Finally, after failure, the fracture surfaces of CT 1, CT 2, and CT 3 specimens were investigated using SEM to identify fatigue and fracture mechanisms. The fatigue striation spacing, observed by SEM and represented by symbols in Figure 6, is compared with the crack propagation rate, da/dN, during the fatigue tests.

For CT 1, the striation spacings represented by rhombi are very similar to the crack growth rates. For CT 2 and CT 3, fatigue striations could be found only in localized areas, as indicated by the symbols in Figure 6. The spacing values for CT 2 (dot) and CT 3 (squares) suggest a larger striation spacing for the hydrided specimens.

The fracture surfaces for the three specimens were investigated at corresponding ΔK values. For the CT 1 specimen, SEM fractographs are presented in Figures 7(a) through (f) for stage I, II, and fast fracture, respectively. In addition, an example of the fatigue striations is shown in the inset of Figure 7(b). This specimen clearly undergoes a ductile behavior. For the CT 2 case, the SEM images (Figures 8(a) through (f)) were taken at ΔK values similar to CT 1. It is worth noting that both Figures 8(a) and (b) correspond to stage I. The fatigue striations observed are presented in the inset of Figure 8(c). The CT 2 specimen displays a brittle behavior with interfacial fracture, dimples, cleavage-like facets, and secondary cracking. The fast fracture surface (Figure 8(f)) reveals a combination of voids and transgranular fracture. The SEM images for the CT 3 specimen are shown in Figures 9(a) through (f) for stages II and III of crack growth and fast fracture, respectively. The inset of Figure 9(b) represents the observed fatigue striations. This specimen also exhibits a brittle behavior with cleavage-like facets and secondary cracking. After failure, the fast fracture surface shows a transgranular fracture (Figure 9(f)).
Fig. 7

SEM fractographs for CT 1 specimen. Images are taken at different ΔK values corresponding to different stages of the crack growth. (a) Stage I, (b) through (e) stage II, and (f) fast fracture surface. An example of the fatigue striations is shown in the inset of (b)
Fig. 8

SEM fractographs for CT 2 specimen. Images are taken at different ΔK values corresponding to different stages of the crack growth. (a) and (b) stage I, (c) and (d) stage II, (e) stage III, and (f) fast fracture surface. An example of the fatigue striations is shown in the inset of (c)
Fig. 9

SEM fractographs for CT 3 specimen after the H charging (FCG 2 in Figure 1(d)). Images are taken at different ΔK values corresponding to different stages of the crack growth (a) through (c) stage II, (d) and (e) stage III, and (f) fast fracture surface. An example of the fatigue striations is shown in the inset of (b)

3.3 Residual Strains Near the Fatigue Crack

The residual strains accompanying the fatigue crack, (a = 22.86 mm as shown in Figure 1(d)) were measured experimentally with neutron diffraction on the CT 3 specimen prior to hydrogen charging. Figure 10 shows the zirconium residual a and c lattice strain profiles from 4 mm behind the crack tip to 20 mm in front. Both a and c axes exhibit compressive (negative) strains at positions near the crack tip (−4 to 1 mm) with balancing tensile (positive) strains, which reach a maximum of 200 με at 2 mm in front of the tip. More neutron diffraction strain results obtained after the hydrogen charging will be presented in a future publication.
Fig. 10

Evolution of zirconium residual lattice strain, εa and εc, for the CT 3 specimen in the as-received condition prior to hydrogen charging. The neutron diffraction lattice strain measurements were conducted from 4 mm behind the crack tip to 20 mm in front of the tip, as indicated in Figure 2, where the crack length a is 22.86 mm. Both a and c axes exhibit a maximum tensile strain in front of the tip

4 Discussion

4.1 Effect of a Homogenous Hydrogen Charging on Fatigue Behavior

The effects of hydride formation on the Zircaloy-4 microstructure, crack growth behavior, and fracture surfaces are discussed by comparing CT 1 and CT 2 specimens.

The zircaloy alloy used for this research exhibits a basketweave morphology, which can form during nuclear reactor’s fuel sheathing brazing cycle when the alloy undergoes cooling from the β-Zr region temperature to room temperature (α-Zr phase).[1,30,31] The presence of the second phase precipitates in the alloy’s matrix can act as nucleation sites for the hydride phase.[34] Other possible trapping sites for hydrides to precipitate can be vacancies, dislocations, and microvoids as well. As is shown in Figure 5, the hydrides are precipitating at boundaries of the α-Zr plates, as also suggested by the EBSD investigation. Moreover, the location of hydrides coincides with the fracture path, as probed by SEM, influencing the fatigue behavior. This trend is consistent with that observed in Ti alloys with Widmanstätten microstructures, which exhibit a crack path that is very dependent on microstructure.[35] It should be pointed out that the hydrides have very low fracture toughness on the order of \( 1\,{\text{MPa}}\sqrt {\text{m}} . \)[8] Therefore, any stress intensity greater than this value will cause fracture of the hydrides. As a result, crack propagates, in part, along the α-plates’ boundaries through hydrides, leading to accelerated crack growth rates.

The comparison of the beginning of the fatigue life (stage I, \( \Updelta K \le 1 4\,{\text{MPa}}\sqrt {\text{m}} , \) Figure 6) for both specimens (CT 1 and CT 2) shows that in the case of CT 2, the crack propagates slower at larger ΔK values. However, the SEM images taken at ΔK within stage I (Figure 7(a), and Figures 8(a) and (b)), indicate that CT 1 has a flat, featureless fracture surface, typical of ductile alloys,[11] whereas CT 2 exhibits a rougher surface. This observation may be explained by the presence of hydrides in the bulk of the CT 2 specimen. Small ΔK values, such as \( \Updelta K = 1 3. 3\,{\text{MPa}}\sqrt {\text{m}} , \) combined with the basketweave microstructure favor a small plastic zone in front of the crack tip.[36] Subsequently, this trend affects the zirconium plates with the hydrides at the boundaries that can be lifted out from the matrix as the crack propagates through it, resulting in the observed rough brittle fracture surfaces. The presence of hydrides in the boundaries between the different interlocking variants of the alpha phase (Figure 3(c)) would cause the advancing crack to move out of its plane by distances equal to some fraction of the variant size. The difference in roughness of these surfaces, across a nominal area of 1 × 4 mm2, was confirmed using a laser scanning microscope. Figure 11 shows the three-dimensional (3-D) images obtained for both CT 1 and CT 2 specimens, and the average surface roughness observed is 17.6 μm and 28.8 μm, respectively. Consequently, a rougher fracture surface can cause an increase in the crack closure effect, which involves larger ΔK values for cracks to grow and slower propagation rates. Information about crack closure phenomenon and implicit calculation of Kclosure, can be obtained by a linear fit of the load vs displacement curve, acquired for each cycle during the fatigue test.[3739] One also can calculate ΔKeffective, as the difference between Kmax and Kclosure.[38,39] Figure 12 shows the crack propagation rates as a function of the ΔKeffective for CT 1 and CT 2 specimens. It is clear that at the beginning of the fatigue life and without the crack closure effect, using the ΔKeffective, the CT 2 specimen exhibits crack propagation at much smaller ΔK values than the CT 1 specimen.
Fig. 11

3-D representation of the surface roughness for CT 1 (top) and CT 2 (bottom) specimens taken from the area corresponding to stage I FCG. The average roughness values obtained are 17.6 μm and 28.8 μm for CT 1 and CT 2, respectively
Fig. 12

Crack propagation rates, da/dN, as a function of the effective stress intensity factor, ΔKeffective, where ΔKeffective = KmaxKclosure for both specimens, CT 1 and CT 2, showing the crack closure effect

However, in Figure 6, at greater ΔK values in stage II for the CT 2 specimen, the crack propagates faster in the presence of hydrides through the bodies of the plates, leading to interfacial fracture and cleavage as observed on the SEM images. This trend is a result of the increase of plastic zone size in front of the crack tip with the increase in ΔK values. The influence of hydrides also could be observed by studying the fatigue striation spacing for the two specimens. The spacing between the fatigue striations is related to the blunting–sharpening sequence that a crack follows during the cyclic fatigue testing.[11] In the presence of zirconium hydrides (CT 2 case) the crack propagates faster. Thus, not enough time is available for a complete blunting–sharpening sequence per fatigue cycle, leading to the observed discrepancy between the crack propagation rate and striation spacing (Figure 6).

At even larger ΔK values, the crack propagation for the CT 2 case becomes unstable, and stage III begins. The propagation rate of the crack is too fast to be affected significantly by the environment. Thus, the presence of the hydride phase plays an important role on the fatigue behavior by enhancing the crack propagation leading to interfacial fracture and reduced toughness.

4.2 Effect of a Preexisting Crack on Hydride Distribution and Subsequent Fatigue Behavior

The effects of a preexisting crack and the associated residual strains on the fatigue behavior of Zircaloy alloy will be discussed by comparing the CT 3 specimen with the CT 2 and CT 1 specimens. The residual strains accompanying the fatigue crack were measured with neutron diffraction on the CT 3 specimen prior to hydrogen charging. The obtained profiles presented in Figure 10 indicate that a and c axes exhibit compressive behavior at positions near the crack tip (−4 to 1 mm) and tensile trends in front of the tip. This observation is consistent with that reported on stainless steel.[12] Furthermore, the observed strain gradient at the crack tip is expected to enhance the hydrogen diffusion during charging toward the crack tip. This observation may be a result of the tensile residual stress exhibited by of both a and c axes in front of the crack tip, which reduces the chemical potential of the hydrogen.[40,41] When the lattice strain gradient and hydrogen diffusion are accompanied by lattice dilation, a mechanical relaxation is associated with it, also known as the Gorsky effect.[42,43] Consequently, the Zr-lattice strain gradient can attract a proportional hydrogen concentration at the crack tip, leading to a higher hydride concentration upon precipitation. Thus, the tensile strain promotes the hydrogen concentration in front of the crack tip but also acts as a promoter for the crack propagation.[14] However, it was observed that under compressive residual stress, zirconium alloys can accept a high hydrogen content without significant damage.[3] This trend may explain the presence of hydrides not only at the crack tip but also along the crack. It is shown in Figure 5 the concentration of hydrides in the CT 3 specimen at the crack is greater than in CT 2, which has no crack. The higher concentration of hydrogen at the crack tip during charging eventually enhances the hydrides effect on the fatigue behavior of the CT 3 specimen, which became more severe than in the CT 2 case. Note that the fatigue behavior curve (Figure 6) showed that after H charging, CT 3 exhibits crack propagation even at \( \Updelta K = 7\,{\text{MPa}}\sqrt {\text{m}} \) under an applied load of 1300 N as well as overall higher crack growth rates.

5 Conclusions

The hydride phase formation and its influence on fatigue behavior of a Zircaloy-4 alloy were investigated. It was shown that the specimen in the as-received condition has a Widmanstätten basketweave morphology that forms during transformation from the β to α phase. The former β-Zr grains have a mean diameter of 700 μm, and each prior grain contains multiple α-Zr plates with several preferred orientations relative to the parent β-grain. The alloy exhibits second-phase precipitates of Fe-Cr in its matrix, and their location and density serve as nucleation sites for the hydride phase. The crack propagation of the as-received specimen remain stable until failure occurs, yielding a ductile behavior. When charged with hydrogen, the Zircaloy forms δ-ZrH2 at the boundaries of α-Zr plates, which is also the path of fracture. The specimen charged with hydrogen and then fatigued to failure underwent a homogenous distribution of hydrides in the bulk, reduced toughness, reduced ductility, and accelerated crack propagation rates. Finally, it was shown that the presence of a gradient in a residual lattice strain associated with a crack tip prior to hydrogen charging plays a critical role in hydride formation. The hydride concentration is higher at the crack leading to a more severe degradation of fatigue properties.


E. Garlea acknowledges the support of the National Science Foundation (NSF) International Materials Institutes (IMI) Program (DMR-0231320) and the Tennessee Advanced Materials Laboratory Fellowship Program. E. Garlea is grateful to Drs. D.A. Smith and S.J. Randolph for valuable suggestions regarding the nickel sputtering. This work has benefited from the use of Lujan Neutron Scattering Center at LANSCE, which is funded by the Office of Basic Energy Sciences (Department of Energy). Los Alamos National Laboratory is operated by Los Alamos National Security LLC under DOE Contract De-AC52-06NA25396. EBSD analysis was conducted at the SHaRE User Facility, which is sponsored at Oak Ridge National Laboratory by the Division of Scientific User Facilities, Office of Science, U.S. Department of Energy.

Copyright information

© The Minerals, Metals & Materials Society and ASM International 2010