Materials Science and Engineering A 528 (2011) 5983–5992

Contents lists available at ScienceDirect

Materials Science and Engineering A

journal homepage: www.elsevier.com/locate/msea

Modeling the fatigue crack growth behavior of Ti-6Al-4V by considering grain

size and stress ratio
Bernd Oberwinkler ∗
Montanuniversität Leoben, Chair of Mechanical Engineering, Franz-Josef-Straße 18, A-8700 Leoben, Austria

a r t i c l e i n f o a b s t r a c t

Article history: Ti-6Al-4V is a commonly used titanium base alloy in aerospace applications. The increasing demand
Received 27 August 2010 for damage-tolerant designs of such components necessitates a detailed knowledge of its crack growth
Received in revised form 12 April 2011 behavior. The aim of this research was the characterization and phenomenological modeling of long crack
Accepted 14 April 2011
growth behavior with respect to microstructure and stress ratio. Therefore, the long crack propagation
Available online 21 April 2011
was characterized for eight different heat treatment conditions and four stress ratios. For comparison,
physically short crack growth tests were also performed. The long crack growth threshold was found
Keywords:
to be dominated by roughness-induced crack closure, and the fracture surface roughness is controlled
Titanium alloys
Fatigue crack growth
by the primary ␣-grain size. The reason for this correlation is a near-threshold crack propagation mode,
Microstructure which is dominated by the transcrystalline fracture of ␣-grains. This correlation was used to model the
Stress ratio crack growth threshold with respect to microstructure. A linear relation was determined between the
stress ratio and the threshold value, which was also found in this approach. Further presented models
cover the crack growth behavior in the near-threshold (Stage I) and mid-growth rate regions (Stage II).
© 2011 Elsevier B.V. All rights reserved.

1. Introduction size due to the crack growth resistance contributions mentioned

above.
Forged components made of the titanium base alloy Ti-6Al- Gray and Lütjering [2] also observed that an increase in the
4V are generally used in the aerospace industry, e.g., for engine equiaxed grain size reduces the fatigue crack propagation rates
mounts, pylon fittings and frame parts. The increasing demand of Ti-6Al-4V for low stress ratios. They attributed the influence of
for damage-tolerant designs in such applications necessitates a grain size and the increase in fatigue crack propagation resistance
detailed knowledge of its crack growth behavior. Several authors with decreasing stress ratio to roughness-induced crack closure.
have investigated the short and long crack growth behavior of Furthermore, they observed that coarse lamellar microstructures
Ti-6Al-4V. Their studies have shown that the crack propagation possess (in comparison to equiaxed microstructures) a higher resis-
behavior is strongly influenced by the stress ratio, microstruc- tance to fatigue crack propagation due to roughness-induced crack
ture, environment and temperature. However, a prediction of crack closure. Exposure to a more aggressive environment (ambient air)
growth behavior, with respect to these influences, for forecasting was found to drastically reduce the crack propagation resistance of
lifetimes is still lacking. Ti-6Al-4V for both equiaxed and lamellar microstructures in com-
Wagner and Gregory [1] reported the crack growth behavior parison to the experimental results in vacuum.
of microstructurally short cracks and long cracks in coarse lamel- Lütjering and Gysler [3] investigated the long crack propaga-
lar and fine equiaxed microstructures, respectively. They found tion behavior of Ti-6Al-4V for varying microstructures. Equiaxed
that the coarse lamellar microstructure exhibited a poorer short and bimodal microstructures were found to have a mixed
crack growth behavior but, due to the additional crack growth basal/transversal texture after rolling. They ascertained that mate-
resistance contributions of the crack front geometry and crack rials with larger ␣-grain sizes exhibited slightly lower long crack
closure, a slower long crack propagation as compared to the propagation rates for both high and low stress ratios. Similar find-
fine equiaxed microstructure. Furthermore, they ascertained a ings were reported by Peters et al. [4].
decrease in the long crack growth rate with increasing ␣-grain Lütjering and Williams [5] presented an example for the fatigue
crack propagation behavior of fully equiaxed microstructures with
an ␣-grain size of 6 ␮m. The propagation rates of small, self-
∗ Tel.: +43 3842 402 1466; fax: +43 3842 402 1402. initiated surface cracks (microcracks) measured on round smooth
E-mail address: [email protected] specimens and large cracks (macrocracks) measured on fracture-
URL: http://www.unileoben.ac.at/amb. mechanics-type specimens for two different stress ratios were

0921-5093/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.msea.2011.04.046
5984 B. Oberwinkler / Materials Science and Engineering A 528 (2011) 5983–5992

compared. Their results indicate that, even for relatively fine- Table 1
Summarized microstructural parameters of the analyzed materials.
grained equiaxed microstructures, a sufficiently large grain size
effect on the crack front profile of macrocracks exists, which Material ˛p [␮m] C␣+␤ [%] Con [%] Col [␮m]
explains the change in grain size ranking between microcracks and MAV 11.5/8.7 20.3 28.7 9.2
macrocracks observed in these experiments. The microcracks in STPQV 8.1 61.4 39.5 13.4
the fully equiaxed microstructure showed the tendency to grow STV 9.0 67.0 78.9 32.4
through areas containing interconnected ␣-grains whereas the MAP 8.2/7.3 30.9 45.2 7.1
STP 8.9 50.9 33.7 16.1
microcracks in the bimodal microstructure showed the tendency to
RAP 10.3 0 99.0 –
propagate through the lamellar (␣ + ␤)-phase. Although the macro- FB 9.8 0 0 –
scopic crystallographic textures measured by XRD were similar for
both microstructures, the slip length appeared to be larger in areas
with interconnected ␣-grains (equiaxed microstructure) than in
the lamellar (␣ + ␤)-phase of the bimodal microstructure.
The initiation and growth behavior of microstructurally short form of Ti-6Al-4V V-shaped (V) and side-pressed pancake (P) forg-
cracks with respect to microstructure are discussed in more detail ings, respectively. Additionally, the feedstock billet (FB, diameter:
in [6]. It was observed that the transition from equiaxed- (primary 230 mm) for the V-shaped forgings was investigated in as-received
␣-grain size controlled) to bimodal-type (colony length controlled) (AR) condition.
fatigue behavior occurs at a relatively low (␣ + ␤)-content of 20%. Different heat treatments were used for the forgings: mill-
Based on linear elastic fracture mechanics, it was further deter- annealing (MA), solution treatment with air cooling (ST), solution
mined that the average short crack growth rate is in accordance treatment with polymer quenching (STPQ), recrystallization-
with the extension of the Paris region for long crack growth data. annealing (RA) and beta-annealing (BA). These different heat
Nalla et al. [7] examined the mixed-mode fatigue crack growth treatments led to a high diversity of equiaxed and bimodal
thresholds of long and short cracks in bimodal Ti-6Al-4V. They microstructures, respectively (Fig. 1). Analysis of the different
found that the large crack growth thresholds were significantly microstructures was performed with a light optical microscope.
influenced by load ratio and mode mixture. These influences were The metallographic sections were grinded, polished and etched
primarily attributed to crack-tip shielding. Furthermore, they found (in HF-HNO3 solution). The microstructure was thus visible in
that the magnitude of physically and microstructurally short crack a light optical microscope, with the ␣-phase appearing bright
thresholds was significantly lower than that measured for large and the ␤-phase appearing dark. The mill-annealed V-shapes
cracks. Moreover, they were essentially independent of the mode and the mill-annealed pancakes showed a typical mill-annealed
mixture and load ratio, again suggesting a dominant role of shield- microstructure. Mill-annealing did not cause complete recrystal-
ing in mixed-mode fatigue threshold behavior. lization and therefore led to a distinct texture of the primary
The crack growth behavior of Ti-6Al-4V with respect to stress ␣-grain shapes, which is representative of the forging process. The
ratio and temperature was reported by Ding et al. [8]. They found no recrystallization-annealing of the pancakes resulted in a coarse
apparent effect of elevated temperature on the fatigue thresholds. equiaxed microstructure. In comparison to the coarse equiaxed
However, elevated temperatures increased the crack growth rates microstructure of the feedstock billet, no continuous ␤-seam was
of the near-threshold region and, depending on the stress ratio, observed at the ␣-grain boundaries. The solution treatment of
even of the Paris regime. the V-shapes and pancakes produced bimodal microstructures.
Dubey et al. [9] studied the crack growth behavior in mill- For comparison, a lamellar microstructure was achieved by beta-
annealed Ti-6Al-4V. They ascertained that the crack growth rates annealing.
increase with increasing stress ratios due to crack closure effects. Due to the chosen forging temperature (between 900 ◦ C
Transmission electron microscopy (TEM) of the crack tip revealed and 930 ◦ C), no crystallographic texture was observed in the
that crack extension in the Paris region occurs by alternating shear- microstructures, cf. [10]. This was confirmed by X-ray diffraction
ing and unzipping along intersecting slip bands. They developed a (XRD) and electron backscatter diffraction (EBSD) measurements.
three-parameter multiple regression model to predict fatigue crack The feedstock billet in the as-received condition also showed
growth data as a function of the stress intensity factor, the closure isotropy in XRD measurements and was therefore comparable to
stress intensity factor level and the stress ratio. However, this is a the forgings. However, macrozones were observed in the feedstock
complex model, which needs adapted fitting parameters for each billet in polished unetched micrographs, analyzed with polarized
microstructure. Hence, it is not easily applicable in practice. light; see Fig. 2 (l). The same structure is observable in the primary
The aim of the present research was to predict the long ␣-grain shape texture, Fig. 2 (r). It is believed that these macrozones
crack growth behavior of Ti-6Al-4V with an equiaxed or bimodal correlate with prior ␤-grains.
microstructure with respect to characteristic microstructural The microstructures were characterized by the parameters
parameters (in this case, as we shall see, the primary ␣-grain size) primary ␣-grain size ˛p , (␣ + ␤)-content C˛+ˇ , connectivity Con
and stress ratio. Furthermore, the model should be easily appli- of ␣-grains, and colony length Col of the (␣ + ␤)-lamellae. With
cable for engineers. Therefore, the long crack growth behavior the exception of the colony length, the parameters were deter-
of Ti-6Al-4V in eight different microstructural configurations was mined by an intercept approach. The (␣ + ␤)-content was defined
investigated with respect to the stress ratio. Physically short crack as the balance of 100% minus the primary ␣-content. The
growth experiments on mill-annealed Ti-6Al-4V were used as a connectivity is defined as the ratio of the mean size of inter-
reference. To gain a better understanding of the fatigue crack prop- connected ␣-regions and the mean primary ␣-grain size. The
agation behavior, light optical and scanning electron microscopy analyses of the different microstructures are summarized in
were additionally performed on the fracture surfaces. Table 1.
For the mill-annealed microstructures, two values were deter-
mined for the primary ␣-grain size, namely the average grain
2. Material characterization size in the direction of the grain shape elongation of three
cuts, and the mean grain size. For the crack growth models
The material used for this research was provided by Böhler discussed below, the mean grain size was determined to be appli-
Schmiedetechnik GmbH & Co KG (Kapfenberg, Austria) in the cable.
 B. Oberwinkler / Materials Science and Engineering A 528 (2011) 5983–5992 5985

Fig. 1. Micrographs of the analyzed materials.

3. Experimental procedure 3.1. Single-edge bending tests

The crack growth behavior of Ti-6Al-4V was characterized The characterization of the long crack growth behavior was
with respect to microstructure and stress ratio (ratio of mini- achieved with V-notched single-edge bending (SEB) specimens
mum and maximum stress or stress intensity factor, respectively) (dimensions: 80 mm × 20 mm × 10 mm) under four-point-bending
with single-edge bending tests. Additionally, physically short crack loading, cf. ASTM E647 [11], using a Rumul Cracktronic resonant
growth tests were performed on the V-shaped mill-annealed mate- testing rig with a frequency of approximately 140 Hz. The crack
rial for two different initial crack lengths and for different stress length was measured with the potential drop method; see Fig. 3.
ratios. All crack growth tests were done at ambient air and room The distance between the contact pins for crack voltage measure-
temperature. ment was 6 mm. Temperature compensations were made with a

Fig. 2. Macrozones in the microstructure of the feedstock billet; unetched (l) and etched (r) micrographs.
5986 B. Oberwinkler / Materials Science and Engineering A 528 (2011) 5983–5992

Fig. 3. Schematic of the potential drop method.

Fig. 4. Single-edge bending long crack growth behavior with respect to the
microstructure and stress ratio.
Pt100 resistance temperature sensor fixed near the crack of the
specimens.
Measurements were collected for four different stress ratios R. rate in the region between 10−9 m/cycle and 10−7 m/cycle. The
The stress intensity factor range K was derived according to lin- ranking of the different microstructures regarding crack propaga-
ear elastic fracture mechanics (the geometry factor according to tion was not affected by the stress ratio.
Murakami [12]) for all stress ratios. The measurements were con-
√
ducted up to 80 MPa m due to testing rig limitations regarding 3.2. Physically short crack growth experiments
the maximum stress intensity factor range. At least two specimens
were tested for each stress ratio and microstructure. The testing If a crack is short (possibly microstructurally short) in the crack
procedure consisted of three stages: constant K testing for crack propagation direction and long perpendicular to the crack propaga-
initiation up to a crack length of 6 mm, K-decreasing testing from tion direction, it can be classified as physically short. In this case, an
the mid-growth rate region down to the threshold, and subsequent influence of local microstructure on the global crack propagation
K-increasing testing with a constant bending moment from the behavior can be excluded. Planar short crack growth (SCG) spec-
beginning of the mid-growth rate region to fracture or the testing imens (gauge section dimensions of 8 mm × 4 mm) were used for
rig limit. Carboni et al. [13] reported that this testing method could this research; the narrow side was provided with a wire-eroded
lead to non-conservative threshold values, mainly due to plasticity- initial crack with a width of 0.04 mm and an initial crack length a0
induced crack closure. Therefore, the size of the plastic zone after of 0.2 or 0.4 mm, respectively.
the constant K testing was determined by both analytical estima- The SCG specimens were tested under tension/compression
tion and FE analysis, cf. [6]. It was found that the size of the plastic loading for three different stress ratios R (−1, 0 and 0.3) on a
zone was in the range of the grain size and that plasticity-induced servo-hydraulic test stand at room temperature and ambient air
crack closure could thus be neglected. with a testing frequency of 30 Hz. The measurement of the short
The measured data points were fitted with a newly developed crack propagation was done with a specially developed camera
four-parameter equation, Eq. (1), where a is the crack length, N is device. A compact, digital progressive charge-coupled device (CCD)
the number of load cycles, K is the range of the stress intensity monochrome area scan camera with 1380 × 1040-pixel resolu-
factor, and p1,2 and q1,2 are parameters resulting from the fitting. tion and eight-fold magnification was used in combination with
The value with the lowest crack growth rate was thereby weighted. dark field illumination. This illumination technique capitalizes on
da p1 · log K + p2 oblique illumination to enhance the contrast of surface defects. The
log = 2
(1) crack thus appears bright with a dark background, cf. [15]. The
dN (log K) + q1 · log K + q2
specimens were polished and ultrasonically cleaned to avoid light
This fit results in an ideal representation of the data points in the scattering due to surface roughness or impurities. The fatigue tests
near-threshold and Paris region and is hence suitable for a visual were interrupted after certain load cycles to image the crack with
comparison of the varying microstructures. Due to the high varia- the camera system; then, the fatigue tests were continued. This
tion and complex interaction of these parameters, they cannot be procedure was repeated until failure.
used for linking crack growth and stress ratio or microstructure. The crack initiation phase was determined for all stress ratios.
Fig. 4 shows the long crack growth behavior of all tested The cycles completed before the first crack propagation detection
microstructures. The equiaxed microstructure of the feedstock in the notch root (crack elongation a ≈ 0.02 mm) were related to
billet exhibited the slowest crack propagation rate, whereas the cycles to failure. It was found that the average crack initiation
the mill-annealed microstructure of the pancake showed rela- phase was about 25% of the total life.
tively fast long crack growth. Crack growth was found down to The plastic zone near the crack tip is small in Ti-6Al-4V, cf. [6] or
propagation rates of 10−12 m/cycle, which was also reported by [16]. The application of linear elastic fracture mechanics is hence
Stanzl-Tschegg [14]. For crack growth rates higher than approx- possible. The distance within which the local notch stress field
imately 10−7 m/cycle, all investigated microstructures showed dominates the stress intensity solution is generally a small fraction
almost the same crack propagation behavior. Distinct differences of the notch radius [17], in this case, approximately 4 ␮m. To deter-
were found in the near-threshold region with respect to the stress mine the short crack growth curves, the stress intensity factor was
ratio. Based on the literature information discussed above, it can be therefore chosen according to Murakami [12], Eq. (2), where  is
assumed that this behavior is caused by crack closure. Both mill- the stress range, a is the crack length, and W is the specimen width.
annealed microstructures had an anomalously high crack growth The crack length was measured with the camera system described
 B. Oberwinkler / Materials Science and Engineering A 528 (2011) 5983–5992 5987

a light optical, a scanning electron, and a laser-scanning micro-

scope.
The influence of roughness-induced crack closure on the near-
threshold region was qualitatively observed by comparing the
different fracture surfaces for a certain stress ratio by light optical
microscopy, cf. Fig. 6 (R = 0, crack propagation direction from left to
right). It was ascertained that those microstructures that led to high
fracture surface roughness also exhibited high crack growth thresh-
olds. It was determined that the highest fracture surface roughness
(darker regions of the fracture surfaces in Fig. 6) generally arises
from crack propagation in the near-threshold region.
Scanning electron microscopy (SEM) was used to clarify the
observed correlations. The presented images (Figs. 7 and 8) were
taken from the middle of the SEB specimens (plane strain state),
loaded in the near-threshold crack propagation region and at the
beginning of the mid-growth rate regime. The long crack propaga-
tion direction relating to these images was from left to right.
The near-threshold crack growth leads to a higher surface
Fig. 5. Comparison of physically short crack growth (SCG) data and long crack roughness for both equiaxed- and bimodal-type microstructures
growth curves. (Fig. 7). Transcrystalline fractures of primary ␣-grains were deter-
mined to be the dominating fracture feature for near-threshold
above. (Stage I) crack propagation, where the quantity of transcrystalline
 a  a 2 fractured ␣-grains decreases with increasing (␣ + ␤)-content. How-
√ ever, a fracture surface analysis of the solution-treated V-shaped
KI =   a · 1.12 − 0.231 + 10.55
W W samples revealed that this type of transcrystalline fracture could
 a 3  a 4  also occur in ␣-lamellae of the (␣ + ␤)-phase, if the thickness is large
−21.72 + 30.39 (2) enough. In this case, these large fracture planes increase both the
W W fracture surface roughness and the long crack propagation thresh-
old. An anomaly was determined for the feedstock material. Large
The physically short crack propagation results were compared fracture features (>200 ␮m) were observed, which was ascribed
with the long crack growth curves obtained from single-edge bend- to the macrozones in the microstructure discussed previously (cf.
ing tests; see Fig. 5. It was found that the physically short crack Fig. 2).
growth data and long crack growth curves match for all three tested Three transcrystalline fracture planes of the recrystallization-
stress ratios. The physically short cracks do not grow faster than annealed microstructure are shown in Fig. 8. The crack propagation
the long cracks, and no crack propagation was observed beneath direction is again from left to right. At first glance, the transcrys-
the long crack growth threshold. A higher scatter of crack growth talline planes seem to exhibit a “river pattern”, typically seen
data was determined for the short cracks. This can be attributed to for cleavage fractures, see, e.g., [21]. However, the observed frac-
the type of crack length measurements employed. The crack length ture plane patterns show their division in the opposite direction.
was measured with the camera system on the surface of the spec- Following again the analogy of a river, the determined fracture
imens. The local microstructure in the surface region hence affects characteristic in Ti-6Al-4V is named a “delta pattern”, much like
the measured crack growth rate, especially in the near-threshold a river opening into the ocean.
region, where the stress intensity factor and crack-driving force are It was assumed that a crack initiates by a prismatic slip (due to
small. Furthermore, it was found that the scatter escalates at the the easier activation, cf. [22]) near a grain boundary (stress raiser),
transition from the mid-growth rate (Stage II) to near-threshold forming the “river” of the delta pattern. Then, the crack propagates
propagation (Stage I), which will be discussed in detail below. along a transcrystalline fracture plane, generating the delta pattern
by a sort of striation formation. Ductile striations are clearly observ-
4. Discussion able in the left image of Fig. 8. With reference to Knobbe et al. [23],
it was assumed that the observed fracture planes are basal planes.
The near-threshold region of the SEB long crack growth curves They concluded that alloying elements such as aluminum or inter-
exhibited distinct differences with respect to the microstructure stitial oxygen lead to a reduction in the stacking fault energy on
(cf. Fig. 4). Regarding Ti-6Al-4V, several authors, e.g., Gray and the basal planes, cf. [24], with resulting dislocation dissociation and
Lütjering [2], have assumed that such differences are caused by restriction of their movement on the basal planes.
roughness-induced crack closure. Some general considerations on The striation dimensions of the delta pattern shown in Fig. 8 (l)
crack closure were discussed by Ritchie and Suresh, cf. [18,19]. were measured to determine the local crack growth rate. The crack
They concluded that roughness-induced crack closures in the near- growth rate within an ␣-grain, 6 × 10−7 m/cyc, was determined to
threshold region result from fatigue crack growth by a single shear be more than a decade higher than the global crack growth rate of
√
decohesion mechanism (Stage I) with associated Mode II + I crack 1 × 10−8 m/cyc (loading conditions: R = 0.3, K = 9.5 MPa m). This
tip displacements. The resulting serrated or faceted fracture sur- means that the crack formation within an ␣-grain accounts for
faces coupled with Mode II displacements thus induce high closure approximately 98% of the time of crack propagation on the scale of
loads by wedging the crack at discrete contact points. Pippan et al. the average grain size. However, it can be assumed that the elon-
[20] attributed the occurrence of roughness-induced crack closure gated ␣-grains of the mill-annealed microstructures account for
far behind the fatigue crack tip to an asymmetric arrangement the same time of crack formation for a crack propagation distance,
of geometrically necessary dislocations in the crack wake, which and hence a crack growth rate, that is approximately two times
induces a mismatch of fracture surfaces. higher. It is believed that this is the reason for the anomalously high
To determine the influence of fracture surface roughness on the crack growth rates of the mill-annealed microstructures mentioned
crack growth behavior, the fracture surfaces were analyzed with previously (cf. Fig. 4).
5988 B. Oberwinkler / Materials Science and Engineering A 528 (2011) 5983–5992

Fig. 6. Fracture surfaces of SEB specimens tested with a stress ratio R = 0.

Fig. 7. Differences in fracture surface roughness: equiaxed-type (l) and bimodal-type (r) microstructures (note the different scales for different microstructures).

With respect to the appearance of and change in fracture surface tified in these fracture surfaces. These delta patterns are located
roughness from near-threshold to Paris-region crack propagation, at different heights (with respect to the fracture surface) and are
it can be assumed that the lower crack-driving force in the near- twisted or tilted against each other, leading to a high fracture
threshold region results in crack growth, especially in preferably surface roughness. In the near-threshold region, the primary ␣-
oriented grains; see Fig. 9 (l). Starting from an existing crack (1), grain size plays a major role with respect to roughness-induced
the stress field in front of the crack tip causes slip and crack for- crack closure. Larger primary ␣-grains lead to more crack deflec-
mation in a preferably oriented grain (2). If the stress is high tion, higher fracture surface roughness and slower crack growth.
enough, the non-preferably oriented in-between grain fails (3). Therefore, the crack propagation behavior of microstructures
This mechanism continues (4), (5). Therefore, many transcrystalline with different primary ␣-grain sizes varies in the near-threshold
(supposably basal) fracture planes (delta patterns) can be iden- region.

Fig. 8. Transcrystalline fracture planes (delta patterns) observed on the fracture surface of pancake RA; near-threshold, R = 0.3.
 B. Oberwinkler / Materials Science and Engineering A 528 (2011) 5983–5992 5989

rate decrease due to an abrupt introduction of roughness-induced

crack closure.
The roughness transition for near-threshold and mid-growth
rate propagation is sharp for the mill-annealed microstructure,
which was also determined at the fracture surfaces of the single-
edge bending specimens. This can be attributed to the high primary
␣-content of the mill-annealed microstructure because the near-
threshold fracture surface roughness is controlled by the primary
␣-grain size. For microstructures with fewer primary ␣-grains,
this transition is fuzzy, as confirmed by the fracture surfaces
Fig. 9. Crack propagation in the near-threshold region (l) and in the Paris region (r). of the single-edge bending specimens for different microstruc-
tures, cf. Fig. 6. The transition from a rough (dark) surface to a
Crack propagation in the Paris regime, shown in Fig. 9 (r), smoother (bright) surface is thereby, for example, sharp for the
involves a higher crack-driving force and higher crack growth mill-annealed pancake (C˛+ˇ = 31%) and fuzzy for the solution-
rates. Therefore, crack growth also occurs in non-preferably ori- treated and polymer-quenched V-shaped sample (C˛+ˇ = 61%).
ented grains (2), (4). The near-threshold propagation mechanism With respect to the crack growth behavior in the Paris region
discussed before is too slow and is outrun by “conventional” tran- (da/dN > 7 × 10−8 m/cyc), it was determined that the microstruc-
scrystalline crack growth (crack extension by alternating shearing ture has no significant influence (cf. Fig. 4). It can be assumed
and unzipping along intersecting slip bands, cf. [7]). This leads to a that this is caused by two facts: (1) there is no formation of delta
decreased fracture surface roughness. patterns and hence no significant influence of grain size on the con-
The fracture surfaces of the flat specimens showed the same comitantly reduced fracture surface roughness, and (2) there is no
characteristics as the single-edge bending specimens. The first mil- roughness-induced crack closure due to Stage II crack growth and
limeter of the fracture surface exhibits a higher fracture surface the omission of Mode II crack tip displacements (cf. [18]). The stress
roughness due to near-threshold crack propagation; see Fig. 10. ratio has also no significant influence on the crack growth behavior
The topography of this transition region was analyzed by confo- in the Paris region, as long as R ≥ 0. This can be attributed to the
cal laser-scanning microscopy. It was found that the center-line fact that no crack closure occurs, neither roughness-induced nor
average roughness Ra decreases abruptly from 4.5 ␮m to 3.4 ␮m. plasticity-induced. However, if load-induced crack closure occurs
The explanation for this transition in the crack growth mode was (R < 0, the minimum stress intensity factor is negative, the load con-
discussed previously. tains a compressive fraction), the crack growth rate is reduced. The
However, the fracture surface roughness of the physically short comparison of the crack growth curves for R = −1 and R = 0 shows
crack growth specimens was significantly reduced at a certain tran- that the effective stress intensity factor range for such load cases
sition crack length, depending on the stress ratio and independent equals the mean stress intensity factor plus the stress intensity fac-
of the initial crack length. It was determined that these transition tor amplitude. This means that the compressive fraction of the load
crack lengths correspond to a certain crack growth rate, namely (which causes load-induced crack closure) has no crack-driving
7 × 10−8 m/cyc. This finding confirms the approach of the crack effect.
growth rate-dependent mode transition discussed above. Further- Roughness measurements were performed with a confocal
more, it was found that the physically short crack growth curves laser-scanning microscope in the region of near-threshold crack
for different stress ratios, cf. Fig. 5, exhibit a knee at this transi- propagation of the SEB fracture surfaces. It was observed that
tion crack growth rate, due to a greater decrease in crack growth the center-line average roughness Ra correlates with the pri-
rate with a decreasing stress intensity factor in the near-threshold mary ␣-grain size of the equiaxed and bimodal microstructures,
crack propagation mode. This abrupt decrease in the crack growth respectively (Fig. 11, open symbols). The microstructure of the mill-
rate can be explained by the findings of Ritchie and Suresh [18] annealed V-shape samples led to an average roughness lower than
in combination with the crack growth mode transition discussed predicted by the linear correlation. In contrast, the solution-treated
above. At a certain crack growth rate, the near-threshold propaga- V-shape specimens exhibited a higher fracture surface roughness,
tion mode is activated and leads to the appearance of delta patterns which can be ascribed to the contribution of the large (␣ + ␤)-
in the fracture surface. These delta patterns are caused by crack colonies to the fracture surface roughness.
growth on one primary slip system in the direction of growth and The average roughness height of the fracture surface can also be
a tensile component normal to this. The resulting crack tip Mode linked to the stress intensity factor threshold (Fig. 11, solid sym-
II displacements are consequently the reason for the crack growth bols). This linear correlation is especially in accordance with the

Fig. 10. Change in fracture surface roughness of a mill-annealed SCG specimen, R = 0, a0 = 0.4 mm.
5990 B. Oberwinkler / Materials Science and Engineering A 528 (2011) 5983–5992

Fig. 11. Correlation between primary ␣-grain size, stress intensity factor threshold,
and center-line average roughness.
Fig. 12. Relationship between the crack growth threshold and stress ratio.

data points for the forged equiaxed and bimodal microstructures;

the maximum deviation is 4%. A higher scatter (10%) was found for The fatigue crack growth behavior in the near-threshold region
the feedstock billet and the beta-annealed pancake samples. It can was described with the stress intensity factor range, which is neces-
be assumed that the observed macrozones, which produce large sary to achieve crack growth rates of 10−9 , 10−8 and 10−7 m/cycle.
features at the fracture surface, are the reason for the uncommonly It was determined that this stress intensity factor range at a cer-
high stress intensity factor threshold of the equiaxed microstruc- tain crack growth rate correlates with the stress intensity factor
ture of the feedstock billet. These features are so large (>200 ␮m) threshold, independent of the stress ratio; see Fig. 14. A discrepancy
that they were not measured as the fracture surface roughness; was only found for R = 0.7 for the crack growth rates of 10−8 and
instead, they correspond to waviness. However, they contribute to 10−7 m/cycle, because these crack growth rates are already located
the roughness-induced crack closure and increase the long crack in the Paris region of the crack growth curves.
growth threshold of the feedstock material. The correlations of the stress intensity factor threshold Kth
Finally, it can be concluded that the stress intensity factor √ √
[MPa m] and the stress intensity factor range K [MPa m] for
threshold is primarily influenced by roughness-induced crack clo- −9 −8 −7
crack growth rates of 10 , 10 and 10 m/cyc are given in Eq.
sure. Therefore, from the dependence of the fracture surface (4)–(6).
roughness on the primary ␣-grain size, there arises a link between
the stress intensity factor threshold and the primary ␣-grain size. 0.89
K1E−9 = 1.57 Kth (4)

1.06
5. Modeling of the crack growth behavior K1E−8 (Kth ≥ 3.8) = 1.82 Kth (5)

1.09
The long crack propagation of all analyzed equiaxed- and K1E−7 (Kth ≥ 4.2) = 3.40 Kth (6)
bimodal-type microstructures exhibited interdependencies on the
primary ␣-grain size due to roughness-induced crack closure.
This grain size effect on the fatigue crack propagation was con-
firmed by fracture surface roughness measurements (cf. Fig. 11).
The feedstock material exhibited anomalously high crack growth
thresholds due to macrozones in the microstructure and was there-
fore excluded from the model.
For a quantitative link between microstructure and crack
growth, the experimental stress intensity factor range thresholds
were plotted as a function of the stress ratio R (Fig. 12). A linear
relationship was chosen to fit the data points. Thereby, a fixed point
was found at a stress ratio of 0.87 and a stress intensity factor range
√
threshold of 1.75 MPa m. Carboni et al. [13] found a similar linear
correlation for the stress ratio and stress intensity factor thresholds
for steel. They reported that the stress intensity factor threshold for
a stress ratio R = −2 matches this relation as well. Therefore, it can
be assumed that the observed linear relation is valid in the region
−2 ≤ R ≤ 0.7.
The slope kth [–] of the fitting lines was linked to the microstruc-
tural parameter primary ␣-grain size ˛p [␮m], cf. Eq. (3) and Fig. 13.
The connectivity of primary ␣-grains does not affect the threshold.
This model leads to a coefficient of determination R2 of 0.97 for the
forged materials.

kth = −0.31 ˛p − 1.4 (3) Fig. 13. Relationship between the slope of the fitting lines and microstructure.
 B. Oberwinkler / Materials Science and Engineering A 528 (2011) 5983–5992 5991

Fig. 14. Correlation of the stress intensity at certain growth rates [m/cyc] and the Fig. 15. Comparison of experimental data and the crack growth model.
stress intensity factor threshold.

Eq. (5) is valid down to a stress intensity factor threshold of

√
3.8 MPa m. For lower thresholds, Eq. (7) applies.

K1E−8 (Kth < 3.8) = 7.5 (7)

Eq. (6) is applicable for stress intensity factor thresholds

√
Kth ≥ 4.2 MPa m. For lower stress intensity factor thresholds, Eq.
(8) applies.

K1E−7 (Kth < 4.2) = 16 (8)

The stress ratio plays a major role in fatigue crack propagation.

However, as mentioned above, crack growth in the Paris regime is
not significantly influenced by microstructure (cf. Fig. 4). Only for
a stress ratio R = 0.7 was an effect of microstructure determinable.
Therefore, a rough model was developed for approximation.
The fitting parameters C and m of the Paris-Erdogan equation, Eq.
(9), were averaged for the analyzed microstructures for low stress
ratios (Table 2).
da
= C · K m (9) Fig. 16. Correlation of the Paris coefficient and Paris exponent.
dN
A polynomial equation was used to determine the Paris expo-
incorporated in the model generation) are necessary for verifica-
nent m with respect to the stress ratio R, Eq. (10). With a knowledge
tion.
of the stress intensity factor corresponding to the crack growth rate
Finally, it was found that C and m can be linked for Ti-6Al-
da/dN = 10−7 m/cyc, which results from the near-threshold model
4V with an exponential correlation (Fig. 16); this has also been
described above, it is possible to predict the Paris region of the crack
reported for other structural alloys, cf. [25] or [26].
growth curve for −1 ≤ R ≤ 0.3.

m = 2.26 + 0.11 R − 0.15 R2 (10) 6. Conclusion

For R > 0.3, the near-threshold model already provides enough
information to approximate the Paris region. This is caused by the It was found that long crack growth thresholds can be linked to
fact that the majority of the Paris region of the crack growth curves the primary ␣-grain size. This correlation arises from the depen-
is affected (for these stress ratios) by the near-threshold crack prop- dence of fracture surface roughness on ␣-grain size. The long crack
agation mode via delta patterns, which was confirmed by SEM. growth threshold is primarily controlled by roughness-induced
The crack growth models described above were applied as crack closure.
examples for two different microstructures and two stress ratios. A higher fracture surface roughness was determined in the
The modeling results were furthermore compared with the exper- near-threshold region. This behavior was explained by crack initi-
imental data; see Fig. 15. A good accordance was observed. ation and propagation in ␣-grains causing transcrystalline fracture
Further experiments for another microstructure (one that was not features, named “delta patterns”, in the fracture surface. These
delta patterns are the reason for the correlation of fracture sur-
face roughness and primary ␣-grain size. The near-threshold crack
Table 2
Paris-parameters for low stress ratios.
propagation mode occurs solely underneath a certain crack growth
√ rate limit.
R [–] C [m/(Cyc. MPa m)] m [–] It was determined that the stress intensity factor range for cer-
−1 4.70E-11 2.00 tain crack growth rates in the near-threshold region correlates with
0 1.23E-10 2.26 the stress intensity factor threshold, independent of the stress ratio.
0.3 2.17E-10 2.28
This correlation was used to estimate the near-threshold crack
5992 B. Oberwinkler / Materials Science and Engineering A 528 (2011) 5983–5992

growth behavior based on the modeled thresholds for stress ratios [3] G. Lütjering, A. Gysler, in: G. Lütjering, U. Zwicker, W. Bunk (Eds.), Titanium  84
between −1 and +0.7. – Science and Technology, vol. 4, DGM, 1985, pp. 2065–2083.
[4] M. Peters, A. Gysler, G. Lütjering, in: H. Kimura, O. Izumi (Eds.), Titanium  80 –
The long crack growth in the Paris regime is almost indepen- Science and Technology, vol. 3, TMS-AIME, Japan, 1980, pp. 1777–1786.
dent of the microstructure. This is due to the missing influence of [5] G. Lütjering, J.C. Williams, Titanium, 2nd ed., Springer, 2007, ISBN 978-3-540-
roughness-induced crack closure. A rough model was developed 71397-5.
[6] B. Oberwinkler, A. Lettner, W. Eichlseder, Int. J. Fatigue 33 (2011) 710–718.
for approximation. [7] R.K. Nalla, J.P. Campbell, R.O. Ritchie, Int. J. Fatigue 24 (2002) 1047–1062.
The stress ratio does not affect the crack growth behavior in the [8] J. Ding, R. Hall, J. Byrne, Int. J. Fatigue 27 (2005) 1551–1558.
Paris region as long as R ≥ 0. For lower stress ratios, load-induced [9] S. Dubey, A.B.O. Soboyejo, W.O. Soboyejo, Acta Mater. 45 (7) (1997) 2777–2787.
[10] G. Lütjering, Mater. Sci. Eng. A 243 (1998) 32–45.
crack closure reduces the applied stress intensity factor range by
[11] ASTM E647 – Standard Test Method for Measurement of Fatigue Crack Growth
the compressive fraction of the load. Rates, ASTM International, West Conshohocken, PA, USA.
[12] Y. Murakami, Stress Intensity Factors Handbook, Pergamon Press, 1987, ISBN
0-08-034809-2.
Acknowledgements
[13] M. Carboni, L. Patriarca, D. Regazzi, Determination of Kth by compression
pre-cracking in a structural steel, J. ASTM Int. 6 (9) (2009).
The author would like to thank the Austrian Federal Ministry for [14] S. Stanzl-Tschegg, Int. J. Fatigue 28 (2006) 1456–1464.
[15] B. Oberwinkler, C. Oberwinkler, S. Redik, H. Leitner, Proceedings of the Twelfth
Transport, Innovation and Technology, the Austrian Federal Min-
International Conference on Fracture (ICF12), Canada, 2009, published on CD-
istry of Economics and Labor, the federal governments of Styria, ROM.
Tyrol, Vienna, Upper Austria and Lower Austria and the Austrian [16] B. Oberwinkler, Fatigue-proof and damage tolerant lightweight design of Ti-
Research Promotion Agency for funding this work, in the frame- 6Al-4V forgings. Doctoral Thesis. Montanuniversität Leoben, 2010.
[17] N.E. Dowling, Fatigue Eng. Mater. Struct. 2 (1979) 129–138.
work of the Austrian Aeronautics Research Program and FFG’s [18] R.O. Ritchie, S. Suresh, Metall. Trans. 13A (5) (1982) 937–940.
BRIDGE program. The support from Böhler Schmiedetechnik GmbH [19] S. Suresh, R.O. Ritchie, Metall. Trans. 13A (9) (1982) 1627–1631.
& Co KG Kapfenberg is also much appreciated. [20] R. Pippan, G. Strobl, H. Kreuzer, C. Motz, Acta Mater. 52 (2004) 4493–4502.
[21] T.L. Anderson, Fracture Mechanics: Fundamentals and Applications, 3rd ed.,
CRC Press, 2005, ISBN 0-84931-656-1.
References [22] F. Bridier, P. Villechaise, J. Mendez, Acta Mater. 56 (2008) 3951–3962.
[23] H. Knobbe, P. Köster, H.-J. Christ, C.-P. Fritzen, M. Riedler, Proc. Eng. 2 (2010)
[1] L. Wagner, J.K. Gregory, in: M. Peters, C. Leyens (Eds.), Titan und Titanlegierun- 931–940.
gen, WILEY-VCH Verlag, 2007, ISBN 978-3-527-30539-1, pp. 163–196. [24] S. Zaefferer, Mater. Sci. Eng. A 344 (2003) 20–30.
[2] G.T. Gray III., G. Lütjering, in: G. Lütjering, U. Zwicker, W. Bunk (Eds.), Titanium [25] K. Tanaka, Int. J. Fracture 15 (1) (1979) 57–68.

84 – Science and Technology, vol. 4, DGM, 1985, pp. 2251–2258. [26] F. Bergner, G. Zouhar, Int. J. Fatigue 22 (2000) 229–239.