The metastable β titanium alloys are considered as attractive materials for aerospace industries due to its high specific strength, excellent hardenability, and high resistance to fatigue crack-initiation (FCI) [1,2]. This excellent strength is associated with a large volume fraction of fine α precipitates in the grains of transformed-β (β_t) [3,4]. However, this microstructure design often leads to failure initiation due to interface cracking between α phase and the β matrix, especially undergoing cyclic loading. The metastable β titanium alloys exhibit higher fatigue performance, causing increased interest in studying the microstructure-fatigue performance relationship, nevertheless, the underlying fatigue damage mechanisms are still lags behind [5,6].

Recent studies indicate that the conventional representation in terms of single mechanism is perhaps an oversimplification as the fatigue data converge on two curves. This phenomenon is the duality in fatigue behavior and the two-stage S-N behavior where the fatigue lifetime data grouped into two separate S-N curves, especially in high-strength alloys including titanium alloys [[7], [8], [9], [10], [11], [12]], high strength steels [13,14], nickel-base superalloys [15], and γ-TiAl based alloy [16]. Typically, the short-life-distribution is generally associated with the surface-crack-initiated failure and the long-life-distribution usually occurs due to the internal-crack-initiated failure. When considered in relation to some recent studies [[17], [18], [19], [20], [21]], the variability in fatigue response often tends to separate into bimodal distributions representing life-limiting behavior and a mean-lifetime dominated behavior. The duality of the S-N curve is not an artifact but perhaps an unusual fatigue phenomenon caused by the material microstructure, such as the volume fraction of primary-α (α_p) phase [7]. The nature of dual fatigue behavior can be termed as “competing failures” as both surface- and internally-initiated cracks can occurred with nearly equal probability. Chandran and Jha [8] reported that these two mechanisms compete to cause fatigue failure of Ti-10V-2Fe-3Al with α + β microstructures. This competing failure phenomenon leads to the unpredictability of how and when a specimen or a component will fail under a given test condition, leading an increased dispersion of fatigue lifetime data. In addition, researches show that the depth of crack-initiation sites show a stress ratio dependence and shifts from surface to interior with increasing of stress ratio [[22], [23], [24]]. In some cases, the apparent shorter- and longer-life-distribution in terms of the S-N curve would not be observed, although surface- and internal-failures are evident [25]. To date, there is no convincing explanation for the presence of dual S-N curves. It is worth noting that some researchers have suggested that competing failures will not occur in notched specimens because the stress is concentrated in the root of the notch [26]. Reportedly, the behavior of the fatigue data of a nickel-based superalloy notched fatigue specimens showed an effect of competing modes, but the fatigue lifetime could be adequately represented by a single S-N curve [27]. Also, the stress state at a notch is multiaxial even under uniaxial loading, leading to large scatter in fatigue lifetime for notched specimens. The failure competition in notched fatigue behavior is still ambiguous and to be explored further.

FCI in metallic materials without significant metallurgical defects occurs due to the cyclic damage accumulation of irreversibly slipping dislocations. Plenty of studies on near-α and α + β titanium alloys evidenced that FCI is accompanied by facet formation across α_p particles [[28], [29], [30], [31], [32], [33], [34]]. Besides, many studies have suggested that the FCI facets form on or very near basal planes [[28], [29], [30], [31], [32],35,36], although FCI by prismatic facet formation has also been reported [28]. In terms of the faceting process, three kinds of mechanisms have been proposed and summarized in Refs. [17,18]. The pure slip is suggested as the most dominant factor in facet formation [30].

In the present work, we have discovered the occurrence of a clear duality of the S-N fatigue curve in the Ti-7Mo-3Nb-3Cr-3Al (Ti-7333 here afterwards) notched specimens. The two competing failure mechanisms are discussed based on microstructure observation and fractographic analysis. Furthermore, high cycle fatigue (HCF) damage mechanism was analyzed by electron back-scattered diffraction (EBSD) in a focused-ion-beam cross-section (FIB-CS) across the faceted grains, emphasizing the slip modes of faceted grains.

The material of interest in the present study is a metastable β titanium alloy Ti-7333. Its chemical composition (wt%) is given in Table 1. The β transus temperature (T_β) was measured to be approximately 850 ℃ by using the metallographic observation method. More specifically, the block specimens extracted from the Ti-7333 forged stocks were solution treated at 860 ℃, 855 ℃, 850 ℃, 845 ℃, 840 ℃ and 830 ℃ for 40 min with subsequent water quenched, then, these specimens were characterized using scanning electron microscope to detect the α phase percentage. Three kinds of forging stocks after different thermo-mechanical processing (TMP) routes were used to evaluate fatigue behavior. On one hand, the Ti-7333 square billets were forged at α + β region (∼790 ℃) and near β region (∼835 ℃), then, the same duplex annealing heat treatment process was conducted on fatigue specimen blanks machined from the forged plates by electric-discharge machining. To be precise, these specimens were solution treated at 820 ℃ for 1 h, furnace cooled, then annealed at 765 ℃ for 1 h, air cooled, and aged at 520 ℃ for 6 h, air cooled. The corresponding microstructures denoted by Ti7333-A and Ti7333-B, respectively. On the other hand, the fatigue specimen blanks machined from Ti-7333 forged bar were solution treated at 820 ℃ for 50 min, air cooled, and then aged at 540 ℃ for 6 h, air cooled, hereafter referred to as Ti7333-C.

The notched cylindrical fatigue specimens with an elastic stress concentration factor of K_t = 3 were machined using electrical discharge machining (EDM) followed by lathe turning. The shape and dimension of the fatigue specimens is shown in Fig. 1. HCF experiments were performed in tension-tension fatigue (R = 0.1) using an electromagnetic resonance high frequency fatigue testing machine (QBG-100B high-frequency fatigue tester) in accordance with the HB 5287 “Metal Material Axial Loading Fatigue Test Method”. The tests were run in stress-controlled at room temperature and a lab-air environment, using a sinusoidal waveform with a frequency of around 120 Hz. Tests were stopped until the specimen failed completely or fatigue lifetime reached 10⁷ cycles (run-out). It should be pointed out that the lifetime data of unbroken specimens do not participate in S-N curve fitting but still show in the S-N scatter plot. The fatigue strength at 10⁷ cycles was determined by staircase test method using 13, 12 and 10 specimens for Ti7333-A, Ti7333-B and Ti7333-C, respectively.

The initial microstructures were characterized by using scanning electron microscope (SEM, Carl Zeiss Microscopy GmbH) and transmission electron microscope (TEM, FEI Themis Z). The samples for SEM were prepared following standard grinding procedures and subsequent electropolished (Struers LectroPol-5) in a solution of 5% perchloric acid in ethanol at 35 V for 40 s. TEM foils were cut perpendicular to the cyclic loading direction by an EDM and were subsequently mechanically ground to ∼60 μm followed by electrolytic polishing in a solution of 5% perchloric acid in methanol at a voltage of 45 V, using a Struers Tenupol-5 twin-jet electropolisher. Post-mortem fractographic analysis were performed using a SEM (TESCAN MIRA3 SEM) in order to determine fracture surface details.

The characterization of FCI site involved measurement of the spatial angles of the faceted planes with respect to the loading direction (LD) and an analysis of crystallographic orientation of crack-initiation grains and their neighbors. The measurement of spatial angle of facets was accomplished by using a quantitative tilt fractography (QTF) technique [30,37]. Stereo-image pairs of facets were acquired in the SEM at tilt angles of 0° and 30°. Crystallographic information of faceted grains and their neighbors was obtained via EBSD (OXFORD NordlysNano) analysis on a 2D FIB-CS across the faceted grains. A FEI Helios G4 CX FIB equipment was used for machining this section. A focused beam of Ga⁺ ions was used at an accelerating voltage of 30 kV and ion beam current of 65 nA, 9.3 nA and 0.23 nA are used in sequence. TEM foil at crack-initiation site was lifted out by FIB. A low Ga⁺ current of 80 pA was used for the final thinning. EBSD analysis was carried out under a voltage of 20 kV and a current of 6.0 nA. The Channel 5 software was used to obtain the EBSD measurements. Following EBSD data acquisition, the slip trace maps of β grain were generated using raw EBSD orientation data [38,39]. The misorientation between a data point and its neighbors was analyzed using the Kernel average misorientation (KAM) map to approximate measure the geometrically necessary dislocation (GND) density [40,41].

As-received microstructures for fatigue tests are shown in Fig. 2 and the corresponding microstructure parameters are listed in Table 2. Apparently, the microstructures consists of different volume fractions of α_p particles distributed in a matrix of β_t structure, and the β_t further consisted of acicular secondary-α (α_s) precipitates. Besides, the globular α_p particles outnumber the elongated α_p particles. In this paper, the α_p particles with an aspect ratio greater than 2.5 are defined as the elongated α_p particles. It should be noted that an almost continuous precipitation of grain boundary α (GBα) phase is observed in Ti7333-C (Fig. 2(i)). The tensile mechanical properties of the three kinds of microstructures are shown in Fig. 3. It shows that the Ti7333-C has the highest strength and give an ultimate tensile strength (UTS) of 1434 MPa, yield strength (YS) of 1385 MPa and the Ti7333-B exhibits the best plasticity with an elongation to fracture of 12.1%. Fig. 4 shows the EBSD results of the initial microstructure of Ti7333-B. As shown in phase map and the corresponding inverse pole figure (IPF) map, the different α_p variants are distributed in the β grain, and three α_p variants prefer to precipitate. Fig. 4(c) represents the pole figures of ten selected α particles and β grain labeled in Fig. 4b. The objective here is not to prove that all of the 12 α variants exist but to verify the Burgers orientation relationship (BOR) between β phase and selected α particles after TMPs. From Fig. 4(c), except for α6 and α9, other eight α particles exhibit strict BOR relation with β grain. This indicates that a majority of intragranular α particles maintain a strict BOR relation with its β phase.

The S-N fatigue curves of three kinds of microstructures are presented in Fig. 5. An interesting phenomenon, called the duality of the S-N fatigue behavior, is observed, in which the data points grouped into two separate curves. Fatigue variability behavior are marked by an increase in the lifetime variability to almost two orders in magnitude as the stress level is decreased, thereby promoting increased fatigue variability. The fatigue data of short-life-distribution and long-life-distribution overlapped each other, indicating a superposition of two mechanisms. Besides, it can be seen that the complete separation of the two failure distributions in terms of fatigue life (Fig. 5(b) and (c)) or they can dominate at high and low stress ranges with step in the mid-stress-range (Fig. 5(a)). It is worth noting that there is very little fatigue data point at the life cycles from 10⁵ to 10⁶. The fatigue lives decrease with steep slope in the short-life-distribution of Ti7333-C (Fig. 5(c)), indicating that the fatigue life decrease rapidly with increasing of stress level. The fatigue strength of 50% survival rate is shown in Fig. 3. It shows that fatigue strength values of three kinds of microstructures are comparable, and the Ti7333-C exhibits the highest fatigue strength, 335 MPa.

The careful fractographic investigation in four representative specimens is presented in Fig. 6. As expected, surface crack-initiation was found in all failed specimens due to the stress concentration at the location of notch root. As shown in Fig. 6(a), (d), (g) and (i), the macroscopic fracture surfaces are relatively flat. For the short-lifetime failed specimen (shown in Fig. 6(a) and (b)), the crack origin can trace back to a surface crack-initiation site without the presence of facets, with deformation accumulation rapidly for crack-initiation in only a few cycles, and the corresponding fatigue lifetime was controlled by the crack-growth process. For the long-lifetime failed specimens (shown in Fig. 6(c-k)), however, the crack origin can be traced back to a surface crack-initiation site with the presence of facets, and the corresponding fatigue lifetime was dominated by the crack-initiation process. These facets were formed across α_p particles, which were confirmed by examining both the fractured specimen and the FIB-CS characterization. It can be seen that an elongated α_p facet (Fig. 6(e) and (k)) or an equiaxed α_p facet (Fig. 6(h) and (k)) locates on or just slightly below the specimen surface. The presence of the elongated α_p facets indicates that the crack-initiation favor in elongated α_p particles although the number of equiaxed α_p particles is much more than the elongated α_p particles. This indicates that the combination of crystallography and morphology of α_p particles plays a role in generating the critical conditions necessary for facet formation. Additionally, the equiaxed dimples and tear ridges (Fig. 6(c)) and sometimes rough fatigue striations (Fig. 6(f)) can be seen in fatigue crack-growth (FCG) region due to the poor plasticity of this alloy.

As mentioned above, all long-lifetime failures occurred by crack-initiation in the specimen surface with the presence of α_p facets. The FCI site in a representative specimen is presented in Fig. 7. This specimen was tested at 330 MPa, and the lifetime was 4,660,000 cycles. The careful fractographic investigation is presented in Fig. 6(e). The inclination of the facet-normal with respect to the LD was about 44°. This suggests a strong role of pure shear deformation along the facet plane in the formation of the α_p facet.

In order to explore the crystallographic plane of the α_p facet and the deformation mode of the faceted grain and the β grain, this specimen was first mechanically grinding to get close to the FCI site, then a 2D section was machined across faceted grains and parallel to the LD by FIB milling, as shown in the secondary electron image of Fig. 7(a). The specimen tilt in this image is 52°. The α_p particles and β+α_s matrix underlying the crack-initiation facet were revealed on this section. It is evident from the figure that the facets (designated as F1, F2 and F3) were formed across three adjacent α_p particles, resulting in an elongated facet. Also labeled are the neighbor α_p particles, designated as N1, N2, N3, N4 and N5. The EBSD-derived IPF//LD map is presented in Fig. 7(b). The corresponding phase map is shown in Fig. 7(c). The 3D-view of crystal of the faceted α_p particles indicates that the edges of the facet planes were close to the basal plane. In order to characterize this phenomenon more accurately, the angles of the basal plane normal of three facets respect to the LD are determined with Euler angles of these faceted grains, which are presented in Table 3 and are about 43°, 38° and 44° for F1, F2 and F3, respectively. This demonstrates that the crystallographic planes of the facets are less than 10° inclination from the basal plane. Schmid factor can be used as a parameter (but not the sole parameter) to determine whether the slip system activation or not in a given grain during fatigue loading [17,33]. The Schmid factors, along the LD, calculated for five slip systems of faceted α_p particles and neighbor α_p particles are shown in Table 4. It can be seen that the basal <a> slip is the dominant deformation mode in each of the faceted α_p particles where the Schmid factor values for basal slip are 0.49, 0.48 and 0.48, respectively, for the three faceted α_p particles, although the facet-normal has an inclination of about 44° to the LD. These results strongly suggest that these facets were formed by almost pure shear deformation.

It can be inferred that the facet formation process ought to be influenced by the neighboring α_p particles and β grain. The slip modes in the first-nearest-neighbor α_p particles (designated as N1 and N2) and β grain were examined with the help of Schmid factor. It can be seen that the N2 was oriented for basal slip, and no clear trend in deformation modes for N1 particle, as shown in Table 4. For a better understanding of the deformation modes of the β grain, the angles between the all possible {110} slip plane traces and the LD and the corresponding Schmid factors values were determined using Euler angles. As shown in Fig. 7(b), the {110} slip traces are superimposed on IPF map, and corresponding Schmid factors are also presented. The crack-initiation site show that the rugged β fracture is roughly parallel to the slip plane with highest Schmid factor, which is (-101)[-11-1] in this specimen, among all slip systems available. It should be noted that the β fracture is very rough due to there is no a dominant slip plane, i.e. the Schmid factor of 0.44 in.(101)[1-1-1], 0.39 in.(110)[1-11] and 0.39 in.(-110)[-1-1-1]. Additionally, there is no clear β facet appear in the notched specimen may be due to the unique stress distribution at the notch root.

The EBSD-derived pole figures of β phase and α_p particles from the Fig. 7(b) IPF map are displayed in Fig. 8 to analyze the orientation relationship between α-β phases. The pole figures illustrate that there is a strict Burgers’ relationship between N3, N5 and β phase, which obeys the Burgers orientation relationship i.e., {110}_β//{0001}_α, <-11-1>_β//<11-20>_α [42,43]. Besides, the faceted α_p particles (F1, F2 and F3) have a near-Burgers orientation relationship with respect to the β grain. However, N1, N2 and N4 exhibit no Burgers’ relationship between α-β phases due to the stress [44].

The TMP routes will affect the microstructure features and further affect the fatigue strength and fatigue S-N behavior. As shown in Fig. 3, three kinds of microstructures exhibit the comparable fatigue strength values, although the tensile strengths are obviously different. Generally, the fatigue strength of the alloys is positively correlated with the yield strength (YS). The differences in the static strength of these three alloys are associated with the α_s phase characteristics, because of the strengthening effect of the metastable β titanium alloys is mainly dependent on the α_s phase characteristics. A majority of researches show that the FCI by facet formation across α_p particles, and that includes the Ti-7333 alloy. The Ti7333-A and Ti7333-B offer very similar α_p grain sizes, shape and distribution, resulting in comparable fatigue strength. Nevertheless, the Ti7333-C also shows the comparable fatigue strength. It can be speculated that the notch effect weakens the effect of microstructure factor, such as α_p phase characteristic, on crack-initiation. These indicate that the fatigue strength of notched specimens at HCF regime is closely related to the notch effect, i.e., the stress state and stress distribution in the vicinity of the notch root. As shown in Fig. 5, compared with the Ti7333-A and Ti7333-B, the Ti7333-C shows a smaller scatter in fatigue data of the short-lifetime failed specimens. Given that the initial microstructure of Ti7333-C contains a smaller α_p phase percentage, the surface crack-initiation in this microstructure is closely associated with the β+α_s matrix due to a small number of α_p particles in the notch root. This weakens the effect of α_p particles, which is closely related to the formation of the facet, on FCI process. Further, the α_p particle characteristics, i.e., the crystallographic orientation, size and the morphology, contribute significantly to variability of fatigue life and minimum fatigue life.

As revealed by fatigue fracture surface and the S-N curves analysis, the dual fatigue behavior is found to result from variability in fatigue life due to a probabilistic selection between two competing fracture mechanisms. Unlike the smooth cylindrical specimen failures where short- and long-life-distributions of S-N curve result from surface- and internal-crack-initiation, respectively [8], the duality of the S-N fatigue behavior of Ti-7333 notched specimens is associated with the surface-failure-without- and with-facets. Moreover, the long-lifetime failed specimens appear to produce fatigue failure after a significant delay (about two orders of magnitude at the lowest applied stresses) with respect to the short-lifetime failed specimens. At a given cyclic stress range, whether a specimen would fail with- or without-facets, it is often not predictable before the test. Given the identical test conditions and the test environment for each of microstructure of Ti-7333, as well as there are no inclusions or pores in the microstructures, except for special forming process such as powder metallurgy [45], the duality of the fatigue behavior reported here is not an artifact but does seem to be an unusual fatigue phenomenon mainly originating from the unique microstructure level, such as α_p and α_s characteristic, and the notch effects. As expected, the stress state and stress distribution in the vicinity of the notch root can significantly affect the process of formation of fatigue micro-cracks. The reduced fatigue lives in surface-without-facet specimens are due to the acceleration of surface crack-initiation process in the lab-air environment as compared to the quasi-vacuum conditions in the interior of the specimens.

For multi-phase alloys, the phase boundaries are the preferential crack-initiation site, resulting in a decline in the resistance of crack-initiation during cyclic loading. This is due to the discontinuity of slip transfer and the difference of elastic modulus between α and β phases. As discussed above, fatigue micro-cracks favorably initiated from an area of microstructural weakness where α_p particles favorably oriented for basal <a> slip. It can be inferred that edges and the sharper corners of α_p particles are the geometrically weak sites for micro-cracks initiation, as observed the elongated α_p facets in FCI site (Fig. 6(e) and (k)). The strain gradient distribution in FCI site was evaluated in terms of KAM map. As shown in Fig. 7(d), the heterogeneously distributed KAM values are observed in α_p particles and β+α_s matrix. The KAM value in β+α_s matrix is higher than that observed in α_p particles. Moreover, the KAM value in faceted α_p particles is slightly higher than that in α_p particles beneath the fracture surface. Fig. 9 shows the EBSD results (50 nm step size) in crack-initiation region of Ti7333-A specimen. The KAM map (Fig. 9(c)) and the corresponding phase map (Fig. 9(b)) also show the heterogeneously distributed KAM values and the KAM value in β phase is higher than that of α phase, resulting in the strain localization within β phase. Interestingly, a high KAM value is observed at the subgrain boundaries, as shown in Fig. 9(a) and (c). This indicates that dislocation slip occurs within the α_p particles during cyclic loading. On the other hand, a TEM foil at the crack-initiation site of the Ti7333-A shows that the significant dislocation activity in the form of dislocation tangles in β+α_s matrix (Fig. 10(a)). Additionally, the faceted α_p particle reveals the presence of the subgrain boundary, as shown in HAADF image (Fig. 10(b)) and the corresponding SAED patterns (Fig. 10(c) and (d)).

The small fatigue cracks revealed by FIB milling were found to form on basal plane of α_p particle. Schematic for phenomenological model of the HCF crack-initiation in Ti-7333 alloy with bimodal microstructure is shown in Fig. 11. Under fatigue loading, cyclic microplastic deformation first occurs in β+α_s matrix result in a mass of GNDs and statistically stored dislocations (SSDs). Besides, α_p particles can act as a barrier to dislocation motion, resulting in the dislocation pileup around α_p particles. Only the basal and prism slip systems most suitably orienting for slip will be activated due to that the stress levels are far below the YS of the alloy. The <c+a> slip are generally inactivated for the stress levels that produce fatigue life greater than 10⁶ cycle [30]. Therefore, the basal dislocation pileup begins to form at the α_p particle boundaries, especially in elongated α_p particles and multiple adjacent α_p particles well orientated for basal slip, until a dislocation through phase boundary. This results in the formation of a residual boundary dislocation. Once the residual boundary dislocations continue to accumulate to a critical value, a fatigue crack nucleus is formed and then propagates back along the basal slip band under the tension stress, leading to the presence of the α_p facet. Then, fatigue crack propagates along (110) slip plane with the high Schmid factor value to form a β facet or not, depending on the crystallographic orientation of β grain. Subsequent crack propagation is no longer sensitive to microstructure and shows the typical equiaxed dimples and fatigue striations in FCG region (Fig. 6(c) and (f)). It should be noted that the size of α_p facet in Fig. 6(e) and (k) is slightly larger than that in Fig. 2. This is because the larger the grain size, the larger the dislocation slip path, the faster the dislocation pileup and the easier the fatigue crack nucleation. In general, fatigue strength depends on the lattice strength and the dislocation slip path of the alloys. Therefore, the large or elongated α_p particles orientating for basal slip often act as the FCI sites. This provides a deep understanding of the fatigue damage mechanisms by correlating crystallographic orientation and morphology of α_p particles. The faceting process and early stages of crack growth are the particularly important points due to the majority of the total fatigue life are typically spent in this regime. However, for the short-lifetime failed specimens, FCI occurred at the specimen surface without the presence of facets. This is because that the fatigue damages accumulation rapidly for surface crack-initiation in only a few cycles, leading to a shorter fatigue lifetime. As shown in Fig. 2, typical microstructure consists of an equiaxed grained β matrix and two populations of α precipitates, including larger equiaxed α_p particles and acicular α_s precipitations. These fine acicular α_s precipitations contribute a lot to the static strength improvement. The fracture observation and FIB-CS analysis suggest that the FCI is strongly related to the α_p particles more than the α_s precipitations. Nevertheless, FCI process is inevitably associated with α_s precipitations due to their significantly strengthening effect and the α_p particles are surrounded by the β+α_s matrix, but unfortunately, we fail to study this α_s effect because they are too small to be characterized and no α_s facets were observed in the FCI sites.

In this work, the duality of the S-N curve in notched specimens of Ti-7333 is reported, and the fatigue damage mechanism was revealed by FIB-CS characterization across the crack-initiation site and surrounding neighborhoods and TEM characterization of crack-initiation site. The following conclusions can be drawn:

(1) The S-N fatigue curves of notched Ti-7333 specimens exhibit duality. The short- and long-life-distribution occurs due to surface failure without and with facets. Fatigue strength at high cycle fatigue regime is associated with static strength and microstructure factor, and strongly dependent on the notch effect.

(2) For long-lifetime failures, crack-initiation occurred by facet formation in α_p particles, and the facet plane was inclined closely to the maximum shear orientation with respect to the loading direction (45°). Faceted α_p particles were found to be well-oriented for basal <a> slip activation, and no apparent β facet appeared in the notched specimen. Besides, the α_p facet planes almost corresponded to the basal plane, and the faceting process was largely controlled by the shear deformation mode.

(3) FCI is strongly related to the α_p particles more than the α_s precipitations. The morphological and crystallographic effects of α_p particles are significantly responsible for facet formation. Elongated α_p particles favorably oriented for basal slip is a preferential fatigue-critical microstructural configuration. FCI process is inevitably associated with α_s characteristic because of its significantly strengthening effect.

(4) The β+α_s matrix has a higher KAM value than the α_p particles in fatigued microstructures. Significant dislocation activity in the form of dislocation tangles is observed in α_p boundaries and leads to crack-initiation along basal plane of α_p particles.

(5) The duality of the fatigue behavior of notched Ti-7333 is found to be resulted from the competing fatigue mechanisms, and mainly originates from the unique microstructure level and the stress state and stress distribution in the vicinity of the notch root.