Dissimilar Friction Stir Welding of Al and Ti: Elucidation of Microstructural Evolution, Material Flow, and Spring-Based Tensile Fracture Behavior

Abstract

Welding aluminum (Al) to titanium (Ti) is particularly challenging because of the large differences in their melting points and the tendency to form cavities and brittle intermetallic compounds. Such issues can be mitigated in friction stir welding (FSW) by understanding the underlying mechanisms of microstructural evolution and tensile fracture behavior. In the present study, FSW was carried out on commercially pure Al and commercially pure Ti. X-ray micro-computed tomography results show that the distribution of Ti fragments depends on their morphology, with fine particles (volume 10³–10⁴ µm³) being distributed homogeneously, while large flakes (10⁷–10⁹ µm³) are concentrated near the joint interface. A three-dimensional analysis of Ti fragment distribution was performed to clarify material flow and particle dispersion within the weld nugget. EDS (Energy-Dispersive Spectroscopy) and EPMA (Electron Probe Microanalysis) composition mapping confirmed the formation of AlTi and Al₃Ti intermetallic phases, with Al₃Ti as the dominant phase (consistent with its lower Gibbs free energy of formation). Because Al is the primary element in the matrix and undergoes the highest degree of deformation, its microstructural evolution in Al was examined using Electron Backscatter Diffraction (EBSD). Grain refinement in Al was attributed to continuous dynamic recrystallization (CDRX). Mechanical mixing and intermetallic formation increased the hardness of the weld, while the tensile response corresponded to a joint efficiency of approximately 77%, alone with an 11% improvement in elongation over base Al. The study further establishes a correlation among Ti particle distribution, local microstructural evolution, and the tensile response of the joint. Fractographic analysis indicates a bimodal fracture mechanism, and failure occurred away from the joint interface, indicating a strong joint. To interpret this behavior, a spring-based model was proposed to relate the fracture location and tensile deformation to the spatial variation in microstructure across the welded zones. This approach provides a conceptual framework that is extendable to other dissimilar material systems with spatially varying microstructures.

1. Introduction

Aluminum (Al) and titanium (Ti) are important engineering materials with a wide range of applications in aerospace, marine, and transportation industries. Hybrid structures made of Al and Ti are attractive because they offer improved fuel efficiency and reduced cost through the combination of their complementary properties, which are useful in weight-critical structural applications [1]. The growing demand for multi-material lightweight assemblies in next-generation aerospace frames, automotive body-in-white structures, and naval hull components has therefore intensified interest in reliable Al–Ti joining technologies [2,3,4]. Therefore, considerable effort has been devoted to fabricating Al–Ti structures. Several methods, including explosive welding [1,5], diffusion bonding [6], friction welding [7], and cold rolling combined with annealing [8], have been recently used for Al/Ti welding.

The main challenge in joining Al to Ti, particularly by fusion processes, is the large difference in their physical and thermal properties and the formation of brittle intermetallic compounds, which degrade the mechanical properties of the joint. Proper selection of process parameters can minimize the formation of these brittle phases and improve joint strength [9]. Van Loo et al. [10] reported that the compound layer of Al₃Ti was the primary intermetallic phase that formed at the Al/Ti diffusion bond. It has been found that brittle intermetallic formation is inevitable during fusion welding and can also occur in solid-state welding of Al to Ti. Although it is well established that Al₃Ti forms in the weld, the exact mechanism and location of its formation have not been studied in sufficient detail. A careful analysis is therefore required to understand intermetallic formation.

The Al–Ti binary system supports the formation of multiple intermetallic phases, such as Al₃Ti (tetragonal, D0₂₂ structure), AlTi (tetragonal L1₀), AlTi₃, and Al₂Ti, among others. Phase selection during welding is governed by the local composition, temperature, and diffusion kinetics. Al₃Ti is the thermodynamically stable phase at Al-rich compositions and forms preferentially because of its low Gibbs free energy of formation (36 kJ/mol at 500 °C) [11,12]. Its hardness (approximately 550 HV) and near-zero ductility make this phase the most detrimental phase for joint mechanical performance when present as a thick and continuous layer along the interface. The challenge, therefore, is not simply to avoid all intermetallics, which is impossible in any thermally activated Al/Ti joining process, but to control their morphology, size, and spatial distribution such that they contribute to local hardness without serving as critical crack initiation sites.

In the present investigation, a friction stir welding (FSW) technique was employed to fabricate a weld between Al and Ti. FSW is a localized solid-state thermo-mechanical joining process that is free from solidification defects. The process and related terminology in friction stir welding were described by Kumar et al. [13] and Mishra et al. [14]. During welding, material is displaced from the leading edge to the trailing edge through the retreating side and is consolidated on the advancing side. A new interface is formed on the advancing side and must also be welded. This is achieved by material flow through the retreating side, where the material forges against the new surface under the combined action of rotation and tool tilt, forming a solid-state bond. The original interface is joined at the trailing edge itself. Because FSW is a friction-based mechanical joining process, the temperature of the weld is self-regulated depending on the process parameters. However, the temperature does not reach the melting point of either material used in dissimilar welding [7]. The sub-solidus processing temperature in FSW fundamentally distinguishes it from fusion welding in terms of the formation of diffusion-controlled thin intermetallic layers and severe plastic deformation imposed by the rotating tool fragments, and disperses any forming intermetallics into the Al matrix [15,16]. These features reduce their deleterious impact. Additionally, the fine-grained, dynamically recrystallized microstructure produced in the stir zone provides additional grain boundary strengthening to offset any intermetallic-induced local weakness. These attributes make FSW uniquely suited to the Al/Ti dissimilar joining challenge.

Tool rotation speed influences defect formation and weld quality. The relationship between tool rotation speed and heat input is complex during welding. In general, increasing the rotation speed increases heat input, which can cause excessive plastic flow in dissimilar FSW. Excessive softening is termed a hot-weld condition and results in excessive flash and weld defects. If the rotational speed is too low, a cold-weld condition results in voids and other defects in the stir zone. The heat input (δ) in the welding system is calculated from the ratio of rpm (ω) of the tool and traverse speed (v) [17]. Thus, process parameters must be controlled to obtain a sound weld suitable for subsequent microstructural and mechanical characterization.

Palanivel et al. [18] investigated the effect of tool rotational speed and tool pin profile on the microstructure and tensile strength of dissimilar AA5083-H111 and AA6351-T6 joints produced by FSW. They reported that a tool rotational speed of 950 rpm and a straight square pin profile yielded the highest tensile strength of 273 MPa. The variation in material flow and dissolution of precipitates caused differences in tensile properties. Bang et al. [19] examined hybrid FSW (HFSW) of AA6061 and Ti-6Al-4V using gas tungsten arc welding preheating as a heat source for the Ti alloy plate. This experiment was performed at tool rotation speeds of 300–450 rpm. They observed that the tensile properties of both FSW and HFSW joints were influenced by tool rotational speed and welding speed. Within this rpm range, 400 rpm in FSW and 350 rpm in HFSW showed superior tensile properties. In addition to process parameters such as tool rotation speed, preheating also influenced the mechanical properties of the weld. In most studies on FSW of dissimilar materials with widely different physical and thermal properties, the selected tool rotation speed has been between 750 rpm and 1000 rpm.

Chen and Nakata [20] characterized the microstructure and mechanical properties of FSW joints between AA1050 Al and cp-Ti. The authors reported that the formation of thin TiAl₃ layers at the interface was unavoidable. However, joint strength was strongly dependent on the Ti fragment distribution in the nugget. Li et al. [21] investigated FSW lap joints of Al and Ti-6Al-4V and demonstrated that intermetallic layer morphology (particularly the continuity and thickness of the Al₃Ti layer) was the primary determinant of fracture mode and joint efficiency. Dressler et al. [16] studied FSW of Ti-6Al-4V to AA2024-T3 and showed that a defect-free, mechanically interlocked joint interface with a thin intermetallic layer (5–10 µm) produced joint efficiencies close to 80%. Shouzheng et al. [22,23] explored pulsed arc welding and GTA welding of Ti/Al dissimilar joints and demonstrated that the composition gradient at the Ti/Al interface governs intermetallic phase selection, consistent with the Al-rich local composition promoting Al₃Ti over AlTi. Yadav et al. [24] processed Al-Ti particulate composites by friction stir processing (FSP) and observed that mechanical mixing of Ti particles into an Al matrix by the FSP tool produces a fragmentation and dispersion sequence similar to that observed in dissimilar FSW, providing mechanistic parallels relevant to the present study.

Despite these advances, several critical aspects of dissimilar cp-Al/cp-Ti FSW remain incompletely understood. First, the three-dimensional spatial distribution of Ti fragments within the weld nugget, which governs local mechanical properties and fracture initiation, has not been systematically quantified; most prior studies depend on two-dimensional cross-sectional micrographs that cannot capture the full volumetric heterogeneity within the weld. Second, the mechanisms governing intermetallic phase selection (AlTi vs. Al₃Ti) under FSW conditions involving both mechanical mixing and thermally-driven diffusion have not been rigorously separated. Third, no predictive tool and understanding currently exists for correlating the spatially varying microstructure across the multi-zone weld cross-section with the macroscopic tensile fracture location and deformation response. The present work addresses these gaps through: (a) 3D X-ray micro-CT quantification of Ti fragment morphology and their distribution, (b) detailed phase analysis using XRD, EDS, and WDS-EPMA; (c) EBSD-based in-depth characterization of grain refinement mechanisms in Al, and (d) a novel spring-based model that provides a mechanistic link between zone-specific microstructure and tensile fracture behavior. The spring-based analytical model is expected to predict fracture location and deformation distribution from zone-specific microstructural stiffness, providing a transparent, physically interpretable alternative to empirical or finite-element simulation approaches. The cp-Al/cp-Ti material combination was selected as a model system because both materials are commercially pure (eliminating alloy-complexity variables), their large property mismatch represents a challenging standard for dissimilar FSW, and their large X-ray attenuation contrast enables 3D-XCT characterization of the weld nugget.

2. Experimental Procedure

2.1. Materials and Methods

In the present study, commercially pure aluminum (Al) and commercially pure titanium (Ti) with a thickness of 4 mm were used. The elemental composition of the materials is shown in

Table 1. Chemical composition of the as-received materials used in friction stir butt welding of Al and Ti.

	Cu	Mg	Si	Fe	Mn	Ti	Zn	C	Al	Other
CP-Al	0.002	0.003	0.170	0.120	0.002	0.009	0.003		99.661	0.030
CP-Ti				0.300		99.510		0.080		0.110

.

The FSW tool used for the experiment was made of tungsten carbide-cobalt (WC-Co). The geometry of the tool was that of a flat-cylindrical shoulder of 18 mm diameter with a cylindrical pin having a diameter of 4.0 mm. The length of the tool pin was 3.2 mm. The FSW experiment was carried out on a custom-built machine in which the tool was kept in a horizontal position and the specimen in a vertical position. The machine was developed with the help of IISc Bangalore, India, and ETA Technology, Bangalore, India. During the experiment, the tool tilt angle was kept constant at 2°, along with a tool offset of 2 mm and a welding speed of 40 mm/min. The tool rotation speed during a single weld continuously varied from 400 rpm to 1000 rpm. The optimal tool rotation speed was selected based on the formation of a defect-free weld. After that, the weld at 900 rpm was characterized in detail. These specific parameters were selected based on preliminary experimental trials. Lower rotational speeds (<800 rpm) resulted in a wormhole defect appearing on the retreating side of the weld (shown in Figure 1a,b), which is the zone subjected to lower stress and temperature than the advancing side. On the other hand, higher speeds (1000 rpm) induced excessive plastic flash and geometric thinning due to over-softening (hot-weld conditions). The weld at 900 rpm shows no visible macro-defect (Figure 1c). Under this specific welding condition, the mechanical heat input and tool-driven material flow kinetics are perfectly balanced and adequate to thoroughly fill the material void left behind the tool pin during traverse consolidation. Consequently, the 900 rpm joint was identified as the optimal configuration and subjected to comprehensive microstructural and mechanical characterization to elucidate mechanisms associated with solid-state bonding and microstructure evolution.

2.2. X-Ray Computed Tomography

X-ray computed tomography (XCT) measurements were performed using the Zeiss Versa 520 system (Carl Zeiss AG, California, USA). This is a 3D non-destructive visualization and quantification technique, which can identify each element present within a weld nugget. It is equipped with a rotating turntable between the X-ray source and the detector.

In the current X-ray computed tomography instrument, an X-ray source was used to generate a cone beam of X-rays, which is focused on the sample. The X-ray was either attenuated or passed through the sample, which developed a 2D gray-scale radiograph on the screen of the detector. The resolution of the scan was determined by the magnification factor of the object that results from the relative position of the source/detector geometry. During the scan, an additional optical magnification of 4× was used to cover the field of view that was achieved by the relative position of the movable X-ray source and detector. Moreover, the sample was placed on the rotating turntable such that the weld zone was positioned with this field of view, the axis of the X-ray path, and the sample transverse cross-section were aligned perpendicular to each other. The combined effect of all these parameters yields a resolution of 6–10 µm, depending on sample dimensions.

2.3. Metallurgical Characterization

The transverse cross-section of the weld was polished by a standard metallographic method. Finely polished samples were etched using Keller’s reagent for Al and Kroll’s reagent for Ti. The macrostructure of the welds was characterized using a Metallovert Carl Zeiss Optical Microscopy (OM, Carl Zeiss AG, Göttingen, Germany) to find out the defects, such as voids, present in the weld. A scanning electron microscope (SEM: ThermoFisher® XL-30 ESEM, Eindhoven, Netherlands) was used to investigate the microstructural evolution at higher magnification. Furthermore, Energy dispersive spectroscopy or EDS (EDAX, AMETEK, Inc., Mahwah, NJ, USA, paired with GENESIS™ microanalysis software) was used to study elemental distribution within the weld nugget and across the Al/Ti interfaces. Electron Back Scatter Diffraction (EBSD) investigations were carried out using a TSL (Tex Sem Lab, OIM 7.1, Draper, UT, USA) data acquisition system fitted to the SEM. Data analyses were carried out using the EDAX Tex-SEM TSL-OIM version 7.1 data analysis software (Tex Sem Lab, OIM 7.1, Draper, UT, USA).

Elemental composition mapping of interfaces and around the particles in the weld nugget was performed using a JEOL JXA-8530F Electron Probe Micro-Analyzer (EPMA; Akishima, Tokyo, Japan). A step of 0.25 µm was used to detect the elements during the scan. The instrument is equipped with 6 Wavelength Dispersive Spectroscopy (WDS) detectors with 2–4 crystals per spectrometer and detectable wavelength range from 0.087 to 9.3 nm. Since this uses WDS, which has higher accuracy for detecting the elements than EDS, this permits high detection sensitivity for trace elements and a high-resolution power for the adjacent X-ray spectra. The phases present in the weld were identified using a Philips X’Pert PRO PANalytical XRD machine (PANalytical, Eindhoven, The Netherlands) equipped with a CuKα source (wavelength 1.541 Å). The operating voltage and current were 40 kV and 30 mA, respectively. Each XRD was performed in θ-2θ mode with a step size of 0.033 degrees and a scan rate of 0.008 degrees/s. The obtained data from XRD measurement were analyzed using X’Pert Highscore PANalytical software (ver. 2.2b).

2.4. Mechanical Testing

Vickers hardness across the weld cross-section was performed using a micro-hardness tester (Model: Zwick/Roell ZHV; ZwickRoell GmbH & Co. KG, Ulm, Germany), applying a load of 200 g for 10 s. Hardness was measured across the stir zone with the distance between the indents being 1.0 mm, which made the distance between two indents more than 5 times the diagonal of the indents.

Tensile test samples were cut from the weld plate, perpendicular to the weld direction, with dimensions as shown in Figure 2. The gauge length of the tensile sample was 16 mm. Bulk tensile testing of the specimens was carried out using the Instron 8032 servo-hydraulic UTM (Instron, Norwood, MA, USA) that has a maximum load capacity of 100 kN. The strain rate of 10⁻³/s was used during testing. It is to be noted that the diameter of the tool shoulder was 18 mm, and hence, the measured tensile property is from the weld zone instead of the weld component. Before testing, a thickness of 0.2 mm was removed from the bottom and top of the weld to smooth the surface, and hence, the surface effect was excluded. Therefore, the approximate thickness of the specimen is within the range from 3.3 to 3.5 mm, and the same was used while calculating the tensile properties and plotting the data.

3. Results and Discussion

3.1. Volumetric Distribution of Particle in the Nugget Zone

The distribution of Ti in the weld nugget is shown in Figure 3. XCT was employed because the large difference in X-ray attenuation between Al (atomic number Z = 13) and Ti (Z = 22) provides natural grayscale contrast that enables unambiguous phase segmentation without chemical etching or sectioning. The spatial distribution of Ti in three dimensions appears heterogeneous, with large flakes noticed adjacent to the joint interface, whereas fine flakes and particles are predominantly deposited homogeneously within the weld nugget (Figure 3a). This heterogeneity is a direct consequence of the FSW material transport mechanism. The detailed distribution of Ti fragments is observed in a cross-sectional view (Figure 3b) and an isometric view (Figure 3c), which depicts the distribution of Ti only. Al is removed from the images by user-defined selection from the absorption distribution obtained from the X-ray micro-CT scan. All the fragments were analyzed and separated into six volume ranges from 10³ µm³ to 10⁹ µm³. The fragments corresponding to each range are shown in Figure 3d. This six-tier classification was adopted to systematically capture the full size spectrum of fragments generated by the FSW process, from nano/sub-micron particles (10³–10⁴ µm³, approximate equivalent diameter 1–3 µm) through intermediate flakes (10⁵–10⁷ µm³, 6–60 µm equivalent diameter) to large primary flakes (10⁷–10⁹ µm³, approximately 60–600 µm equivalent diameter). It is important to note here that the larger fragments are markedly non-spherical (platy/flake morphology), which has direct consequences for their drag behavior and transport mechanism during stirring. Fine particles, as observed, are homogeneously distributed, whereas the distribution of large flakes appears non-uniform in the weld nugget. The fragments developed in the weld nugget are moved from the leading edge- advancing side to the trailing edge-advancing side with Al by the rotational movement imposed due to stirring by the tool. The flowability of these fragments depends on the drag force as well as the size of the fragments. Large flakes (volume range 10⁷ µm³ to 10⁹ µm³) are subjected to a higher drag force compared to fine particles (volume range 10³ µm³ to 10⁴ µm³) that are not carried away with semisolid Al and primarily deposited adjacent to the joint interface and/or retreating side heterogeneously.

It should be noted that a fragment-free zone is observed near the joint interface (Figure 3b). This appears to be due to the severe rotational motion of Ti fragments alongside the tool. During rotation, Ti fragments hit against the Ti interface and get deflected away from the joint interface, leaving a fragment-free zone. The width of this fragment-free zone is estimated from Figure 3b to be approximately 0.3–0.5 mm. It is worth noting that very fine particles are not detected due to the resolution limit of the scan, which is about 6–10 µm. However, these particles might be present in the zone as very fine particles are subjected to less deflection than larger particles, and their smaller inertia means they can re-enter the interface region after deflection. These observations show that particle morphology strongly influences transport during stirring, resulting in a heterogeneous three-dimensional distribution of Ti within the weld nugget. The fragment-free zone is microstructurally significant for several reasons. First, it represents a region of near-pure Al matrix immediately adjacent to the Ti faying interface. Therefore, it creates a compositional boundary that restricts intermetallic formation locally compared to fragment-rich zones elsewhere in the nugget zone. Second, the absence of hard Ti or intermetallic particles in this zone indicates a relatively lower local hardness, which influences local deformation and fracture behavior under tensile loading. Third, the sharp transition from the fragment-free zone to the adjacent fragment-rich nugget creates a local microstructural and compositional gradient that contributes to the spatial hardness heterogeneity.

3.2. Microstructural Characterization

Elemental distribution in the weld nugget is characterized using backscatter electron mode (BSE) in SEM, as shown in Figure 4. Distribution of fragments adjacent to the interface reveals severe mechanical mixing of different-sized fragments in the weld nugget and complex material flow of Al. Very fine particles are uniformly mixed with Al during welding, and the flow of material is identified by the image contrast, as observed in Figure 4a. A distinct flow is visible near the joint interface from top to bottom of the weld nugget, as reported previously [25]. However, a similar flow is absent in the center of the weld nugget, which comprises particles of different sizes. Alignment and arrangement in the weld nugget help identify the trajectory of material flow in the weld, as it is presumed that flakes are aligned with the direction of material flow to reduce drag force at the interface of the solid flakes and semisolid aluminum. Based on this understanding, lines corresponding to material flow are drawn in Figure 4a. This self-alignment behavior is analogous to the alignment of elongated reinforcement particles observed in FSP of MMCs [7] and provides an independent, SEM-based confirmation of the material flow trajectories inferred from the 3D XCT data (Figure 3).

A joint interface in the middle of the weld nugget was selected for interface characteristics, as shown in Figure 4b. The interface is defect-free, which indicates a sound joint. In addition, the interface is observed to be wavy. This is attributed to shear deformation imposed by the tool. The wavy interface morphology arises from the periodic nature of the tool’s rotational and translational motion. The combined motion deposits a discrete layer of plasticized material that, under the combined action of forging pressure and tool tilt, creates an undulating interface due to shear deformation. Mechanical interlocking created by the wavy interface geometry contributes to joint strength through two mechanisms: geometric interlocking and increased interfacial area relative to a flat interface. The first mechanism resists shear-mode crack propagation along the interface plane. The second mechanism distributes any residual stress concentration over a larger bonded area.

Figure 4c shows a fragmented particle present in the weld nugget. Initially, such particles are developed from the Ti interface. While moving with the tool, these particles are subjected to continuous deformation, leading to fragmentation. Therefore, the fine particles seen in the weld nugget are generated by continuous fragmentation of Ti. It leads to the formation of particles with variation in size and volume, as seen in Figure 3. The fragmentation mechanism can be described in three stages. In the first stage, the rotating tool pin contacts the Ti base material at the advancing side and mechanically abrades and shears the Ti surface, generating coarse primary fragments (corresponding to the 10⁷–10⁹ µm³ volume range in Figure 3d). In the second stage, these primary fragments are subjected to continued shear and compressive stresses within the stirring zone as they circulate with the tool. It causes progressive fracture at stress concentration sites (e.g., sharp fragment corners and Al/Ti interface stress concentrations), progressively reducing fragment size. In the third stage, the smallest fragments (10³–10⁵ µm³) are reduced below the critical size for further brittle fracture and are instead deformed plastically and dispersed homogeneously throughout the Al matrix by the continuing stirring action. The intercalated particle shown in Figure 4c represents a second-stage particle.

The observed flow lines, wavy interface, and fragmented Ti particles collectively indicate severe mechanical mixing and complex material transport during FSW.

3.3. Composition Analysis

The EDS (SEM) line scan across this interface is shown in Figure 5a. The line scan was performed perpendicular to the joint interface (scan path indicated in Figure 4b). The interface does not show any evidence of elemental composition corresponding to plausible intermetallic phases. However, the continuous slope in elemental composition at the interface, without any horizontal step, is attributed to mechanical mixing and/or physical phenomena occurring within the material interaction volume. This observation is physically consistent with the FSW process. The macroscopic joint interface is a region of intense material flow, which prevents the local composition from dwelling at any fixed stoichiometry long enough for a stable intermetallic layer to nucleate and grow. The continuous compositional slope instead reflects the combined effect of mechanical intermixing within the EDS interaction volume and expected sub-resolution compositional gradients arising from elemental diffusion across the freshly created Al/Ti contact during the FSW thermal cycle, especially the cooling cycle [16,26]. This observation contrasts sharply with fusion-welded Al/Ti joints, where thick (>50 µm) continuous Al₃Ti layers at the joint interface are universally reported [20,21]. The difference in formation mechanism of the interface structure demonstrates that the solid-state FSW process successfully prevents the formation of a mechanically detrimental continuous intermetallic layer at the primary bonding interface.

It is already known that local composition is responsible for the formation of intermetallic compounds. The composition ratio at different interfaces changes with the morphology of Ti interfaces and mechanical mixing. Fine particles mix homogeneously in Al, and locally, they have a higher Al to Ti composition ratio than that of large particles. Fine particles have a proportionally larger surface-area-to-volume ratio than large flakes, meaning that diffusion-driven composition changes occur throughout their entire volume rather than just at the surface layer. Therefore, fine particles and flakes are more prone to intermetallic phase formation due to higher diffusion and reaction kinetics. Additionally, temperature evolution during welding leads to thermal diffusion, which, along with stress-assisted diffusion, further enhances the overall elemental interdiffusion at the interface. The chemical reaction between Ti and Al is also likely to occur. A similar particle is shown in Figure 4c. The EDS line scan across this particle is shown in Figure 5b. The scan path is indicated by a directional arrow in Figure 4c. Two locations in this plot are further magnified and shown in Figure 5c,d. The constancy of composition and the stoichiometry indicate the formation of Al₃Ti and AlTi intermetallics in these regions. The width of each identified intermetallic region from Figure 5c,d is estimated at 1–3 µm for AlTi and 2–5 µm for Al₃Ti, both below the 5 µm threshold below which intermetallics are generally considered too thin to initiate critical brittle cracking under typical tensile loading conditions [16,21,27].

The elemental mapping in the particles (same as seen in Figure 4c) also demonstrates the formation of these intermetallics within. The particle is composed of multiple layers with a variation in elemental composition. To understand the complete elemental spatial distribution around the fragmented particle, Wavelength Dispersive Spectrometry (WDS) in an EPMA was carried out on a fragmented particle similar to Figure 4c, and it is presented in Figure 6. Mechanical mixing due to stirring by tool rotation develops stress and strain inhomogeneity across the interface. It aids in the fragmentation of Ti interfaces and further fragments into fine particles, as discussed earlier. This leads to different deformation mechanisms and mechanical mixing across the particles, and the same is observed in Figure 4c and Figure 6, respectively. Therefore, the composition of this particle is non-uniform, and hence intermetallic phases are primarily developed in the regions of maximum deformation and fragmentation across the interfaces of the particles. There is a spatial correlation and interlinking between high-deformation regions and regions of intermetallic formation. The interlink arises from deformation and fresh formation of intermetallics, local temperature rise from deformation (adiabatic heating), accelerating diffusion kinetics, and evolution of stress and strain gradients around deforming Ti fragments, creating local lattice distortions that reduce the activation energy for nucleation of the ordered intermetallic phases [7,28].

The compositional gradient is observed across all the interfaces of the particle. The color code corresponding to atomic percentage suggests that very fine particles in the weld nugget and interface within the fragmented particle transform into intermetallics, as indicated in Figure 5. A layer of AlTi intermetallic surrounds the Ti, and an additional layer of Al₃Ti covers the AlTi layer. These layers of intermetallics develop due to local composition and enhanced reaction kinetics. Local composition leads to the formation of AlTi, and the presence of additional Al away from the joint interface produces Al₃Ti. The compositional gradients and localized stoichiometry variations confirm that intermetallics form preferentially within fragmented Ti-rich regions and at interfaces experiencing intense deformation.

3.4. Phase Analysis

To identify the evolution of phases in the weld nugget, X-ray diffraction (XRD) of the weld nugget was carried out (Figure 7). The result shows the evolution of intermetallic Al₃Ti and AlTi. The two phases are predicted to be thermodynamically stable in the Al-rich composition range at temperatures below 600 °C [12]. The cumulative peak intensity corresponding to the phases illustrates that Al₃Ti is the primary intermetallic phase that grows in the weld nugget. The relative peak intensities of Al₃Ti and AlTi can be used to estimate their volume fractions semi-quantitatively using the relative intensity ratio (RIR) method: I(Al₃Ti)/I(AlTi) >> 1 from Figure 7, indicating that Al₃Ti constitutes the majority (estimated >70 vol.% of the total intermetallic fraction) while AlTi is the minority phase. This is consistent with the WDS-EPMA observations (Figure 5 and Figure 6), which show a thicker Al₃Ti outer layer relative to the AlTi inner layer in the reaction product sequence. This occurs because the Gibbs free energy of phase formation for Al₃Ti is lower at welding temperatures [12]. The evolution of AlTi is preferred over AlTi₃ in phase evolution at temperatures below 600 °C. The chemical reaction between Al and Ti, leading to the formation of Al₃Ti, is exothermic and is self-propagating [29,30,31,32]. Therefore, the mechanism of phase evolution during FSW of Al with Ti is a complex phenomenon that depends on mechanical mixing, resulting in local variations in particle composition and morphology. The phase evolution, in general, governs the properties of the interface, which eventually influences the mechanical properties of the weld [20,21,22,33]. XRD confirms the presence of Al₃Ti and AlTi, with Al₃Ti as the dominant intermetallic phase, consistent with the localized composition and deformation conditions identified by EDS and EPMA.

While bulk intermetallic phases like Al3Ti and AlTi are characteristically brittle, their controlled, discontinuous formation as thin interfacial layers (<5 μm) surrounding fragmented Ti particles acts effectively as a dispersion-strengthening phase within the compliant aluminum matrix [34]. This multi-layered structure strengthens the mechanical bond between the dissimilar metallic matrices and elevates the local load-carrying capacity of the nugget zone [35]. Provided the thickness of these brittle layers remains below a critical threshold, they prevent premature macro-cracking while actively contributing to local matrix strengthening through load-sharing kinetics [15,36].

3.5. Micro-Mechanism and Microstructural Evolution in Aluminum

EBSD scanning at the center of the weld was carried out, and the result is shown in Figure 8. The black spots in the Inverse Pole Figure (IPF) map (Figure 8a) represent the regions of lower confidence index (CI < 0.05) value, which corresponds to Ti distributed in the weld nugget. Figure 8b shows the grain boundary character distribution (GBCD) of the weld sample, with colour code referring to very low angle grain boundary (VLAGB, 2–5°), low angle grain boundary (LAGB, 5–15°), and high angle grain boundary (HAGB, 15–65°) grain boundaries. GBCD reveals a large fraction of VLAGB and HAGB in the weld nugget, indicating the occurrence of dynamic recovery (DRV) and dynamic recrystallization (DRX) [23]. The important observation is that fine grains present adjacent to the flow line (marked by the red arrow in Figure 8a) are surrounded by a high fraction of HAGB, and the presence of VLAGB and LAGB is not substantial. A Grain Orientation Spread (GOS) map, which is generally used to describe the average deviation of misorientation of grain, shows a lower grain orientation spread of this region, as shown in Figure 8c. Careful observation reveals the intermixing of boundary characters at the boundary of a few grains, which are indicated by a red arrow in Figure 8b. These grains also show a higher GOS value, indicating a higher orientation distribution within the grain that resembles a higher dislocation density inside the grain. The continuous evolution of grain boundaries occurs through grain rotation, which means continuous dynamic recrystallization leading to grain refinement [37,38]. The GOS map can be used to separate the recrystallized and deformed grain. Grains with GOS value with 0–2° are commonly considered as low strain-induced and recrystallized grains, whereas grains having orientation spread more than 2° are regarded as deformed grains [39]. Recrystallized grains are distributed in the weld nugget in a non-uniform manner. A high fraction of such grains is primarily observed close to the flow zone and in the region where Ti particles are deposited. In addition to the heat generated as a result of the friction stir process, the exothermic evolution of heat due to the interaction of Al and Ti (leading to the formation of Al₃Ti) may also contribute to the critical conditions for dynamic recrystallization.

Figure 8a shows a normal distribution of grain size for the characterized weld. The mean grain size determined from the EBSD scan is measured as 10 µm (10.08 ± 5.41 μm). Since Al is a high-stacking-fault-energy material, generated dislocations arrange in a sequence to reduce stored energy through dislocation climb and glide. It results in a higher fraction of VLAGB and a lower fraction of HAGB (Figure 8b). The presence of Ti flakes restricts the conventional flow of Al, leading to higher deformation around the particles. Furthermore, the deformation depends on the morphology of Ti particles and flakes. The degree of deformation around Ti particles is less than that of Ti flakes. Thus, the flow discontinuity results in strain inhomogeneity and dislocation generation in the Al matrix. The EBSD results indicate that strain heterogeneity induced by Ti fragments promotes dynamic recovery and continuous dynamic recrystallization, leading to localized grain refinement in the Al matrix (Figure 9a). This localized grain refinement by CDRX is driven by the severe shear strain gradients inherent to FSW, further aggravated by the severe pinning effect of hard Ti or intermetallic fragments. The accumulation of high dislocation densities promotes dynamic recovery (DRV) by inducing the formation of low-angle subgrain boundaries. Under continuous plastic deformation, these subgrains systematically accommodate strains (dislocations and subgrains), converting progressive low-angle misorientations into stable high-angle grain boundaries (HAGBs), which is supported by the pronounced bimodal misorientation angle distribution (Figure 9b).

A semi-quantitative assessment of the relative contributions of DRV and CDRX can be derived from EBSD data using three independent analyses. (i) GBCD-based boundary ratio: The VLAGB fraction (0.397) represents the sub-grain boundary network generated by DRV, while the HAGB fraction (0.395) represents boundaries that have undergone sufficient progressive rotation to qualify as fully recrystallized products of CDRX. The ratio HAGB/(VLAGB + LAGB) = 0.395/(0.397 + 0.207) = 0.65 indicates that approximately 65% of grain boundaries in the nugget originate due to CDRX, while the remaining ~35% retains a DRV sub-structured character. (ii) GOS-based grain area fraction: Applying the GOS < 2° criterion for recrystallized grains (Figure 8c), approximately 44% of grains are fully recrystallized and have direct correlation with CDRX, with a further ~30–35% in the DRV sub-structured state (GOS 2–5°) and the remainder in a deformed state where DRV is incomplete. (iii) Misorientation angle distribution: The bi-modal misorientation distribution (Figure 9b) exhibits a dominant sub-5° DRV peak, a reduced 5–15° LAGB range, and a secondary HAGB contribution; this is the characteristic of active CDRX in high-SFE, such as Al. The reduction in the 5–15° LAGB range directly evidences that CDRX is continuously converting from LAGB to HAGB, resulting in a decrease in the intermediate misorientation range. This considers the evolution of a purely DRV microstructure, and this range would remain well-populated.

The DRV and CDRX contributions are spatially heterogeneous as fragment and particle-rich zones near Ti deposits and flow lines show CDRX dominance with finer grain size (~5–8 µm, low GOS) due to higher local stored energy and exothermic Al₃Ti heat contributions, while fragment-poor interior regions show a more prominent DRV state (~8–12 µm, GOS 2–5°). Integrating these three analyses, the analysis and prediction estimate is that CDRX accounts for approximately 44–65% of the nugget microstructure by boundary area and grain area fraction, while DRV accounts for the remaining ~35–50%. This analysis confirms that while DRV is active and provides the sub-boundary precursor network essential to CDRX initiation, CDRX is the dominant final grain-refining mechanism responsible for the observed mean grain size of 10.08 ± 5.41 µm (Figure 9a).

It is important to note that the microstructural evolution mechanisms outlined here, such as severe mechanical fragmentation of the Ti plate edges and CDRX within the Al matrix, are fundamentally applicable across the working window of sound FSW parameters rather than being restricted to the 900 rpm condition. Changing the tool rotation speed in this window does not alter the weld peak temperature, as it is determined by the recrystallization temperature of Al [36,40]. However, it is expected to influence the particle size of Ti as it is dictated by the shear strain and flow strain of Ti [41,42,43]. Lower speeds decrease heat input and raise flow stress, generating coarser Ti fragments and thinner interfacial reaction boundaries, whereas higher speeds accelerate thermal interdiffusion, leading to thickened intermetallic layers [4,44]. The 900 rpm condition represents an intermediate kinetic state where structural fragmentation and phase growth are optimized.

3.6. The Mechanical Properties of the Weld

3.6.1. Hardness Distribution of the Weld

Figure 10 shows the hardness profile across the stir zone for the weld at 900 rpm. A significant variation in hardness can be seen from the Ti to the Al side. It varies from 144 HVN (Vickers hardness number) on the Ti side to 31 HVN on the Al side. The variation in hardness observed in the weld nugget is attributed to the non-uniform distribution of Ti fragments with different sizes, Al matrix, and the intermetallic phase having a thickness less than 5 µm, which is much smaller than that of the indentation dimension, 40–110 µm. Such a partial distribution is unable to measure the hardness corresponding to intermetallics. Moreover, the presence of the intermetallic phase and Ti fragments leads to a higher deviation in hardness at any particular position across the weld. For example, at the location of 1 mm toward Ti from the faying interface (“−1”), the mean hardness is 86 HVN with a standard deviation (SD) of 77 HVN. Such a high SD in the weld is due to the presence of Ti and intermetallic phases with varying morphologies and mixing ratios. The hardness of the weld nugget seems to be higher than that of base Al (28 HVN). The increase in hardness can be attributed to the combined effects of grain refinement, mechanical mixing, and the presence of intermetallics [45].

3.6.2. Tensile Properties of the Optimized Weld

The engineering stress-strain curve for the weld at 900 rpm is shown in Figure 11. The tensile properties of the base Al and the weld are summarized in

Table 2. Tensile test results of the as-received base Al and weld sample.

	UTS		Ductility
	Absolute (MPa)	Relative to cp-Al (%)	Absolute (%)	Relative to cp-Al (%)
As-received cp-Al	106	100	19	100
Weld	82	77	21	111

. The UTS of the weld is 82 MPa, which corresponds to a joint efficiency of approximately 77% with respect to base cp-Al. For clarity, the joint efficiency is calculated relative to the base cp-Al (the weaker of the two base materials, UTS = 106 MPa), as is standard practice in dissimilar FSW studies. This value of ~77% compares favorably with joint efficiencies of 60–85% reported in the literature for dissimilar Al–Ti FSW joints [27,36,46,47,48,49]. The weld exhibited 11% higher elongation compared to the base cp-Al (Table 2). The tensile sample failed on the retreating side of the weld. This indicates a sound bond at the joint interface. The sample underwent extensive neck formation before fracture. The fracture surface of the weld is examined under SEM (Figure 11). The fractography shows that the sample fails by both ductile and brittle fracture modes.

The brittle fracture occurs at the Ti particle and intermetallic (Al₃Ti, AlTi) interfaces (indicated by red arrows in Figure 12a), where limited slip systems and stress concentrations promote debonding or cleavage under tensile load. In contrast, the Al-rich zone matrix regions between particles fail by classical void nucleation, growth, and coalescence, generating the dimpled topography (indicated by green arrows in Figure 12). Fractography from other regions of this fracture surface shows the bimodal distribution of dimples, indicating the ductile mode of fracture, and it is consistent with the non-uniform particle size distribution across this zone (Figure 12b). The fine particles nucleate fine dimples and coarser microstructural heterogeneities, which, in turn, nucleate larger voids. A single dominant failure mode would be expected only in a microstructurally homogeneous material; the inherent heterogeneity of the RS transition zone, as reflected by the measured hardness gradient (Figure 10), necessitates the observed cooperative bimodal failure. This bimodal fracture illustrates that rupture has occurred within the process zone, but on the RS of the weld, where microstructural evolution is inhomogeneous in morphology and size due to a gradual transition from the weld nugget to the base Al.

A closer examination of the hardness profile (Figure 10) reveals that fracture did not occur at the absolute minimum-hardness point (~31 HVN at ~4 mm from the interface on the RS) but rather within the zone of steepest hardness gradient (~75 HVN to ~31 HVN across a ~2 mm span at the weld nugget/processed-Al boundary on the RS). This distinction is mechanistically significant as under a uniform applied tensile load, the steepest stiffness gradient, rather than the absolute weakest point, generates the highest strain localization, which is consistent with the incompatibility-induced necking mechanism [50].

The weld therefore exhibits a hardness gradient and a joint efficiency of approximately 77%, while the increased elongation and fracture away from the interface indicate effective bonding despite the lower absolute tensile strength than base Al.

3.6.3. Strengthening Mechanism of the Weld Nugget

UTS of the weld (82 MPa) is lower than that of the base cp-Al (106 MPa), the local hardness within the weld nugget (up to ~86 HVN near the interface) substantially exceeds that of base Al (~28 HVN). This local hardness enhancement is attributed to the combined effects of grain refinement (CDRX-induced grain size of 10.08 ± 5.41 µm), Ti particle reinforcement (Orowan looping by sub-micron particles), and the presence of dispersed intermetallic phases.

It was also found that many particles are distributed within the weld nugget, and the fracture occurred adjacent to the weld nugget. Therefore, the tensile properties of the weld depend on the comparative mechanical properties of the weld nugget and transition zone adjacent to the base material, and the strengthening of the weld nugget. The strength of the weld nugget depends on the uniform distribution of second-phase particles [28,51] and the evolution of the recrystallized microstructure of aluminum with substantially refined grains [24,52,53]. Figure 3 and Figure 4 exhibited the inhomogeneous distribution of Ti particles with variation in their size. Furthermore, Figure 4 shows that micron- and nano-sized Ti and intercalated/IMC particles are distributed in the weld nugget. The role of Al₃Ti and AlTi intermetallics in improving local properties is size- and distribution-dependent. Depending on their size, particles can strengthen by two primary mechanisms: particle cutting and Orowan’s strengthening. The weld with micron-sized and sub-micron particles illustrates strengthening due to particle shearing and the Orowan looping mechanism, respectively. Particles exceeding a specific critical size lose coherency with the matrix due to flow discontinuities at the interfaces. This effectively reduces the pinning ability of the coarse particles, leading to a reduction in strength and promoting the formation of a coarse-grained structure in the matrix material.

On the other hand, Kumar et. al. [54] incorporated fine ceramic particles in the Cu matrix by Friction Stir Processing (FSP) and proposed that the improvement of tensile properties of the processed zone containing nano-particles (smaller than 100 nm) was due to the Orowan strengthening mechanism. To develop an aluminum-based composite, Madhu et al. [28] used FSP to homogenize particle distribution in aluminum-based composites. Due to the nano-scale dispersion, the UTS of the composite substantially increased when compared to the base Al. The increase in UTS was observed due to the combined effect of grain refinement and Orowan strengthening. The strengthening of the weld nugget is therefore attributed to the combined effects of grain refinement, particle reinforcement, and localized intermetallic formation. In the current investigation, the thin (<5 µm) intermetallic layers observed here are below the critical cracking size under the applied strain rate, and nanoscale Al₃Ti particles dispersed in the Al matrix act as effective Orowan obstacles. This locally reinforced nugget is stronger than the processed Al transition zone on the RS, which is why fracture occurs on the RS rather than at the Al/Ti interface. This is a key indicator of sound bonding. The lower UTS of the weld relative to the base Al reflects the constraint imposed by the mechanically weaker transition zone rather than the deficiency of the weld nugget itself.

3.6.4. Spring Model of Tensile Fracture

The spring model presented here is a conceptual analytical framework that provides a physically motivated, qualitative prediction of fracture location by treating each microstructurally distinct weld zone as a spring with stiffness proportional to its local elastic modulus and hardness. To understand the mechanism of the fracture, a cross-section of the welds is schematically shown in Figure 13a. The figure has been introduced to present a schematic layout of the series-spring configuration across the different microstructural zones of the joint. This diagram illustrates how the weld is discretized into series components. The weld cross-section consisted of titanium, an intercalated layer at the interface, a mechanically mixed zone in the weld nugget, and processed aluminum. The mechanically mixed layer forms at the interface of the titanium and the weld nugget. The intercalated particles are distributed in the weld nugget. During tensile testing, the mechanical properties of the weld samples depend on the mechanical properties of each layer developed during welding. The tensile sample consists of all the segments mentioned above, as shown in Figure 13a. Each layer has a different load-carrying capacity and elongation during a tensile test. It clarifies the localized compliance assumptions and strain partitioning under tensile loading.

To understand the deformation of each layer during tensile testing, the deformation of the tensile sample under loading conditions can be analyzed using a system of springs with different stiffnesses (Figure 13b). In Figure 13b, spring stiffness constants k₁, k₂, k₃, and k₄ correspond to Ti (1), interlayer (2), weld nugget (3), and Al (4), respectively. The corresponding elongations of each spring are X₁, X₂, X₃, and X₄, respectively. Since all other springs are connected in series, the total elongation (X) produced by the springs is equal to the sum of the elongations of the individual springs. Therefore, the following equation can be written,

$$\begin{array}{l}x=x_1+x_2+x_3+x_4\\ x={\displaystyle \frac{W}{k_1}}+{\displaystyle \frac{W}{k_2}}+{\displaystyle \frac{W}{k_3}}+{\displaystyle \frac{W}{k_4}}\\ x=W({\displaystyle \frac{1}{k_1}}+{\displaystyle \frac{1}{k_2}}+{\displaystyle \frac{1}{k_3}}+{\displaystyle \frac{1}{k_4}})\end{array}$$

It is assumed that the stiffness of the Ti and interlayer is superior to that of the weld nugget and base Al. The evolution of a defect-free interface with a strong diffusion layer and interlayer containing intermetallics [16] leads to superior joint quality and hence, is expected to exhibit higher stiffness than that of titanium as well. On the other hand, the stiffness of the weld nugget is expected to be higher than that of Al, and hence, the interface of the weld nugget and Al is expected to be the weak zone to failure. Therefore, the critical load (W) is determined by the stiffness of these zones, including the weld nugget (k₃) and the Al (k₄), resulting in a higher UTS for the joint. Since the microstructure of the weak zone was also refined during processing, the joint exhibited a superior UTS compared to the base Al. It is important to note that all the layers (Figure 10 and Figure 13a) are subjected to elongation (Δ) until the sample reaches the UTS corresponding to the critical load. Cumulative elongation leads to higher ductility (percent elongation) of the joint. The spring model provides a useful conceptual framework for relating zone-specific stiffness and elongation to the observed fracture location and tensile response of the joint.

The existing fracture prediction approach in dissimilar welds uses empirical parameter-property correlations [40,55,56], hardness-based weakest-zone criteria [57,58], and finite-element (FE) modeling [59,60]. Empirical correlations predict joint strength from process parameters but lack mechanistic insight into the microstructure. Hardness-based criteria identify the minimum-hardness location as the failure site but do not predict elongation or deformation distribution. FE modeling can capture full stress-strain fields but requires complete constitutive laws for each weld zone, which typically require extensive additional experimental characterization and testing. The spring model proposed here occupies a complementary role as it requires only qualitative or semi-quantitative stiffness ranking of the weld zones and explicitly predicts both fracture location and cumulative elongation from first principles of series-spring mechanics. The quantitative stiffness ranking of the weld zones can be derived from microstructural observations and hardness measurements. The limitations of the current model are (i) the current model does not account for cross-sectional area variations between zones; (ii) zone stiffnesses are ranked qualitatively rather than measured by nano-indentation or DIC; and (iii) damage mechanics, interface debonding, and strain localization are not explicitly incorporated. These extensions are identified as valuable directions for future model development.

The conceptual framework presented here is structurally general and may serve as a starting point for predicting fracture behavior in other multi-zone dissimilar welded systems. However, extending the quantitative analysis to additional material combinations will require independent experimental validation of zone-specific stiffness values, which is identified as an important direction for future work.

4. Conclusions

FSW of Al to Ti has been successfully carried out. The microstructure and mechanical properties of the weld produced at 900 rpm have been characterized. Based on the observations, the following conclusions can be drawn:

• The distribution of Ti particles depends on the morphology of the particles. Fine particles (volume 10³–10⁴ µm³) are homogeneously distributed in the weld nugget, unlike large particles and flakes (10⁷–10⁹ µm³), due to the effect of the drag force. Large Ti particles and flakes align along the direction of material flow due to high drag force, thereby resembling the trajectory of material flow in the weld nugget.
• The intermetallic phases (Al₃Ti and AlTi) appear in the weld due to metallurgical reasons. High-stress-assisted deformation and thermal evolution lead to the formation of AlTi along with Al₃Ti. Deformation of fragments, mechanical mixing, and particle size influence stress and compositional inhomogeneity across the particles and flakes, leading to the formation of intermetallics. The evolution of a high fraction of Al₃Ti over AlTi is attributed to the lower Gibbs’s free energy of Al₃Ti at the estimated weld temperatures (<600 °C).
• Al in the weld nugget is dynamically recrystallized. Continuous Dynamic Recrystallization (CDRX) is the dominant recrystallization mechanism. Dynamic recovery (DRV) is the driving force behind CDRX in Al during high-strain-rate deformation owing to its high stacking-fault energy. The coupled mechanism yields a refined microstructure with a mean recrystallized grain size of 10.08 ± 5.41 µm. GOS analysis confirms that 44% of grains have undergone full recrystallization.
• The weld achieved a UTS of 82 MPa (joint efficiency of approximately 77% relative to base cp-Al) with an 11% improvement in elongation (21% vs. 19% for base cp-Al), and the variation in hardness in the weld nugget (up to 86 HVN vs. 28 HVN for base cp-Al) is attributed to the contribution from mechanical mixing and the formation of intermetallics. The fracture surface reveals a bimodal mode of fracture, which illustrates that the tensile sample fractured on the RS or weld nugget of the weld, indicating superior joint characteristics.
• A spring model has been introduced to correlate the microstructure evolution with springs connected in series. The model is used to predict the fracture location and mechanical properties of the weld. This method and mechanism can be an attractive option to predict the properties of a complex system having variations in the microstructure.