Crack-Resistance Behavior of an Encapsulated, Healing Agent Embedded Bu ﬀ er Layer on Self-Healing Thermal Barrier Coatings

: In this work, a novel thermal barrier coating (TBC) system is proposed that embeds silicon particles in coating as a crack-healing agent. The healing agent is encapsulated to avoid unintended reactions and premature oxidation. Thermal durability of the developed TBCs is evaluated through cyclic thermal fatigue and jet engine thermal shock tests. Moreover, artiﬁcial cracks are introduced into the bu ﬀ er layer’s cross section using a microhardness indentation method. Then, the indented TBC specimens are subject to heat treatment to investigate their crack-resisting behavior in detail. The TBC specimens with the embedded healing agents exhibit a relatively better thermal fatigue resistance than the conventional TBCs. The encapsulated healing agent protects rapid large crack openings under thermal shock conditions. Di ﬀ erent crack-resisting behaviors and mechanisms are proposed depending on the embedding healing agents. ‐

During the actual operating service, TBCs are exposed to severe conditions with the following complex phenomena [14,15]: thermomechanical stresses, erosion, corrosion by foreign objects, diffusion, oxidation, phase transformation, and sintering. In particular, abrupt thermomechanical stress during starting up and shutting down is one of the key factors for damage accumulation, which can lead to formation and propagation of cracks and ultimately result in the failure of the TBCs [16][17][18]. Because

Healing Agent Powder Encapsulation Process
To produce a protective shell for the Si healing agents, the binary glassification process from our previous studies was employed [34][35][36]. This method involved the formation of a glass phase, which was synthesized through the sol-gel reaction and glassification between silicate and sodium alkoxide, resulting in a glassificated SiO 2 shell around the particle surface [36]. The detailed encapsulation process is described in Figure 1. Prior to the encapsulation process, the healing agent-silicon powder (Si, 325 mesh, 99% trace metal basis, Sigma-Aldrich, Saint Louis, MO, USA)-was dispersed in methyl alcohol (methanol, CH 3 OH, 99.5%, Samchun Chemicals, Seoul, Korea) using an ultrasonic bath (Branson 3510, Branson Ultrasonics Corporation, Danbury, CT, USA) for 30 min at 60 • C. The binary precursor solution was composed of tetraethyl orthosilicate (TEOS, Si(OC 2 H 5 ) 4 , 98%, Sigma-Aldrich) and sodium methoxide solution (NaOMe, CH 3 ONa, 25 wt% in methanol, Sigma-Aldrich) as SiO 2 and Na 2 O precursors, respectively. The composition of the binary precursor solution followed our previous studies [34][35][36] with the molar ratio of 0.18:1.5 of TEOS and NaOMe, which was stirred for just a few minutes at room temperature. Then, the dispersed healing agent was added to the solution, and the mixture was left in an oven at 80 °C for 24 h to undergo the hydrolysis reaction. The resulting gel was vacuum-filtered using filter paper (F1001, CHMLAB, Barcelona, Spain) and washed to remove unreacted precursors with deionized water. Finally, to produce a uniform glassificated SiO2 shell, the coated Si powders were heat-treated at 1000 °C for 60 min in air.
The morphology of the encapsulated healing agents was investigated using scanning electron microscopy (SEM, JSM-5610, JEOL, Tokyo, Japan). In particular, to observe the cross-section of the microscale powder, the encapsulated powder was cold vacuum mounted by resin and finely polished with 1 μm diamond paste and polishing cloth. The polished area was Au-coated by a sputtering system (108 Auto Sputter Coater, Cressington, UK), which prevented charging during SEM observation. The composition was also analyzed by an energy dispersive spectrometer (EDS, Oxford Instruments, Abingdon, UK) and X-ray diffractometer (XRD, Mini Flex II, Rigaku, Japan) using Cu Kα radiation (λ = 1.5418 Å) with the operation condition of a step size 0.02° and a scan speed of 3°/min. The particle sizes of healing agents before and after encapsulation were analyzed by laser diffraction (Mastersizer 2000 Hydro 2000S, Malvern Instruments, Herrenberg, Germany). The encapsulated particles were gently ground to prevent the encapsulation shell from crushing to submicron particles, then each sample (100 mg) was dispersed in deionized water using an ultrasonic bath.

Coating Fabrication
In this study, the healing agents' embedded buffer layer was introduced between the bond coat and conventional top coat as a crack-resisting layer, following the coating designs in Figure 2. Three types of coating specimens were prepared: each structure of a conventional TBC with a YSZ top coat (sample A), a crack-healing TBC with healing agents embedded in a buffer layer (sample B), and a crack-healing TBC with encapsulated healing agents (sample C), which are described in Figure 2A-C, respectively. A disk-shaped Ni-based superalloy (Nimonic 263, a nominal composition of Ni-20Cr-20Co-5.9Mo-0.5Al-2.1Ti-0.4Mn-0.3Si-0.06C, in wt%, Thyssen Krupp VDM, Werdohl, Germany) was used as a substrate, with dimensions of 25 mm in diameter and 5 mm in thickness. The substrate was sandblasted using Al2O3 powder (particle size: ≈ 420 μm) before bond coat fabrication, and the bond coat was deposited by high-velocity oxygen fuel (Diamond Jet-2600 DJM, Sulzer Metco Holding AG, Winterthur, Switzerland) methods with the thickness of 150-200 μm using AMDRY 9954 (Oerlikon Metco AG, Pfäffikon, Switzerland, nominal composition of Co-32Ni-21Cr-8Al-0.5Y in wt% and particle size of 11-63 μm). The buffer layer and top coat were fabricated by atmospheric plasma spray (APS, 9MB coating system, Sulzer Metco Holding AG) methods for thicknesses of 100-150 and 400-450 μm, respectively. METCO 204 C-NS (Oerlikon Metco AG, 8 wt% Y2O3 doped in ZrO2, particle size of 45-140 μm) feedstock was used for the top coat material. For the buffer layer, YSZ feedstock was blended with each normal Si and encapsulated Si powder. The healing agent and YSZ with a volume ratio of 10:90 were mechanically mixed through ball milling for 24 h. All employed top and bond coat fabrication parameters as suggested by the manufacturer; see Table 1. The composition of the binary precursor solution followed our previous studies [34][35][36] with the molar ratio of 0.18:1.5 of TEOS and NaOMe, which was stirred for just a few minutes at room temperature. Then, the dispersed healing agent was added to the solution, and the mixture was left in an oven at 80 • C for 24 h to undergo the hydrolysis reaction. The resulting gel was vacuum-filtered using filter paper (F1001, CHMLAB, Barcelona, Spain) and washed to remove unreacted precursors with deionized water. Finally, to produce a uniform glassificated SiO 2 shell, the coated Si powders were heat-treated at 1000 • C for 60 min in air.
The morphology of the encapsulated healing agents was investigated using scanning electron microscopy (SEM, JSM-5610, JEOL, Tokyo, Japan). In particular, to observe the cross-section of the microscale powder, the encapsulated powder was cold vacuum mounted by resin and finely polished with 1 µm diamond paste and polishing cloth. The polished area was Au-coated by a sputtering system (108 Auto Sputter Coater, Cressington, UK), which prevented charging during SEM observation. The composition was also analyzed by an energy dispersive spectrometer (EDS, Oxford Instruments, Abingdon, UK) and X-ray diffractometer (XRD, Mini Flex II, Rigaku, Japan) using Cu Kα radiation (λ = 1.5418 Å) with the operation condition of a step size 0.02 • and a scan speed of 3 • /min. The particle sizes of healing agents before and after encapsulation were analyzed by laser diffraction (Mastersizer 2000 Hydro 2000S, Malvern Instruments, Herrenberg, Germany). The encapsulated particles were gently ground to prevent the encapsulation shell from crushing to submicron particles, then each sample (100 mg) was dispersed in deionized water using an ultrasonic bath.

Coating Fabrication
In this study, the healing agents' embedded buffer layer was introduced between the bond coat and conventional top coat as a crack-resisting layer, following the coating designs in Figure 2. Three types of coating specimens were prepared: each structure of a conventional TBC with a YSZ top coat (sample A), a crack-healing TBC with healing agents embedded in a buffer layer (sample B), and a crack-healing TBC with encapsulated healing agents (sample C), which are described in Figure 2A-C, respectively. A disk-shaped Ni-based superalloy (Nimonic 263, a nominal composition of Ni-20Cr-20Co-5.9Mo-0.5Al-2.1Ti-0.4Mn-0.3Si-0.06C, in wt%, Thyssen Krupp VDM, Werdohl, Germany) was used as a substrate, with dimensions of 25 mm in diameter and 5 mm in thickness. The substrate was sandblasted using Al 2 O 3 powder (particle size: ≈ 420 µm) before bond coat fabrication, and the bond coat was deposited by high-velocity oxygen fuel (Diamond Jet-2600 DJM, Sulzer Metco Holding AG, Winterthur, Switzerland) methods with the thickness of 150-200 µm using AMDRY 9954 (Oerlikon Metco AG, Pfäffikon, Switzerland, nominal composition of Co-32Ni-21Cr-8Al-0.5Y in wt% and particle size of 11-63 µm). The buffer layer and top coat were fabricated by atmospheric plasma spray (APS, 9MB coating system, Sulzer Metco Holding AG) methods for thicknesses of 100-150 and 400-450 µm, respectively. METCO 204 C-NS (Oerlikon Metco AG, 8 wt% Y 2 O 3 doped in ZrO 2 , particle size of 45-140 µm) feedstock was used for the top coat material. For the buffer layer, YSZ feedstock was blended with each normal Si and encapsulated Si powder. The healing agent and YSZ with a volume ratio of 10:90 were mechanically mixed through ball milling for 24 h. All employed top and bond coat fabrication parameters as suggested by the manufacturer; see Table 1.

Thermal Durability Evaluation and Analysis
Cyclic Thermal Fatigue (CTF) and Jet Engine Thermal Shock (JETS) tests were performed for the prepared specimens to investigate the crack-resisting behavior under thermal durability tests. First, for the CTF tests, the specimens were held in the furnace with a dwell time of 40 min at 1100 °C to impose a thermal fatigue condition and then naturally cooled in air for 20 min as one cycle [37]. The CTF tests were performed until the TBCs failed. The failure criterion was defined as more than 50% spallation area of the top coat. The JETS test was conducted to evaluate the thermal shock resistance using a system designed with a jet engine motif; the top coat surface was exposed for 25 s to a direct flame with a temperature of about 1400 °C, directly quenched by N2 gas for 25 s, and then followed by natural cooling for 1 min as one cycle [38]. The JETS tests were performed with the target of 2000 cycles, and the failure criteria were defined as a spallation area of more than 20% of the top coat or any cracking specifically at the interface between top and bond coats. At least three samples were tested for each thermal durability test. The microstructure before and after thermal durability tests was observed using SEM and EDS.
To investigate the crack-resisting behavior in detail, the as-coated specimens were cut and polished using SiC paper, and they were finely polished with 1 and 3 μm diamond pastes and polishing sheets. The polished cross-sectional area was artificially cracked using a microindenter (HM-114, Mitutoyo Corp., Kawasaki, Japan) with a Vickers tip with the loading level of 10 N for 30 s. The artificial cracks were located above the interface between the top and bond coats within the buffer layer. Then, they were heat-treated at 1000 °C for 100 h, and their healing reaction and detailed crack-resisting behavior were analyzed by SEM and EDS.

Protection of Healing Agents and Resistance Against Premature Oxidation
The microstructure of the Si healing agent powder before and after encapsulation is shown in Figure 3. Figure 3A,B shows the normal Si and encapsulated Si powder surface morphologies, respectively. Each number indicated the components' intensity of Si, O, and Na from EDS mapping. Compared with the normal Si powder, the encapsulated powder had a relatively smooth surface morphology, which was evidence of a successful glassification process [36]. This morphology of

Thermal Durability Evaluation and Analysis
Cyclic Thermal Fatigue (CTF) and Jet Engine Thermal Shock (JETS) tests were performed for the prepared specimens to investigate the crack-resisting behavior under thermal durability tests. First, for the CTF tests, the specimens were held in the furnace with a dwell time of 40 min at 1100 • C to impose a thermal fatigue condition and then naturally cooled in air for 20 min as one cycle [37]. The CTF tests were performed until the TBCs failed. The failure criterion was defined as more than 50% spallation area of the top coat. The JETS test was conducted to evaluate the thermal shock resistance using a system designed with a jet engine motif; the top coat surface was exposed for 25 s to a direct flame with a temperature of about 1400 • C, directly quenched by N 2 gas for 25 s, and then followed by natural cooling for 1 min as one cycle [38]. The JETS tests were performed with the target of 2000 cycles, and the failure criteria were defined as a spallation area of more than 20% of the top coat or any cracking specifically at the interface between top and bond coats. At least three samples were tested for each thermal durability test. The microstructure before and after thermal durability tests was observed using SEM and EDS.
To investigate the crack-resisting behavior in detail, the as-coated specimens were cut and polished using SiC paper, and they were finely polished with 1 and 3 µm diamond pastes and polishing sheets. The polished cross-sectional area was artificially cracked using a microindenter (HM-114, Mitutoyo Corp., Kawasaki, Japan) with a Vickers tip with the loading level of 10 N for 30 s. The artificial cracks were located above the interface between the top and bond coats within the buffer layer. Then, they were heat-treated at 1000 • C for 100 h, and their healing reaction and detailed crack-resisting behavior were analyzed by SEM and EDS.

Protection of Healing Agents and Resistance Against Premature Oxidation
The microstructure of the Si healing agent powder before and after encapsulation is shown in Figure 3. Figure 3A,B shows the normal Si and encapsulated Si powder surface morphologies, respectively. Each number indicated the components' intensity of Si, O, and Na from EDS mapping. Compared with the normal Si powder, the encapsulated powder had a relatively smooth surface morphology, which was evidence of a successful glassification process [36]. This morphology of glassificated shell implied that some agglomerated healing particles were covered by the glass shell immediately. The EDS mapping image shows that the encapsulated powder contained Na, which consisted of the shell from the glassification process. Further, encapsulated powders were mounted and fine-polished to observe the cross-sectional microstructure and more detailed shell morphology. Figure 3C shows the obtained microstructure and EDS mapping analysis. Encapsulated Si powder (light area) was surrounded by mounting resin (dark gray), showing unremarkable differences in the morphologies. The existence of the oxygen intensity indicated that the encapsulation process successfully produced the sub-micron-scale SiO 2 shell (see Figure 3C-1,C-2). This suggested that the healing agents had a glassificated shell, which could protect the healing agent from premature oxidation during coating fabrication. Figure 4 shows the XRD analysis of the healing agents before and after the encapsulation processes. The cubic phase Si was observed in both specimens. After the encapsulation processes, a small cristobalite SiO 2 peak was observed as a shell component. The results of microstructural and phase analyses indicated that both crystalline and amorphous SiO 2 shells were well glassificated in the encapsulation process. glassificated shell implied that some agglomerated healing particles were covered by the glass shell immediately. The EDS mapping image shows that the encapsulated powder contained Na, which consisted of the shell from the glassification process. Further, encapsulated powders were mounted and fine-polished to observe the cross-sectional microstructure and more detailed shell morphology. Figure 3C shows the obtained microstructure and EDS mapping analysis. Encapsulated Si powder (light area) was surrounded by mounting resin (dark gray), showing unremarkable differences in the morphologies. The existence of the oxygen intensity indicated that the encapsulation process successfully produced the sub-micron-scale SiO2 shell (see Figure 3(C-1,C-2)). This suggested that the healing agents had a glassificated shell, which could protect the healing agent from premature oxidation during coating fabrication. Figure 4 shows the XRD analysis of the healing agents before and after the encapsulation processes. The cubic phase Si was observed in both specimens. After the encapsulation processes, a small cristobalite SiO2 peak was observed as a shell component. The results of microstructural and phase analyses indicated that both crystalline and amorphous SiO2 shells were well glassificated in the encapsulation process.   The particle size of the healing agents was analyzed for both the normal and encapsulated powders, the results of which are shown in Figure 5. The cumulative mass of the particles are shown in the Supplementary Table S1. Both samples showed similar size distributions, even after the encapsulation process. The encapsulated sample B showed that it contained more particles, with a particle size below 4 μm and above 20 μm, which resulted from the grinding after encapsulation   The particle size of the healing agents was analyzed for both the normal and encapsulated powders, the results of which are shown in Figure 5. The cumulative mass of the particles are shown in the Supplementary Table S1. Both samples showed similar size distributions, even after the encapsulation process. The encapsulated sample B showed that it contained more particles, with a particle size below 4 µm and above 20 µm, which resulted from the grinding after encapsulation and agglomeration, respectively. Similar particle size distributions indicated that the encapsulation process produced only a sub-micron-scale shell as well as the agglomerated particles, which corresponded to the microstructural observations in Figure 3B,C. In the APS process, the Si healing agent can be more easily oxidized than YSZ because it has a lower melting point and plasma-spray-forming temperature (10,000-15,000 K), resulting in the loss of its function [39]. However, the successfully produced submicron shell can protect healing agents, so it was expected that premature oxidation could be prevented. With the above discussion, the encapsulated healing agent was capable of being employed in the actual coating of specimens.

Thermal Durability Evaluation
All specimens were fabricated with the coating design shown in Figure 2. The cross-sectional microstructure of the as-coated specimens is shown in Figure 6. Each healing agent was embedded differently within the buffer layer, depending on the encapsulation process.

Thermal Durability Evaluation
All specimens were fabricated with the coating design shown in Figure 2. The cross-sectional microstructure of the as-coated specimens is shown in Figure 6. Each healing agent was embedded differently within the buffer layer, depending on the encapsulation process.
In Figure 6A-1,A-2, the conventional YSZ (sample A) showed a typical TBC microstructure fabricated by the APS method, showing splat boundaries, microcracks, and pores. All samples showed a sound bond coat and an interfacial condition without noticeable growth of a thermally grown oxide (TGO) layer.
The Si healing agent without encapsulation (sample B) had the morphology of an undulating shape and was embedded between the YSZ splats in the form of particles ( Figure 6B-1,B-2). However, fewer Si healing agents were observed than the blended volume ratio in the as-prepared coating, which resulted from the evaporation of Si because of its low melting point [40].
The microstructure of the encapsulated healing-agent-embedded specimen (sample C) is shown in Figure 6C-1,C-2. The healing agents were embedded in the shape of the particles, which were similar to sample B. However, some of the healing agents were observed in the agglomerated condition, resulting from the encapsulation process (shown in Figure 6C-2). EDS-point mapping analysis of Figure 6C Figure S1 to prove the existence of the healing agent within the buffer layer. There was a relatively large number of embedded healing agents in comparison to normal Si (See Figure 6B-1,C-1. This result indicated that the encapsulation shell and larger particle size, as a result of agglomeration, reduced their evaporation [40].

Thermal Durability Evaluation
All specimens were fabricated with the coating design shown in Figure 2. The cross-sectional microstructure of the as-coated specimens is shown in Figure 6. Each healing agent was embedded differently within the buffer layer, depending on the encapsulation process. In Figure 6(A-1,A-2), the conventional YSZ (sample A) showed a typical TBC microstructure fabricated by the APS method, showing splat boundaries, microcracks, and pores. All samples showed a sound bond coat and an interfacial condition without noticeable growth of a thermally grown oxide (TGO) layer.   Figure 7A, the healing agents in sample B were located between the YSZ splats, filling the empty splat boundaries. However, the encapsulated Si was in the shape of agglomerated powder as a kind of a cluster, which formed during heat treatment of the encapsulation process. The agglomerated healing particle consisted of pores, Si (light gray), partially oxidized Si (SiO x , gray), and shell component SiO 2 (dark gray), which were based on the EDS-point analysis in Supplementary Figure S2. These results confirmed that the glassificated shell covered the healing agents well and protected them from premature oxidation during the APS process. The effects of healing agent distribution and morphology on the thermal durability and crack-healing behavior were investigated according to the encapsulation. The Si healing agent without encapsulation (sample B) had the morphology of an undulating shape and was embedded between the YSZ splats in the form of particles ( Figure 6(B-1,B-2)). However, fewer Si healing agents were observed than the blended volume ratio in the as-prepared coating, which resulted from the evaporation of Si because of its low melting point [40].
The microstructure of the encapsulated healing-agent-embedded specimen (sample C) is shown in Figure 6(C-1,C-2). The healing agents were embedded in the shape of the particles, which were similar to sample B. However, some of the healing agents were observed in the agglomerated condition, resulting from the encapsulation process (shown in Figure 6(C-2)). EDS-point mapping analysis of Figure 6(C-1) is attached in Supplementary Figure S1 to prove the existence of the healing agent within the buffer layer. There was a relatively large number of embedded healing agents in comparison to normal Si (See Figure 6(B-1,C-1). This result indicated that the encapsulation shell and larger particle size, as a result of agglomeration, reduced their evaporation [40]. Figure 7 shows the magnified microstructures of samples B and C with the component analysis around the healing agents. In Figure 7A, the healing agents in sample B were located between the YSZ splats, filling the empty splat boundaries. However, the encapsulated Si was in the shape of agglomerated powder as a kind of a cluster, which formed during heat treatment of the encapsulation process. The agglomerated healing particle consisted of pores, Si (light gray), partially oxidized Si (SiOx, gray), and shell component SiO2 (dark gray), which were based on the EDS-point analysis in Supplementary Figure S2. These results confirmed that the glassificated shell covered the healing agents well and protected them from premature oxidation during the APS process. The effects of healing agent distribution and morphology on the thermal durability and crack-healing behavior were investigated according to the encapsulation.  Figure 8 shows the cross-sectional microstructure around the delaminated area after CTF tests. During thermal cycling, different thermo-mechanical properties between the matrix and healing agent could develop stress, so only 10 vol% of healing agents was blended with YSZ to minimize the development of stress, which was founded in other previous studies [27,29].
Samples A, B, and C were delaminated after 1067, 1143, and 1103 cycles of CTF tests on average, respectively. Healing-agent-embedded TBCs (samples B and C) showed comparable thermal fatigue resistance to conventional YSZ. The continuous growth of TGO and some spinel   Figure 8 shows the cross-sectional microstructure around the delaminated area after CTF tests. During thermal cycling, different thermo-mechanical properties between the matrix and healing agent could develop stress, so only 10 vol% of healing agents was blended with YSZ to minimize the development of stress, which was founded in other previous studies [27,29]. extrinsic toughening mechanism [41]. During thermal exposure, the healing agents were oxidized to SiO2, which had a glass transition temperature of about 1200 °C. Then, the shell components were glassificated and acted as ductile particles, resulting in the enhancement of fracture toughness by plastic deformation [42,43]. Although they showed similar thermal fatigue resistance, the slightly enhanced thermal durability in sample B can be explained by increased toughening probability, which resulted from a relatively uniform distribution and the fine size of the healing agents. However, sample C with encapsulated Si showed a slight decrease in thermal fatigue lifetime compared to the normal Si. This was because the concentration of agglomerated Si particles rather limited its healing function. To achieve noticeable enhancement of thermal fatigue lifetime, these limitations were countered by optimization of encapsulation process or postprocessing for uniform powder without agglomeration. The cross-sectional microstructures after the JETS tests are shown in Figure 9. All the specimens survived 2000 cycles of JETS tests. The noticeable delamination or failure of TBCs was not observed in common. Even though they had similar lifetime performances, they had quite different microstructural degradation behaviors. In common, the bond coats showed sound conditions without evident oxidation and TGO growth, while severe TGO growth (including spinel phase) was observed after the CTF test. The most severe microstructural degradation was observed Samples A, B, and C were delaminated after 1067, 1143, and 1103 cycles of CTF tests on average, respectively. Healing-agent-embedded TBCs (samples B and C) showed comparable thermal fatigue resistance to conventional YSZ. The continuous growth of TGO and some spinel oxides was observed at the interface between the top and bond coats, showing similar, common interfacial conditions. In sample A, small microcracks and pores were generated compared with the as-coated condition, while relatively thickened and enlarged cracks were generated within both buffer layers in samples B and C. Most of the cracks just above the bond coat grew horizontally because of the stress caused by CTE mismatch, which resulted in delamination, while the propagation of some cracks was inhibited by the healing agents. During the temperature change, the stress from CTE mismatch dominantly accelerated the thickening, linkage, and coalescence of cracks, resulted in the propagation of cracks. However, the healing agents inhibited propagation of these cracks because of their existence, which functioned as a ductile particle elaborated with an extrinsic toughening mechanism [41]. During thermal exposure, the healing agents were oxidized to SiO 2 , which had a glass transition temperature of about 1200 • C. Then, the shell components were glassificated and acted as ductile particles, resulting in the enhancement of fracture toughness by plastic deformation [42,43]. Although they showed similar thermal fatigue resistance, the slightly enhanced thermal durability in sample B can be explained by increased toughening probability, which resulted from a relatively uniform distribution and the fine size of the healing agents. However, sample C with encapsulated Si showed a slight decrease in thermal fatigue lifetime compared to the normal Si. This was because the concentration of agglomerated Si particles rather limited its healing function. To achieve noticeable enhancement of thermal fatigue lifetime, these limitations were countered by optimization of encapsulation process or postprocessing for uniform powder without agglomeration.
The cross-sectional microstructures after the JETS tests are shown in Figure 9. All the specimens survived 2000 cycles of JETS tests. The noticeable delamination or failure of TBCs was not observed in common. Even though they had similar lifetime performances, they had quite different microstructural degradation behaviors. In common, the bond coats showed sound conditions without evident oxidation and TGO growth, while severe TGO growth (including spinel phase) was observed after the CTF test. The most severe microstructural degradation was observed in sample A, which showed large crack formations because of the repeated thermal shocks. Vertical cracks initiated from the surface because of rapid volume shrinkage during N 2 quenching, and they grew into large cracks through preexisting splats or pores under repeat of thermal shock, which was observed similarly in previous work [44]. Large horizontal cracks were generated near the interface because there was thermal stress from CTE mismatch between the top and bond coats, which could result in the failure of TBCs. Crack formation around the surface of the top coat was similarly observed in samples B and C. However, a different crack propagation was observed near the interface. Horizontal cracks formed and propagated to the healing agents during the JETS test. Although horizontal cracks were not clearly observed, the sub-micron-healing agent in sample B did not show noticeable crack-resisting behavior. This was because the crack scale had obviously larger openings and propagations than microscale Si healing particles, which embedded in the form of particles. On the other hand, the embedded Si particle, while agglomerated as healing agents in sample C, inhibited the rapid crack opening and propagation as a ductile particle under thermal shock conditions, which was not fully oxidized because of the relatively short thermal exposure period compared with the CTF test. Detailed crack-resisting mechanisms during the tests are elaborated on in Section 3.3. Figure 10 shows the cross-sectional microstructure around the indentation imprints, which induced artificial cracks within the crack-healing buffer layer in sample B. The crack propagation-resisting behavior related to healing agents was investigated. The typical self-healing system was designed to seal nano-or microcracks before the occurrence of extensive damage. However, in the air-plasma sprayed TBC system, there were preexisting defects such as splat boundaries and microcracks, so crack-filling behavior against the opened defects was mainly investigated [27]. In Figure 10A-C, the dotted rhombus box indicated indentation imprints (about 50 µm diagonal), and emerged cracks were indicated by white arrows. Crack-filling behavior was observed in Figure 10E,F by the healing agents, which showed volume expansion because of oxidation. The formation of ZrSiO 4 (zircon) evidenced new solid bonds between the YSZ matrix and Si healing agents, which could enhance the interfacial stability as well as heal the cracks by filling (see the magnified image in Figure 10D and EDS-point and mapping analysis in Supplementary Figure S3). Moreover, crack growth, like coalescence and/or linkage, was inhibited by the embedded healing agents, which can be observed within the YSZ splat boundaries in Figure 10G,H. larger openings and propagations than microscale Si healing particles, which embedded in the form of particles. On the other hand, the embedded Si particle, while agglomerated as healing agents in sample C, inhibited the rapid crack opening and propagation as a ductile particle under thermal shock conditions, which was not fully oxidized because of the relatively short thermal exposure period compared with the CTF test. Detailed crack-resisting mechanisms during the tests are elaborated on in Section 3.3.  The cross-sectional microstructure of sample C is shown in Figure 11 to investigate the crack-resisting behavior of encapsulated healing agents. In Figure 11A-C, the dotted rhombus box and white arrows indicated the indentation imprints (about 50 µm diagonal) and emerged cracks around the healing agents, respectively. In the magnified microstructure of Figure 11D, solid bonding was formed, the behavior of which was similar to the unencapsulated healing agents in sample B. The cracks around the healing agents were filled and propagated no further, especially around the unagglomerated particles ( Figure 11E,F). However, some small microcracks were not fully filled around the agglomerated healing agents because the protective shell accommodated the volume expansion by oxidation. Even though the microcracks were partially filled, it could still prevent severe crack propagation that could lead to catastrophic failure [45]. In Figure 11G, large cracks were initiated because of the abrupt stress resulting from the CTE mismatch and TGO growth between the top and bond coats. The volume expansion of healing agent could generate stress, but the agglomerated healing agents contained some pores as a kind of a cluster, which accommodated some extent of stress (see the Figure 11F). Therefore, it was considered that stress by expansion was not dominant compared to the stress by CTE mismatch and TGO growth. As shown in Figure 11H,I, the microcracks and splat boundary were thickened from the interface. However, these cracks, which can finally cause delamination, stopped growing by branching into small cracks within the agglomerated healing agents, observed commonly in sample C. microcracks, so crack-filling behavior against the opened defects was mainly investigated [27]. In Figure 10A-C, the dotted rhombus box indicated indentation imprints (about 50 μm diagonal), and emerged cracks were indicated by white arrows. Crack-filling behavior was observed in Figure  10E,F by the healing agents, which showed volume expansion because of oxidation. The formation of ZrSiO4 (zircon) evidenced new solid bonds between the YSZ matrix and Si healing agents, which could enhance the interfacial stability as well as heal the cracks by filling (see the magnified image in Figure 10D and EDS-point and mapping analysis in Supplementary Figure S3). Moreover, crack growth, like coalescence and/or linkage, was inhibited by the embedded healing agents, which can be observed within the YSZ splat boundaries in Figure 10(G,H). The cross-sectional microstructure of sample C is shown in Figure 11 to investigate the crackresisting behavior of encapsulated healing agents. In Figure 11A-C, the dotted rhombus box and white arrows indicated the indentation imprints (about 50 μm diagonal) and emerged cracks around the healing agents, respectively. In the magnified microstructure of Figure 11D, solid bonding was formed, the behavior of which was similar to the unencapsulated healing agents in sample B. The cracks around the healing agents were filled and propagated no further, especially around the unagglomerated particles ( Figure 11E,F). However, some small microcracks were not fully filled around the agglomerated healing agents because the protective shell accommodated the volume expansion by oxidation. Even though the microcracks were partially filled, it could still prevent severe crack propagation that could lead to catastrophic failure [45]. In Figure 11G, large cracks were initiated because of the abrupt stress resulting from the CTE mismatch and TGO growth between the top and bond coats. The volume expansion of healing agent could generate stress, but the agglomerated healing agents contained some pores as a kind of a cluster, which accommodated some extent of stress (see the Figure 11F). Therefore, it was considered that stress by expansion was not dominant compared to the stress by CTE mismatch and TGO growth. As shown in Figure 11H,I, the microcracks and splat boundary were thickened from the interface. However,  these cracks, which can finally cause delamination, stopped growing by branching into small cracks within the agglomerated healing agents, observed commonly in sample C. The crack-resisting behavior of Si healing agents can be elaborated on with the following three mechanisms: (i) healing agents that are located between splat boundaries form zircon as a new solid bonding, which enhances the resistance of tensile stress and results in less delamination of TBC; (ii) healing agents, with their partially unmelted shapes, can reduce the near-crack-tip stress as ductile The crack-resisting behavior of Si healing agents can be elaborated on with the following three mechanisms: (i) healing agents that are located between splat boundaries form zircon as a new solid bonding, which enhances the resistance of tensile stress and results in less delamination of TBC; (ii) healing agents, with their partially unmelted shapes, can reduce the near-crack-tip stress as ductile particles and fill small cracks, preventing microcrack opening; and (iii) agglomerated healing agents above the interface between the top and bond coats can inhibit large crack propagation by branching into small cracks within the healing agents, although they cannot stop small crack openings through crack filling.

Crack-Resisting Mechanisms
Based on the results above, repeating out-of-plane tensile stress can accumulate damage and cause coalescence and linkage of discontinuities such as microcracks and splat boundaries. Sample B has a relatively uniform distribution of healing agents with a fine particle size, so the healing agent can effectively stop the crack opening and form new solid bonding all over the buffer layer, which results in thermal durability and, especially, thermal fatigue resistance. On the other hand, large crack propagations, because of the abrupt stress from rapid temperature changes during the JETS test, were blocked by the agglomerated healing agent above the interface in sample C. This elaborates on the enhanced thermal shock resistance of sample C, as it has a relatively more sound microstructure than sample B without noticeable large cracks after the JETS tests.
However, even though the healing agents filled some cracks and blocked the crack opening successfully, they cannot fully fill the microcracks around them nor can they fill large cracks because of their poor fluidity [46,47]. This can be overcome by increasing the fluidity or decreasing the viscosity at the operating temperature by doping some appropriate contents or modifying the encapsulation process [27]. Furthermore, the optimization of particle size distribution and volume ratio of healing agents will be studied in future work.

Conclusions
In this work, a novel TBC system with a buffer layer embedded with Si-based crack-healing agents has been developed and evaluated. Using thermal durability tests, the crack-resisting behaviors and mechanisms of Si, and encapsulated Si, and the related mechanisms are presented. The main contributions are as follows.

•
An encapsulation process is developed to cover the healing agents, preventing them from premature oxidation or evaporation during coating fabrication.

•
The TBCs with a crack-healing buffer layer show comparable thermal durability in both the CTF and JETS tests.

•
Crack-resisting mechanisms are elaborated on depending on the distribution and shape of healing agents through the observation of crack-healing behaviors.
Supplementary Materials: The following are available online at http://www.mdpi.com/2079-6412/9/6/358/s1, Figure S1: EDS mapping analysis around the healing agents in as-coated samples from Figure 6C-1; Figure S2: EDS point-analysis around the healing agents in as-coated samples from Figure 7A,B; Figure S3: EDS pointand mapping analysis around the healing agents after heat treatment from the magnified image in Figure 10C; Table S1: Particle diameters at 10%, 50%, and 90% of the cumulative mass of (A): normal Si, and (B) encapsulated Si from Figure 5.