Introduction

Today, there is no need to elaborate on the reasons why cold spray is an attractive and powerful process with a high potential to meet industrial demands. The best evidence can be found in the just-published “The 2016 Thermal Spray Roadmap” by Vardelle et al. (Ref 1), which ranked cold spray as a leading thermal spray process. Further evidence does exist in various overviews of cold spray such as in Ref 2 recently. Most popular materials used in cold spray are aluminum (Al) and its alloys due to their low density, deformation rate and ductility (Ref 3,4,5,6,7,8). Cold spray process parameters can be easily optimized to achieve spray velocities required to produce dense Al coatings (Ref 3, 4) with high bonding on different substrate materials. Because of their specific properties, they are commonly used for automotive and aircraft applications on a wide range of substrates. In particular, they are expected to be beneficial for coating steel using cold spray (Ref 5), for wear protection for example.

Aluminum coating properties can be improved by mixing ceramic particles such as B4C (Ref 4), TiN (Ref 9), SiC (Ref 10) or Al2O3 (Ref 11,12,13,14,15,16,17), with aluminum powder. Process parameters can be easily optimized to produce dense Al-Al2O3 coatings. These coatings exhibited higher mechanical properties such as wear resistance and bond strength, compared to that of pure Al coatings (Ref 11,12,13,14). Alumina (Al2O3) addition is known to reduce coating porosity and increase deposition efficiency and hardness. These improvements result from the well-known hammering effect of ceramic particles during spraying (Ref 15, 16). However, the benefit of Al2O3 addition, through hammering effect, is not really understood today because it requires observation of Al/Al2O3 interfaces down to a very fine scale, i.e., that of transmission electron microscopy (TEM) typically. In Ref 18, for example, some Al particles were cold-sprayed onto a sapphire substrate, which resulted in Al/Al2O3 interfaces with a high grain refinement.

Some Al and Al-Al2O3 powders are commercially available for cold spray today. However, these were not made specifically and tailored for cold spray, due to the lack of results on the effect of powder characteristics on cold spray coatings (Ref 7, 8). Powder parameters are numerous, e.g., grain size, particle shape, particle microstructure or oxide content. A few studies only focused on powder morphology and its effect on cold spray properties. The well-known influence of powder particle shape results in the higher in-flight velocity of irregular particles compared to that of spherical particles (Ref 2, 7, 19). The influence of Al2O3 morphology on cold spray Al-based Al2O3-added composite coating microstructure and related dry-sliding wear behavior was also investigated (Ref 13). Al-Al2O3 coating needs a lower ceramic content for spherical alumina than for angular alumina to reach a similar tribological behavior.

In this article, a multiscale experimental and numerical approach to powder particle shape effect on the Al-Al2O3 coating build-up mechanism is proposed. Two aluminum and two alumina powders with either spherical or irregular shape were characterized and successfully cold-sprayed onto a grit-blasted steel substrate. The same cold spray parameters were used for all powders and mixtures to focus on the effect of the powder particle shape. These parameters were not optimized deliberately to be sure that the influence of powder particle shape could be observed. Several effects of the Al and Al2O3 particle shape on the coating build-up were highlighted from scanning electron microscopy (SEM) observation. Numerical simulation of multiple particle impact was undertaken to elucidate the densification of Al coating due to Al2O3 addition, observed conventionally (Ref 11, 13) but never clarified. Using TEM, Al/Al and Al/Al2O3 interfaces were compared to understand the strengthening effects caused by embedded hard particles within a composite coating. This study succeeded in understanding cold spray composite coating build-up mechanisms and optimizing them to improve properties such as coating–substrate bond strength.

Experimental

Materials

Two commercial aluminum powders, one with a spherical particle shape (Poudres Hermillon, France) and the other one with an irregular particle shape (ECKA Granules, Germany), were cold-sprayed onto grit-blasted steel (200 HV0.1). The nominal particle size for these two powders is, respectively, [−55 + 5 µm] and [−45 µm]. These aluminum powders were mixed with two alumina powders (Saint-Gobain, France) using a Turbula® T2A blender (Willy A. Bachofen, Switzerland) during 5 h. The Al2O3 content in powder mixtures was fixed at 15 wt.%, which was supposed to be the best compromise to observe Al and Al2O3 particle shape effects in a powder mixture (Ref 11, 12). The first type of alumina powder showed angular particles, and the second type showed spherical particles. The nominal particle size for these two powders is, respectively, [−30 µm] and [−26 + 12 µm]. Depending on their particle shape, the two Al powders will be referred, respectively, as spherical Al powder and irregular Al powder and the two Al2O3 powders as spherical Al2O3 and angular Al2O3 in this article. Each mixture will be referred by the name of the Al and Al2O3 powders. For instance, a mixture made with the spherical Al powder and the angular Al2O3 powder is referred in this article as sph. Al + 15 wt.% ang. Al2O3.

Particle size distributions of aluminum and alumina powders were obtained with a laser particle size analyzer Mastersizer 3000® (Malvern).

Cold Spray Process

Al powders and Al-15 wt.% Al2O3 mixtures were cold-sprayed using CGT Kinetics 3000M facilities with, respectively, a PBI nozzle (Oerlikon, Switzerland) and a MOC nozzle (Oerlikon, Switzerland) perpendicular to the substrate. Different nozzles were used for spraying Al and Al-Al2O3 powders because of aluminum clogging in MOC nozzle and alumina abrasion of PBI nozzle. Nitrogen was used as spraying gas. The process gas pressure was 2.5 MPa with a gas temperature of 400 °C. The powder feed rate was equal to 7.1 g min−1. The stand-off distance between the nozzle exit and the substrate, was kept at 30 mm. The substrate remained fixed, and the nozzle moved at 200 mm s−1. All the operating conditions were exactly the same for aluminum powders and aluminum–alumina mixtures. Consequently, the effect of powder morphology on cold spray coating microstructure could be studied. Coatings made of irregular aluminum powder and those made of spherical aluminum powder will be termed, respectively, as “irregular aluminum coating” and “spherical aluminum coating” in the rest of the article.

In-flight particle velocities were measured with a DPV-2000® (Tecnar, Canada). Particles with a diameter between 1 and 150 µm and an in-flight velocity between 50 and 1200 m s−1, were detected. Four samples (25 × 25 × 6 mm3) were weighed before m i and after deposition m f . Deposition efficiency was calculated according to Eq 1, where D was the powder flow rate and t was the spraying duration.

$${\text{DE}} = \frac{{m_{f} - m_{i} }}{D \cdot t} \cdot 100$$
(1)

Microstructure

Coatings and powder particles were cold-mounted and cross-sectioned. They were then ground with SIC paper down to 1200 grit paper and polished down to 1-µm diamond paste followed by finishing step using colloidal silica. They were then metallized with Au-Pd prior to scanning electron microscope observations with a SEM LEO 1450VP (Zeiss, UK). Ten images per coating were processed using the ImageJ® software. Average porosity, alumina weight content and thickness were calculated from ten SEM pictures for every coating. XRD measurements (copper cathode, λ = 1.540 Å) were used to check the results from image analysis for the alumina weight content in starting Al-Al2O3 powder mixtures and in Al-Al2O3 coatings.

TEM thin foils were prepared for microstructure examination using the focused ion beam (FIB) lift-out technique (Fig. 1) with a dual-column scanning electron microscope associated with a focused ion beam SEM-FIB Helios 660 (FEI, USA). This technique allows a targeted preparation of relatively large electron-transparent areas, which is particularly useful to observe specific areas in composite materials. Prior to this preparation, the surface of a coating made of sph. Al + 15 wt.% sph. Al2O3 powder was polished to remove the coating surface roughness. A given Al-Al2O3 interface was selected, and its surface was protected by an ion beam-induced carbon layer (Fig. 1a and b). Two large stair-step trenches were milled on both sides of the area of interest with a 30-kV FIB Ga+ beam (Fig. 1c). Subsequently, the resulting thin foil was lifted out and placed on a TEM copper lift-out grid (Omniprobe, USA) and FIB polished (Fig. 1h and i). TEM observation and analyses were carried out at 200 kV using a Tecnai F20-ST field-emission gun microscope (FEI, USA) equipped with an energy-dispersive x-ray spectroscopy (EDX) device and a high-angle annular dark field (HAADF) detector. The latter associated with probe scanning (STEM) capabilities of the microscope allows acquiring of Z-contrast images. EDX analysis in line-scan mode using a ~1-nm probe size was combined to STEM-HAADF imaging to determine the oxygen content at Al/Al and Al/Al2O3 interfaces. In addition, high-resolution transmission electron microscopy (HRTEM) was performed to investigate structural and crystallographic features at these interfaces.

Fig. 1
figure 1

FIB lift-out method: (a) Al2O3 particle selection, (b) carbon coating deposition, (c) and (d) coarse ion beam polishing, (e) ion beam milling, (f) and (g) sample lift-out, (h) sample on copper holder and (i) fine polishing

Full size image

Numerical Simulation

A two-dimensional numerical simulation of high-velocity Al and Al2O3 particle impacts onto Al coating was developed to study interactions between particles during spraying as a function of particle shape. This study focused on coating build-up mechanisms during spraying. The first stages in coating build-up were not taken into account in the simulations. This means that steel substrate was not involved. Consequently, coating build-up started from already-deposited Al-Al2O3, assuming that this substrate had the same properties as the bulk Al material. The simulations were carried out using ABAQUS/Explicit® software (Ref 20) with a Lagrangian formulation.

Two different approaches were used to model the real particle shape of Al and Al2O3 particles. Spherical particles were simply modeled as disks with a diameter of 20 µm. Irregular Al and angular Al2O3 particle shapes were implemented using an in-house image extraction method, as described in Ref 21. First, several SEM micrographs were taken from polished cross sections of irregular Al and angular Al2O3 powders. Then, all the particles were isolated individually in micrographs. The coordinates of their outlines were extracted using the open-source “Simple Morphological Image Library (SMIL)” software (Ref 22). Using these coordinates, the real profile of irregular particles was involved in simulations with ABAQUS®. Only one particle shape of irregular Al and one of angular Al2O3 are shown in this article from series of various tested particle shapes. No alumina layer on the Al particle surface was considered in contrast with Ref 5.

Two equations of state were used for the aluminum material model. The Mie–Grüneisen equation of state (EOS), describing the material state in the high-pressure domain and shock wave propagation, was used in the Hugoniot formulation (Ref 21) and the viscoplastic deformation by the widely used Johnson–Cook plasticity model (Ref 2, 21, 23, 24). For the alumina material model, the Johnson–Holmquist model (Ref 25,26,27) was selected. The material parameters for the two material models are given in Table 1. They were optimized using single splat experiments of Al substrate impacted by Al2O3 particles and Al2O3 substrate impacted by Al particles.

Table 1 Material parameters for numerical simulation taken from (Ref 24) for Al and from (Ref 26, 27) for Al2O3
Full size table

Four finite element impact configurations were studied (Fig. 2). Each configuration simulated the impact of one alumina particle (in dark gray) and two aluminum particles (in light gray) onto an Al-Al2O3 substrate which was assumed to be an Al substrate. The in-flight particle velocities were set as their values measured with a DPV-2000® device. This is explained latter in this article. Particles and substrate were meshed with CPE4RT elements which are bilinear displacement and temperature quadratic elements with reduced integration with hourglass control. A mesh size of 0.6 µm was used in the simulations. Arbitrary Lagrangian Eulerian (ALE) adaptive remeshing was also performed in this numerical simulation to avoid excessive element distortion.

Fig. 2
figure 2

Numerical simulation cases: (a) impact of spherical particles, (b) impact of one angular Al2O3 particle and two spherical Al particles, (c) impact of one spherical Al2O3 particle and two irregular Al particles and (d) impact of one angular Al2O3 particle and two irregular Al particles

Full size image

Mechanical Characterization

A Vickers indenter applied during a hold time of 10 s was used for microhardness measurements. A 10-g and a 100-g loads were applied to measure powder and coating microhardnesses, respectively. The hardness value was an average of at least ten indents taken at a random location. The bond strength of each coating was determined by pull-off testing which was conducted according to ASTM C633 (Ref 28). Five test specimens for each aluminum powder or aluminum–alumina powder mixtures were assembled by glue DELOMONOPOX© AD286. The glue needed heat treatment for curing during 1 h at 150 °C with air convection oven. No influence of this heat treatment on hardness of the coating was observed. These were pulled using a Zwick 1474© (Zwick/Roell) tensile test machine.

Results

Feedstock Powder Characterization

Figure 3 shows the particle shape of the four powders used in this work. Irregular aluminum powder (Fig. 3a) was atomized in nitrogen, which explains its spheroidal rather elongated shape as in Ref 8. Spherical Al powder (Fig. 3b) is conventionally used in cold spray (Ref 3, 7, 11, 12, 15) and was also atomized in nitrogen. For the two aluminum powders, some small particles were agglomerated onto large ones. These small satellites were smaller in the spherical Al powder than in the irregular Al powder. No Al oxidized layer was observed at the Al particle surface of the two powders with SEM, and no Al oxide was detected with XRD. The two Al powders were therefore not or little oxidized. Microhardnesses of the irregular and spherical powders were very close (respectively, 28 ± 1.8 HV0.01 and 28 ± 2.3 HV0.01). Almost the same hardness was measured in Ref 11.

Angular Al2O3 (Fig. 3c) is conventionally used in cold spray (Ref 11, 13, 15). All the particles in this powder were plain angular. Spherical Al2O3 powder (Fig. 3d) was obtained by plasma spheroidization of angular Al2O3 powder. Most of the particles were plain spherical. However, a small number of angular alumina particles can be found due to the powder manufacturing process. Few spherical alumina particles were hollow. Due to powder production, transport and handling, some of these hollow particles broke and fragments could be observed in Fig. 3(d). The two alumina powders showed the same hardness (more than 2000 HV0.01) as in Ref 11.

Fig. 3
figure 3

SEM micrographs of feedstock aluminum powders with irregular (a) and spherical (b) particle shapes and alumina powders with angular (c) and spherical (d) particle shapes

Full size image

Figure 4 shows the powder size distribution and the powder size range of the four powders. All the powders had rather conventional powder size range, i.e., <45 µm in diameter, for cold spray (Ref 2). The two Al powder size ranges were [−45 + 11 µm] for irregular aluminum powder and [−32 + 11 µm] for spherical aluminum powder. The powder size distribution of the irregular Al powder was large and asymmetrical with a high fraction of coarse particles. The powder size distribution of the spherical Al powder was rather narrow and symmetrical with a significant fraction of very fine particles (<5 µm) which were probably detached satellites (Fig. 3b). The particle size range was [−34 + 9 µm] for angular Al2O3 and [−26 + 10 µm] for spherical Al2O3. Even if there is a slight difference in the two particle sizes, the latter were considered to be similar. The powder size distribution of spherical Al2O3 was narrow and symmetrical centered on the mean particle size value (Fig. 4). In contrast, the powder size distribution of angular alumina was less symmetrical with a higher fraction of coarse particles. A significant fraction of the two Al2O3 powders was composed of very small particles (<5 µm). For spherical alumina, these small particles were assumed to be fragments of hollow particles which broke off. For angular alumina, these were assumed to be fragments due to crushing at the powder production stage.

The mean diameters for the spherical Al, irregular Al, spherical Al2O3 and angular Al2O3 were, respectively, 20, 24, 17 and 19 µm. A 20-µm particle size seemed to be the best compromise to study the particle shape effect only but not the particle size effect. This particle size had been fixed for the in-flight velocity measurements and for the simulations.

Fig. 4
figure 4

Powder size distributions and powder size ranges of the four tested powders

Full size image

Process

Deposition efficiency for Al and Al-Al2O3 coatings can be almost compared to that which had been found in earlier works (Ref 2, 13). Despite the use of the same spraying conditions, the deposition efficiency varied depending on the type of aluminum and alumina powder particle shape (Fig. 5). Irregular aluminum powder showed a higher deposition efficiency than spherical aluminum. Spherical Al2O3 addition increased the deposition efficiency of aluminum powders. Angular Al2O3 addition increased the deposition efficiency of irregular aluminum powder but decreased that of spherical aluminum powder. Higher deposition efficiency was achieved for two powder mixtures i.e., that made of irregular Al powder with angular Al2O3 and that of spherical Al powder with spherical Al2O3. Similar particle shapes for aluminum and alumina powders led to cold spray coatings with a very high deposition efficiency.

Fig. 5
figure 5

Deposition efficiency for Al and Al-Al2O3 cold-sprayed coatings

Full size image

In-flight particle velocities were acquired for the two types of Al powders and the two Al2O3 powders as in Ref 11. Similar particle size ranges and distributions were found using a conventional particle size analyzer (Fig. 4) and DPV-2000® for each powder. Particle velocities for a particle size between 15 and 20 µm were extracted from these data. For each powder, a velocity map with a size of 16 × 14 mm2 (Fig. 7) was acquired, and from each map, the average particle velocity (Fig. 6) was calculated. The latter was used for the numerical simulations.

Spherical Al particles had a higher average velocity (578 ± 150 m s−1) than irregular Al particles (569 ± 150 m s−1) which was not expected. However, this difference was rather low and not very relevant because of scattering (Fig. 6). The two Al powders can be assumed to have almost the same average in-flight velocity which was not really expected (Ref 19). In contrast, velocity differences between alumina and aluminum powders were significant. Alumina particles had a higher velocity than aluminum powders. Spherical Al2O3 particles velocity (614 ± 79 m s−1) was lower than that of angular particles (678 ± 102 m s−1) as expected (Ref 2).

Fig. 6
figure 6

In-flight particle velocities of different powders used in this work

Full size image

Al and Al2O3 velocity maps (Fig. 7) cannot be compared because these powders were not sprayed with the same nozzle. The exit diameter (10 mm2) of the PBI nozzle used for Al powders was higher than that of MOC nozzle (6 mm2) for Al2O3 powders. Consequently, the velocity maps of Al powders were larger than those of Al2O3 powders. Velocity maps (Fig. 7a and b) showed that the jet for irregular Al (12 × 10 mm2) was smaller than that for spherical Al (12 × 12 mm2). The center zone with the highest particle velocities was smaller for spherical Al particles (2 × 1 mm2) than for irregular Al particles (2 × 2 mm2). The center of the particle flow of irregular Al particles was rich in high-speed particles. This was the sign of a velocity effect due to Al particle shape. The same effect was observed for alumina powders. The jet for angular Al2O3 powder was smaller than that for spherical Al2O3 powder. Moreover, the center zone with the highest particle velocities was smaller for spherical Al2O3 particles (2 × 1 mm2) than for irregular Al2O3 particles (2 × 2 mm2). The angular alumina particle jet was more concentrated in high-velocity particles, and the spherical alumina particle jet was more spread out. All these observations ascertained a velocity effect due to powder particle shape.

Fig. 7
figure 7

In-flight particle velocity map for different powders, (a) spherical Al, (b) irregular Al, (c) sph

Full size image

Microstructure

Coating microstructures were studied using SEM. In Fig. 8, the steel substrate is in white and the coating in gray. All the coatings seemed to be well bounded to the steel substrate, except coatings made with the irr. Al powder and irr. Al + 15 wt.% sph. Al2O3 powder mixture (Fig. 8a and c). The small delamination observed between this coating and the substrate came from the polishing. Irregular aluminum coatings (Fig. 8a) were slightly thicker in average than those made of spherical aluminum powder (Fig. 8b). Spherical alumina addition in the starting powder increased coating thickness, regardless of the aluminum powder particle shape (Fig. 8c and d). Angular alumina addition increased coating thickness for irregular aluminum but not that for spherical aluminum (Fig. 8e and f). Spherical aluminum mixed with spherical alumina resulted in the thickest and the densest coating.

Fig. 8
figure 8

Cross-sectional SEM backscattered electron micrographs Al and Al-Al2O3 coatings: (a) irr. Al, (b) sph. Al, (c) irr. Al + 15 wt.% sph. Al2O3, (d) sph. Al + 15 wt.% sph. Al2O3, (e) irr. Al + 15 wt.% ang. Al2O3, (f) sph. Al + 15 wt.% ang. Al2O3

Full size image

Angular and spherical alumina particles were homogenously distributed in the Al matrix of the Al-Al2O3 coatings. The content of spherical Al2O3 in coatings was smaller than that of angular Al2O3. In coatings with spherical Al2O3, plain and hollow spherical particles and fragments could be observed (Fig. 8c and d). The number of alumina fragments was higher in the Al-Al2O3 coatings than in the starting alumina powder or in Al-Al2O3 powder mixtures. As for starting spherical Al2O3 powder (Fig. 3d), some plain angular particles were observed in coatings with spherical alumina addition. In coatings with angular Al2O3 addition, most of the angular alumina particles were intact (Fig. 8e and f), although a few fragments of broken particles were observed. Al-Al2O3 coating micrographs showed porosities at the interface between Al and added angular Al2O3 particle or between Al and Al2O3 fragments, as already observed in Ref 14. In contrast, the contact between Al matrix and spherical Al2O3 particles seemed almost perfect without porosity. This corresponded to a better composite cohesion, i.e., to a better embedding effect.

Average coating porosity results are summarized in Fig. 9. The errors bars correspond to minimum and maximum porosity values measured for each coating. After optimization of spraying conditions, cold spray coating porosity is classically <1% (Ref 2, 11, 15). The high porosity values obtained in this work were due to unoptimized spraying parameters. Aluminum coatings without alumina addition had almost the same average porosity, i.e., 7% for irregular Al coating and 6% for spherical Al coating (Fig. 8a and b). However, porosity was more homogeneous in spherical Al coating than in irregular Al coating. The latter was actually composed of low porosity regions (4-6%) and high porosity regions (8-13%). Dense regions were more numerous in the coating. These two types of regions were observed within the coating but also at the coating–substrate interface due to substrate roughness. Spherical aluminum coating porosity was less discrepant from the average value (from 5 to 9%). Alumina addition reduced the porosity of aluminum coatings (Fig. 9) as shown in Ref 11. Irregular aluminum coating porosity decreased with alumina addition (down to around 5% for the two aluminas) whatever the alumina particle shape. The heterogeneous microstructure of irregular Al vanished with alumina addition. Spherical Al2O3 addition decreased more the porosity of spherical Al coating (from 6 to 2 %) than angular Al2O3 addition (from 6 to 4 %).

Fig. 9
figure 9

Porosity of Al and Al-Al2O3 cold-sprayed coatings

Full size image

Alumina weight contents were determined using XRD analysis (Table 2) and confirmed using image analysis. No Al oxide was found in aluminum coatings as in aluminum starting powders (from XRD measurements and SEM observation). This means that there was no oxidation of the aluminum powders during cold spraying. This confirms that cold spray process did not oxidize the powder during spraying (Ref 2, 14). Despite an initial alumina content of 15 wt.% in all Al-Al2O3 powder mixtures, the alumina content in the coatings was less than 15 wt.%. Similar trends were already reported (Ref 11, 12, 17). The spherical alumina content in the coating was smaller than the angular alumina content, regardless of the aluminum powder used in the mixture (Ref 13). Alumina contents in coatings were also smaller when using spherical aluminum rather than irregular aluminum as starting powder. The highest alumina content in coatings was achieved with a mixture of irregular aluminum with angular alumina.

Table 2 Alumina weight content in pure aluminum and aluminum–alumina coatings
Full size table

Simulation of Particle Impact

To understand interactions between particles during spraying, four numerical simulations of high-velocity Al and Al2O3 particle impacts onto Al coating were made as a function of particle shape. The plastic equivalent strain (PEEQ) contour for these four cases is shown in Fig. 10. For example, case (a) referred to the impact of spherical particles and case (b) referred to the impact of one angular Al2O3 particle and two spherical Al particles. For each case, an experimental observation of an Al2O3 particle embedded in Al coating was also added as an inset photograph in Fig. 10 to validate the simulation. The Johnson–Holmquist model was implemented in ABAQUS/Explicit® software as a user material. For this type of material properties, output variables were calculated separately from the classical output variables as PEEQ. Consequently, the PEEQ variable cannot be shown for alumina particles (in gray in Fig. 10). The deletion of damage elements was not implemented to represent the fragmentation of alumina. The plastic deformation of Al2O3 particles was not conformed to the experimental observations (in Fig. 10b and d).

Fig. 10
figure 10

PEEQ contours for the four cases: (a) impact of spherical particles, (b) impact of one angular Al2O3 particle and two spherical Al particles, (c) impact of one spherical Al2O3 particle and two irregular Al particles and (d) impact of one angular Al2O3 particle and two irregular Al particles

Full size image

For the four studied cases, the plastic strain of Al particles was located near the particle surface. This confirmed that bounding initiates at impact very locally. With high plastic strain and temperature increase, a metallurgical bonding could be probably created between Al particles and substrate and maybe between aluminum and alumina particles. However, the latter need observation at a very fine scale which justified the need of TEM observation of Al/Al and Al/Al2O3 interfaces. Mechanical interlocking was only observed in case (b). Almost no porosity was observed in case (a) in keeping with experimental observation. This corresponded to the best composite cohesion (among the four cases), i.e., the best embedding. For the three other cases, some porosity remained around the alumina particle. It was more difficult to embed an angular alumina particle than a spherical particle. Moreover, spherical shape of Al particles promoted embedding in comparison with that could be observed with irregular Al particles. All these numerical simulations were validated experimentally in Al-Al2O3 coatings. In a nutshell, embedding was better with spherical Al particles than with irregular particles and when surrounding spherical Al2O3 particles than angular particles.

Characterization of the Al-Al2O3 Interface

The embedding of the spherical alumina particles in the aluminum matrix at a microscale requires the study of the Al/Al and Al/Al2O3 interfaces at the nanoscale due to low-scale involved mechanisms. During particle impact, two mechanisms compete, i.e., adiabatic shear instabilities and dynamic recrystallization. Adiabatic shear instabilities are caused by severe plastic deformation at high strain rate (Ref 2, 29). Heterogeneous local temperature rise can therefore occur. Dynamic recrystallization conduct to nanosubgrains without melting of the particle (Ref 30) and grain refinement down to nanoscale (Ref 31, 32).

Although several interfaces between different Al particles could be observed at the nanoscale, only one is shown in Fig. 11. Aluminum particles are composed of microsized grains, which are in turn divided into subgrains as observed in Ref 29. Near the Al/Al interfaces, a grain refinement was observed (Fig. 11a) which comes from dynamic recrystallization (Ref 3, 30). The Al/Al interfaces were composed of an amorphous phase (not yet identified), aluminum nanograins and nanoporosities (in black in Fig. 11a and in white in Fig. 11b). The latter did exist along the interface and seem to form a porosity network. Aluminum, oxygen, iron and silicon concentration profiles obtained by EDX are shown in Fig. 11(c). Iron and silicon came from pollution during powder production process. Their small amounts were notified in the powder quality control certificate from the powder supplier. Although no aluminum oxide was detected in Al powders and coatings by SEM and XRD, a small layer of Al oxide was detected by line-scan EDX analysis of several Al/Al interfaces (Ref 5). This layer had a heterogeneous thickness of almost a few dozen of nanometers. However, the oxygen content in this layer was too low (<20 at.%) to be Al2O3. Even taking into account the oxygen absorption effect which affects its reliable quantification, the low oxygen content confirms that cold spray is a really protective process which only oxidized at a very fine scale.

Fig. 11
figure 11

Al/Al interface: (a) STEM-HAADF image, (b) high-magnification TEM image and (c) EDX concentration profiles through the interface

Full size image

Several Al/Al2O3 interfaces were observed by TEM. The one shown in Fig. 12 is representative of a typical observation. In Al particle near Al/Al2O3 interface, a grain refinement was observed as for Al/Al interface (Ref 18). The Al/Al2O3 interface was made of nanosized aluminum grains and an amorphous phase (not yet identified). It seems to be free of nanoporosity. The Al2O3 particle was composed of α-Al2O3 surrounded by a γ-Al2O3 layer. No deformation of the Al2O3 particle could be observed in contrast with the high deformation of Al particle. The alumina particle can be supposed to be well bounded to the Al matrix. Moreover, alumina addition improved interfaces between particles because these were free of porosities. The intimate contact between Al and angular Al2O3 particle had been observed in Ref 15. The fast Fourier transformation (FFT) of the HRTEM image equivalent to a diffraction pattern revealed an orientation relationship of (110) γ-Al2O3//(111) Al and [1-10] γ-Al2O3//[1-10] Al which is consistent with the cubic crystallographic structure of both Al and γ-Al2O3. In the EDX line-scan analysis profiles in Fig. 12, only aluminum and oxygen concentration profiles are shown. Neither silicon nor iron was detected in aluminum particle and at the interface. Only aluminum was detected in the aluminum particle. At the interface, the oxygen content increased at the expense of that of Al. Near the γ-Al2O3 layer, these two contents matched with alumina (40 at.% of Al and 60 at.% of O).

Fig. 12
figure 12

Al/Al2O3 interface: (a) STEM-HAADF image, (b) HRTEM image and (c) EDX concentration profiles through the interface

Full size image

Mechanical Properties

Hardness values (Fig. 13) had the same order of magnitude than those of Al and Al-Al2O3 coatings in Ref 2, 6, 11. Hardness differences were relatively small (only 10 HV0.1 between the softest and the hardest coating). Although the error was quite important for hardness and alumina weight content, several series of hardness indents had been done and the error was reproducible. Despite similar initial powder hardness (28 HV0.01), the irregular Al coating was harder (60 HV0.1) than the spherical Al coating (55 HV0.1). Microhardness testing values of irregular Al coating were more discrepant than that for spherical Al coating due to higher heterogeneity of the microstructure of irregular Al coating. Alumina addition hardened aluminum coatings (Ref 11, 12, 17), whatever the aluminum powder used. Microhardness of coatings with angular alumina addition was higher than that for coatings with spherical alumina. Hardness of the Al-Al2O3 coatings was independent of alumina morphology and depended on the Al matrix and on Al2O3 content (Fig. 13). Irregular Al coating was harder than spherical Al coating. Consequently, Al-Al2O3 coatings made of irregular Al were harder than those made of spherical Al. The more numerous the entrapped alumina particles were, the harder the coating was. The content of spherical Al2O3 in the coating was smaller than that of angular Al2O3 (Table 2). Consequently, coatings with spherical alumina were softer than coatings with angular alumina, whatever the aluminum powder used in the mixture.

Fig. 13
figure 13

Hardness (HV0.1) of Al and Al-Al2O3 coatings

Full size image

Bond strength tests according to ASTM C633 (Ref 28) were performed on five specimens for each coating. Specimens exhibited three types of fracture surfaces (Fig. 14). Most of the specimens broke at the coating–substrate interface (Fig. 14a). The coating remained mostly bounded to the pull-off bar on which the glue was applied. This failure type was characteristic of an adhesive rupture between the coating and the substrate. These specimens could be considered to study the bond strength of the coating. Some of the specimens broke in the glue (Fig. 14b), as observed in Ref 33. Glue failure was the sign that the bond strength of these specimens was higher but could not be determined exactly due to the glue. These specimens therefore led to a bond strength threshold only. A few specimens exhibited failures that occurred partly in the glue and partly at the coating–substrate interface (Ref 4). After SEM examination, the coating which remained on the substrate showed glue penetration (arrowed in Fig. 14c). These glue penetration spots seemed to increase the coating bond strength. However, coating thickness in this work was lower than that specified in ASTM C633. This could explain glue penetration in some specimens. These were therefore not kept for assessing bond strength.

Fig. 14
figure 14

Fracture surfaces after pull-off testing: (a) coating–substrate adhesive fracture, (b) glue fracture and (c) mixed fracture. Glue penetration spots are arrowed

Full size image

Figure 15 gives the bond strength for every coating. The failure strength of the glue was also tested separately. It was measured at 81 ± 13 MPa, which was really high for a glue (Ref 28). Bond strengths of the two aluminum coatings were also very high (more than 60 MPa), compared to that in Ref 2, 6, 11. The bond strength of the irregular Al coating was higher (77 MPa) than that of the spherical Al coating (65 MPa). Most of the specimens broke at the coating–substrate interface (Fig. 14a). Ruptures were adhesive, which was in keeping with bond strength measurements. The porous/dense regions were observed in the coating but also at the substrate–coating interface due to substrate roughness. Dense regions can be assumed to be more adhesive to the substrate due to anchoring effect, which improved the coating–substrate bond strength. These two types of regions only existed in irregular Al coating and not in spherical Al coating. Dense regions were more numerous than porous regions. Consequently, for a given steel substrate, the bond strength of pure irregular Al coating was higher than that of pure spherical Al coating.

Specimens with a coating made of mixtures of irregular Al and Al2O3 powders broke partly at the coating–substrate interface and partly in the glue. The resulting measured bond strengths (Fig. 15) have therefore to be considered as lower limits only. Consequently, results from these two powders could therefore not be compared with those from the others. However, they showed that there was still open porosity in coatings, even if alumina was added to the aluminum powder. Alumina addition increased the bond strength of spherical aluminum coating (Fig. 15). Spherical alumina addition increased more the bond strength (from 65 to 78 MPa) than angular alumina addition (from 65 to 70 MPa). Most of the specimens broke at the coating–substrate interface (Fig. 14a). However, some of them broke partly at the coating–substrate interface and partly in the glue, due to glue penetration into the substrate (Fig. 14c).

Fig. 15
figure 15

Bond strength of the Al and Al-Al2O3 coatings onto steel substrate

Full size image

Discussion

Aluminum Coating Build-Up Mechanisms

General coating build-up mechanisms were governed by the Al particle shape. Two effects of the aluminum particle shape could be shown in this study, i.e., impinging heterogeneity and in-flight velocity effect depending on particle shape.

An irregular particle shape led to heterogeneous impinging therefore to an heterogeneous microstructure. For suitable impinging, sprayed particles highly deformed at impact and adhered to the coating. Successive impacts densified the previously deposited layers, which further densified the coating. This resulted in so-called dense regions (Fig. 9). For unsatisfactory impinging conditions, sprayed particles did not deform and possibly rebounded. This left “porous regions” in the coating (Fig. 9). On the contrary, spherical aluminum particles showed a more homogeneous microstructure due to homogeneous impinging. For all spherical aluminum particles, impinging conditions were nearly the same. Porosity was almost homogenously distributed in the sph. Al coating (Fig. 9). This heterogeneous and homogeneous microstructure led to heterogeneous and homogeneous hardness (Fig. 13), respectively, for irregular Al and spherical Al. A given dense region acted as an anchoring point, which increased the adhesion of irregular Al in comparison with that of spherical Al. The bond strength of irregular Al was therefore higher than that of spherical Al (Fig. 15).

The second effect of the Al particle shape seemed to be less significant than the first effect on coating build-up. Despite similar average in-flight velocities (Fig. 6), the particle jets of the two aluminum powders were slightly different. Spherical Al particle flow was more expanded than that of irregular Al (Fig. 7a and b). Moreover, the center of the particle jet which concentrated the fastest particles was bigger for irregular Al than for spherical Al. Consequently, the center of the particle jet of irregular Al particles was more concentrated in high-speed particles, which resulted in the so-called velocity effect. This effect led to a higher strain of irregular particle and therefore to a higher hardness for irregular Al coating (Fig. 13).

Influence of Alumina Powder Addition

Al particle shape governed the general build-up mechanism for both Al and Al-Al2O3 coatings. For the latter, it was due to the predominant role of the Al matrix which embedded the Al2O3 particles. However, three effects of the alumina particle shape were also shown to contribute to the composite coating build-up mechanism.

As for aluminum powders, alumina particle shape differences led to particles in-flight velocity differences. However, the latter were more obvious between Al2O3 particle shapes than between Al particle shapes. Angular Al2O3 particles had a higher average in-flight velocity than spherical Al2O3 particles. Like Al powders, spherical Al2O3 particle flow was more expanded than that of angular Al2O3. Consequently, the angular Al2O3 particle jet was more concentrated in high-speed particles, which is called the “velocity effect.”

Angular alumina particles were plain or fragmented in the initial angular Al2O3 powder batch (Fig. 3c) and in the coatings (Fig. 8d and f). When angular alumina particles deposited on the coating, they could either rebound on already-deposited particles, fragment or be entrapped. The number of fragmented particles was higher in the coating than in the starting powder due to high particle velocities when spraying. In the starting alumina powder, spherical particles were plain, hollow or made of fragments from broken hollow particles (Fig. 3d). A few angular particles were also found. In coating with spherical Al2O3 powder, alumina particles could be plain angular or spherical particles (Fig. 8). Although spherical hollow particles were numerous in the Al-Al2O3 powder mixtures, these remained rather scarcely dispersed in the coatings. At the impact, plain spherical alumina particles showed the same behavior as plain angular alumina particles. They could either rebound on already-deposited particles, fragment or be entrapped. However at impact, due to the alumina brittleness, hollow spherical particles were prone to fragment. Fragments could be expelled in all directions or, for part of them, entrapped in the coating. Consequently, a few hollow particles only and particle fragments were found in the coating with starting spherical Al2O3 batch. This effect can be called “alumina fragmentation effect” (Ref 13, 34). Due to fragmentation of hollow particles, spherical alumina content was smaller than that of angular alumina in Al-Al2O3 coatings (Table 2). As hardness in Al-Al2O3 coatings depended mainly on alumina content, coatings with spherical alumina were softer than coatings with angular alumina addition.

From SEM observation (Fig. 8), some porosity was detected at Al-Al2O3 interfaces for angular Al2O3 but not for spherical Al2O3. As shown by numerical simulation, the embedding of spherical alumina particle was easier than that of angular particle (Ref 13, 34). This merely resulted from the particle shape of aluminum and alumina particles. Consequently, spherical Al2O3 particles showed better adhesion to the Al matrix than angular Al2O3 particles. TEM study exhibited Al/Al interfaces with nanopores but sound (i.e., free of any porosity) Al/Al2O3 interface. Moreover, alumina addition improved interfaces between particles because these were free of porosities (Fig. 11 and 12). Porosity reduction can be assumed to be the actual main mechanism for interface improvement due to alumina addition which resulted in the so-called hammering effect.

Conclusion

Pure aluminum and composite Al-Al2O3 coatings were successfully cold-sprayed onto grit-blasted steel. The very low oxygen content detected by TEM confirms that cold spray is a really protective process which only oxidized at a very fine scale. Alumina addition decreased porosity in aluminum coatings. Several Al-Al2O3 composite coating build-up mechanisms were proposed from the study of the particle shape for both Al and Al2O3 powders. Numerical simulation of multi-impact particles showed the higher embedding in Al matrix of spherical alumina particles compared to that of angular particles. From a through TEM study, interface improvement was shown to result from alumina addition. Al/Al2O3 interfaces showed no porosity actually, unlike Al/Al interfaces were somewhat nanoporous. Spherical alumina addition strengthened the coating at nanoscale. Al-Al2O3 coating hardness depended on the alumina content only and not on alumina particle shape. Spherical Al powder can therefore be claimed to be more suitable to obtain better coatings with lower porosity. Angular Al2O3 enhanced hardness of spherical aluminum coatings. Spherical alumina addition increased the coating–substrate bond strength and improved interface soundness in the coatings.