2016 Volume 63 Issue 7 Pages 504-510
In cold spray deposition, powder particles accelerated to several hundred meters a second impact the substrate and deform into splats that are stacked up to form dense coatings on the substrate. These deposited splats, however, are usually not fully bonded to each other, making deposited materials less ductile than wrought materials. An ultrasonic washing test, applied to Cu coatings sprayed on Al substrates, revealed poor splat bonding along spray boundaries and spray layer boundaries. Thickness measurements of coatings deposited with a single linear spray traverse yielded profiles that are approximately Gaussian curves but with reduced edges. We attribute the reduced edges of the thickness profiles to lower particle impact velocities and temperatures in the outer regions of the spray cone. A simulation model was developed that can predict locations of poor-bonded regions in coatings produced by multi-traverse cold-spray deposition from experimentally determined thickness profiles of single-traverse coatings.
Multi-traverse cold spray (CS) deposition provides a means for the additive manufacturing of thick coatings and bulk materials and may also be used to repair parts1–3). An important concern in the thick coating/bulk material fabrication by CS is the extent to which powder particles in the deposited material are bonded metallurgically. Although there have been several reports claiming the production of coatings with good mechanical, electrical and thermal properties comparable to those of wrought materials4–7), fabrication of truly defect-free materials by CS has not been realized to date.
The most fundamental barrier to producing metallurgically sound materials by CS arises from the difficulty in achieving full bonding at all positions of the particle-substrate and particle-particle interfaces for all particles in the spray. While multiple mechanisms may be involved in the particle bonding in CS8–11), it is generally accepted that the metallurgical bonding of impacting particles in a CS is promoted by the local shear instability that occurs at the particle-substrate and particle-particle interfaces as the particles deform into splats11–16). The shear instability, promoted by local heating at the impact interface, is manifested by the occurrence of a viscous flow of softened solid in a thin, localized channel along the interface11,12). This produces oxide-free contact and hence promotes metallurgical bonding at the interface12). While the exact nature of the shear instability still remains to be understood, it is known that shear instability does not occur at all positions of the particle surface, i.e., good bonding may not be achieved at positions in a particle that do not experience a large shear. Thus, defect-free CS is inherently difficult.
Moreover, since the powders used in CS normally consist of particles of non-uniform sizes and shapes, the particles in a CS may not all reach the critical velocity above which they have a sufficient kinetic energy to cause the shear instability needed for bonding12,15,17,18). Particles with insufficient kinetic energy may either bounce off or be trapped in the deposited material; the latter creating poor bonded splat boundaries in the deposited material. While such particles may occur at any positions in a CS, particles carried in the outer regions of the spray cone are of particular concern since both the particle velocity and temperature must necessarily decrease to low values in the outer layer of the spray cone. This would create poor-bonded regions at the boundaries between successive spray layers in the coatings. At present, however, few reports are found on this important problem in the fabrication of thick coatings and bulk materials. This paper addresses the occurrence of such poor-bonded regions and presents methods for revealing and predicting the locations of poor-bonded regions in thick copper coatings fabricated by multi-traverse, multi-spray layer CS deposition.
To investigate the locations of poor splat bonding in multi-traversed CS, copper coatings 4 mm in thickness were sprayed with N2 on an aluminum (1050 Al = JIS H4000) substrate at gas temperatures of 873 K and 1073 K under otherwise identical spray conditions shown in Table 1. The spray gun was traversed linearly back and forth with a 1 mm transverse shift after each traverse over the length of the 50 × 50 × 5 mm substrate at a constant velocity of 50 mm/s as shown in Fig. 1. A total of 30 traverses from A to B were made to deposit the first spray layer over a width of 30 mm. Then, the gun was traversed back along the exact reverse path from B to A to lay the second spray layer on top of the first layer. This operation created a 39 mm wide, 50 mm long and 4 mm thick copper coating on the aluminum substrate.
Spray path (schematic) used to fabricate thick coating specimens.
Polished cross sections of as-sprayed coatings were first examined under the optical microscope to investigate the extent of densification in them. The coating specimens were then subjected to an ultrasonic washing test19,20) in which a polished cross section of a specimen is subjected to 42 kHz ultrasonic waves in water, Fig. 2, to cause loose powder particles to come off the polished surface. The extent and locations of poorly bonded splats were determined on the ultrasonically washed cross sections of the coatings under the optical microscope.
Single-traverse coating specimens were sprayed on 100 × 100 × 5 mm Al substrates using CS parameters identical to those used for spraying the thick (two spray-layer) coatings (Table 1). The cross sectional thickness profiles of the single-traverse specimens were measured under the optical microscope to investigate the effect of the radial distribution of particle impact velocity in the spray cone on the bonding in the sprayed material.
Fig. 3 shows as-polished cross sections of the coatings sprayed at gas temperatures of 873 K and 1073 K about 1 mm above the interface with the aluminum substrate. Both coatings exhibit only a small amount of pores, indicating a high degree of densification (~100 %) achieved in them. The second (top) spray layers of the coatings (not shown in Fig. 3) were also as dense as their first (bottom) spay layers.
Cross sections of as-sprayed coatings sprayed at (a) 873 K and (b) 1073 K. Very few pores are found in the coatings.
Fig. 4 shows the same cross sections of the two coatings after 35 minutes of ultrasonic washing, at and just below the boundary between the first (bottom) and second (top) spray layers. Both normal and inverted images are shown to help visualize the surface pores left by the particles that came off during ultrasonic washing. In the cross section of the coating sprayed at 873 K, the surface pores are clearly seen to be concentrated along both the spray boundaries between successive traverses (parallel, inclined strings) and the spray layer boundary between the first and second spray layers (top horizontal band). The coating sprayed at 1073 K exhibits less surface porosity, indicating stronger particle bonding than in the coating sprayed at 873 K. However, traces of pores are still visible along both the spray and spray layer boundaries, though subtle.
Cross sections of coatings sprayed after 35 minutes of ultrasonic washing. (a) and (c) are in normal contrast. (b) and (d) are inverted images of (a) and (c), respectively. (a) and (b): Coating sprayed at 873 K. Porosity is noted in the coating sprayed at 873 K at both spray boundaries (parallel inclined strings) and the spray layer boundary (top horizontal band). (c) and (d): Coating sprayed at 1073 K. Porosity at spray boundaries is subtle.
The poor bonding that occurred at the spray and spray layer boundaries in the multi-traverse coatings implies that the powder particles carried at the outer surface of the spray cone did not stick to the substrate, or the prior deposited material, as much as the particles carried within the spray cone did. This may be verified by studying the thickness profiles of single-traverse coatings sprayed under the same conditions as those used for the thick coatings, Table 1. Fig. 5 shows the measured thickness profiles, together with the Gaussian distribution curves to which the measured profiles were fitted. Since the mass distribution in a particle spray, such as that of a thermal spray, is normally well-approximated with an axisymmetric Gaussian distribution21–23), cross sections of single-traverse coatings deposited with such a spray would ideally have a Gaussian curve as well. (See Section 4.2.) Thus, good fitting of a measured thickness profile to a Gaussian curve implies a high, uniform deposition efficiency (DE) at all positions on the substrate. This is seen for most part of the profiles in Fig. 5 except at the edges where the measured thickness clearly falls below the Gaussian curves. Thus, the particles that landed in the edge regions did not stick to the Al substrate as well as those that arrived inside the edge regions.
Measured thickness profiles of single-traverse Cu coating on Al substrate and Gaussian distribution curves to which the measurements are fitted. (a) Gas temperature = 873 K, (b)1073 K. Other parameters are found in Table 1.
The low DE at the edges of the profiles occurred because the powder particles carried in the outer regions of the spray cone were slower than those carried inside the cone. The low velocity in the spray cone surface has been shown to occur by two-dimensional axisymmetric computation of particle velocities in the nozzle and at downstream positions of the cold spray24). This was also verified by a laser imaging technique18) and dual-slit velocimetry24). Therefore, the particles carried in the spray cone periphery probably did not have a sufficient velocity required for good bonding, while the velocity of most of the particles carried in the inner part of the spray cone exceeded the required critical velocity. Moreover, the particles near the spray cone surface were likely colder than those inside the cone, which would raise the bar for good particle sticking in the edge areas. Fig. 6 illustrates the above effects of particle velocity and temperature on the bonding of particles in the edge areas of a cold spray.
Effects of particle velocity and temperature on the bonding of particles in the edge areas of a cold spray. Critical velocity νc may not be exceeded in the outer part of the spray.
When a cold spray with insufficient particle bonding at its periphery is traversed linearly on a substrate, Fig. 7 (a), the first layer of material must necessarily be laid by the leading periphery of the spray where particle bonding is poor. As the main part of the spray moves in, material with good particle bonding is deposited on top of the poor-bonded first layer. However, another poor-bonded layer is laid on top of the coating as the trailing periphery passes over, Fig. 7 (b).
Single linear traverse of a cold spray with low particle-velocity periphery on a substrate lays a coating with poor-bonded top and bottom layers.
Gaussian curve-fitted thickness profiles of single-traverse coatings may be used to simulate the coating thickening in multi-traverse, multi-spray layer cold spray deposition and predict the locations of poor-bonded regions in the resultant coating.
When an axisymmetric spray with a Gaussian mass flow rate at the substrate
is traversed linearly in the y direction at a constant velocity ν, and with a uniform deposition efficiency η at all positions, the thickness of the deposited material, τ(x, y, t), increases as25),
where μm is the mass flow rate in kg/m2·s at the center of the spray and ρ is the density of the material. Since the terms in the braces in Eq. (2) together have a constant value of 2 except in the initial and final transient regions, Fig. 8, the deposited material has a constant cross section:
Longitudinal cross section of a single-traversed coating calculated by Eq. (2). (t = 4 s, ν = 50 mm/s)
where . Thus, if the spray used to deposit a single-traverse coating has an axisymmetric Gaussian mass distribution, i.e., Eq. (1), then the transverse cross section of a single-traverse coating also is a Gaussian curve of the same width, provided that η is the same at all positions of the substrate.
The transverse cross section, τm(x, y, t), of a coating in multi-linear traverse spray deposition where the axisymmetric spray is traversed at a constant velocity with a constant successive transverse shift of Δx after each traverse, e.g., the path from A to B in Fig. 1, is given by:
where m is the number of spray traverses. If (n − 1) spray layers are added on top of the initial spray layer, the cumulative deposite thickness, Tn (x, y, t), is calculated by:
Eqs. (4) and (5) were used to simulate the thickening of the two spray-layer coatings sprayed at 873 K and 1073 K. The Gaussian curve-fitted thickness profiles of the single-traversed specimens shown in Table 2 were used in the calculation. The constant η approximation was justified as the overall deposition efficiencies of the thick (two spray-layer) coatings, determined from their deposited mass, were high; 98.0 % and 98.5 % at 873 K and 1073 K, respectively, although the local deposition efficiency, η(x, y), should be addressed in a stricter calculation.
The calculated thicknesses of the two coatings sprayed at 873 K and 1073 K are 4.039 mm and 4.141 mm, respectively, Fig. 9, which agree with their measured thicknesses of 4.104 and 4.174 mm. Fig. 9 also shows the spray boundaries between successive spray traverses and the boundary between the first and second spray layers. These are the places where poor bonding is expected since the material at the spray and spray-layer boundaries were deposited with the particles in the spray periphery as illustrated in Fig. 7. This is actually confirmed in Fig. 4, particularly for the coating sprayed at 873 K. Less surface porosity was revealed at the spray boundaries in the coating sprayed at 1073 K despite the comparably low deposition yields at the edges of the thickness profiles of the corresponding single-traverse specimens sprayed on aluminum, Fig. 5. Thus, at 1073 K, Cu powder particles in the spray periphery stuck more firmly to the prior Cu splats, while their bonding to the aluminum substrate might not have improved significantly over that of the coating sprayed at 873 K. This is because the bonding between dissimilar metals is more challenging metallurgically. Moreover, the bonding at the coating-substrate interface is intrinsically weak since the first layer of material deposited on the substrate is necessary produced by the poor-sticking particles carried in the spray periphery as illustrated in Fig. 7 (b).
Calculated cross sections of two-layer coatings deposited by multi-linear traverse cold spray deposition. (a) Gas temperature = 873 K, (b) 1073 K. Other parameters are found in Table 1.
Occurrence of poor-bonded spray and spray layer boundaries is an important concern in the commercial fabrication of thick coatings and bulk materials by cold spray deposition. Fig. 10 shows cross sections of a CS coating of a high-strength brass (Cu-33Zn-2.8Mn-2Al-1.45Fe) sprayed with N2 at 873 K, after 30 minutes of ultrasonic washing. The porosity patterns exactly coincide with those of spray boundaries and are even more sharply discernible than those in the copper coatings in Fig. 4. Possible oxidation of the aluminum in the high-strength brass may have further decreased the bonding along the spray boundaries in the brass coating. Thick coatings and bulk materials of commercial alloys with improved integrity may be fabricated by optimizing process parameters for low powder loss at the spray periphery through single-traverse spraying experiments.
Cross sections of a Cu-33Zn-2.8Mn-2Al-1.45Fe coating sprayed with N2 at 873 K, after ultrasonic washing for 30 minutes. Porosity patterns match those of spray boundaries in (a) mid-portion and (b) corner of the cross section of the coating.
Multi-traverse CS provides a means for the fabrication of thick coatings and bulk materials from powders. However, fabrication of truly defect-free materials has not been realized. This is because achieving metallurgical bonding at all positions of particle surface is inherently difficult. Moreover the low particle velocity and temperature in the outer surface of the spray cone promote poor bonding in the deposited material. This effect, confirmed by thickness profile measurements of single-traversed copper coatings, creates poor-bonded material along the spray and spray-layer boundaries in thick coatings. Poor-bonded spray and spray-layer boundaries were observed in thick (4 mm) copper coatings sprayed with nitrogen gas heated at 873 K and 1073 K, with the aid of an ultrasonic washing test. A model was developed to simulate the coating thickening and predict the locations of poor-bonded regions in multi-traverse, multi spray-layer CS deposition from Gaussian curve-fitted thickness profiles of single-traverse sprayed coatings.