This article is one of eight papers to be presented exclusively on the web as part of the January 2000 JOM-e—the electronic supplement to JOM.
JOM-e Logo
The following article appears as part of JOM-e, 52 (1) (2000),
http://www.tms.org/pubs/journals/JOM/0001/Pei/Pei-0001.html

JOM is a publication of The Minerals, Metals & Materials Society

Functional Coatings: Research Summary

Producing Functionally Graded Coatings by Laser-Powder Cladding

Y.T. Pei and J.Th.M. De Hosson


JOM-e Logo

TABLE OF CONTENTS

Al-40Si functionally graded coatings produced by a one-step laser powder cladding process on cast aluminum-alloy substrate is a possible solution for avoiding the interfacial problems often present in laser coatings. The microstructure of the coatings consists of a large amount of silicon-primary particles surrounded by a-aluminum dendritic halos and Al/Si eutectic. The silicon particles exhibit a continuous increase in both size and volume fraction, from 8.5 mm and 22.7% at the bottom to 52 mm and 31.4% at the top of the coating layer, respectively. The morphology of the silicon particles changes accordingly from a small polygon shape to a coarse, branched, equiaxial shape. The a-aluminum halos and eutectic areas show less change over the same distance. From a qualitative analysis of the temperature field of the laser pool, silicon particles heterogeneously nucleate on incompletely melted silicon particles. The number density of silicon particles is most likely controlled by the nonhomogeneous temperature field of the pool that determines the decomposition of original phase in the Al-40Si powder. The final size of the silicon particles is mainly affected by the growth rate and time available at different depths of the laser pool.

INTRODUCTION

The laser-surface processing of aluminum alloys has drawn a lot of interest for enhancing the mechanical and chemical resistance of aluminum-alloy components.1 Examples of laser-surface processing techniques include laser-surface melting (LSM), laser-surface alloying (LSA), and laser-surface cladding (LSC).2 In LSM, the surface performance of the aluminum-alloy substrate is modified mainly by the homogenization and refinement of the microstructure. In addition, different precipitates may be formed and the supersaturation of the a-aluminum phase increased due to nonequilibrium solidification.3-5 However, the modification is limited, because the composition of the melted layer is the same with respect to the substrate.

The employment of appropriate alloying additions with LSA can lead to the formation of hard phases that reinforce and enhance the properties of the alloyed layer.6-8 Elements used to date have been restricted to the transition metals, such as Ni, Cr, Mo, W, Ti, and Zr, that react with aluminum to form intermetallic aluminide.1 The LSA technique works on the principle of mixing an alloying element in the melt pool of the substrate created by the laser beam. The amount of alloying addition is usually minor compared with the total amount of material that has melted. Therefore, local thermal distortions may be introduced to the substrate, resulting in severe residual stresses.

Previous studies9,10 have shown that LSC is the most flexible technique. In this process, only a thin surface layer of aluminum-alloy substrate melts together with the additive material to form a coating. By minimizing the thickness of the melted substrate layer, the desired properties of the coating are realized while minimizing the changes to the substrate. Consequently, advanced coatings produced by LSC can be designed independent of the composition of the substrate material. Therefore, LSC has the potential to be an effective and economic technique to improve the surface properties with the view of significantly extending the performance of the material as a whole.

A major drawback in laser cladding is that a sharp interface usually forms between the coating and the substrate that is often a potential source of weakness. For example, if the coating material has a very different thermal-expansion coefficient than the substrate, there is the possibility of severe stresses building at the interface and resulting in a crack. A common way to circumvent this problem is to optimize the coating thickness or to introduce a compliant interlayer for the reduction of the thermal stress. Unfortunately, most compliant films also melt at lower temperature. A recent development by Jasim et al.11 is a functionally graded coating (FGC) built up by three overlaid laser tracks in which the proportion of SiC reinforcement increased in steps from 10 vol.% to 50 vol.%. Their work showed the possibility of laser processing to deposit a thick multilayer of essentially discrete composition rather than a gradual composition change.

The term "functionally graded materials" (FGMs) is now widely used by the materials community for a class of materials exhibiting spatially inhomogeneous microstructures and properties. Graded materials in themselves are not something new, but what is exciting about them is the realization that gradients can be designed at microstructural level to tailor specific materials for the functional and performance requirements of an intended application. Therefore, a possible approach for eliminating the sharp interface present in LSC is to introduce the concept of FGMs into the design of coating structure.


EXPERIMENTAL PROCEDURES

Substrates were cut from cast rods of commercial aluminum alloy with a nominal composition of Al-6.3Si-4.0Cu-1.0Fe-0.3Mg (in wt.%). Standard face milling finished the surfaces of the flat substrate specimens in dimensions of 100 mm3 × 50 mm3 × 10 mm3. The specimen was mounted on a cooling block with the underside in direct contact with flowing water or oil maintained at a constant temperature. In this way, the substrate was kept at a constant temperature as much as possible with a change of less than 5C during laser cladding. This ensured that the geometry and dilution degree of laser-clad tracks depended only on the processing parameters. In addition, setting different substrate temperatures from 15C to 200C made it possible to adjust the thermal gradient of the specimens that affected the solidification of the laser pool.


Figure A

Figure A. An optical micrograph of Al-40Si powder showing the original silicon primary particles inside the powder.

Al-40Si alloy powder was used as the coating material for three reasons. First, using the same alloy system results in similar thermal properties. Second, even a high degree of local dilution can create a composition gradient for the desired FGCs. Third, the primary silicon particles may serve as hard reinforcement to FGCs, and the solidification process can control their size. This is very important for the in-situ formation of FGCs during laser cladding. The powder produced with spray-atomizing technology exhibited a globular geometry with a particle size of 50 mm to 125 mm (Figure A).

The powder feeder used a Perkin-Elmer-Metco Co (MFP-I type) commercial instrument. The cladding set-up was equipped with a specially developed powder-addition module.12 The powder nozzle had an integrated coaxial shielding gas flow to buffer the laser-molten pool from the atmosphere. The shielding gas was also used to converge the powder stream into the laser-molten pool, leading to a powder efficiency higher than 90% as well as a more homogeneous cladding track with a smooth surface.

A HAAS HL3006D-type 3kW Nd:YAG laser was used. The laser beam was transported by means of a 0.6 mm fiber, resulting in a homogeneous intensity distribution. The focal length of the focusing lens was 140 mm operated at a 25 mm defocusing distance with a 3.34 mm spot on the surface of the substrate for cladding. A numerically controlled four-axial machine executed the specimen movement. The processing parameters varied in the region of 2,000 ~ 3,000 W laser power, 8.3 ~ 26.7 mm/s beam speed, and 10 ~ 30 g/m powder-feeding rate. The shielding gas was helium with a flow rate of 0.167 l/s.

The transverse sections of the clad tracks were cut for microstructural studies. A Philips XL-30 FEG scanning electron microscope (SEM) equipped with energy-disperse x-ray analysis (EDAX) and a standard optical microscope were employed for the microstructure study. The samples were etched with 2% NaOH solution for 5 ~ 10 s at 40C. A Shimadzu HMV-2000 type micro-Vickers was used for hardness measurements. The used load was 200 grams, and loading time was set at 15 seconds.


EXPERIMENTAL RESULTS

Figures 1a, 1b, and 1c present the cross sections of laser-clad single Al-40Si FGC tracks produced by different beam speeds. All of the tracks have a good fusion bond with the substrate. Within the substrate there are casting defects in the form of large holes. If the melt boundary of the laser pool reached the holes, then the trapped gases of the hole were brought into the melt. Consequently, some pores appear in the clad layer produced with faster beam speeds, since the gas bubbles did not have enough time to escape from the melt. However, the FGC tracks were always porosity-free if there was not a hole inside the melted layer of the substrate.


Figure 1a
 
Figure 1b
 
Figure 1c

Figure 1. Optical micrographs of Al-40Si FGCs clad at 3,000 W laser power and different beam speeds of (a-left) 10 mm/s, (b-center) 20 mm/s, and (c-right) 26.7 mm/s.

A large amount of primary silicon particles (Sip) is observed over the cross sections of FGC tracks. These fine silicon particles act as reinforcement and are expected to improve the tribological properties of the FGCs significantly. Note that the primary silicon particles gradually increase in size with distance from the bottom of FGC tracks. Moreover, the scanning speed of the laser beam exerts an obvious influence on the distribution of silicon particles. The difference in Sip size between the top and the bottom of the tracks decreases with decreasing beam speed.


Figure 2a
 
Figure 2b
 
Figure 2c

Figure 2. The (a-left) top (b-center) intermediate, and (c-right) bottom of a graded microstructure of Al-40Si FGC produced at 3,000 W laser power and 26.7 mm/s beam speed. The arrows indicate five-branch silicon particles.

The graded microstructure of Al-40Si FGC is more clearly shown in Figures 2a, 2b, and 2c. The FGC layer consists of primary silicon crystals surrounded by a-aluminum dendritic halos and branched Al/Si eutectic adjacent to the a-aluminum halos. From the bottom to the top surface of the clad layer, the discrete silicon particles increase in apparent size by a factor of five. In contrast, the a-aluminum halos and the adjacent eutectic cells exhibit less change in size over the same distance. These observations suggest that the primary silicon particles were nucleated from the liquid state rather than from a solid-state precipitate. In addition, a thin Sip-free zone composed of a-aluminum dendrites and Al/Si eutectic can be seen at the melted boundary. The thickness of this Sip-free zone reduces from 30 mm to 10 mm with increasing beam speed. It is predicted that this zone will be a beneficial feature by being a region where possible thermal stresses may be relieved.


Figure 3a
 
Figure 3b
 
Figure 3c

Figure 3. Quantitative metallography results of the FGC track revealed in Figure 2 on the graded change of silicon particles as a function of depth about (a-left) apparent size and mean intercept length, (b-center) volume fraction and interparticle spacing, and (c-right) areal density (NA) and volumetric density (NV).

The changes in apparent size, volume fraction, and number density of silicon particles as a function of the distance (z) from the molten boundary are given in Figures 3a, 3b, and 3c, using quantitative metallographic analysis.13 The size of the silicon particles is evaluated in Figure 3a in two terms (i.e., apparent size [d] and mean intercept length [L3]). The apparent size of the Sip is taken as the tip-to-tip distance of the star-shaped particles and represents the maximum distance of growth for a particle along certain preferred directions. The mean intercept length denotes the average size of particles and is a unique assumption-free value that is valid for particles of any size and configuration. Both parameters exhibit an obvious increase over the thickness of the FGC (i.e., six times in d and 2.7 times in L3), and their difference is due to the development in shape of Sip. Accordingly, the volume fraction of Sip varies continuously from 22.7% at the bottom to 31.4% at the top of the track (Figure 3b). Note that the numbers of Sip observed per unit area and subsequently calculated per unit of volume decreases obviously with z (Figure 3c). In other words, the interparticle spacing (s) increases significantly with z according to

s = 12.97 + 8.83 × 10-2z - 1.63 × 10-6z2
(1)


which is obtained by polynomial regression of the profile present in Figure 3b.

Figure 4

Figure 4. An SEM micrograph showing the growth feature of the five-branch silicon particle and the surrounding a-aluminum dendritic halos as well as the eutectic adjacent to a-aluminum.

Figure 5

Figure 5. The hardness distribution of laser-clad Al-40Si FGCs produced with different powder feeding rates.

In addition to the change in size with depth, silicon particles also display a morphological transition from polygon at the bottom to an equiaxially branched shape at the top region of the FGC. An example of the equiaxially branched shape can be seen as the five-fold branched grains indicated by arrows in Figures 2a and 2b. Such five-fold geometry of silicon crystals has been previously reported as a rare morphology that forms at a slow cooling rate,14 but in our case it seems to be very often observed in the upper region of the FGC. It has been known that they grow from a twinned decahedral nucleus, which is an assembly of five silicon tetrahedrons in twinned orientation.

The typical growing feature of the five-fold silicon particles is shown in Figure 4. The obtuse facet angles at the ends of each branch, as well as the twin planes along the branches and through the intersection points of the facets, suggests a twin-plane re-entrant edge growth mechanism.15

The typical hardness profile of the FGC tracks is presented in Figure 5. The graded microstructure leads to a gradual hardness distribution of the FGC from HV0.2 180 down to HV0.2 80. Sometimes it was impossible to avoid making indentations close to the primary silicon particles, the result of which is a much higher hardness. This resulted in some fluctuations on the hardness curves, despite that an average of five measurements was taken. It is interesting to note that the transition from the coatings to the substrate exhibits a gradual change in the hardness, which indicates the absence of a sharp demarcation in materials properties across the interface.

DISCUSSION

The graded microstructure of Al-40Si FGC is formed during the solidification of the laser pool. Therefore, factors that affect the nucleation and growth of silicon particles and a-aluminum halos will play a role in controlling the formation of the graded structure. The observed microstructures suggest the following solidification process of Al-40Si FGCs. The silicon particles nucleate from the liquid by heterogeneous mechanisms, grow into the surrounding undercooled melt, and reject the aluminum solvent until the local concentration is sufficient to nucleate a-aluminum phase. This a-aluminum phase appears as halos surrounding the silicon particles, which arrest the growth of silicon particles. The growth of the a-aluminum halos results in an increasing silicon content of the remaining liquid phase to the extent that eventually the composition of the liquid phase lies in the coupled zone. This results in the cooperative-growth mechanism between silicon and a-aluminum that yields the eutectic phase in the latter stages of solidification. The gradual change in size and morphology of Sip as well as a-aluminum halos is a consequence of their local growth conditions and nucleation environments.

During laser cladding at constant power and beam speed, a steady-state melt pool is created after the first few millimeters of the track. Taking a longitudinal section through the centerline of the track, the speed of the solidification front, Vs, is correlated to the beam speed, Vb, via the expression3

Vs = Vb cosq
(2)



Figure 6

Figure 6. The temperature field of the laser pool in the case of cladding Al-40Si FGCs. Isotherms I1 and I2 represent the liquidus of Al-40Si alloy and the Al/Si eutectic temperature, respectively; Vb is the beam scanning speed; and Vs is the speed of eutectic growth front.

where q is the angle between Vs and Vb (Figure 6). This equation is based on geometrical considerations of the melt pool-the motion rate of its isotherms follows exactly the beam speed. Vs varies from zero at the bottom of the melt pool to a maximum approaching the value of Vb at the top of the melt pool. In the case of laser cladding Al-40Si FGCs that involve the formation of primary particles ahead of the solidification front, Vs represents the speed of eutectic growth front, rather than the growth rate of the silicon particles and a-aluminum halos. This is different compared to the case of epitaxial regrowth from a substrate during laser remelting.

Although it is difficult to know the actual value of the Sip growth rate (VsSi), it can be expected to take a similar trend with depth comparable to the Vs of the eutectic, as required by the steady-state movement of the laser pool. For a given track speed, relating the number of Sip observed at different depths to the estimated VsSi suggests that the Sip number density decreases with increasing growth-front speed. This differs from the results reported by Gremaud et al.16 and contradicts classical solidification theory. Generally, for faster solidification rates more particles nucleate. This is related to the undercooling, DT, via the two equations17

(3)


and

DT = Kva
(4)


where NV is volumetric density of particles; v is growth rate; and , b, K, and a are all positive constants, with a = 0.25 ~ 0.5 depending on the actual growth mechanism. Therefore, there appear to be additional factors that determine the solidification microstructure.

To understand the formation of the graded microstructure, it is necessary to know the temperature field of the melt pool in which the Al-40Si powder is melted together with the surface layer of the substrate.

Following the numerical model for laser cladding created by Hoadley and Rappaz,18 the temperature field of the laser pool in the case of a clad Al-40Si FGC may be described as in Figure 6. It is composed of two isotherms, I1 and I2, that represent the liquidus of the Al-40Si alloy (1,220 K) and the eutectic temperature of Al-Si alloy (850 K), respectively. The molten region of the boundary of the pool (i.e., denoted as line OA) corresponds to the liquidus of the aluminum-alloy substrate (about 900 K). The pool is divided into three parts by drawing two horizontal lines, a, as a tangent to the lowest point of the isotherm I1 and through the surface of the substrate (b). The part of the pool above line b is in a superheated condition--the temperature of this liquid exceeds the liquidus temperature (1,220 K). Evidence in support of this was given by optical-pyrometer measurements of the mean surface temperature of the melt, and it was found that the temperature of a 2 mm diameter area of the melt-pool surface was 1,650 ~ 1,800 K.

The temperature of the melt below line a never exceeds the liquidus of Al-40Si alloy and cools down quickly to the liquidus temperature of the substrate alloy due to mixing of the melt. In this region of the melt, it is expected that the decomposition of the original silicon primary particles from the Al-40Si powder is incomplete due to premature mixing with the relatively cool melt from the substrate. Consequently, there would be many silicon particles available that are above the critical size for heterogeneous nucleation and which become active sites for growth.

By decreasing the beam speed, the dissolution of the original silicon primary particles from the Al-40Si powder is more complete, and, therefore, the number density of these incompletely melted silicon particles decreases, as shown in Figures 1a, 1b, and 1c. However, there will be fewer silicon particles inside the superheated melt behind isotherm I1 in the upper region of the melt pool, such that fewer particles can obtain the critical size for continued growth. The portion between the lines a and b can be considered a transient region and presents a gradual increase both in temperature and in number of silicon particles. It can be assumed that the final number of silicon particles observed in a particular part of the track is similar to the original number of incompletely melted silicon particles in the melt. The observed variation of Sip number density (0.63 ~ 4.95 × 1014 Sip/m3 shown in Figure 3c) over the entire cross section of the FGC supports the above hypothesis of heterogeneous nucleation on the incompletely melted particles as it is too low for a homogeneous-nucleation mechanism.

The primary silicon particles nucleate and grow within the undercooled melt ahead of the main solidification interface and continue to grow as long as the imposed local undercooling permits. The latent heat that is released makes the solidified silicon particles somewhat hotter than the surrounding liquid and establishes a negative thermal gradient around the silicon particles. This leads to difficulties in determining the amount of local undercooling at different depths in the laser pool and, hence, in relating the growing rate of silicon crystals to the degree of undercooling. Nevertheless, it is possible in this case to estimate the amount of time available for Sip to grow before the a-aluminum halos nucleate. From the above discussion, Sip must nucleate ahead of the eutectic interface at a distance about the same as the interval between two isotherms, which is proportional to the mean particle interspacing s. Therefore, the growing time (t) for Sip can be estimated by the expression

t = js/Vb
(5)


where j is a proportionality factor of the interval between the isotherms to the Sip interspacing. This factor is larger than unity and can be affected by the geometry of the clad tracks or by the heat-conducting condition.

The studies of Gremaud et al.16 assumed that during the laser remelting of Al-26Si (wt.%) the primary silicon particles nucleated ahead of the eutectic interface by a distance approximately equal to the average particle spacing, which means j is unity. Their assumption is not appropriate to explain our results.

Because silicon particles nucleate at a temperature below the liquidus (1,220 K) of Al-40Si, for the observed maximum average particle spacing of 40 mm in Al-40Si FGCs, the thermal gradient between two isotherms will reach 9,250 K/mm. Such a large thermal gradient seems to be unlikely for a thick laser-clad AlSi FGC on an aluminum-alloy substrate. Taking a 2 mm thick Al-40Si FGC, the maximum surface temperature measured is never greater than 2,000 K. Assuming that the temperature at the molten boundary of the laser pool takes the liquidus temperature (900 K) of the substrate, the average thermal gradient over the whole pool is less than 1,000 K/mm.

It is reasonable to expect that the interval between the isotherms is proportional to, but larger than the particle spacing. This proportional factor is represented by j in Equation 5 and is assumed to be constant with depth. The reason for using Vb rather than Vs is to consider a steady movement of laser pool. By substituting Equation 1 into Equation 5, the growth time is found to be t = j(12.97 + 8.83 × 10-2z - 1.63 × 10-6 z2)/Vb. It can be seen that the silicon particles in the upper region of the melt have a significantly longer time for growth than those particles at the bottom of the track; hence, the silicon particles in the upper region of the melt are larger in size.

The liquid phase surrounding the silicon particles becomes enriched in aluminum by the growth of the silicon particles, which involves the rejection of aluminum solute. As described by Barclay et al.,19 the liquid composition moves down along the depressed liquidus and passes by the coupled zone and into the hypoeutectic side. Once the undercooling at the solid/liquid interface is large enough, a-aluminum will nucleate on the facets of the silicon crystals; consequently, the growth of the latter is halted. As seen in Figures 2a, 2b, and 2c, the a-aluminum halos surrounding the Sip in the lower part of the track are thinner than those in the upper part. Another notable feature is that the dendritic nature of the a-aluminum halos is more fully developed further from the bottom of the track. This is considered to be additional evidence that their growth rate increases with decreasing depth in the melt pool. These observed characteristics suggest that the size of the halos is limited by both their growth rate and by the time necessary to reach the temperature at which the eutectic finally forms. Obviously, because of the smaller thermal gradient, the halos in the upper part have a greater amount of time to grow. In addition, the impingement between the silicon particles relatively close to each other in the lower part restricts the halos from fully developing.

ACKNOWLEDGEMENT

The Netherlands Institute for Metals Research and the Foundation for Fundamental Research on Matter (FOM-Utrecht) are acknowledged for their support.

References
1. K.G. Watkins, M.A. McMahon, and W.M. Steen, Mater. Sci. Eng., A231 (1997), p. 55.
2. J.Th.M. De Hosson, Intermetallic and Ceramic Coatings, ed. N.B. Dahotre and T.S. Sudarshan (New York: Marcel Dekker, 1999).
3. M. Gremaud, M. Carrard, and W. Kurz, Acta Metall. Mater., 38 (1990), p. 2587.
4. J.L. De Mol van Otterloo, D. Bagnoli, and J.Th.M. De Hosson, Acta Metall. Mater., 43 (1995), p. 2649.
5. J. Noordhuis and J.Th.M. De Hosson, Acta Metall. Mater., 41 (1993), p. 1989; H.J. Hegge and J.Th.M. De Hosson, Acta Metall., 38 (1990), p. 2471.
6. A. Almeida et al., Surf. Coat. Technol., 70 (1995), p. 221.
7. H.J. Hegge and J.Th.M. De Hosson, J. Mater. Sci., 26 (1991), p. 711.
8. U. Luft, H.W. Bergmann, and B.L. Mordike, Laser Treatment of Materials, ed. B.L. Mordike (Oberursel, Germany: DGM, 1987), p. 147.
9. P. Sallamand and J.M. Pelletier, Mater. Sci. Eng., A171 (1993), p. 263.
10. Y. Liu, J. Mazumder, and K. Shibata, Metall. Mater. Trans., 25B (1994), p. 749.
11. K.M. Jasim, R.D. Rawlings, and D.R.F. West, J. Mater. Sci., 28 (1993), p. 2820.
12. R. Volz et al., Proc. 30th ISATA: Rapid Prototyping/Laser Applications in the Automotive Industries, ed. D. Roller (London: Automotive Automation Ltd., 1997), p. 393.
13. E.E. Underwood, Metals Handbook, 9th ed., vol. 9, ed. K. Mills et al. (Materials Park, OH: ASM, year), p. 123.
14. K. Kobayashi and L.M. Hogan, J. Cryst. Growth, 40 (1979), p. 399.
15. M.C. Flemings, Solidification Processing (New York: McGraw-Hill, 1974).
16. M. Gremaud et al., Acta Metall. Mater., 44 (1996), p. 2669.
17. D. Turnbull, Acta Metall., 1 (1953), p. 8.
18. A.F.A. Hoadley and M. Rappaz, Metall. Trans., 23B (1992), p. 631.
19. R.S. Barclay, P. Niessen, and H.W. Kerr, J. Cryst. Growth, 20 (1973), p. 175.

Y.T. Pei and J.Th.M. De Hosson are with the Department of Applied Physics, Materials Science Center and Netherlands Institute for Metals Research, University of Groningen.

For more information, contact J.Th.M. De Hosson, Department of Applied Physics, Materials Science Center and Netherlands Institute for Metals Research, University of Groningen, Nijenborgh 4, 9747 AG Groningen, Netherlands; e-mail hossonj@phys.rug.nl.


Copyright held by The Minerals, Metals & Materials Society, 2000
Direct questions about this or any other JOM page to jom@tms.org.

SearchTMS Document CenterSubscriptionsOther Hypertext ArticlesJOMTMS OnLine