Functionally graded ceramics by a new in situ processing route: Residual stress and wear resistance

Available online at www.sciencedirect.com

ScienceDirect Journal of the European Ceramic Society 35 (2015) 2693–2698

Short Communication

Functionally graded ceramics by a new in situ processing route: Residual stress and wear resistance Chen Xu, Richard I. Todd ∗ University of Oxford, Department of Materials, Parks Road, Oxford OX1 3PH, UK Received 26 January 2015; received in revised form 24 February 2015; accepted 25 February 2015 Available online 18 March 2015

Abstract A new in situ method of producing ceramic functionally graded materials (FGMs) with continuously changing microstructure by pressureless sintering in combination with chemical reactions and evaporation is reported. Solid solutions of Fe2 O3 in Al2 O3 were first sintered in air and then reduction aged in 4 vol% H2 /N2 . Metallic Fe formed near the surface and was able to evaporate, producing a surface layer with graded composition. The thermal expansion mismatch led to compressive residual stresses in the surface region of ∼500 MPa and the graded structure avoided delamination. The abrasive wear resistance of the FGM ceramics was improved by a factor of 5 times compared with monolithic alumina. The improvement was attributed to the suppression of crack propagation normal to the surface by the compressive surface stresses. © 2015 Elsevier Ltd. All rights reserved. Keywords: Abrasive wear; FGM; Alumina; Pressureless sintering; Residual stress

1. Introduction Laminated ceramics have been widely used as a method to improve properties of monolithic ceramics such as flaw tolerance and wear resistance.1–3 However, a major problem in ceramic laminates is that the high thermal residual stresses between the layers which are often responsible for the property improvements can also lead to failure by delamination and edge cracking, especially when the interlaminar bonding is weak.4 Functionally graded materials (FGMs) have been developed to reduce the residual stress concentrations and enhance the bonding strength of laminates, by using a systematically varied microstructure and/or composition to produce gradual changes in mechanical and thermal properties across the geometry.5,6 Several studies have been reported about ceramic FGMs, among which hot pressing has been widely used6,7 because of the difficulty of co-sintering mixtures of powders with different composition owing to the mismatch in shrinkage rates.4 Hot pressing is

∗

Corresponding author. Tel.: +44 1865 273718; fax: +44 1865 273783. E-mail address: [email protected] (R.I. Todd).

http://dx.doi.org/10.1016/j.jeurceramsoc.2015.02.032 0955-2219/© 2015 Elsevier Ltd. All rights reserved.

expensive, however, and also limits the layered structure to planar geometries. In this work, we investigated the residual stresses and wear resistance of new in situ ceramic FGMs with continuously changing microstructure developed by pressureless sintering in combination with chemical reactions and evaporation. This offers a more economical and versatile way of producing ceramic FGMs. This work is based on the precipitation in Al2 O3 –10 wt% Fe2 O3 solid solutions under reducing atmospheres when sintered pressurelessly at high temperatures which has been used previously by Mukhopadhyay and Todd8 to produce ceramic composites. Although Mukhopadhyay and Todd8 aimed to produce homogeneous bulk composites, the presence of a graded structure due to the different degrees of chemical reaction near the surface suggests the possibility of making FGMs through pressureless sintering using a related route. 2. Experimental methods During the fabrication, high purity (99.99%), fine (particle size 200 nm) ␣-Al2 O3 powders (AKP-50, Sumitomo, Japan) were mixed with Fe(NO3 )3 ·9H2 O (purity >98%, Sigma Aldrich, UK) dissolved in ethanol to form slurries that would achieve

2694

C. Xu, R.I. Todd / Journal of the European Ceramic Society 35 (2015) 2693–2698

Al2 O3 –10 wt% Fe2 O3 after calcination. 250 ppm (by weight) MgO powder was added to the slurries to prevent abnormal grain growth during sintering8 and 850 ppm Y2 O3 in the form of Y(NO3 )3 ·9H2 O was added to limit the growth of large FeAl2 O4 particles at a later stage of the process.9 The slurry was dispersed using an ultrasonic probe, and then ball milled for 24 h in polyethylene bottles using high purity (99.99%) alumina balls. After ball milling, the slurries were dried in a rotatory evaporator at a temperature of 60 ◦ C in vacuum. The dried powders were heated at 10 ◦ C/min up to 600 ◦ C and held there for 1.5 h to decompose the Fe(NO3 )3 to Fe2 O3 completely. The calcined powder was then mixed with ethanol and ball milled for another 24 h. The final slurry was dried completely by rotatory evaporator under the same conditions as those used for the first drying step and then gently ground using an alumina mortar and pestle and passed through a 150 ␮m sieve. Green compacts were fabricated by cold isostatic pressing at a pressure of 150 MPa and sintered pressurelessly in air with an alumina tube furnace at 1450 ◦ C for 5 h. The heat treatment was designed to achieve full dissolution and homogeneous distribution of Fe3+ in the cation sub-lattice of Al2 O3 according to the Fe2 O3 –Al2 O3 phase diagram.10 The second step in the heat treatment scheme (aging treatment) was carried out by annealing Al2 O3 –10 wt% Fe2 O3 solid solutions at 1350 ◦ C for 20 h under 4 vol% H2 /N2 . The resultant specimen was 2.5 mm thick and is denoted the FGM sample in what follows. Discs of both pure alumina and solid solution were made for comparison with similar processing except for the omission of the aging step. The sintered densities of all samples were measured in distilled water according to Archimedes’ principle, and were all above 98% of the theoretical density. Back Scattered Electron (BSE, JEOL JSM 840A, Japan) images of a polished cross section were used to examine the microstructure of the FGM sample. Energy Dispersive X-Ray

Spectroscopy (EDX) attached to the SEM was used for elemental analysis. The residual stresses were measured by piezospectroscopic microscopy using a Renishaw 1000 Series Raman microscope, which relates the R1 Cr3+ fluorescence peak position to the sample stress.3,6 Given the uncertainties involved, two distinct methods were used. In Method 1 a line scan (step size 50 ␮m) was made through a polished cross section of the specimen (scan 1 in Fig. 1). To provide a stress free reference, a second line scan was made on the bulk part of the same polished cross section but with the surface regions on both sides removed by grinding (scan 2 in Fig. 1). The two line scans were made on the same position in the bulk part. The shifts of R1 Cr3+ fluorescence peak position between the two line scans were measured. The stresses along the lines scanned before removal of the surface regions should be approximately uniaxial with the stress direction lying in the polished cross-section and parallel to the surface regions. However, the surface regions must have approximately the same in-plane strain across the cut surface as in the bulk, where the stress should be biaxial. Thus the biaxial residual stress in the bulk part can be calculated using Eq. (1) σ=

(1)

where σ is the biaxial stress in the centre of the bulk, ␯ is the frequency shift of the R1 peak compared with the stress-free sample, γ is Poisson’s ratio and the hydrostatic piezospectroscopic coefficient Π H is 7.59 cm−1 GPa−1 for texture-free polycrystalline alumina.11 The main problem with Method 1 is the uncertainty due to the stress relaxation at the cut surface. To give confidence in the stress levels, an alternative Method 2 was therefore used to measure the residual stress avoiding this stress relaxation issue. This method involved two steps (Fig. 1). In step 1, an FGM sample was polished on one large surface until the midpoint of the cross-section was reached. Cr3+ fluorescence spectra were recorded from at least 100 points on the polished bottom surface of each sample. Spectra were taken from points at least 2 mm from the sample edges to avoid edge related effects. In step 2, the top surface layer was removed and the measurements on the polished bottom surface were repeated to provide a stress free reference. The shift of the average R1 Cr3+ fluorescence peak position between the two steps was measured. The polished bottom surface should have a biaxial stress given by:11 σr =

Fig. 1. The geometry of the cross sections and the position of fluorescence measurements for (top) Method 1; (bottom) Method 2.

3 ν ΠH (1 − γ)

3 v 2ΠH

(2)

The hardnesses and wear resistances and of pure alumina, the solid solution and the FGM sample were compared. Vickers hardness indentations (AVK-C1/C2) were made on the lightly polished surfaces (the thickness removed was around 5 ␮m) with indentation loads (P) 5 kg applied for 15 s to evaluate the hardness (Hv) of all three samples. A micro-scale abrasive wear tester (TE 66, Phoenix Tribology, UK) was used to measure the wear resistance of all samples. Samples were loaded by a preconditioned chrome steel ball (25 mm in diameter) with a dead weight of 5 N. The abrasive slurry consisted of silicon carbide

C. Xu, R.I. Todd / Journal of the European Ceramic Society 35 (2015) 2693–2698

2695

Fig. 2. BSE images of the FGM sample (a) iron depleted zone, 50 ␮m away from the gas/solid interface; (b) 200 ␮m away from the gas/solid interface; (c) bulk, 1000 ␮m beneath the gas/solid interface and (d) the volume fraction of second phase particles and chemical level of Fe from the gas/solid interface to the middle of the sample thickness.

particles (4.3 ␮m, grade F1200, Washington Mills, UK) suspended in distilled water (concentration 0.35 g/cm3 ) and doped with Dispex A40 dispersant. A magnetic stirrer was used in the slurry supply to prevent particle sedimentation, and the slurry was dispersed drop-wise at a constant rate onto the surface of the rotating ball. The total sliding distance used in each test was 58.9 m (750 ball rotations) and the relative sliding speed was 0.2 ms−1 . The tests were interrupted every 125 ball rotations to record the wear scar diameters using an optical microscope attached to the wear tester. The final wear craters were also examined after the test using scanning electron microscopy (JEOL JSM 840A). The calculated wear crater depths were all in the range of 10–40 ␮m. Therefore the wear tests measured the wear properties of the surface region (250 ␮m) of the FGM sample. 3. Results and discussion Starting with the microstructural development, the microstructure of the FGM sample had an approximately ‘sandwich structure’. According to Mukhopadhyay and Todd,8 two reactions occur during the aging treatment: Al1.868 Fe0.132 O3 + 0.066H2 → 0.132FeAl2 O4 + 0.802Al2 O3 + 0.066H2 O(g)

FeAl2 O4 + H2 → Fe + Al2 O3 + H2 O(g)

(3)

(4)

Both FeAl2 O4 and Fe were also observed in the FGM sample cross section using BSE images on the polished cross-sections of the FGM sample (Fig. 2). The bright particles in Fig. 2(a)

are metallic Fe while the light gray phase in Fig. 2(a)–(c) is FeAl2 O4 .8 The microstructure and composition varied continuously from the gas/solid interface to the middle of the specimen. The volume fraction of second phase precipitates (both FeAl2 O4 and Fe), measured using ImageJ image analysis software, increased from zero at the gas/solid interface to a maximum at about 250 ␮m below the surface and then reduced again towards the centre of the cross section. The chemical level of Fe was estimated by combining EDX results of the matrix and calculations based on the volume fractions of precipitates. Fig. 2(d) shows that the chemical content of Fe gradually increased from zero to a maximum at about 250 ␮m below the surface and then leveled off at about 7 wt% (as calculated corresponding to 10 wt% of Fe2 O3 ). It is therefore evident that the low particle volume fraction near the surface is the result of the loss of Fe by evaporation, while the lack of particles in the central part is because the oxygen vacancies responsible for the reduction of the solid solution had not had sufficient time to diffuse from the surface to the central part of the specimen.12,13 The loss of Fe is presumably also responsible for the near-surface porosity evident in Fig. 1(a). The kinetics of this process will be explored in a future publication, but it is noted here that the reduction reactions (1) and/or (2) are a necessary precursor to the loss of Fe because Fe2 O3 has a high melting point and a low vapour pressure at the temperatures used. Turning now to consider the residual stresses in the FGM sample, the thermal expansion coefficient (CTE) values in Table 1 show that the main factor influencing the local CTE is the Fe content. Since the average chemical concentration of Fe near the surface was much lower than in the bulk (Fig. 2d), there is expected to be a compressive residual stress in the

2696

C. Xu, R.I. Todd / Journal of the European Ceramic Society 35 (2015) 2693–2698

Table 1 Values and sources of material properties used in calculation of residual stress. The values for the solid solution were calculated according to the rule of mixtures. E is Young’s modulus, α is thermal expansion coefficient and γ is Poisson’s ratio. Alumina [14]

Fe [15]

FeAl2 O4 [16,17]

Fe2 O3 [18]

Solid solution

E (GPa) α (K−1 ) γ

402 9 × 10−6 0.23

211 14 × 10−6 0.29

222 13 × 10−6 0.32

359 16 × 10−6 0.12

398 9.7 × 10−6 0.22

Residual stress in the bulk part (MPa)

Property

The residual stress on the polished bottom surface of the halfFGM sample by Method 2 was measured as −75 ± 20 MPa. The stress-free thermal expansion mismatch strain εM between the surface layer and the bulk could then be estimated as 8.13 × 10−4 using Eq. (6), which can be derived from simple beam theory for a two layer structure:

450 400 350 300 250 200 150

σr =

100 50 0

0

500

1000

1500

2000

Fig. 3. Residual stress in the bulk part of the FGM sample.

surface region after cooling from the aging temperature. The continuously changing microstructure of the iron-depleted surface zone should lead to a gradually changing residual stress. This FGM quality is beneficial in avoiding cracking or delamination but makes the residual stresses difficult to measure accurately. Approximate methods of measurement and analysis are therefore used here, in which the FGM sample is assumed to consist of two 250 ␮m iron-depleted surface zones bounding a central, 2.0 mm thick bulk part. The Fe concentration is assumed to be constant within each of the layers but is lower in the surface layers than in the bulk. The resulting measurements and estimates of the residual stress thus refer to the mean stresses in each layer. The residual stress distribution measured in the bulk part of the FGM sample by Method 1 is shown in Fig. 3 confirming that the bulk part was in tension as expected. The steeply increasing stresses close to the surface region/bulk interface are attributed to the complex relaxation of stress close to the interface. The average biaxial residual stress measured in the central part of the bulk (900–1500 ␮m away from the gas/solid interface) in Fig. 3 was 140 ± 19 MPa (the error is from the scatter in the measurements only) using Eq. (1). The mean stress in the surface layers can be calculated using the fact that the mean force across any section is zero19 : tb 2ts

Es 2 ts4 +4Eb Es tb ts3 + 6Eb Es tb2 ts2 +4Eb Es tb3 ts +Eb Eb 2 tb4 (6)

where Es is the average biaxial stiffness (Es /(1 − γ s )) of the surface layer (calculated using the rule of mixtures at the midpoint of the surface layer where the volume fraction of FeAl2 O4 is 7%), and Eb is the average biaxial stiffness of the bulk part. Applying the measured εM in Eq. (7),20

2500

Distance from the gas/solid interface (μm)

σs = −σb

εM Eb Es ts (Es ts3 − 3Eb ts tb2 − 2Eb tb3 )

(5)

where σ s is the average stress in the iron depleted zone, σ b is the average stress in the bulk part of the FGM sample, ts is the thickness of iron depleted zone (250 ␮m) and tb is the thickness of the bulk part (2.0 mm). The resulting average residual stress in the surface layer of the FGM sample was −560 ± 76 MPa.

σs = −

εM Es 1 + 2 × ts Es /tb Eb

(7)

the residual stress σ s in the surface layer of the FGM can be calculated as −332 ± 88 MPa. This was lower than the value estimated by Method 1 but of the same order. Although Method 2 avoids the stress relaxation problem of Method 1, the difference term (Es ts3 − 3Eb ts tb2 − 2Eb tb3 ) in the numerator is small, which results in a relatively large error, in addition to the quoted scatter in the results, from the uncertainty in the physical values it contains. The theoretical average residual stress in the surface zone can be estimated by approximating the FGM to a sandwich structure, as for the experimental analysis. An average CTE could be calculated for the iron depleted surface zone using the rule of mixtures. εM is given by:20  T0 εM = (αb − αs )dT (8) T

where T is room temperature and T0 is the temperature below which elastic stress develops, αb is the average CTE of the bulk part and αs is the average CTE of the iron-depleted surface zone. Using Eqs. (5), (7) and (8) and the values in Table 1, T0 − T = 1330 ◦ C, tb = 2.0 mm and ts = 250 ␮m, the theoretical residual stress in the surface layer and bulk part of the FGM sample can be obtained as σ s = −440 MPa, σ b = 110 MPa. This is of similar magnitude to the experimental values. It can be concluded that the mean compressive residual stress in the surface region is ∼500 MPa, and presumably even higher values exist at the surface, where the iron depletion is most pronounced. No evidence of cracking or delamination was observed, demonstrating the success of the FGM structure in suppressing these failure mechanisms.

C. Xu, R.I. Todd / Journal of the European Ceramic Society 35 (2015) 2693–2698 Table 2 Vickers hardness (Hv5 ) and wear properties of samples. For hardness the surfaces of as-aged samples were lightly polished by 3 ␮m diamond suspension before indentation so that the subsequent indentation could be seen clearly. Properties

Alumina

Solid solution

FGM

Hardness (GPa) Wear resistance, 1/k (N m−2 ) Pull-out percentage (%) Pull-out size (␮m)

16.3 ± 0.1 1.5 × 1013 34 3.9 ± 0.2

13.7 ± 0.2 1.2 × 1013 36 4.0 ± 0.2

11.9 ± 0.2 7.3 × 1013 9 4.7 ± 0.3

Table 2 summarizes the hardness results. The pure alumina had the highest hardness, followed by the solid solution and then the FGM sample. The average diagonal radius of indentations on the FGM sample (around 35 ␮m) can be regarded as the plastic zone depth, which was much less than the thickness of the iron depleted surface zone (around 250 ␮m). Thus the hardness measured represents the properties of the iron depleted surface zone. The low hardness value is presumed to be a consequence of the porosity apparent in Fig. 1(a). The worn surfaces of the three samples were observed by SEM. The area fractions and size of pull-outs caused by brittle fracture were measured from SE-SEM images of the worn surfaces and analysed by ImageJ software (Table 2). All the worn surfaces comprised two distinct features: (1) areas showing material removal by cutting and ploughing by plastic deformation, and (2) areas showing pull-outs by brittle fracture.21 The worn surfaces of both monolithic alumina and solid solution (Fig. 4a and b) showed the classical feature of extensive pullout, whereby a significant amount of material was removed by intergranular fracture. By contrast, the worn surfaces of the FGM sample (Fig. 4c) were smoother than both monolithic alumina and the solid solution (Fig. 4a and b), and also showed much

2697

less pull-out. The plastic deformation induced material removal is dominant in the FGM sample. The wear volumes of samples were calculated using the following equation22 : V =

πb4 64R

(9)

where b is the crater diameter and R is the ball radius, assuming that the shape of the crater is conformal to the shape of the ball. The Archard wear law22 relates the wear volume V to the sliding distance S and applied load N as follows: V = kSN

(10)

where k is the wear coefficient. A linear plot of V against SN can therefore be plotted based on Eq. (10), with gradient equal to the wear coefficient k (Fig. 4(d)). It can be seen that the wear resistances, 1/k, of monolithic alumina and the solid solution were comparable, while the FGM sample exhibited a wear resistance around 6 times of that of the solid solution (Table 2). Table 2 shows that the main reason for the improved wear resistance of the FGM is a marked reduction in surface pullout by brittle fracture. The other parameters relevant to wear are either similar or, in the case of the modest reduction in hardness, would lead to a lower wear resistance. Since the mean size of the pullouts is similar in all three ceramics, the reduction of pullout in the FGM must be caused by a reduction in the pullout formation rate.21 This is thought to be a result of the compressive residual stress in the surface layer. It has been demonstrated previously that residual compressive stress in the surface layer of a laminate sample leads to an improvement in wear resistance,1–3 and the surface stresses here are very high. The rate at which cracks are initiated during wear depends upon extremely high

Fig. 4. SEM images obtained from the worn surfaces of (a) monolithic alumina; (b) solid solution; (c) the FGM sample; (d) wear volume determined as a function of sliding distance for selected samples. The wear coefficient k is equal to the slope of the corresponding straight line for points after the initial bedding in period.

2698

C. Xu, R.I. Todd / Journal of the European Ceramic Society 35 (2015) 2693–2698

stress concentrations at the head of twins and dislocation pile ups during deformation and thus seems unlikely to be affected significantly by the residual stresses of several hundred MPa measured here. However, the arrest of cracks normal to the surface under the action of the compressive stresses in the surface region could reduce the pull-out formation rate.3 This method of improving the wear resistance of Al2 O3 –Fe2 O3 solid solutions by using reduction aging to induce residual stresses is more effective and easier to produce than an alternative method,23 in which the processing conditions are adjusted to precipitate intragranular FeAl2 O4 nanoparticles, which hinder crack initiation by suppressing twinning and dislocation motion. 4. Conclusions In summary, this work provides a new method for fabricating ceramic FGMs which avoids using hot pressing and allows the production of highly compressive surface residual stresses in non-planar, pressurelessly sintered components. The FGM architecture of the surface layer also avoids the ‘delamination problem’ which is common in laminates. The greatly improved wear resistance of the FGM sample (by 5 times) compared with the monolithic alumina is attributed to crack arrest by the compressive residual stresses and indicates an opportunity for these ceramic FGMs to be used in industrial wear resistant applications. Acknowledgement Chen Xu is grateful for the support of her studentship by an Oxford University-Chinese Education Ministry Joint Scholarship. References 1. de Portu G, Micele L, Prandstraller D, Palombarini G, Pezzotti G. Abrasive wear in ceramic laminated composites. Wear 2006;260:1104–11. 2. Toschi F, Melandri C, Pinasco P, Roncari E, Guicciardi S, de Portu G. Influence of residual stresses on the wear behavior of alumina/alumina–zirconia laminated composites. J Am Ceram Soc 2003;86:1547. 3. Dancer CEJ, Yahya NA, Berndt T, Todd RI, de Portu G. Effect of residual compressive surface stress on severe wear of alumina–silicon carbide twolayered composites. Tribol Int 2014;74:87.

4. Erdogan F. Fracture mechanics of functionally graded materials. MRS Bull 1995;20:43. 5. Calis N, Kushan SR, Kara F, Mandal H. Functionally graded SiAlON ceramics. J Eur Ceram Soc 2004;24:3387. 6. Dancer CEJ, Achintha M, Salter CJ, Fernie JA, Todd RI. Functionally graded alumina–silicon carbide components. Scr Mater 2012;67:281. 7. Lee CS, Ahn SH, DeJonghe LC, Thomas G. Effect of functionally graded material (FGM) layers on the residual stress of polytypoidally joined Si3 N4 –Al2 O3 . Mater Sci Eng A 2006;434:160. 8. Mukhopadhyay A, Todd RI. Microstructure and mechanical properties of Al2 O3 matrix nanocomposites produced by solid state precipitation. J Eur Ceram Soc 2010;30:1359. 9. Mukhopadhyay A, Todd RI. Effect of yttria doping on the microstructure and mechanical properties of Al2 O3 –FeAl2 O4 nanocomposites developed via solid state precipitation. J Eur Ceram Soc 2010;30: 2905. 10. Raghavan V. Al–Fe–O (aluminum–iron–oxygen). J Phase Equilib Diffus 2010;4:367. 11. He J, Clarke DR. Determination of the piezospectroscopic coefficients for chromium-doped sapphire. J Am Ceram Soc 1995;78:1347. 12. Porter DA, Easterling KE. Phase transformations in metals and alloys. London: Chapman & Hall; 1992. 13. Kingery WD, Bowen HK, Uhlmann DR. Introduction to ceramics. WileyInterscience; 1976. 14. Todd RI, Bourke MAM, Borsa CE, Brook RJ. Neutron diffraction measurements of residual stresses in alumina/SiC nanocomposites. Acta Mater 1997;45:1791. 15. Ledbetter HM. Reference database: elastic properties of metals and alloys. J Phys Chem 1973;3. 16. Tromans D, Meech JA. Fracture toughness surface energies of minerals: theoretical estimates for oxides. Miner Mater 2002;15: 1027. 17. Fujimura T, Tanaka SI. In-situ high temperature X-ray diffraction study of Fe/Al2 O3 interface reactions. J Mater Sci 1999;34:425. 18. Chicot D, Mendoza J, Zaoui A, Louis G, Lepingle V, Roudet F, et al. Mechanical properties of magnetite (Fe3 O4 ), hematite (␣-Fe2 O3 ) and goethite (␣-FeO·OH) by instrumented indentation and molecular dynamics analysis. Mater Chem Phys 2011;129:862. 19. Sanchez-Herencia AJ, Baudin C. Ceramic laminates with tailored residual stress. Bol Soc Esp Ceram 2009;48:311. 20. Ho S, Hillman C, Lange FF, Suo Z. Surface cracking in layers under biaxial, residual compressive stress. J Am Ceram Soc 1995;78: 2353. 21. Ortiz Merino JL, Todd RI. Relationship between wear rate, surface pullout and microstructure during abrasive wear of alumina and alumina/SiC nanocomposites. Acta Mater 2005;53:3345. 22. Rutherford KL, Hutchings IM. Theory and application of a micro-scale abrasive wear test. J Test Eval 1997;25:259. 23. Mukhopadhyay A, Todd RI. Relationship between microstructure and abrasive wear resistance of Al2 O3 –FeAl2 O4 nanocomposites produced via solid-state precipitation. J Eur Ceram Soc 2011;31:339–50.