- Email: [email protected]
Simulation of erosive smoothing in the abrasive jet micro-machining of glass
Journal of Materials Processing Technology 213 (2013) 2254–2261
Contents lists available at ScienceDirect
Journal of Materials Processing Technology journal homepage: www.elsevier.com/locate/jmatprotec
Simulation of erosive smoothing in the abrasive jet micro-machining of glass R. Haj Mohammad Jafar a , M. Papini b,a,∗ , J.K. Spelt a,b,∗∗ a b
Department of Mechanical and Industrial Engineering, Ryerson University, 350 Victoria Street, Toronto, ON, Canada M5B 2K3 Department of Mechanical and Industrial Engineering, Ryerson University, 350 Victoria Street, Toronto, ON, Canada M5B 2K3
a r t i c l e
i n f o
Article history: Received 12 May 2013 Received in revised form 23 June 2013 Accepted 26 June 2013 Available online 4 July 2013 Keywords: Abrasive jet micro-machining Surface roughness Smoothing Glass
a b s t r a c t Abrasive jet micro-machining (AJM) is a promising technique to machine micro-features in brittle and ductile materials. However, the roughness of micro-channels machined using AJM is generally greater than that from other methods of micro-machining such as wet etching. Previous investigators have suggested that the surface roughness resulting from AJM can be reduced by post-blasting with abrasive particles at a relatively low kinetic energy. This approach was investigated in the present work by measuring the roughness reduction of a reference unmasked channel in borosilicate glass as a function of post-blasting particle size, velocity, dose, and impact angle. Post-blasting the reference channels reduced the roughness by up to 60%. It was observed that post-blasting at shallower angles was more efficient, probably due to the increased amount of edge chipping as opposed to cratering, which contributed to the enhanced removal of profile peaks, leaving a smoother surface. Moreover, post-blasting with smaller particles ultimately resulted in smoother surfaces, but at the penalty of requiring a relatively large particle dose, and consequently a significantly increased channel depth, before reaching the steady-state roughness. Hence, finishing with smaller particles until reaching the steady-state roughness may not be practical when a shallow channel is desired. A previously developed numerical model was modified and used to simulate the post-blasting process leading to the creation of smooth channels as a function of particle size, velocity, dose, impact angle, and target material properties. The model simulated both crater formation (due to growth of lateral cracks) and the chipping of facet edges. Comparisons with centerline roughness measurements for channels in borosilicate glass showed that the model can predict the transient roughness reduction with post-blasting particle dose with a 7% average error. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Abrasive jet micro-machining (AJM) utilizes an abrasive particle jet to mechanically etch micron-sized features into various materials for micro-systems fabrication. A possible disadvantage of AJM in micro-fluidic applications is the relatively high resulting surface roughness that can affect fluid flow phenomena. Ghobeity et al. (2012) compared the separation efficiency of glass channels machined using AJM and wet etching and found that the separation efficiency in AJM channels (Ra ∼ 0.4–0.6 m) was significantly lower (0.2–0.25 times) than that in wet-etched channels
∗ Corresponding author at: Department of Mechanical and Industrial Engineering, Ryerson University, 350 Victoria Street, Toronto, ON, Canada M5B 2K3. Tel.: +1 416 979 5000x7655; fax: +1 416 979 5265. ∗∗ Corresponding author at: Department of Mechanical and Industrial Engineering, University of Toronto, King’s College Road, Toronto, ON, Canada M5S 3G8. Tel.: +1 416 978 5435; fax: +1 416 978 7753. E-mail addresses: [email protected] (M. Papini), [email protected] (J.K. Spelt). 0924-0136/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jmatprotec.2013.06.022
(Ra ∼ 2–5 nm). Similarly, Solignac et al. (2001) observed that the electro-osmotic mobility in AJM channels in soda-lime glass is less than 50% of that in HF-etched channels. Moreover, Ladouceur (1997) has shown that roughness is a constraining factor in optoelectronics devices where it scatters light and attenuates power. Therefore, methods to reduce roughness resulting from AJM operations are highly desirable. The mechanics of cracking in brittle materials due to Vickers indentation was investigated by Marshall et al. (1982) who derived equations to predict the length and depth of lateral cracks as a function of the material properties and indentation force. Utilizing fracture mechanics models of Vickers indentations (Marshall et al., 1982), Slikkerveer et al. (1998) estimated the erosion rate and steady-state roughness of channels in borosilicate glass by assuming that each particle impact removed a spherical cap of material with a radius equal to that of the predicted lateral crack length and a depth equal to that of the plastic zone radius and by assuming that there was no overlap among impact sites. It was concluded that the only important parameter affecting the roughness was the kinetic energy of the impinging particles associated with the velocity
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261
2255
Table 1 Mechanical properties of the borosilicate glass targets. Elastic modulus (GPa)
Fracture toughness (MPa
63
0.76
√
m)
component normal to the target. The present authors investigated the accuracy of this model by measuring the dimensions of individual impact sites resulting from the AJM of borosilicate glass by using aluminum oxide particles of various sizes (Jafar et al., 2013a). It was found that lateral crack initiation was better approximated as originating from the indentation depth rather than the bottom of the plastic zone, as was assumed in the model of Slikkerveer et al. (1998). Roughness measurements for various impact angles also showed that Ra increased with tangential velocity component; i.e. Ra was not a function of only the kinetic energy due to the normal velocity. This observation was not explored further in Jafar et al. (2013a) where the analytical model was compared only with steady-state roughness measurements made at 90◦ impact angle. The present authors (Jafar et al., 2013b) also developed a numerical model of the erosion process that was used to predict the steady-state roughness and erosion rate of unmasked channels resulting from AJM on borosilicate glass as a function of particle size, velocity, dose, and impact angle. It was assumed that two brittle damage mechanisms contributed to erosion: (1) crater removal due to the initiation and growth of lateral cracks and (2) edge chipping due to the propagation of cracks through a facet in the direction of the impact. The model predicted the steady-state roughness and the erosion rate of unmasked channels machined in borosilcate glass with average errors of 9% and 29% respectively, when compared with experimental data over a wide range of machining conditions. Moreover, the model predicted profile shape parameters such as waviness, skewness, and kurtosis with less than 15% error compared with experimental data. The results showed that, although the number of edge chipping events (the second erosion mechanism) increased with decreasing angle of attack, the main damage mechanism was crater removal, even at shallow angles. The roughness of a typical borosilicate glass channel machined with AJM depends strongly on the particle kinetic energy, varying from 0.1 m to 9 m using alumina particles with average sizes from 5 m to 200 m, respectively (Slikkerveer et al., 1998). The smoothing of glass surfaces after AJM has not been investigated extensively in literature. Wensink et al. (2002) achieved a 100% reduction in the roughness of two reference channels machined in borosilicate glass using either 9 m or 29 m alumina by annealing the eroded glass at a temperature just below its softening point. In contrast, a post-blast wet etching of the reference channels with hydrofluoric acid produced a rougher surface, probably because of the opening of cracks generated by the AJM. Post-blasting the reference channels with 3 m alumina decreased the roughness of both reference channels by about 25% while the depth of the channels increased approximately 10 m. Mineta et al. (2009) examined the reduction of roughness due to post-blasting using a wet abrasive on a borosilicate glass (PyrexTM ) substrate. Both the original and the post-blasted channels were machined using an aqueous abrasive slurry of 4 m alumina, however the later were machined
Vickers hardness (GPa)
Density (g/cm3 )
Poisson’s ratio
5.2
2.2
0.2
with a lower pressure so that a ductile-mode erosion was dominant, resulting in a roughness reduction of about 50%. The present work investigated the role of post-blasting particle size, velocity, dose, and impact angle on roughness reduction of unmasked channels machined in borosilicate glass using AJM. A numerical model, which was previously developed to predict the steady-state roughness resulting from the AJM of an initially un-eroded surface (Jafar et al., 2013b), was used to predict both the transient and steady-state roughnesses resulting from postblasting smoothing operations. 2. Experiments 2.1. Apparatus and target material The experiments were conducted using an AccuFlo abrasive blaster from Comco Inc. (Burbank, CA, USA) with a blasting nozzle having an inner diameter of 1.5 mm that was held stationary at a nozzle-to-surface stand-off distance of 10 mm (centerline distance between the nozzle exit and the target). The target material was 3 mm thick Borofloat® 33 borosilicate glass (Schott Inc., NY, USA) with the mechanical properties given in Table 1. To create unmasked channels, the glass specimen was clamped to a computer-controlled stage and moved across the blast zone of a stationary nozzle at a scan speed, Vs , of 0.25–20 mm/s using various angles of nozzle inclination, measured with respect to the target surface, . Varying the scan speed provided a convenient way of changing the particle dose. The surface roughnesses were reported as the average of three repeat measurements along a 5 mm length of the centerline of the machined channels using an optical profilometer (Nanovea Inc., Irvine, CA) with a 0.2 m step size and a cut-off length of 800 m. A more detailed discussion of the roughness measurements can be found in Jafar et al. (2013a). To provide a baseline channel roughness from which various post-blasting scenarios could be evaluated, unmasked reference channels were machined at normal incidence (i.e. = 90◦ ), using 150 m alumina particles at a scan speed of 4 mm/s and a pressure of 200 kPa. The average depth of three machined reference channels measured with the optical profilometer was 165 m, with a standard deviation of 7.6 m, while the average width was approximately 3 mm. The steady-state roughness of these same reference channels was Ra = 5.0 m, with a standard deviation for three channels of 0.27 m. The reference channels were then post-blasted with alumina particles of various sizes (50, 100 and 150 m) at pressures of 100 and 200 kPa, and impact angles of 30◦ , 60◦ and 90◦ . The standard deviations over three measurements of the centerline roughness and centerline depth (or erosion rate) of the post-blasted channels were less than 10% of the mean values. Table 2 summarizes the AJM process parameters used in the machining of the reference and the post-blasted channels.
Table 2 Process parameters used to machine the reference and the post-blasted channels. Channel
Particle size (m)
Pressure (kPa)
Impact angle (◦ )
Scanning speed (mm/s)
Stand-off distance (mm)
Reference Post-blasted
150 50, 100, 150
200 100, 200
90 30, 60, 90
4 0.25–20
10 10
2256
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261
Table 3 Measured particle mean diameter, jet focus coefficients, particle velocities at the jet center 10 mm from the nozzle exit, and mass flow rates. Velocity range corresponds to 95% of population (Jafar et al., 2013b). Nominal particle size (m)
Mean diameter (m)
Pressure (kPa)
ˇw
Velocity range (m/s)
Mean velocity (m/s)
Mass flow rate (g/min)
50 100 150 150
41 138 182 182
100 100 100 200
10.9 13.9 15.3 9.6
56–107 45–86 45–66 62–97
80 57 52 78
7.2 9.8 15.8 22.3
2.2. Particle size, velocity and mass flux characterization The particle size, velocity and mass flux distributions were measured in a previous work (Jafar et al., 2013b), and thus only a short summary has been given here. The size distributions of the alumina powders were determined using an automated image analysis system (Clemex Vision PE, Clemex Technologies Inc., QC, Canada) and fitted with either a log-normal distribution for the 50 m particles, or a Weibull distribution for the 100 and 150 m particles. The mean diameter of the particles is given in Table 3. The distributions of particle velocity and spatial density as a function of radius from the jet axis, at pressures of 100 and 200 kPa and 10 mm from the nozzle exit, were measured using high resolution imaging with laser-pulsed backlit illumination called shadowgraphy (Dehnadfar et al., 2012). The alumina mass flow rates were measured by weighing the mass of abrasive particles blasted for 2 min into a closed container having a filter at the end to permit the air to escape. Table 3 gives these data for the abrasive jets along with the values of the jet focus coefficient, ˇw , which is a measure of jet divergence, with higher values indicating a more focused stream (Ghobeity et al., 2008).
3. Erosion simulation A full description of the model is presented in Jafar et al. (2013b) where it was used to predict the steady-state roughness of channels machined by an abrasive jet on an initially flat brittle target. In the present work, the model was adapted to simulate the transient development of the roughness of a pre-eroded profile as it approached a steady-state topography. This smoothing process was modeled as a function of particle size, velocity, dose, and impact angle. The simulation was implemented in MATLABTM incorporating the two erosion mechanisms of crater formation and edge chipping (Jafar et al., 2013b). The measured centerline profile of a 5 mm long reference channel was entered into the model as the initial profile, along with the desired total number of impacts and the nozzle inclination angle. Other inputs were the measured distributions of particle size and velocity, as well as the particle spatial distributions. To simulate the erosion of the centerline of the reference channel, impacts were randomly located over an area of 5 mm long and 6dnom wide (6 times the nominal particle diameter) until a portion of the particle projected area crossed the strip centerline. The model then examined the possibility of both crater removal and edge chipping occurring in the profile facet that was in contact with the impacting particle. In the case of a crater removal event, a V-shaped crater was formed with a depth equal to that of the predicted indentation and a radius equal to the lateral crack length, both of which are functions of particle kinetic energy associated with the normal velocity component and target material properties (Slikkerveer et al., 1998). The crater was formed in the profile only if the lateral crack was predicted to reach the surface of the profile. Edge chipping was examined using a 2D model developed by Chai and Lawn (2007), in which the edge chip geometry is a function of the plane-strain fracture toughness (Kc ), the indentation distance from the edge, the indentation force magnitude and angle, and the edge geometry. If the particle impact force was lower
than that required for both chipping and cratering, it was assumed that the impacting particle did not damage the surface. Once the required total number of impacts was reached and the final surface profile was obtained, the average channel centerline depth was calculated along with the roughness, Ra , employing the same 800 m cut-off length that was used in the experimental measurements. 4. Results and discussion 4.1. Experimental results Previous studies have shown that although the tangential component of velocity can affect the surface roughness of channels machined using AJM (Jafar et al., 2013a), the roughness most strongly depends on the ‘normal component’ kinetic energy; i.e. that associated with the component of particle velocity perpendicular to the target (Slikkerveer et al., 1998). This implies that the roughness of an AJM channel can be reduced by blasting it with a jet having a lower normal component kinetic energy, achieved using shallow impact angles, low blasting pressures, and/or smaller particles. The following describes the effect that each of these process parameters had on the smoothing of the reference channels. The results of the numerical simulation are then described in Section 4.2. 4.1.1. Impact angle To investigate the effect of the impact angle on roughness reduction, the reference channels were finished (i.e. ‘post-blasted’) using the same particles and pressure (150 m, 200 kPa) that were used to make the reference channel, but at impact angles of 30◦ and 60◦ to the surface, which decreased the normal component of kinetic energy relative to the original 90◦ blasting by 75% and 25%, respectively. Fig. 1a shows the measured and predicted decrease in surface roughness and the increase in channel depth with particle dose, and Fig. 1b depicts the measured roughness reduction and the depth increase as a function of “normal energy density” (i.e. product of the number of impacts and the normal component of kinetic energy per unit area) of the post-blasted particles. In each case, the starting roughness was that of the reference channel (Ra = 5 m). Also shown on each graph is the steady-state roughness achieved on an initially smooth glass surface eroded under the same conditions used for post-blasting. With increasing particle dose (Fig. 1a) or normal energy density (Fig. 1b), the channel depth increased almost linearly (slight curve with log scale) as expected for a constant erosion rate. The R2 of the least-squares linear fits for 30◦ and 60◦ were 0.99 and 0.98, respectively. At 30◦ , a gradual decrease in roughness was observed with increasing dose leading to a steady-state roughness at approximately 2.51 × 103 impacts/mm2 or 2.35 × 10−2 J/mm2 . The depth increase exhibited the same trend at 60◦ , but the roughness transient period was very much shorter. Several other observations can be made from Fig. 1a: (1) the channels at a given dose were deeper at 60◦ than 30◦ ; (2) the steady-state roughness at 30◦ was lower than that at 60◦ ; (3) a higher dose was needed to reach the steady-state roughness at a shallower angle of attack. These are all attributed to the decrease in kinetic energy due to the impact velocity
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261
Measured depth increase (30º)
Predicted roughness (60º)
Predicted depth increase (60º)
Predicted roughness (30º)
Predicted depth increase (30º)
4.5
160
5
140
4.5
60
Ra (µm)
4
120
3.5
100
30
3
80
2.5
2
1.5
Measured roughness (30º)
Measured depth increase (30º)
Measured roughness (60º)
Measured depth increase (60º)
180
5.5
180
160
60
140
120
4
Ra (µm)
5
Measured roughness (30º)
Depth increase (µm)
5.5
Measured depth increase (60º)
200
(b) 6
200 Measured roughness (60º)
3.5
100
30
3
80
60
2.5
60
40
2
40
20
1.5
20
150 m, P=200 kPa 1 1
10
100
0 1000
Particle dose (impacts/mm2)×102
Depth increase (µm)
(a) 6
2257
150 m, P=200 kPa 1 1
10
100
0 1000
Normal kinetic energy density (J/mm2)×10-3
Fig. 1. Variation of roughness and depth of a post-blasted borosilicate glass channel with (a) particle dose and (b) normal kinetic energy density (log scale) for 150 m alumina particles at P = 200 kPa and = 30◦ , 60◦ . Horizontal lines indicate the steady-state roughness achieved by blasting a smooth glass surface under the same conditions as those used in post-blasting; solid lines are measured values and dashed lines are model predictions.
normal to the surface at shallower impact angles, which leads to the removal of smaller chips. This in turn causes a lower erosion rate and steady-state roughness. The depth changes (erosion) for both impact angles in Fig. 1b roughly follow a single curve, confirming that the erosion rate is a function of particle normal kinetic energy. The predictions of roughness and depth will be discussed in Section 4.2. 4.1.2. Blasting pressure The effect of the pressure and hence velocity on roughness reduction and depth increase is shown in Fig. 2a and b, as the original channels made with150 m alumina at 200 kPa and 90◦ were post blasted using the same particles but at 100 kPa and at impact angles of 30◦ , 60◦ and 90◦ . Reducing the blasting pressure from 200 to 100 kPa resulted in a jet with 56% less kinetic energy. The trends of Fig. 2a are similar to those in Fig. 1a, with the dose corresponding to the onset of steady-state roughness increasing as the kinetic energy due to the normal velocity decreased, leading to removal of smaller chips per impact. A comparison between Figs. 1a and 2a shows that this also led to smoother surfaces; e.g. at 30◦ the steadystate Ra was approximately 2.7 m, at 100 kPa while it was 3.3 m at 200 kPa. Fig. 2b shows many of the same trends seen in Fig. 1b: (1) the depth increase for all angles followed approximately a single curve, indicating that the erosion was only function of the normal kinetic energy; (2) during the transient period, a lower roughness was achieved at a given normal energy density when post-blasting at shallower angles, which was probably due to the increased amount of edge chipping as opposed to cratering (chipping percentage) at lower angles, especially during the early stages of smoothing where the rate of smoothing (the roughness reduction per normal energy density) was higher, which contributed to the enhanced removal of profile peaks, leaving a smoother surface. This is supported by the absence of an increased initial smoothing rate in Fig. 3b (discussed below), which shows additional data at 90◦ where little or no edge chipping occurs (this hypothesis will be further discussed in Section 4.2); (3) the normal energy density required to reach steady state increased as the angle decreased; and (4) the channel depth corresponding to the onset of steady-state increased as the post-blasting impact angle decreased; e.g. the depth increase for
90◦ was only approximately 10 m, for 60◦ it was 35 m, and for 30◦ it was 65 m. The reason for observations (3) and (4) is simply that, since the steady-state roughness at shallower angles is lower, the amount of required smoothing is greater; thus a greater normal kinetic energy density is required in order to reach the steady-state, which consequently increases the channel depth. 4.1.3. Particle size Fig. 3a shows the changes in roughness and depth with dose during post-blasting of the reference channels with 50 m, 100 m and 150 m alumina at 100 kPa and 90◦ . As expected, smoother surfaces were obtained at steady state with the 50 m particles, because of their lower normal kinetic energy compared with 100 and 150 m particles. As in Figs. 1a and 2a, this also meant that a higher dose was required to reach steady-state. For example, the dose to reach steady state with the 50 m particles was about 45 times more than with the 100 m particles and about 450 times more than that with the 150 m particles (Fig. 3a). As discussed below, these large differences were attributable to a reduced normal velocity. Fig. 3b shows that, as expected, the normal kinetic energy density to achieve a steady-state roughness increased when postblasting with smaller particles. For example, at P = 100 kPa and = 90◦ , the required energy density to reach steady state with the 50 m particles was about 2.3 times more than with the 100 m particles and 12 times more than that with the 150 m particles (Fig. 3b). This was simply because the steady-state roughness was lower with smaller particles, and so a greater amount of smoothing was required to reach the steady-state; i.e. a higher normal kinetic energy density was required. Fig. 3a and b shows that at a steady-state roughness, more material was removed from the reference channel when smaller particles were used; e.g. at the beginning of the steady state, the depth increase of the reference channels blasted at 90◦ was 11 m for 150 m particles, 125 m for 100 m particles, and 215 m for 50 m particles (Fig. 3a and b). Thus, post-blasting until reaching the steady-state roughness using small particles may not be practical when a shallow channel is required. To aid in a comparison among all the studied post-blasting conditions, Fig. 4a and b shows the depth increase and roughness,
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261 6
(a)
200 Measured depth increase (90º)
Measured roughness (60º)
Measured depth increase (60º)
Measured roughness (30º)
Measured depth increase (30º)
200
(b) 6
180
5.5
Measured roughness (90º)
Measured depth increase (90º)
Measured roughness (60º)
Measured depth increase (60º)
Measured roughness (30º)
Measured depth increase (30º)
180
160
5
160
4.5
140
4.5
140
4
120
4
120
90
3.5
100
60
3
80
30
Ra (µm)
5
Depth increase (µm)
Ra (µm)
5.5
Measured roughness (90º)
3.5
100
90 60
3
80
30
2.5
60
2.5
60
2
40
2
40
20
1.5
1.5
20 150 m, P=100 kPa
150 m, P=100 kPa 0 1000
1 10
1
100
Depth increase (µm)
2258
0 1000
1 1
10
Particle dose (impacts/mm2)×102
100
Normal kinetic energy density (J/mm2)×10-3
Fig. 2. Variation of roughness and depth of a post-blasted borosilicate glass channel with (a) particle dose and (b) normal kinetic energy density (log scale) for 150 m alumina particles at P = 100 kPa and = 30◦ , 60◦ , 90◦ . Horizontal lines indicate the measured steady-state roughness achieved by blasting a smooth glass surface under the same conditions as those used in post-blasting.
respectively, for all the conditions of Figs. 1–3 as a function of normal energy density and Fig. 4c shows the roughness versus the depth increase. Fig. 4a confirms that the erosion rate is only a function of the normal energy density. However, Fig. 4b shows that the transient roughness behavior is affected by the impact angle so that post-blasting is generally more efficient when performed at lower angles for a given particle size and pressure; i.e. less normal kinetic energy density is required to reach a given roughness at shallower angles. Although a unique trend cannot be seen in Fig. 4c, it may be concluded that for a given channel depth generally using the smallest particle available at the lowest pressure and impact angle possible could result in a lower transient roughness.
700
5.5
Measured roughness (150 µm)
Measured depth increase (150 µm)
Measured roughness (100 µm)
Measured depth increase (100 µm)
Measured roughness (50 µm)
Measured depth increase (50 µm)
Predicted roughness (100 µm)
5
300
2.5
200 50 µm
2
Measured roughness (100 µm)
Measured depth increase (100 µm)
Measured roughness (50 µm)
Measured depth increase (50 µm)
500
4
Ra (µm)
100 µm
Measured depth increase (150 µm)
4.5
Depth increase (µm)
400
3
Measured roughness (150 µm)
600 5
4 150 µm
700
5.5
500
3.5
6
600
Predicted depth increase (100 µm)
4.5
Ra (µm)
(b)
400 3.5
150 µm
3
100 µm
300
2.5
Depth increase (µm)
(a) 6
Fig. 5a shows the total roughness reduction that is possible and the per particle normal kinetic energy versus the normal energy density required to reach the steady-state roughness, es , for all the studied post-blasting conditions (except 150 m particles at 200 kPa and 60◦ where steady-state occurred almost immediately). It is seen that when post-blasting with a particle of lower normal kinetic energy, Un , the roughness reduction (Ra ) is higher, but a greater normal energy density is required to reach steadystate. Fig. 5b depicts the total roughness reduction versus the increase in depth at the steady-state, ds , showing that to obtain a greater roughness reduction more material must be removed from the surface. Fig. 5 can be used as a guide to find a post-blasting
200 50 µm
2
100
100 1.5
1.5 P =100 kPa,
1 1
10
100
1000
Particle dose (impacts/mm2)×102
=90° 10000
0
P =100 kPa,
1 1
10
100
=90° 0 1000
Kinetic energy density (J/mm2)×10-3
Fig. 3. Variation of roughness and depth of a post-blasted borosilicate glass channel with (a) particle dose and (b) normal kinetic energy density (log scale) for 50 m, 100 m and 150 m alumina particles blasted at P = 100 kPa and = 90◦ . Horizontal lines indicate the steady-state roughness achieved by blasting a smooth glass surface under the same conditions as those used in post-blasting; solid lines are measured values and the dashed line is model prediction for 100 m particles.
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261
50
(a) 4 Roughness reduction
50 µm-100 kPa-90º
450
3.5
100 µm-100 kPa-90º
400
40
150 µm-100 kPa-90º 350
45
Per particle kinetic energy
3 35
150 µm-100 kPa-60º
300
2.5
250
150 µm-200 kPa-60º
200
150 µm-200 kPa-30º
Ra (µm)
150 µm-100 kPa-30º
150
30 Ra = 0.0062es + 1.65 R² = 0.79
2
25 20
1.5
15
100 1
50
Un = -0.0658es + 13.44 R² = 0.65
0.5
0 1
10
100
10 5
1000
0
0
Kinetic energy density (J/mm2)×10-3
Per particle kinetic energy, Un (µJ)
(a) 500
Depth increase (µm)
2259
0
50
100
150
200
250
Normal kinetic energy density to reach steady-state, es (J/mm2)×10-3
(b) 5
(b) 4
4.5
3.5 4
3
Ra = 0.0053ds + 1.66 R² = 0.65
2.5
50 µm-100 kPa-90º
3
Ra (µm)
Ra (µm)
3.5
100 µm-100 kPa-90º 2.5
150 µm-100 kPa-90º 150 µm-100 kPa-60º
2
2 1.5
150 µm-100 kPa-30º
1
150 µm-200 kPa-60º
1.5
150 µm-200 kPa-30º
0.5
1 1
10
100
1000
0
Kinetic energy density (J/mm2)×10-3 (c)
0
5
50
100
150
200
250
Depth increase at steady-state, ds (µm)
4.5
Fig. 5. Measured roughness reduction versus (a) normal energy density to reach steady-state roughness for a given post-blasting particle normal kinetic energy and (b) depth increase at steady state.
4
Ra (µm)
3.5 3
50 µm-100 kPa-90º 100 µm-100 kPa-90º
2.5
150 µm-100 kPa-90º 150 µm-100 kPa-60º
2
150 µm-100 kPa-30º 1.5
150 µm-200 kPa-60º 150 µm-200 kPa-30º
1 1
10
100
1000
Depth increase (µm) Fig. 4. (a) Depth increase and (b) roughness versus normal energy density, and (c) roughness versus depth increase for all post-blasting conditions of Figs. 1–3.
condition to achieve the minimum possible steady-state roughness for a given depth increase. For example, the ultimate roughness reduction for a given depth increase can be found using Fig. 5b, and the corresponding normal kinetic energy density follows from the upper curve of Fig. 5a. The lower curve of Fig. 5a then gives the required kinetic energy per particle which then can be used to determine the post-blasting condition (particle size, pressure and impact angle). It remains to be seen whether the approximately linear trend of Fig. 5b will hold for larger Ra , although the present range will be typical of many applications, since the initial roughness of 5 m was quite high, resulting from blasting with relatively large particles at high pressure. Recall that the data of Figs. 1–3 showed transient periods of decreasing roughness starting from an initially rough surface with
Ra = 5.0 m. In contrast, when blasting began on a smooth surface (as-received Ra = 0.05 m), a transient period could not be recorded for any of the blasting conditions in Figs. 1–3, even at the lowest dose corresponding to a scanning speed of Vs = 20 mm/s; i.e. the steady-state roughness was reached in a single pass of the nozzle. This implies that both the particle dose and the normal kinetic energy density required to reach the steady-state roughness depend not only on the processing conditions, but also strongly on the initial surface condition (roughness and waviness) of the target. This contrasts with the erosion rate, which was almost the same when blasting on a smooth surface (as-received glass) or on a rough surface (Figs. 1–3). For example, when the scanning speed was 2 mm/s, the average erosion rate of channels machined on a smooth, flat surface for all the conditions of Figs. 1–3 was 2.9 mg/g while it was 3.3 mg/g when starting from an eroded surface (reference channel). This difference of less than 14% was not statistically significant, confirming that the erosion rate was not a function of the initial surface condition. 4.2. Numerical simulation Fig. 6 illustrates a section of the scanned initial centerline profile of a reference channel and the predicted profile after post-blasting with 100 m alumina particles at P = 100 kPa, = 90◦ and with a particle dose of 104 impacts/mm2 , which is equal to a normal energy density of 8.8 × 10−2 J/mm2 and corresponds to the onset of steady state (Fig. 3). It is seen that the model has preserved realistic
2260
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261
Table 4 Average roughness, Ra , root-mean-square roughness, Rq , skewness, Rsk , kurtosis, Rku , and waviness, Wa , of the initial glass reference channel centerline profile and the measured and predicted profiles at the onset of steady-state for different blasting conditions. (%) is the percentage error between the predicted and measured values. Profile
Blasting condition
Initial measured profile Smoothed profiles Measured
Predicted
Ra (m) (%)
Rq (m) (%)
Rsk
Rku (%)
Wa (m) (%)
90
5.0
6.1
−0.2
2.8
4.5
60 30 90 60 30 90 90 60 30 90 60 30 90 90
4.4 3.3 3.4 3.0 2.7 3.1 2.0 4.2 (−4) 3.2 (3) 3.7 (9) 3.5 (17) 3.0 (11) 3.2 (3) 2.4 (20)
5.3 4.1 4.3 3.8 3.7 4.0 2.6 5.0 (−6) 4.4 (7) 4.6 (7) 4.3 (13) 3.8 (3) 3.9 (−2) 3.3 (−34)
−0.3 0.1 −0.3 −0.1 −0.3 −0.2 0.0 0.4 0.4 0.2 −0.1 0.3 −0.1 0.2
2.8 3.4 3.1 3.3 3.5 3.2 2.9 2.9 (4) 3.0 (−12) 2.7 (−13) 2.8 (−15) 3.0 (−14) 2.6 (−19) 3.0 (3)
4.5 3.5 4.4 2.7 2.7 3.3 3.7 4.2 (−7) 4.2 (20) 3.9 (−11) 4.2 (55) 4.1 (52) 2.7 (−18) 3.7 (−11)
Particle size (m)
Pressure (kPa)
Nozzle angle (◦ )
150
200
150 150 150 150 150 100 50 150 150 150 150 150 100 50
200 200 100 100 100 100 100 200 200 100 100 100 100 100
small-scale roughness features while simulating the decrease in Ra . Table 4 gives measured profile parameters of the initial reference channel along with the measured and predicted profile parameters for this case. Also included in Table 4 are the measured and predicted profile parameters for the smoothed channels at the onset of steady state for the conditions of Figs. 1–3. The predicted parameters after smoothing to a steady-state value of Ra agreed quite well with the measured values, with a maximum error of 20% (Table 4). Both the roughness and waviness decreased during the post-blasting, while the change in the measured and predicted skewness (Rsk ) and kurtosis (Rku ) was insignificant. The predicted and measured skewness were effectively zero, indicating that the surface profile shape was essentially symmetric relative to the surface mean; i.e. the shapes of the valleys and peaks were the same. The predicted and measured kurtosis values were close to 3, indicating a Gaussian distribution of peak heights in the eroded surface profiles. The results of the simulation indicated that the main erosion mechanism was chip removal due to the initiation and propagation of lateral cracks rather than chipping of the edges of existing facets. For example, the percentage of the total erosion events that -125 Initial profile
were edge chipping, ε, was below 7% when 50, 100 and 150 m particles were used as the finishing media at 90◦ . The highest values of ε were seen when small particles impacted relatively large facets at shallow angles where the resistance to edge chipping was lower. This is evident in Fig. 7 which shows ε for 50 m particles at impact angles of 30◦ , 60◦ , 90◦ and 100 kPa as a function of particle dose. During the early stages of smoothing (low doses), a higher fraction of these relatively small particles were able to penetrate into valleys and strike the sides of peaks, creating edge chips and leading to highest ε values for each impact angle. As the particle dose increased, the topography of the channel became smoother with smaller peaks so ε decreased to a steady-state value. At any particle dose, ε was higher for shallower angles since the chipping resistance is lower at shallower angles (Chai and Lawn, 2007). The above discussion explains two key features of Fig. 2b: (1) at a given normal energy density, lower roughnesses were achieved when finishing was performed at shallower angles, because of the higher values of ε at any particle dose at these lower angles (Fig. 7) and (2) the rate of smoothing was higher at the early stages of finishing at shallow angles, because of the decreasing trend of ε with particle dose (Fig. 7). For 100 m particles at 30◦ and 100 kPa, the model predicted similar trends over the same dose range as in Fig. 7, but ε decreased a smaller amount, from 16% to 4%, probably because the larger particles could not penetrate as far into valleys to strike the sides of peaks. These trends were continued with the 150 m particles at
-150
30º 35
60º
30
Smoothed profile
-200
90º
25 (%)
Depth ( m)
40
-175
20 15
-225
10 5
-250 0
1000
2000
3000
4000
Length ( m)
0 0
50
100
150
200
Particle dose (impacts/mm2)×103 Fig. 6. The scanned centerline profile of a reference channel used as the initial profile in the simulation and the predicted smoothed profile after finishing with 100 m alumina particles at 100 kPa, 90◦ and with a dose of 104 impacts/mm2 .
Fig. 7. Percentage of edge chipping events (ε) as a function of particle dose of 50 m alumina particles at 100 kPa.
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261
100 kPa (conditions of Fig. 2), where the model predicted that ε decreased from 12% to 3% for 30◦ and from 7% to 2% at 60◦ . Thus, in all cases, the prevalence of edge chipping decreased with dose, particle size and increasing nozzle angle. As an example of typical results of the numerical simulation, Figs. 1a and 3a illustrate the predicted roughness and depth of the reference channels as a function of particle dose for different post-blasting conditions. In all cases, the actual trends of roughness and depth variation and also the effects of particle size and impact angle on the dose required to reach the steady-state were successfully predicted. The average error for the prediction of all transient and steady-state roughnesses in Figs. 1–3 was 7% while the change in the centerline depth during smoothing for all the conditions of Figs. 1–3 was predicted with an average error of approximately 19%. 5. Conclusions It was found that a practical and effective method to reduce the surface roughness of AJM channels is post-blasting with a particle jet of low kinetic energy. Three methods to reduce the jet energy (lowering the pressure and the impact angle, and using smaller particles) were investigated and the effect of post-blasting particle size, velocity, dose, and angle of attack on the resulting reduction of roughness of reference channels in a borosilicate glass was demonstrated. Using post-blasting, the roughness of a reference channel could be decreased by up to 60% of its initial value. It was demonstrated that the erosion is a function of normal energy density alone. The transient roughness, however, is also affected by the impact angle, with shallower impact angles resulting in a lower transient roughness at a given normal energy density. This was demonstrated to be due to the increased number of chipped edges at lower angles, especially at the early stages of smoothing where the highest rate of roughness reduction per normal energy density was observed. Although a lower ultimate steady-state roughness could be achieved using smaller particles, a higher normal energy density was required to achieve it. To simulate the post-blasting process, a numerical model was used to predict the depth increase and the transient roughness leading to steady-state as a function of particle size, velocity, dose, and impact angle. The results showed that, although the number of
2261
edge chipping events increased with decreasing the angle of attack, the main damage mechanism was crater removal even at shallow angles. The model could accurately predict the channel depth and profile parameters such as roughness, skewness, kurtosis, and waviness of the transient and steady state profiles. Acknowledgments The authors would like to acknowledge the Natural Sciences and Engineering Research Council of Canada (NSERC), the Canada Research Chairs Program, Micralyne Inc. and BIC Fuel Cells for financial and technical support. References Chai, H., Lawn, B., 2007. Edge chipping of brittle materials: effect of side-wall inclination and loading angle. International Journal of Fracture 145, 159–165. Dehnadfar, D., Friedman, J., Papini, M., 2012. Laser shadowgraphy measurements of abrasive particle spatial, size and velocity distributions through micro-masks used in abrasive jet micro-machining. Journal of Materials Processing Technology 212, 137–149. Ghobeity, A., Krajac, T., Burzynski, T., Papini, M., Spelt, J.K., 2008. Surface evolution models in abrasive jet micromachining. Wear 264, 185–198. Ghobeity, A., Crabtree, H.J., Papini, M., Spelt, J.K., 2012. Characterisation and comparison of microfluidic chips formed using abrasive jet micromachining and wet etching. Journal of Micromechanics and Microengineering 22, 025014 (10 pp.). Jafar, R.H.M., Spelt, J.K., Papini, M., 2013a. Surface roughness and erosion rate of abrasive jet micro-machined channels: experiments and analytical model. Wear 303, 138–145. Jafar, R.H.M., Spelt, J.K., Papini, M., 2013b. Numerical simulation of surface roughness and erosion rate of abrasive jet micro-machined channels. Wear 303, 302–312. Ladouceur, F., 1997. Roughness, inhomogeneity, and integrated optics. Journal of Lightwave Technology 15, 1020–1025. Marshall, D.B., Lawn, B.R., Evans, A.G., 1982. Elastic/plastic indentation damage in ceramics: the lateral crack system. Journal of American Ceramic Society 65, 561–566. Mineta, T., Takada, T., Makino, E., Kawashima, T., Shibata, T., 2009. A wet abrasive blasting process for smooth micromachining of glass by ductile-mode removal. Journal of Micromechanics and Microengineering 19, 15031–15038. Slikkerveer, P.J., Bouten, P.C.P., Veld, F.H., Scholten, H., 1998. Erosion and damage by sharp particles. Wear 217, 237–250. Solignac, D., Sayah, A., Constantin, S., Freitag, R., Gijs, M.A.M., 2001. Powder blasting for the realization of microchips for bio-analytic applications. Sensors and Actuators A 92, 388–393. Wensink, H., Schlautmann, S., Goedbloed, M.H., Elwenspoek, M.C., 2002. Fine tuning the roughness of powder blasted surfaces. Journal of Micromechanics and Microengineering 12, 616–620.
Contents lists available at ScienceDirect
Journal of Materials Processing Technology journal homepage: www.elsevier.com/locate/jmatprotec
Simulation of erosive smoothing in the abrasive jet micro-machining of glass R. Haj Mohammad Jafar a , M. Papini b,a,∗ , J.K. Spelt a,b,∗∗ a b
Department of Mechanical and Industrial Engineering, Ryerson University, 350 Victoria Street, Toronto, ON, Canada M5B 2K3 Department of Mechanical and Industrial Engineering, Ryerson University, 350 Victoria Street, Toronto, ON, Canada M5B 2K3
a r t i c l e
i n f o
Article history: Received 12 May 2013 Received in revised form 23 June 2013 Accepted 26 June 2013 Available online 4 July 2013 Keywords: Abrasive jet micro-machining Surface roughness Smoothing Glass
a b s t r a c t Abrasive jet micro-machining (AJM) is a promising technique to machine micro-features in brittle and ductile materials. However, the roughness of micro-channels machined using AJM is generally greater than that from other methods of micro-machining such as wet etching. Previous investigators have suggested that the surface roughness resulting from AJM can be reduced by post-blasting with abrasive particles at a relatively low kinetic energy. This approach was investigated in the present work by measuring the roughness reduction of a reference unmasked channel in borosilicate glass as a function of post-blasting particle size, velocity, dose, and impact angle. Post-blasting the reference channels reduced the roughness by up to 60%. It was observed that post-blasting at shallower angles was more efficient, probably due to the increased amount of edge chipping as opposed to cratering, which contributed to the enhanced removal of profile peaks, leaving a smoother surface. Moreover, post-blasting with smaller particles ultimately resulted in smoother surfaces, but at the penalty of requiring a relatively large particle dose, and consequently a significantly increased channel depth, before reaching the steady-state roughness. Hence, finishing with smaller particles until reaching the steady-state roughness may not be practical when a shallow channel is desired. A previously developed numerical model was modified and used to simulate the post-blasting process leading to the creation of smooth channels as a function of particle size, velocity, dose, impact angle, and target material properties. The model simulated both crater formation (due to growth of lateral cracks) and the chipping of facet edges. Comparisons with centerline roughness measurements for channels in borosilicate glass showed that the model can predict the transient roughness reduction with post-blasting particle dose with a 7% average error. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Abrasive jet micro-machining (AJM) utilizes an abrasive particle jet to mechanically etch micron-sized features into various materials for micro-systems fabrication. A possible disadvantage of AJM in micro-fluidic applications is the relatively high resulting surface roughness that can affect fluid flow phenomena. Ghobeity et al. (2012) compared the separation efficiency of glass channels machined using AJM and wet etching and found that the separation efficiency in AJM channels (Ra ∼ 0.4–0.6 m) was significantly lower (0.2–0.25 times) than that in wet-etched channels
∗ Corresponding author at: Department of Mechanical and Industrial Engineering, Ryerson University, 350 Victoria Street, Toronto, ON, Canada M5B 2K3. Tel.: +1 416 979 5000x7655; fax: +1 416 979 5265. ∗∗ Corresponding author at: Department of Mechanical and Industrial Engineering, University of Toronto, King’s College Road, Toronto, ON, Canada M5S 3G8. Tel.: +1 416 978 5435; fax: +1 416 978 7753. E-mail addresses: [email protected] (M. Papini), [email protected] (J.K. Spelt). 0924-0136/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jmatprotec.2013.06.022
(Ra ∼ 2–5 nm). Similarly, Solignac et al. (2001) observed that the electro-osmotic mobility in AJM channels in soda-lime glass is less than 50% of that in HF-etched channels. Moreover, Ladouceur (1997) has shown that roughness is a constraining factor in optoelectronics devices where it scatters light and attenuates power. Therefore, methods to reduce roughness resulting from AJM operations are highly desirable. The mechanics of cracking in brittle materials due to Vickers indentation was investigated by Marshall et al. (1982) who derived equations to predict the length and depth of lateral cracks as a function of the material properties and indentation force. Utilizing fracture mechanics models of Vickers indentations (Marshall et al., 1982), Slikkerveer et al. (1998) estimated the erosion rate and steady-state roughness of channels in borosilicate glass by assuming that each particle impact removed a spherical cap of material with a radius equal to that of the predicted lateral crack length and a depth equal to that of the plastic zone radius and by assuming that there was no overlap among impact sites. It was concluded that the only important parameter affecting the roughness was the kinetic energy of the impinging particles associated with the velocity
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261
2255
Table 1 Mechanical properties of the borosilicate glass targets. Elastic modulus (GPa)
Fracture toughness (MPa
63
0.76
√
m)
component normal to the target. The present authors investigated the accuracy of this model by measuring the dimensions of individual impact sites resulting from the AJM of borosilicate glass by using aluminum oxide particles of various sizes (Jafar et al., 2013a). It was found that lateral crack initiation was better approximated as originating from the indentation depth rather than the bottom of the plastic zone, as was assumed in the model of Slikkerveer et al. (1998). Roughness measurements for various impact angles also showed that Ra increased with tangential velocity component; i.e. Ra was not a function of only the kinetic energy due to the normal velocity. This observation was not explored further in Jafar et al. (2013a) where the analytical model was compared only with steady-state roughness measurements made at 90◦ impact angle. The present authors (Jafar et al., 2013b) also developed a numerical model of the erosion process that was used to predict the steady-state roughness and erosion rate of unmasked channels resulting from AJM on borosilicate glass as a function of particle size, velocity, dose, and impact angle. It was assumed that two brittle damage mechanisms contributed to erosion: (1) crater removal due to the initiation and growth of lateral cracks and (2) edge chipping due to the propagation of cracks through a facet in the direction of the impact. The model predicted the steady-state roughness and the erosion rate of unmasked channels machined in borosilcate glass with average errors of 9% and 29% respectively, when compared with experimental data over a wide range of machining conditions. Moreover, the model predicted profile shape parameters such as waviness, skewness, and kurtosis with less than 15% error compared with experimental data. The results showed that, although the number of edge chipping events (the second erosion mechanism) increased with decreasing angle of attack, the main damage mechanism was crater removal, even at shallow angles. The roughness of a typical borosilicate glass channel machined with AJM depends strongly on the particle kinetic energy, varying from 0.1 m to 9 m using alumina particles with average sizes from 5 m to 200 m, respectively (Slikkerveer et al., 1998). The smoothing of glass surfaces after AJM has not been investigated extensively in literature. Wensink et al. (2002) achieved a 100% reduction in the roughness of two reference channels machined in borosilicate glass using either 9 m or 29 m alumina by annealing the eroded glass at a temperature just below its softening point. In contrast, a post-blast wet etching of the reference channels with hydrofluoric acid produced a rougher surface, probably because of the opening of cracks generated by the AJM. Post-blasting the reference channels with 3 m alumina decreased the roughness of both reference channels by about 25% while the depth of the channels increased approximately 10 m. Mineta et al. (2009) examined the reduction of roughness due to post-blasting using a wet abrasive on a borosilicate glass (PyrexTM ) substrate. Both the original and the post-blasted channels were machined using an aqueous abrasive slurry of 4 m alumina, however the later were machined
Vickers hardness (GPa)
Density (g/cm3 )
Poisson’s ratio
5.2
2.2
0.2
with a lower pressure so that a ductile-mode erosion was dominant, resulting in a roughness reduction of about 50%. The present work investigated the role of post-blasting particle size, velocity, dose, and impact angle on roughness reduction of unmasked channels machined in borosilicate glass using AJM. A numerical model, which was previously developed to predict the steady-state roughness resulting from the AJM of an initially un-eroded surface (Jafar et al., 2013b), was used to predict both the transient and steady-state roughnesses resulting from postblasting smoothing operations. 2. Experiments 2.1. Apparatus and target material The experiments were conducted using an AccuFlo abrasive blaster from Comco Inc. (Burbank, CA, USA) with a blasting nozzle having an inner diameter of 1.5 mm that was held stationary at a nozzle-to-surface stand-off distance of 10 mm (centerline distance between the nozzle exit and the target). The target material was 3 mm thick Borofloat® 33 borosilicate glass (Schott Inc., NY, USA) with the mechanical properties given in Table 1. To create unmasked channels, the glass specimen was clamped to a computer-controlled stage and moved across the blast zone of a stationary nozzle at a scan speed, Vs , of 0.25–20 mm/s using various angles of nozzle inclination, measured with respect to the target surface, . Varying the scan speed provided a convenient way of changing the particle dose. The surface roughnesses were reported as the average of three repeat measurements along a 5 mm length of the centerline of the machined channels using an optical profilometer (Nanovea Inc., Irvine, CA) with a 0.2 m step size and a cut-off length of 800 m. A more detailed discussion of the roughness measurements can be found in Jafar et al. (2013a). To provide a baseline channel roughness from which various post-blasting scenarios could be evaluated, unmasked reference channels were machined at normal incidence (i.e. = 90◦ ), using 150 m alumina particles at a scan speed of 4 mm/s and a pressure of 200 kPa. The average depth of three machined reference channels measured with the optical profilometer was 165 m, with a standard deviation of 7.6 m, while the average width was approximately 3 mm. The steady-state roughness of these same reference channels was Ra = 5.0 m, with a standard deviation for three channels of 0.27 m. The reference channels were then post-blasted with alumina particles of various sizes (50, 100 and 150 m) at pressures of 100 and 200 kPa, and impact angles of 30◦ , 60◦ and 90◦ . The standard deviations over three measurements of the centerline roughness and centerline depth (or erosion rate) of the post-blasted channels were less than 10% of the mean values. Table 2 summarizes the AJM process parameters used in the machining of the reference and the post-blasted channels.
Table 2 Process parameters used to machine the reference and the post-blasted channels. Channel
Particle size (m)
Pressure (kPa)
Impact angle (◦ )
Scanning speed (mm/s)
Stand-off distance (mm)
Reference Post-blasted
150 50, 100, 150
200 100, 200
90 30, 60, 90
4 0.25–20
10 10
2256
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261
Table 3 Measured particle mean diameter, jet focus coefficients, particle velocities at the jet center 10 mm from the nozzle exit, and mass flow rates. Velocity range corresponds to 95% of population (Jafar et al., 2013b). Nominal particle size (m)
Mean diameter (m)
Pressure (kPa)
ˇw
Velocity range (m/s)
Mean velocity (m/s)
Mass flow rate (g/min)
50 100 150 150
41 138 182 182
100 100 100 200
10.9 13.9 15.3 9.6
56–107 45–86 45–66 62–97
80 57 52 78
7.2 9.8 15.8 22.3
2.2. Particle size, velocity and mass flux characterization The particle size, velocity and mass flux distributions were measured in a previous work (Jafar et al., 2013b), and thus only a short summary has been given here. The size distributions of the alumina powders were determined using an automated image analysis system (Clemex Vision PE, Clemex Technologies Inc., QC, Canada) and fitted with either a log-normal distribution for the 50 m particles, or a Weibull distribution for the 100 and 150 m particles. The mean diameter of the particles is given in Table 3. The distributions of particle velocity and spatial density as a function of radius from the jet axis, at pressures of 100 and 200 kPa and 10 mm from the nozzle exit, were measured using high resolution imaging with laser-pulsed backlit illumination called shadowgraphy (Dehnadfar et al., 2012). The alumina mass flow rates were measured by weighing the mass of abrasive particles blasted for 2 min into a closed container having a filter at the end to permit the air to escape. Table 3 gives these data for the abrasive jets along with the values of the jet focus coefficient, ˇw , which is a measure of jet divergence, with higher values indicating a more focused stream (Ghobeity et al., 2008).
3. Erosion simulation A full description of the model is presented in Jafar et al. (2013b) where it was used to predict the steady-state roughness of channels machined by an abrasive jet on an initially flat brittle target. In the present work, the model was adapted to simulate the transient development of the roughness of a pre-eroded profile as it approached a steady-state topography. This smoothing process was modeled as a function of particle size, velocity, dose, and impact angle. The simulation was implemented in MATLABTM incorporating the two erosion mechanisms of crater formation and edge chipping (Jafar et al., 2013b). The measured centerline profile of a 5 mm long reference channel was entered into the model as the initial profile, along with the desired total number of impacts and the nozzle inclination angle. Other inputs were the measured distributions of particle size and velocity, as well as the particle spatial distributions. To simulate the erosion of the centerline of the reference channel, impacts were randomly located over an area of 5 mm long and 6dnom wide (6 times the nominal particle diameter) until a portion of the particle projected area crossed the strip centerline. The model then examined the possibility of both crater removal and edge chipping occurring in the profile facet that was in contact with the impacting particle. In the case of a crater removal event, a V-shaped crater was formed with a depth equal to that of the predicted indentation and a radius equal to the lateral crack length, both of which are functions of particle kinetic energy associated with the normal velocity component and target material properties (Slikkerveer et al., 1998). The crater was formed in the profile only if the lateral crack was predicted to reach the surface of the profile. Edge chipping was examined using a 2D model developed by Chai and Lawn (2007), in which the edge chip geometry is a function of the plane-strain fracture toughness (Kc ), the indentation distance from the edge, the indentation force magnitude and angle, and the edge geometry. If the particle impact force was lower
than that required for both chipping and cratering, it was assumed that the impacting particle did not damage the surface. Once the required total number of impacts was reached and the final surface profile was obtained, the average channel centerline depth was calculated along with the roughness, Ra , employing the same 800 m cut-off length that was used in the experimental measurements. 4. Results and discussion 4.1. Experimental results Previous studies have shown that although the tangential component of velocity can affect the surface roughness of channels machined using AJM (Jafar et al., 2013a), the roughness most strongly depends on the ‘normal component’ kinetic energy; i.e. that associated with the component of particle velocity perpendicular to the target (Slikkerveer et al., 1998). This implies that the roughness of an AJM channel can be reduced by blasting it with a jet having a lower normal component kinetic energy, achieved using shallow impact angles, low blasting pressures, and/or smaller particles. The following describes the effect that each of these process parameters had on the smoothing of the reference channels. The results of the numerical simulation are then described in Section 4.2. 4.1.1. Impact angle To investigate the effect of the impact angle on roughness reduction, the reference channels were finished (i.e. ‘post-blasted’) using the same particles and pressure (150 m, 200 kPa) that were used to make the reference channel, but at impact angles of 30◦ and 60◦ to the surface, which decreased the normal component of kinetic energy relative to the original 90◦ blasting by 75% and 25%, respectively. Fig. 1a shows the measured and predicted decrease in surface roughness and the increase in channel depth with particle dose, and Fig. 1b depicts the measured roughness reduction and the depth increase as a function of “normal energy density” (i.e. product of the number of impacts and the normal component of kinetic energy per unit area) of the post-blasted particles. In each case, the starting roughness was that of the reference channel (Ra = 5 m). Also shown on each graph is the steady-state roughness achieved on an initially smooth glass surface eroded under the same conditions used for post-blasting. With increasing particle dose (Fig. 1a) or normal energy density (Fig. 1b), the channel depth increased almost linearly (slight curve with log scale) as expected for a constant erosion rate. The R2 of the least-squares linear fits for 30◦ and 60◦ were 0.99 and 0.98, respectively. At 30◦ , a gradual decrease in roughness was observed with increasing dose leading to a steady-state roughness at approximately 2.51 × 103 impacts/mm2 or 2.35 × 10−2 J/mm2 . The depth increase exhibited the same trend at 60◦ , but the roughness transient period was very much shorter. Several other observations can be made from Fig. 1a: (1) the channels at a given dose were deeper at 60◦ than 30◦ ; (2) the steady-state roughness at 30◦ was lower than that at 60◦ ; (3) a higher dose was needed to reach the steady-state roughness at a shallower angle of attack. These are all attributed to the decrease in kinetic energy due to the impact velocity
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261
Measured depth increase (30º)
Predicted roughness (60º)
Predicted depth increase (60º)
Predicted roughness (30º)
Predicted depth increase (30º)
4.5
160
5
140
4.5
60
Ra (µm)
4
120
3.5
100
30
3
80
2.5
2
1.5
Measured roughness (30º)
Measured depth increase (30º)
Measured roughness (60º)
Measured depth increase (60º)
180
5.5
180
160
60
140
120
4
Ra (µm)
5
Measured roughness (30º)
Depth increase (µm)
5.5
Measured depth increase (60º)
200
(b) 6
200 Measured roughness (60º)
3.5
100
30
3
80
60
2.5
60
40
2
40
20
1.5
20
150 m, P=200 kPa 1 1
10
100
0 1000
Particle dose (impacts/mm2)×102
Depth increase (µm)
(a) 6
2257
150 m, P=200 kPa 1 1
10
100
0 1000
Normal kinetic energy density (J/mm2)×10-3
Fig. 1. Variation of roughness and depth of a post-blasted borosilicate glass channel with (a) particle dose and (b) normal kinetic energy density (log scale) for 150 m alumina particles at P = 200 kPa and = 30◦ , 60◦ . Horizontal lines indicate the steady-state roughness achieved by blasting a smooth glass surface under the same conditions as those used in post-blasting; solid lines are measured values and dashed lines are model predictions.
normal to the surface at shallower impact angles, which leads to the removal of smaller chips. This in turn causes a lower erosion rate and steady-state roughness. The depth changes (erosion) for both impact angles in Fig. 1b roughly follow a single curve, confirming that the erosion rate is a function of particle normal kinetic energy. The predictions of roughness and depth will be discussed in Section 4.2. 4.1.2. Blasting pressure The effect of the pressure and hence velocity on roughness reduction and depth increase is shown in Fig. 2a and b, as the original channels made with150 m alumina at 200 kPa and 90◦ were post blasted using the same particles but at 100 kPa and at impact angles of 30◦ , 60◦ and 90◦ . Reducing the blasting pressure from 200 to 100 kPa resulted in a jet with 56% less kinetic energy. The trends of Fig. 2a are similar to those in Fig. 1a, with the dose corresponding to the onset of steady-state roughness increasing as the kinetic energy due to the normal velocity decreased, leading to removal of smaller chips per impact. A comparison between Figs. 1a and 2a shows that this also led to smoother surfaces; e.g. at 30◦ the steadystate Ra was approximately 2.7 m, at 100 kPa while it was 3.3 m at 200 kPa. Fig. 2b shows many of the same trends seen in Fig. 1b: (1) the depth increase for all angles followed approximately a single curve, indicating that the erosion was only function of the normal kinetic energy; (2) during the transient period, a lower roughness was achieved at a given normal energy density when post-blasting at shallower angles, which was probably due to the increased amount of edge chipping as opposed to cratering (chipping percentage) at lower angles, especially during the early stages of smoothing where the rate of smoothing (the roughness reduction per normal energy density) was higher, which contributed to the enhanced removal of profile peaks, leaving a smoother surface. This is supported by the absence of an increased initial smoothing rate in Fig. 3b (discussed below), which shows additional data at 90◦ where little or no edge chipping occurs (this hypothesis will be further discussed in Section 4.2); (3) the normal energy density required to reach steady state increased as the angle decreased; and (4) the channel depth corresponding to the onset of steady-state increased as the post-blasting impact angle decreased; e.g. the depth increase for
90◦ was only approximately 10 m, for 60◦ it was 35 m, and for 30◦ it was 65 m. The reason for observations (3) and (4) is simply that, since the steady-state roughness at shallower angles is lower, the amount of required smoothing is greater; thus a greater normal kinetic energy density is required in order to reach the steady-state, which consequently increases the channel depth. 4.1.3. Particle size Fig. 3a shows the changes in roughness and depth with dose during post-blasting of the reference channels with 50 m, 100 m and 150 m alumina at 100 kPa and 90◦ . As expected, smoother surfaces were obtained at steady state with the 50 m particles, because of their lower normal kinetic energy compared with 100 and 150 m particles. As in Figs. 1a and 2a, this also meant that a higher dose was required to reach steady-state. For example, the dose to reach steady state with the 50 m particles was about 45 times more than with the 100 m particles and about 450 times more than that with the 150 m particles (Fig. 3a). As discussed below, these large differences were attributable to a reduced normal velocity. Fig. 3b shows that, as expected, the normal kinetic energy density to achieve a steady-state roughness increased when postblasting with smaller particles. For example, at P = 100 kPa and = 90◦ , the required energy density to reach steady state with the 50 m particles was about 2.3 times more than with the 100 m particles and 12 times more than that with the 150 m particles (Fig. 3b). This was simply because the steady-state roughness was lower with smaller particles, and so a greater amount of smoothing was required to reach the steady-state; i.e. a higher normal kinetic energy density was required. Fig. 3a and b shows that at a steady-state roughness, more material was removed from the reference channel when smaller particles were used; e.g. at the beginning of the steady state, the depth increase of the reference channels blasted at 90◦ was 11 m for 150 m particles, 125 m for 100 m particles, and 215 m for 50 m particles (Fig. 3a and b). Thus, post-blasting until reaching the steady-state roughness using small particles may not be practical when a shallow channel is required. To aid in a comparison among all the studied post-blasting conditions, Fig. 4a and b shows the depth increase and roughness,
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261 6
(a)
200 Measured depth increase (90º)
Measured roughness (60º)
Measured depth increase (60º)
Measured roughness (30º)
Measured depth increase (30º)
200
(b) 6
180
5.5
Measured roughness (90º)
Measured depth increase (90º)
Measured roughness (60º)
Measured depth increase (60º)
Measured roughness (30º)
Measured depth increase (30º)
180
160
5
160
4.5
140
4.5
140
4
120
4
120
90
3.5
100
60
3
80
30
Ra (µm)
5
Depth increase (µm)
Ra (µm)
5.5
Measured roughness (90º)
3.5
100
90 60
3
80
30
2.5
60
2.5
60
2
40
2
40
20
1.5
1.5
20 150 m, P=100 kPa
150 m, P=100 kPa 0 1000
1 10
1
100
Depth increase (µm)
2258
0 1000
1 1
10
Particle dose (impacts/mm2)×102
100
Normal kinetic energy density (J/mm2)×10-3
Fig. 2. Variation of roughness and depth of a post-blasted borosilicate glass channel with (a) particle dose and (b) normal kinetic energy density (log scale) for 150 m alumina particles at P = 100 kPa and = 30◦ , 60◦ , 90◦ . Horizontal lines indicate the measured steady-state roughness achieved by blasting a smooth glass surface under the same conditions as those used in post-blasting.
respectively, for all the conditions of Figs. 1–3 as a function of normal energy density and Fig. 4c shows the roughness versus the depth increase. Fig. 4a confirms that the erosion rate is only a function of the normal energy density. However, Fig. 4b shows that the transient roughness behavior is affected by the impact angle so that post-blasting is generally more efficient when performed at lower angles for a given particle size and pressure; i.e. less normal kinetic energy density is required to reach a given roughness at shallower angles. Although a unique trend cannot be seen in Fig. 4c, it may be concluded that for a given channel depth generally using the smallest particle available at the lowest pressure and impact angle possible could result in a lower transient roughness.
700
5.5
Measured roughness (150 µm)
Measured depth increase (150 µm)
Measured roughness (100 µm)
Measured depth increase (100 µm)
Measured roughness (50 µm)
Measured depth increase (50 µm)
Predicted roughness (100 µm)
5
300
2.5
200 50 µm
2
Measured roughness (100 µm)
Measured depth increase (100 µm)
Measured roughness (50 µm)
Measured depth increase (50 µm)
500
4
Ra (µm)
100 µm
Measured depth increase (150 µm)
4.5
Depth increase (µm)
400
3
Measured roughness (150 µm)
600 5
4 150 µm
700
5.5
500
3.5
6
600
Predicted depth increase (100 µm)
4.5
Ra (µm)
(b)
400 3.5
150 µm
3
100 µm
300
2.5
Depth increase (µm)
(a) 6
Fig. 5a shows the total roughness reduction that is possible and the per particle normal kinetic energy versus the normal energy density required to reach the steady-state roughness, es , for all the studied post-blasting conditions (except 150 m particles at 200 kPa and 60◦ where steady-state occurred almost immediately). It is seen that when post-blasting with a particle of lower normal kinetic energy, Un , the roughness reduction (Ra ) is higher, but a greater normal energy density is required to reach steadystate. Fig. 5b depicts the total roughness reduction versus the increase in depth at the steady-state, ds , showing that to obtain a greater roughness reduction more material must be removed from the surface. Fig. 5 can be used as a guide to find a post-blasting
200 50 µm
2
100
100 1.5
1.5 P =100 kPa,
1 1
10
100
1000
Particle dose (impacts/mm2)×102
=90° 10000
0
P =100 kPa,
1 1
10
100
=90° 0 1000
Kinetic energy density (J/mm2)×10-3
Fig. 3. Variation of roughness and depth of a post-blasted borosilicate glass channel with (a) particle dose and (b) normal kinetic energy density (log scale) for 50 m, 100 m and 150 m alumina particles blasted at P = 100 kPa and = 90◦ . Horizontal lines indicate the steady-state roughness achieved by blasting a smooth glass surface under the same conditions as those used in post-blasting; solid lines are measured values and the dashed line is model prediction for 100 m particles.
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261
50
(a) 4 Roughness reduction
50 µm-100 kPa-90º
450
3.5
100 µm-100 kPa-90º
400
40
150 µm-100 kPa-90º 350
45
Per particle kinetic energy
3 35
150 µm-100 kPa-60º
300
2.5
250
150 µm-200 kPa-60º
200
150 µm-200 kPa-30º
Ra (µm)
150 µm-100 kPa-30º
150
30 Ra = 0.0062es + 1.65 R² = 0.79
2
25 20
1.5
15
100 1
50
Un = -0.0658es + 13.44 R² = 0.65
0.5
0 1
10
100
10 5
1000
0
0
Kinetic energy density (J/mm2)×10-3
Per particle kinetic energy, Un (µJ)
(a) 500
Depth increase (µm)
2259
0
50
100
150
200
250
Normal kinetic energy density to reach steady-state, es (J/mm2)×10-3
(b) 5
(b) 4
4.5
3.5 4
3
Ra = 0.0053ds + 1.66 R² = 0.65
2.5
50 µm-100 kPa-90º
3
Ra (µm)
Ra (µm)
3.5
100 µm-100 kPa-90º 2.5
150 µm-100 kPa-90º 150 µm-100 kPa-60º
2
2 1.5
150 µm-100 kPa-30º
1
150 µm-200 kPa-60º
1.5
150 µm-200 kPa-30º
0.5
1 1
10
100
1000
0
Kinetic energy density (J/mm2)×10-3 (c)
0
5
50
100
150
200
250
Depth increase at steady-state, ds (µm)
4.5
Fig. 5. Measured roughness reduction versus (a) normal energy density to reach steady-state roughness for a given post-blasting particle normal kinetic energy and (b) depth increase at steady state.
4
Ra (µm)
3.5 3
50 µm-100 kPa-90º 100 µm-100 kPa-90º
2.5
150 µm-100 kPa-90º 150 µm-100 kPa-60º
2
150 µm-100 kPa-30º 1.5
150 µm-200 kPa-60º 150 µm-200 kPa-30º
1 1
10
100
1000
Depth increase (µm) Fig. 4. (a) Depth increase and (b) roughness versus normal energy density, and (c) roughness versus depth increase for all post-blasting conditions of Figs. 1–3.
condition to achieve the minimum possible steady-state roughness for a given depth increase. For example, the ultimate roughness reduction for a given depth increase can be found using Fig. 5b, and the corresponding normal kinetic energy density follows from the upper curve of Fig. 5a. The lower curve of Fig. 5a then gives the required kinetic energy per particle which then can be used to determine the post-blasting condition (particle size, pressure and impact angle). It remains to be seen whether the approximately linear trend of Fig. 5b will hold for larger Ra , although the present range will be typical of many applications, since the initial roughness of 5 m was quite high, resulting from blasting with relatively large particles at high pressure. Recall that the data of Figs. 1–3 showed transient periods of decreasing roughness starting from an initially rough surface with
Ra = 5.0 m. In contrast, when blasting began on a smooth surface (as-received Ra = 0.05 m), a transient period could not be recorded for any of the blasting conditions in Figs. 1–3, even at the lowest dose corresponding to a scanning speed of Vs = 20 mm/s; i.e. the steady-state roughness was reached in a single pass of the nozzle. This implies that both the particle dose and the normal kinetic energy density required to reach the steady-state roughness depend not only on the processing conditions, but also strongly on the initial surface condition (roughness and waviness) of the target. This contrasts with the erosion rate, which was almost the same when blasting on a smooth surface (as-received glass) or on a rough surface (Figs. 1–3). For example, when the scanning speed was 2 mm/s, the average erosion rate of channels machined on a smooth, flat surface for all the conditions of Figs. 1–3 was 2.9 mg/g while it was 3.3 mg/g when starting from an eroded surface (reference channel). This difference of less than 14% was not statistically significant, confirming that the erosion rate was not a function of the initial surface condition. 4.2. Numerical simulation Fig. 6 illustrates a section of the scanned initial centerline profile of a reference channel and the predicted profile after post-blasting with 100 m alumina particles at P = 100 kPa, = 90◦ and with a particle dose of 104 impacts/mm2 , which is equal to a normal energy density of 8.8 × 10−2 J/mm2 and corresponds to the onset of steady state (Fig. 3). It is seen that the model has preserved realistic
2260
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261
Table 4 Average roughness, Ra , root-mean-square roughness, Rq , skewness, Rsk , kurtosis, Rku , and waviness, Wa , of the initial glass reference channel centerline profile and the measured and predicted profiles at the onset of steady-state for different blasting conditions. (%) is the percentage error between the predicted and measured values. Profile
Blasting condition
Initial measured profile Smoothed profiles Measured
Predicted
Ra (m) (%)
Rq (m) (%)
Rsk
Rku (%)
Wa (m) (%)
90
5.0
6.1
−0.2
2.8
4.5
60 30 90 60 30 90 90 60 30 90 60 30 90 90
4.4 3.3 3.4 3.0 2.7 3.1 2.0 4.2 (−4) 3.2 (3) 3.7 (9) 3.5 (17) 3.0 (11) 3.2 (3) 2.4 (20)
5.3 4.1 4.3 3.8 3.7 4.0 2.6 5.0 (−6) 4.4 (7) 4.6 (7) 4.3 (13) 3.8 (3) 3.9 (−2) 3.3 (−34)
−0.3 0.1 −0.3 −0.1 −0.3 −0.2 0.0 0.4 0.4 0.2 −0.1 0.3 −0.1 0.2
2.8 3.4 3.1 3.3 3.5 3.2 2.9 2.9 (4) 3.0 (−12) 2.7 (−13) 2.8 (−15) 3.0 (−14) 2.6 (−19) 3.0 (3)
4.5 3.5 4.4 2.7 2.7 3.3 3.7 4.2 (−7) 4.2 (20) 3.9 (−11) 4.2 (55) 4.1 (52) 2.7 (−18) 3.7 (−11)
Particle size (m)
Pressure (kPa)
Nozzle angle (◦ )
150
200
150 150 150 150 150 100 50 150 150 150 150 150 100 50
200 200 100 100 100 100 100 200 200 100 100 100 100 100
small-scale roughness features while simulating the decrease in Ra . Table 4 gives measured profile parameters of the initial reference channel along with the measured and predicted profile parameters for this case. Also included in Table 4 are the measured and predicted profile parameters for the smoothed channels at the onset of steady state for the conditions of Figs. 1–3. The predicted parameters after smoothing to a steady-state value of Ra agreed quite well with the measured values, with a maximum error of 20% (Table 4). Both the roughness and waviness decreased during the post-blasting, while the change in the measured and predicted skewness (Rsk ) and kurtosis (Rku ) was insignificant. The predicted and measured skewness were effectively zero, indicating that the surface profile shape was essentially symmetric relative to the surface mean; i.e. the shapes of the valleys and peaks were the same. The predicted and measured kurtosis values were close to 3, indicating a Gaussian distribution of peak heights in the eroded surface profiles. The results of the simulation indicated that the main erosion mechanism was chip removal due to the initiation and propagation of lateral cracks rather than chipping of the edges of existing facets. For example, the percentage of the total erosion events that -125 Initial profile
were edge chipping, ε, was below 7% when 50, 100 and 150 m particles were used as the finishing media at 90◦ . The highest values of ε were seen when small particles impacted relatively large facets at shallow angles where the resistance to edge chipping was lower. This is evident in Fig. 7 which shows ε for 50 m particles at impact angles of 30◦ , 60◦ , 90◦ and 100 kPa as a function of particle dose. During the early stages of smoothing (low doses), a higher fraction of these relatively small particles were able to penetrate into valleys and strike the sides of peaks, creating edge chips and leading to highest ε values for each impact angle. As the particle dose increased, the topography of the channel became smoother with smaller peaks so ε decreased to a steady-state value. At any particle dose, ε was higher for shallower angles since the chipping resistance is lower at shallower angles (Chai and Lawn, 2007). The above discussion explains two key features of Fig. 2b: (1) at a given normal energy density, lower roughnesses were achieved when finishing was performed at shallower angles, because of the higher values of ε at any particle dose at these lower angles (Fig. 7) and (2) the rate of smoothing was higher at the early stages of finishing at shallow angles, because of the decreasing trend of ε with particle dose (Fig. 7). For 100 m particles at 30◦ and 100 kPa, the model predicted similar trends over the same dose range as in Fig. 7, but ε decreased a smaller amount, from 16% to 4%, probably because the larger particles could not penetrate as far into valleys to strike the sides of peaks. These trends were continued with the 150 m particles at
-150
30º 35
60º
30
Smoothed profile
-200
90º
25 (%)
Depth ( m)
40
-175
20 15
-225
10 5
-250 0
1000
2000
3000
4000
Length ( m)
0 0
50
100
150
200
Particle dose (impacts/mm2)×103 Fig. 6. The scanned centerline profile of a reference channel used as the initial profile in the simulation and the predicted smoothed profile after finishing with 100 m alumina particles at 100 kPa, 90◦ and with a dose of 104 impacts/mm2 .
Fig. 7. Percentage of edge chipping events (ε) as a function of particle dose of 50 m alumina particles at 100 kPa.
R.H.M. Jafar et al. / Journal of Materials Processing Technology 213 (2013) 2254–2261
100 kPa (conditions of Fig. 2), where the model predicted that ε decreased from 12% to 3% for 30◦ and from 7% to 2% at 60◦ . Thus, in all cases, the prevalence of edge chipping decreased with dose, particle size and increasing nozzle angle. As an example of typical results of the numerical simulation, Figs. 1a and 3a illustrate the predicted roughness and depth of the reference channels as a function of particle dose for different post-blasting conditions. In all cases, the actual trends of roughness and depth variation and also the effects of particle size and impact angle on the dose required to reach the steady-state were successfully predicted. The average error for the prediction of all transient and steady-state roughnesses in Figs. 1–3 was 7% while the change in the centerline depth during smoothing for all the conditions of Figs. 1–3 was predicted with an average error of approximately 19%. 5. Conclusions It was found that a practical and effective method to reduce the surface roughness of AJM channels is post-blasting with a particle jet of low kinetic energy. Three methods to reduce the jet energy (lowering the pressure and the impact angle, and using smaller particles) were investigated and the effect of post-blasting particle size, velocity, dose, and angle of attack on the resulting reduction of roughness of reference channels in a borosilicate glass was demonstrated. Using post-blasting, the roughness of a reference channel could be decreased by up to 60% of its initial value. It was demonstrated that the erosion is a function of normal energy density alone. The transient roughness, however, is also affected by the impact angle, with shallower impact angles resulting in a lower transient roughness at a given normal energy density. This was demonstrated to be due to the increased number of chipped edges at lower angles, especially at the early stages of smoothing where the highest rate of roughness reduction per normal energy density was observed. Although a lower ultimate steady-state roughness could be achieved using smaller particles, a higher normal energy density was required to achieve it. To simulate the post-blasting process, a numerical model was used to predict the depth increase and the transient roughness leading to steady-state as a function of particle size, velocity, dose, and impact angle. The results showed that, although the number of
2261
edge chipping events increased with decreasing the angle of attack, the main damage mechanism was crater removal even at shallow angles. The model could accurately predict the channel depth and profile parameters such as roughness, skewness, kurtosis, and waviness of the transient and steady state profiles. Acknowledgments The authors would like to acknowledge the Natural Sciences and Engineering Research Council of Canada (NSERC), the Canada Research Chairs Program, Micralyne Inc. and BIC Fuel Cells for financial and technical support. References Chai, H., Lawn, B., 2007. Edge chipping of brittle materials: effect of side-wall inclination and loading angle. International Journal of Fracture 145, 159–165. Dehnadfar, D., Friedman, J., Papini, M., 2012. Laser shadowgraphy measurements of abrasive particle spatial, size and velocity distributions through micro-masks used in abrasive jet micro-machining. Journal of Materials Processing Technology 212, 137–149. Ghobeity, A., Krajac, T., Burzynski, T., Papini, M., Spelt, J.K., 2008. Surface evolution models in abrasive jet micromachining. Wear 264, 185–198. Ghobeity, A., Crabtree, H.J., Papini, M., Spelt, J.K., 2012. Characterisation and comparison of microfluidic chips formed using abrasive jet micromachining and wet etching. Journal of Micromechanics and Microengineering 22, 025014 (10 pp.). Jafar, R.H.M., Spelt, J.K., Papini, M., 2013a. Surface roughness and erosion rate of abrasive jet micro-machined channels: experiments and analytical model. Wear 303, 138–145. Jafar, R.H.M., Spelt, J.K., Papini, M., 2013b. Numerical simulation of surface roughness and erosion rate of abrasive jet micro-machined channels. Wear 303, 302–312. Ladouceur, F., 1997. Roughness, inhomogeneity, and integrated optics. Journal of Lightwave Technology 15, 1020–1025. Marshall, D.B., Lawn, B.R., Evans, A.G., 1982. Elastic/plastic indentation damage in ceramics: the lateral crack system. Journal of American Ceramic Society 65, 561–566. Mineta, T., Takada, T., Makino, E., Kawashima, T., Shibata, T., 2009. A wet abrasive blasting process for smooth micromachining of glass by ductile-mode removal. Journal of Micromechanics and Microengineering 19, 15031–15038. Slikkerveer, P.J., Bouten, P.C.P., Veld, F.H., Scholten, H., 1998. Erosion and damage by sharp particles. Wear 217, 237–250. Solignac, D., Sayah, A., Constantin, S., Freitag, R., Gijs, M.A.M., 2001. Powder blasting for the realization of microchips for bio-analytic applications. Sensors and Actuators A 92, 388–393. Wensink, H., Schlautmann, S., Goedbloed, M.H., Elwenspoek, M.C., 2002. Fine tuning the roughness of powder blasted surfaces. Journal of Micromechanics and Microengineering 12, 616–620.