Improvement in Sintered Tungsten Carbide Tool Surface Integrity Using Femtosecond Pulse Laser

In this paper, we propose a surface integrity technique referred to as pulse laser grinding (PLG). This method is a combination of laser ablation and averaging processes, such as grinding, and has two characteristic features. The rst feature is that the irradiation area consists of a long-focus lens and a pulse laser with a Gaussian prole, and this method has a laser irradiation area equivalent to that of the grinding wheel. However, owing to the difference in the removal process of PLG and that of conventional grinding, the surface roughness after processing is expected to be different. The second feature is that the workpiece is placed at an approximately parallel angle between the laser axis and the work surface to undergo laser ablation on the surface. The characteristic piece placement restrains laser-specic problems such as debris and redeposition. This study targets sintered tungsten carbide (WC-Co), for which it is particularly dicult to form low-roughness surfaces. Binder removal followed by WC grain detachment caused by conventional grinding contributes to the increase in roughness and deterioration of corner sharpness. The experimental results of PLG with a suciently averaged surface of a WC-Co tool conrmed that the parallel roughness reached an arithmetical mean roughness (Ra) of 0.025 μm, and the perpendicular roughness reached Ra below 0.006 μm, in agreement with the aforementioned considerations.


Introduction
Various nishing techniques have been evaluated and their characteristics have been classi ed because the material nishing processes in uence the product performance [1]. Surface roughness, which re ects geometrical accuracy, is a basic criterion for evaluating these processes. It is important to determine the tribological properties and sharpness of the cutting tool edge comprising the two surfaces. Sharpness improves the cutting performance of the tool and reduces damage to the workpiece [2]. However, highhardness tool materials such as WC, cBN, and diamond are di cult to nish with low roughness owing to the hardness or toughness of the materials. Grinding methods using grindstones or abrasive grains for these materials have been studied. Hegeman et al. reported that, when diamond-grinding wheels were used to grind WC-Co, the arithmetical mean roughness (Ra) was 0.77-0.28 µm [3]. A higher wear resistance of WC-Co was observed with high loads and coarse abradants of a few hundred micrometers [4,5], and binder removal followed by WC grain detachment also contributes to the deterioration of corner sharpness [6]. Mogilnikov et al. reported that the main methods to obtain smoother surfaces involved polishing with diamond powders or electrochemical grinding [7]. However, in both techniques, the polishing time and cost of the abrasives posed inherent limitations.
Pulsed laser ablation can be applied as a new alternative technique to grinding to fabricate surfaces with greater integrity. We proposed pulse laser grinding (PLG), a combination of characteristic laser ablation and averaging processes, such as grinding. The sharpening of the tool edge enhances the cutting performance [8][9][10][11][12]. However, these reports only mention the effects of smoothing and cutting performance on certain tools. In this paper, we clari ed the roughness factors that should be considered for PLGs to process various materials. The WC-Co tool roughness obtained by PLG using a femtosecond laser was experimentally evaluated, parallel and perpendicular to the infeed direction, which demonstrated the advantage of PLG. Based on the experimental results, we identi ed the factors that determine the nished surface roughness.

Motivation for pulse laser processing
A pulse laser with high energy facilitates the processing of materials [13][14][15]. Ultrashort pulse lasers with an instantaneous application of a high laser uence enable ablation without a heat-affected zone on the surface [16]. Low-surface-roughness fabrication and application to hard and di cult-to-process materials with pulse lasers have been investigated. Dumitru et al. con rmed the ablation threshold of ultra-hard metals and reported the fundamentals of laser ablation [17,18]. Eberle and Wegener used WC-Co and polycrystalline diamond (PCD) to demonstrate that Ra obtained using ultrashort pulses was signi cantly lower than that obtained using nanosecond pulses [19]. However, Fang et al. observed that femtosecond multi-pulse irradiation caused repetitive melting and redeposition of cemented tungsten carbides [20].
Denkena et al. reported the formation of B 2 O 3 at a higher laser uence using picosecond and femtosecond lasers on a polycrystalline cubic boron nitride (PcBN) surface, distinguishing it from thermal effects such as melting using a nanosecond laser [21]. Therefore, a low uence is recommended, and the number of pulses, pulse energy, and laser spot size should be controlled. Dumitru et al. demonstrated in a study pertaining to the laser processing of holes or grooves that the discharge performance of debris decreased and the roughness was not improved [22]. Re ection at the side wall fabricated by laser processing was used to investigate the variations in hole shapes [23], effect of laser drill welding [24], and keyhole welding model [25], which inhibits the fabrication of at surfaces. Hence, the control of debris and shape during processing is important for perpendicular laser irradiation.

PLG mechanism
PLG is a laser processing method that has a laser irradiation area equivalent to that of the grinding wheel. However, its features differ from that of conventional laser processing methods such as drilling, cutting, and laser polishing, which are perpendicular laser irradiation methods. The rst feature is that the irradiation area consists of a long-focus lens and a pulse laser with a Gaussian pro le. Therefore, it is necessary to construct a three-dimensional uence distribution. The Gaussian uence F(r) is expressed as follows [26]: where denotes the single-pulse energy, r denotes the spatial coordinate with the peak intensity as the origin, and w denotes the radius, referred to as beam waist, de ned by the beam intensity decreased to 1/e 2 (13.5%) of its peak. When the laser is focused with the lens, the beam radius, w(z), at the propagation distance from the focusing point, z, is expressed as follows: where λ denotes the wavelength, and w 0 denotes the beam waist at the focus point determined by the laser spot diameter and focal length. The spatial distribution of the laser intensity, I(r, z), is obtained by assigning w = w(z) in Eq. (1). Therefore, the laser power density can be calculated from the squared laser intensity. The contour pro le of the calculated power density is shown in Fig. 1.
There is a long, narrow power-density area. This power density has a cylindrical distribution in the threedimensional PLG area. The length along the z-axis is determined by the focal length of the lens, pulse energy, and laser spot size. Therefore, by determining the laser uence such that the inside of the PLG area is equal to or above the processing threshold of the material, only the material overlapping with the PLG area can be removed. The second feature is that the workpiece is placed at an approximately parallel angle between the laser axis and the work surface to undergo laser ablation on the surface (Fig. 2). This piece placement restrains the residual debris because the discharge of the ablation particle is perpendicular to the surface. Processing owing to the re ection laser on the wall surface formed by the previous processing is avoided. However, the angle of incidence based on the irradiated surface is often larger than Brewster's angle, and the re ectance is larger than that of perpendicular irradiation. Therefore, high-energy and P-polarized pulses are required.
The PLG area is multi-scanned in the ridge of the cutting tool edge. The processing volume is larger near the cutting tool edge, where the laser uence is higher. However, the processed surface gradually deviates from the center of the Gaussian beam at each scan. All positions on the surface are nished with the same uence, which is slightly higher than the processing threshold. Furthermore, as most laser pulses after nishing pass through the nished surface after nishing the surface, it is possible to prevent melting or redeposition. This implies that it is not necessary to strictly determine the number of irradiations as long as the number of irradiations is substantial. For example, even if the operator decides that the process is complete by the end of the plasma emission, the same quality is always obtained.
The nished PLG roughness is decreased using the same averaging method as that of plunge grinding. However, the function of the PLG area differs from that of the grindstone. Finished surfaces obtained by plunge grinding have been reported [3,[27][28][29][30][31]. The surface pro le parallel to the infeed direction is formed by different rotation trajectories consisting of abrasive grains with irregular heights (Figs. 3(a) and (b)). Multi-scanning averages these trajectories and the vibration of the scanning movement. The elastic de ection by the vertical load is completely removed in the nishing stage, which is referred to as spark-out [29,32,33]. However, the shapes of the tall abrasive grains remain on the nished surface owing to abrasive grain wear before averaging is completed [34,35]. In addition, the linear and angular motion errors attributed to the guideway accuracy are not averaged. However, there is no averaging effect in the perpendicular direction. The abrasive grain pro le directly determines the perpendicular surface roughness (Fig. 3(c)). Therefore, the perpendicular surface roughness is higher than the parallel surface roughness, which shows that the pro le of the grindstone and effect of wear have a strong in uence on the roughness.
The nished PLG roughness is de ned as the variance of the laser ablation depth in the surface direction.
The ablation depth has been reported as a function of the laser uence, pulse width, and ablation threshold [16,36]. The variances of the laser uence and pulse width are averaged by multi-scanning ( Fig.  4(a)). Unlike abrasive grain wear, these averages do not shift. The ablation threshold depends on the material used. Its variance depends on the defect variance that affects the absorption coe cient. In addition, the ablation threshold for an ultrashort pulse laser decreases owing to accumulation defects, referred to as incubation effects [37][38][39][40]. Structural defects, crystal lattices, and thermal effects have been reported as incubation factors [38,39,41]. The variance in these effects is unclear. However, if the low and averaged laser uence do not affect the variance of the ablation threshold, the parallel surface roughness is determined only by the material-speci c defect variance and motion error ( Fig. 4(b)). Furthermore, the laser uence pro le in the perpendicular surface direction without the averaging effect and motion error is the uence trajectory of one pulse; there is no laser uence variance in this direction ( Fig. 4(c)). Therefore, the perpendicular surface roughness depends only on the material defect variance and is smaller than the parallel surface roughness; a result that is opposite to that of plunge grinding, even though the infeed directions are the same.

Experimental Setup
We considered the roughness convergence irrespective of the position and difference depending on the direction and conducted experiments to con rm the surface formation with overlapping PLG areas. The overlap depth was determined by the radius of the PLG area and the distance between the surface and the laser axis. The adjustment of the spot size with optical components, such as a lens, was challenging because of the complexity of the system. A simple method was to increase the pulse energy. However, the increase in the PLG radius was not proportional to the pulse energy. When the beam waist, w 0 , matches the PLG radius, the threshold uence is obtained using Eq. (1).
If the PLG radius at α times the pulse energy is r α , the uence of that circumference also has the same threshold uence, For example, when the beam waist w 0 = 5 µm and the pulse energy factor is 5, the radius expansion is only 0.81 µm. Therefore, it is easier to adjust the overlap width based on the machining depth. The PLG surface roughness with different pulse energies and machining depths was con rmed in this study.
A WC-Co cutting tool of grade "G10E" with an average grain size of 2.5 µm, 6-wt% Co, and edge line with right-angle corners (Sumitomo Electric Hardmetal Corp.) was selected. The ank surface was completed by grinding only in the horizontal direction. Based on the measurements using a laser microscope, Ra = 0.087 and 0.137 in the horizontal and vertical directions, respectively. Because Ra differs according to the grinding direction, we de ned the initial surface roughness and compared it with that of a pulse laser ground surface.
A schematic of the PLG system is presented in Fig. 5. A Yb:KGW chirped pulse ampli cation system (PH2-10 W, Light Conversion) with a nonlinear optical unit at a wavelength of 257 nm (PH1F4, Light Conversion) with a pulse frequency of 25 kHz was used. The scanning stage (L-731, PI) had a scanning speed and pulse-to-pulse distance of 2.5 mm/s and 0.1 µm, respectively. The laser shutter opening was controlled at an arbitrary position by the laser control module connected to the laser and scanning stage. When a mirror is inserted in front of the focus lens, the focus position is adjusted by the Z-stage and CCD camera. In this experiment, the deviation between the camera focus and laser focus was veri ed in advance and corrected. All focal points were adjusted to the tool edge line.
An ultrashort pulse was focused using a plano-convex lens with f = 100 mm. The calculated beam waist diameter was 12.6 µm, whereas the processing hole diameter with 8 µJ/pulse was 10.5 µm, as previously veri ed. The ank surface was tilted at a processing angle of 8° using a p-polarized pulse laser. In this system, the tool surface and laser axis were maintained at an appropriate distance to prevent laser ablation. Subsequently, the tool approached the laser axis gradually by 5 µm such that the PLG start position was smaller than the spot size from the tool edge line (Fig. 6). We de ned the rst pass for which plasma with laser ablation was visually observed. Pulse energies of 4, 6, 8, 10, and 26 µJ were used to demonstrate the relationship with the pulse energy. The samples were processed using the rst-pass method. The depth of the machining from the rst to the second pass was 10 µm. Furthermore, to con rm that the nished PLG surface affected the defect generation during PLG processing, the depths of machining from the rst to the second pass were 1, 2, 5, and 10 µm with a pulse energy of 4 µJ. The number of scans per pass was set to 50 for all passes. This implied that a larger machining depth corresponded to higher-uence processes.

Results And Discussion
Scanning electron microscopy (SEM) was used to observe the nished surfaces. Using a laser microscope, roughness was evaluated with Ra along two axes: one parallel to both the infeed direction and the nished surface and another perpendicular to the infeed direction and parallel to the surface. The measurement positions of the parallel roughness were 20 and 40 µm from the tool edge line.
The nished PLG surfaces and pro les with different pulse energies observed using SEM and a laser microscope with an inclination of 45° are shown in Fig. 7. The Ra of the PLG-processed surface was measured, as shown in Fig. 8. The perpendicular surface roughness, Ra, did not vary with the pulse energy and was in the range of 0.011-0.012 µm. This implied that the surface was processed by a uniform laser pro le, irrespective of the pulse energy. The nished surface roughness was formed depending only on the variance defects of the WC-Co tools. The parallel surface roughness had no correlation with the pulse energy and was in the range of 0.041-0.054 µm. The roughness at the position of 40 µm was lower than that at 20 µm, except for the result at 8 µJ. The effect on the intensity and gradient of the pro le near the circumference of the beam waist was assumed to be small, even if the pulse energy was increased. Moreover, because the parallel surface has more variance factors than the perpendicular surface, su cient irradiation pulses are required to average these factors. Therefore, the nished surface under each condition had a small difference in the removal depth. The nished surface was treated with the same uence. The variation in roughness occurred because of the variability owing to incomplete PLG with the lack of averaging and removal.
The SEM images and pro les with different machining depths are presented in Fig. 9. The measured roughness values are shown in Fig. 10. The perpendicular surface roughness, Ra, was in the range of 0.006-0.011 µm. The roughness depending on the material defect variance was formed irrespective of the irradiation conditions, as in the previous experiment. The parallel surface roughness decreased as the depth of machining decreased and was in the range of 0.025-0.038 µm, with a minimum roughness of 0.025 µm and a machining depth of 1 µm. This experiment had the same machining depth of 10 µm and pulse energy as that in the previous experiment at 4 µJ/pulse. However, the parallel surface roughness was smaller than that in the previous experiment, and there was a low variation depending on the measurement position, probably because the overlap in this experiment was lower than that in the previous experiment because the rst pass was de ned with a gradual approach of 5 µm. Moreover, a smaller machining depth corresponded to a smaller parallel surface roughness. A smaller removal volume corresponded to a lower and more converged surface roughness by a su cient averaging process. In addition, the parallel convergence roughness was larger than the perpendicular surface roughness. This indicated that the parallel surface had non-averaging effects, such as motion errors, in addition to the defect variance. This implied that PLG could produce low-roughness surfaces without a laser pro le shift during processing, such as grindstone wear.

Conclusions
In this study, the PLG method and WC-Co surface obtained by PLG using an ultraviolet femtosecond laser were evaluated. The results of this study can be summarized as follows.
The PLG system included an approximately parallel irradiation and averaging process based on conventional grinding. This irradiation angle restrained laser-speci c problems such as debris and redeposition. A low surface roughness was realized by averaging the vibration of the processing system and the variance of the laser pulses. In addition, the PLG with a uniform and non-shifted laser pro le during processing was explained to achieve a lower roughness than that of conventional grinding with grindstone wear.
The experimental investigation with WC-Co also showed a lower roughness than that of conventional grinding. The perpendicular surface had a minimum roughness, Ra, of 0.006 μm, whereas the parallel surface had a minimum roughness, Ra, of 0.025 μm.
However, su cient averaging by multi-scanning with a low uence was required. The uence and number of irradiation pulses differed depending on the material.
The variance of material-speci c defects and motion error was not considered to restrain even if the averaging process was su cient. The parallel surface roughness was higher than the perpendicular surface roughness owing to the motion error. This is important when parallel surface roughness is required owing to irradiation direction restrictions (e.g., tool cutting edges).     Parallel and perpendicular roughnesses with different pulse energies PLG with different depths of machining