Laser Finishing of Ti6Al4V Additive Manufactured Parts by Electron Beam Melting

: In this work, the feasibility of laser surface ﬁnishing of parts obtained by additive manufacturing (AM) was investigated. To this end, a 450 W ﬁber laser (operating in continuous wave, CW) was adopted to treat the surface of Ti-6Al-4V samples obtained via electron beam melting (EBM). During the tests, di ﬀ erent laser energy densities and scanning speeds were used. In order to assess the quality of the treatment, either the as-built or the treated samples were analyzed by means of a three-dimensional (3D) proﬁlometer, digital microscopy, and scanning electron microscopy. Analysis of variance (ANOVA) was performed to check which and how process parameters a ﬀ ected the ﬁnishing. The results show that, in the best conditions, the laser treatment reduced surface roughness by about 80%.


Introduction
Initially, three-dimensional (3D) printing was adopted just for rapid prototyping (RP); recently, it has been used as a real production technology, starting from the three-dimensional mathematical model, realized with CAD (computer aided design) software. The technique allows creating components by adding material "layer-by-layer". This is the reason why it is also known as additive manufacturing (AM) [1]. By using different AM techniques, it is possible to obtain customized and complex components of different materials (metals, plastic or ceramic) with competitive costs and times [2]. In the aeronautic field, the most attractive materials are definitely metals, such as titanium and aluminum alloys, as they allow to design components with very complex shapes that are hardly obtained when using conventional technologies [2][3][4]. Metal parts can be obtained using different AM techniques and are all based on power bed fusion (PBF) technology [1], such as electron beam melting (EBM) [5], direct metal laser sintering (DMLS) [4,6], and selective laser melting (SLM) [7,8]. However, the surfaces of the printed parts all have a low quality. The typical arithmetic average of the roughness profile (Ra) varies in the range 8-25 µm [2,6,9] depending on the adopted technology. Even if lower values of roughness are obtainable by laser sintering/melting, the EBM process is generally preferred for the production of aerospace components as it can produce high-quality products with excellent mechanical properties and fewer defects compared to those achieved by using laser technologies. Thus, it is possible to obtain almost fully-dense parts [3]. This is due to the EBM process: the parts are printed in a vacuum environment which eliminates all impurities with high build temperatures, and thus low residual stress. In addition, EBM has faster build rates and all elements do not necessarily require any support [9].

Laser Finishing Setup
A 450 W fiber laser (IPG YLR-450/4500-QCW-MM-AC-Y14) was used to finish the surface of the samples. The system can operate either in continuous wave (CW) or in a modulated wave known as "quasi continuous wave" (QCW). The laser radiation is transferred via 50 µm core fiber, with a length of 5 m, to a collimator with a lens of 85 mm, mounted onto a micro cutting head (IPG FLC D 25) with a focusing lens of 100 mm. Thus, a beam diameter of about 60 µm is achieved. The main laser source characteristics are reported in Table 2. During all tests, the CW mode was adopted in order to achieve a smoother surface. Nitrogen gas with a flow rate of 12 l/min at 2 bar was adopted to shield the samples, thus avoiding oxidation. The samples were rotated using a stepping motor (SANYODENKI 103H7126) at 200 rpm, while the laser was switched on outside the sample. The laser beam was then moved onto the samples at different speeds, with a length of about 50 mm, as reported in Figure 1. After the laser treatment, the samples were sandblasted with pink corundum (mesh size 120) at a pressure of 4 bar, with the aim to remove any incoherent material, and then cleaned in an ultrasonic bath with acetone to remove any traces of sand.

Laser Finishing Setup
A 450 W fiber laser (IPG YLR-450/4500-QCW-MM-AC-Y14) was used to finish the surface of the samples. The system can operate either in continuous wave (CW) or in a modulated wave known as "quasi continuous wave" (QCW). The laser radiation is transferred via 50 μm core fiber, with a length of 5 m, to a collimator with a lens of 85 mm, mounted onto a micro cutting head (IPG FLC D 25) with a focusing lens of 100 mm. Thus, a beam diameter of about 60 µ m is achieved. The main laser source characteristics are reported in Table 2.
During all tests, the CW mode was adopted in order to achieve a smoother surface. Nitrogen gas with a flow rate of 12 l/min at 2 bar was adopted to shield the samples, thus avoiding oxidation. The samples were rotated using a stepping motor (SANYODENKI 103H7126) at 200 rpm, while the laser was switched on outside the sample. The laser beam was then moved onto the samples at different speeds, with a length of about 50 mm, as reported in Figure 1. After the laser treatment, the samples were sandblasted with pink corundum (mesh size 120) at a pressure of 4 bar, with the aim to remove any incoherent material, and then cleaned in an ultrasonic bath with acetone to remove any traces of sand.

Experimental Procedures
The work was divided into two phases. In the first phase, preliminary tests were performed aimed at identifying the laser parameters (i.e., the control factors) to allow the treatment (i.e., the process windows) and the levels adopted in the second phase in which a 2 × 3 factorial plan was defined and tested.
It is known that the laser energy density (i.e., the so called fluence) affects the interaction between laser and material. When the density increases, it is possible to pass from heating up to melting or even vaporizing the material. In other words, the fluence affects the amount of treated

Experimental Procedures
The work was divided into two phases. In the first phase, preliminary tests were performed aimed at identifying the laser parameters (i.e., the control factors) to allow the treatment (i.e., the process windows) and the levels adopted in the second phase in which a 2 × 3 factorial plan was defined and tested.
It is known that the laser energy density (i.e., the so called fluence) affects the interaction between laser and material. When the density increases, it is possible to pass from heating up to melting or even vaporizing the material. In other words, the fluence affects the amount of treated surface, the final roughness, and the heat affected zone (HAZ). The adopted solution to decrease the energy density is to vary the focal distance. At the increase of the focal distance, an increase of the laser footprint/track on the surface of the specimens occurs, and accordingly, a decrease of the fluence.
Preliminary tests were carried out keeping the laser radiation on the samples at maximum power (i.e., 450 W) with three different focal distances (+5, +7.5, and +10 mm) on the surface of the samples during rotation.
In the second phase, a 2 × 3 factorial plan was defined and tested based on the results of the first phase. Table 3 shows the control factors and their levels. The plan was repeated three times; thus, a total of 18 tests were performed. The influence of the process parameters on the roughness of the samples was assessed via analysis of variance (ANOVA). The main effect and the interaction plots were plotted to see the effect of the control factors on the laser finishing. In order to assess the quality of the treatment, the surface morphology of as-built and treated samples were studied by means of a digital microscope (Hirox KH 8700). In addition, roughness measurements were performed on all samples using a 3D surface profiling system (Taylor Hobson Talysurf CLI 2000). More specifically, in order to characterize the surface topography [24], the amplitude roughness parameters were calculated on the acquired profile: the arithmetic mean surface roughness (Ra) and the root mean squared surface roughness (Rz), skewness (Rsk), and kurtosis (Rku) were calculated in height; the mean width of profile elements (RSm) was calculated horizontally. In addition, the root mean square slope (RDq), defined as a hybrid parameter (i.e., a combination of amplitude and spacing), was calculated. The latter parameter was studied as it is known to influence tribological properties.
For what concerns roughness analysis, the cut-off and evaluation length were chosen as foreseen by the UNI EN ISO 4288:2000 standard reported in Table 4. Accordingly, for the as-built sample, the cut-off was set at 8, while for the finished samples, when Ra was in the range 2-10 µm, it was set at 2.5. In addition, 3D maps of 4 × 4 mm 2 areas of the samples were also acquired using a resolution of 2 µm along both directions. After the acquisitions, the samples were cut with a diamond saw, embedded in an epoxy resin and polished. Then, micro-indentation tests (CSM Instruments Micro-Combi) were performed on the cross-sections, using a load of 20 N, starting at 0.75 mm from the edge up to the center of the samples with a step of 0.2 mm. Then, a chemical etching was performed in HF solution to analyze the microstructure of the specimens in order to assess the effect of the laser treatments. and material. When the density increases, it is possible to pass from heating up to melting or even vaporizing the material. If we consider the tracks, laser power, and treatment time, we can achieve an estimation of the adopted fluences. It is worth noting that these values differ from the theoretical ones but are more suitable for industrial applications. The measurements are reported in Table 5, while in Figure 3 the contour plot of the estimated fluence was traced vs. the track width and focal distance.  a-state-the laser track appears regular with the typical V-shape of a deep marking. This means that the adopted fluence is so high as to vaporize the little ridges and part of the bulk (Figure 2a).

Preliminary Tests
b-state-the laser track is just visible. The fluence is high enough to melt and vaporize only the little ridges while the bulk is light blue (Figure 2b). c-state-the laser track is not visible. Just a dark shade is visible on the ridges and no melted material ( Figure 2c).
According to these observations, the c-state was not desired for finishing purposes; consequently, focal distance (Fd) = 10 mm was excluded from the second phase.
The screening tests show that, as expected, the fluence affects the interaction between the laser and material. When the density increases, it is possible to pass from heating up to melting or even vaporizing the material. If we consider the tracks, laser power, and treatment time, we can achieve an estimation of the adopted fluences. It is worth noting that these values differ from the theoretical ones but are more suitable for industrial applications. The measurements are reported in Table 5, while in Figure 3 the contour plot of the estimated fluence was traced vs. the track width and focal distance.    Figure 4 shows the surface morphology of an as-built sample. As expected, the surface appears very irregular, with high peaks and valleys very close to one other. This is due to the melting of the smaller powder particles near the surface of the component, as observable in Figure 5a,b reporting respectively the cross-section surface and its magnification. From the analysis of the cross-section, it clearly appears that the outline is very jagged and the profilometer inductive stylus, with a 2 μm radius, cannot penetrate in the very narrow valleys. Thus, the roughness parameters calculated on the acquired profiles of the as-built surface may be underestimated.   Figure 4 shows the surface morphology of an as-built sample. As expected, the surface appears very irregular, with high peaks and valleys very close to one other. This is due to the melting of the smaller powder particles near the surface of the component, as observable in Figure 5a,b reporting respectively the cross-section surface and its magnification. From the analysis of the cross-section, it clearly appears that the outline is very jagged and the profilometer inductive stylus, with a 2 µm radius, cannot penetrate in the very narrow valleys. Thus, the roughness parameters calculated on the acquired profiles of the as-built surface may be underestimated.  Figure 4 shows the surface morphology of an as-built sample. As expected, the surface appears very irregular, with high peaks and valleys very close to one other. This is due to the melting of the smaller powder particles near the surface of the component, as observable in Figure 5a,b reporting respectively the cross-section surface and its magnification. From the analysis of the cross-section, it clearly appears that the outline is very jagged and the profilometer inductive stylus, with a 2 μm radius, cannot penetrate in the very narrow valleys. Thus, the roughness parameters calculated on the acquired profiles of the as-built surface may be underestimated.   Figure 4 shows the surface morphology of an as-built sample. As expected, the surface appears very irregular, with high peaks and valleys very close to one other. This is due to the melting of the smaller powder particles near the surface of the component, as observable in Figure 5a,b reporting respectively the cross-section surface and its magnification. From the analysis of the cross-section, it clearly appears that the outline is very jagged and the profilometer inductive stylus, with a 2 μm radius, cannot penetrate in the very narrow valleys. Thus, the roughness parameters calculated on the acquired profiles of the as-built surface may be underestimated.    (Figures 6f and 7c,f,i); Type 3: Visible laser melting tracks (Figure 6i).

Morphological Analysis
in Figure 6i the laser tracks are visible. This is due to the Gaussian shape of the beam: the central part of the laser beam results in material vaporization, while the tails just allow the melting. This effect is not visible in Figure 6c,d because of the low scan speed (i.e., very high COP).
To better understand the phenomena occurring during the laser finishing, the treated surfaces were also analyzed by scanning electron microscopy (SEM). Figure 8 depicts the SEM images of the surface after laser polishing at different scan speed (Ss) and Fd. From the image, the lower focal distance (Fd = 5 mm) allows obtaining the smoothest surfaces by vaporizing the higher peaks and melting the greater ones. On the contrary, at a higher focal distance (Fd = 7.5 mm), the higher peaks are just melted and thus the surface appears more irregular. Consequently, at a lower focal distance, a higher diameter reduction of the sample is expected. In Figure 9, the sample's diameter and its ablation are reported vs. the laser parameters. The figure represents a confirmation of the aforementioned mechanism (i.e., higher ablation at Fd = 5 mm).    These morphologies are the consequence of the laser spot, rotating, and scan speed. A value that considers these parameters is the circumferential overlap percentage (COP) [25], which is defined as the following ratio: where N, D, and Ss represent the workpiece rotating speed in rpm, the laser spot in mm, and the scan speed in mm/s. Then, based on the preliminary results of Table 5, it is possible to calculate the COP for each test condition by replacing the laser footprint at different focal distances (the values are reported in Table 6). The results show that at the minimum focal distance, the COP is in the range 84%-97%. However, in Figure 6i the laser tracks are visible. This is due to the Gaussian shape of the beam: the central part of the laser beam results in material vaporization, while the tails just allow the melting. This effect is not visible in Figure 6c,d because of the low scan speed (i.e., very high COP). To better understand the phenomena occurring during the laser finishing, the treated surfaces were also analyzed by scanning electron microscopy (SEM). Figure 8 depicts the SEM images of the surface after laser polishing at different scan speed (Ss) and Fd. From the image, the lower focal distance (Fd = 5 mm) allows obtaining the smoothest surfaces by vaporizing the higher peaks and melting the greater ones. On the contrary, at a higher focal distance (Fd = 7.5 mm), the higher peaks are just melted and thus the surface appears more irregular. Consequently, at a lower focal distance, a higher diameter reduction of the sample is expected. In Figure 9, the sample's diameter and its ablation are reported vs. the laser parameters. The figure represents a confirmation of the aforementioned mechanism (i.e., higher ablation at Fd = 5 mm).       Figures 10 and 11 depict the comparison of 3D maps and surface profiles acquired by the profilometer respectively on the as-built and on the laser-polished samples. The analysis confirms the smoother surface when the lower focal distance was adopted (Fd = 5 mm). In order to better analyze the improvement, all the acquired roughness parameters are reported in Figure 12 vs. the laser ones. For the as-built samples, an arithmetic mean surface roughness (Ra) of about 27 µm is observable; while for the treated sample, in the best conditions, a Ra value of about 5 µm can be obtained, allowing an 80% reduction in the roughness surface.

Surface Roughness
ANOVA was performed to assess the influence of the process parameters on sample roughness; the residual distribution was previously checked and a confidence level of 95% was adopted (α = 0.05). The results for Ra are reported in Table 7, where the model summary is also reported in terms of R 2 . It is worth noting that the R 2 value is very high, this means that the model has a good fit; the predicted R 2 is high, thus the model has a good predictive ability. Based on these assumptions, both the control factors (Fd and Ss) and their interaction are significant (i.e., p-value less than 0.05). However, the most affecting factor is surely the focal distance (Fd), since it shows the greatest F-value. Figure 13 depicts the main effect plot and interaction plot for Ra. Figure 13a shows that the effect of the control factors (i.e., the laser parameters). The roughness increases when the focal distance and scan speed increases (i.e., when energy density and laser-material interaction decrease); this result is consistent in literature. The effect of the interaction between the focal distance and scan speed can be seen in Figure 13b: at Fd = 7.5 mm the roughness parameter is not affected by a significant variation, while at Fd = 5 mm, Ra changes from 5 up to 20. This aspect can be clarified by considering the surface morphologies of Figures 6 and 7.

Surface Roughness
Figures 10 and 11 depict the comparison of 3D maps and surface profiles acquired by the profilometer respectively on the as-built and on the laser-polished samples. The analysis confirms the smoother surface when the lower focal distance was adopted (Fd = 5 mm). In order to better analyze the improvement, all the acquired roughness parameters are reported in Figure 12 vs. the laser ones. For the as-built samples, an arithmetic mean surface roughness (Ra) of about 27 µ m is observable; while for the treated sample, in the best conditions, a Ra value of about 5 µ m can be obtained, allowing an 80% reduction in the roughness surface.
ANOVA was performed to assess the influence of the process parameters on sample roughness; the residual distribution was previously checked and a confidence level of 95% was adopted (α = 0.05). The results for Ra are reported in Table 7, where the model summary is also reported in terms of R 2 . It is worth noting that the R 2 value is very high, this means that the model has a good fit; the predicted R 2 is high, thus the model has a good predictive ability. Based on these assumptions, both the control factors (Fd and Ss) and their interaction are significant (i.e., p-value less than 0.05). However, the most affecting factor is surely the focal distance (Fd), since it shows the greatest Fvalue. Figure 13 depicts the main effect plot and interaction plot for Ra. Figure 13a shows that the effect of the control factors (i.e., the laser parameters). The roughness increases when the focal distance and scan speed increases (i.e., when energy density and laser-material interaction decrease); this result is consistent in literature. The effect of the interaction between the focal distance and scan speed can be seen in Figure 13b: at Fd = 7.5 mm the roughness parameter is not affected by a significant variation, while at Fd = 5 mm, Ra changes from 5 up to 20. This aspect can be clarified by considering the surface morphologies of Figures 6 and 7.        Figure 14 shows microhardness distribution along the cross-sections. No trend with edge distance is visible (about 330 HV for all samples, including the as-built one). As can be inferred, this outcome was not expected since laser treatment is a thermal interaction and therefore generally affects mechanical characteristics [13,16]. Then, a chemical etching was performed to analyze the microstructure of the specimens. In Figure 15 the microstructure analysis of an as-built and a laserpolished sample (Fd = 5 mm, Ss = 3.6 mm/min) are reported: the as-built sample shows the typical microstructure of Ti alloy (lamellar α + β phase); the laser finished sample shows a polished layer (HAZ, heat affected zone) with a thickness of about 25 μm. This is consistent [16,26,27] and related to melting and rapid cooling during laser finishing. It also explains the distribution of the microhardness distribution: the polished layer is too thin to allow a microhardness test (starting at 0.75 mm from the edge). Probably, a nanoindentation test would allow a measurement of the change in mechanical characteristics.  Figure 14 shows microhardness distribution along the cross-sections. No trend with edge distance is visible (about 330 HV for all samples, including the as-built one). As can be inferred, this outcome was not expected since laser treatment is a thermal interaction and therefore generally affects mechanical characteristics [13,16]. Then, a chemical etching was performed to analyze the microstructure of the specimens. In Figure 15 the microstructure analysis of an as-built and a laser-polished sample (Fd = 5 mm, Ss = 3.6 mm/min) are reported: the as-built sample shows the typical microstructure of Ti alloy (lamellar α + β phase); the laser finished sample shows a polished layer (HAZ, heat affected zone) with a thickness of about 25 µm. This is consistent [16,26,27] and related to melting and rapid cooling during laser finishing. It also explains the distribution of the microhardness distribution: the polished layer is too thin to allow a microhardness test (starting at 0.75 mm from the edge). Probably, a nanoindentation test would allow a measurement of the change in mechanical characteristics.   . Figure 14. Microhardness along the cross section.

Conclusions
In this paper, laser surface finishing of components obtained by electron beam melting (EBM) was performed. A high-efficiency fiber laser operating in continuous wave was used to treat the surface of Ti-6Al-4V samples. During the tests, the laser energy density (by way of focal distance) and the scanning speed were varied. The following findings were found: The laser treatment can be adopted as a polishing treatment of Ti-6Al-4V samples. In the best conditions, a Ra value of about 5 µ m can be obtained with a reduction of 80% in roughness.
The main mechanism consists of melting and/or vaporization (depending on the laser energy) of the highest peaks.
The treatment involves material ablation and, therefore, the increase in dimensions must be considered when the components are designed.
Author Contributions: All the authors contributed equally to the various aspect of this work. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.

Conflicts of Interest:
The authors declare no conflicts of interest.

Conclusions
In this paper, laser surface finishing of components obtained by electron beam melting (EBM) was performed. A high-efficiency fiber laser operating in continuous wave was used to treat the surface of Ti-6Al-4V samples. During the tests, the laser energy density (by way of focal distance) and the scanning speed were varied. The following findings were found: The laser treatment can be adopted as a polishing treatment of Ti-6Al-4V samples. In the best conditions, a Ra value of about 5 µm can be obtained with a reduction of 80% in roughness.
The main mechanism consists of melting and/or vaporization (depending on the laser energy) of the highest peaks.
The treatment involves material ablation and, therefore, the increase in dimensions must be considered when the components are designed.