*3.2. Surface Roughness Evaluation*

As discussed in the introduction, the surface roughness of an additively manufactured part tends to play a crucial role in determining its overall performance. A smoother surface will usually imply an enhanced mechanical performance (superficial defects tend to act as crack nucleation sites) and will also be better received by the end consumer as parts resemble their conventionally manufactured counterparts.

In this sense, the analysis of the surface roughness parameters *Ra* and *Rz* presented in Figure 1 shows that all finishing techniques herein presented have lowered the surface roughness of FFF Ultem parts to a certain degree. However, an exception should be made in the chemical treatment to remove Ultem's support material, as changes in surface roughness are almost negligible. This result was expected as the purpose of this treatment is to affect Ultem's integrity as little as possible.

**Figure 1.** (**a**) Changes in average surface roughness; (**b**) Changes in average peak-to-valley height of the roughness profile.

Results obtained from the other chemical treatment, namely vapor smoothing with chloroform, are drastically different. Improvements on Ultem's surface roughness are as remarkable as 90–95% (*Ra* = 1.4 μm, *Rz* = 3.57 μm) after 180 min of treatment. Interestingly, there is no further improvement after this point (surface roughness after 270 min is similar to the one obtained after 180 min), while chloroform continues to be absorbed, as discussed in the mass and dimensional analysis section. For this reason, 180 min should be considered the maximum treatment time to achieve an excellent surface finish. The magnitude of the roughness improvements reported by Chohan et al. [17] (99.62%) is comparable to the results of the present study. It should be noted, though, that the initial surface roughness of their studied hip replica was two-times lower than the one of the Ultem parts used in this study and that their total contact time with the vaporized solvent was considerably lower (in the order of a 1 min) than in the present case. The roughness improvements reported by Rajan et al. [22] are more moderate (around 80%) when they used tetrahydrofuran (THF) vapors during a total time of 5 min to smooth FFF PLA parts. They also observed that the initial build orientation (and thus the initial surface roughness) plays a crucial role in the outcome of the vapor smoothing process. Considering that all consulted treatment times are lower than the ones needed to obtain similar results with Ultem, it is safe to affirm that Ultem's chemical resistance to chloroform is higher than the chemical resistance of ABS to acetone and of PLA to THF.

Concerning thermal annealing, no significant differences in surface roughness have been identified, regardless of the temperature of the treatment. Nevertheless, dimensional changes are higher at higher temperatures, implying that there is no benefit in increasing the temperature of the treatment from 210 ◦C to 240 ◦C. The obtained 15% improvement in the surface roughness of the upper part (*Ra* = 12.60 μm, *Rz* = 62.28 μm) is probably due to the partial fusion of adjacent filaments. Interestingly, the lower part of the samples, which was in direct contact with the thermal chamber's tray, presented a glass-like finish, with improvements in *Ra* higher than 90% (1.05 μm) and higher than 80% in and *Rz* (8.69 μm). This suggests that a physical contact that favors heat transfer by conduction is noticeably more effective in improving the surface finish of the parts than heat transfer by convection, as it happens on the air-exposed faces of the part.

Surface roughness obtained after mechanical post-processing of the parts differs considerably. When samples are ball burnished, *Ra* and *Rz* diminish by about 70% (5.06 μm) and 50% (34.19 μm), respectively, due to the plasticization of the outer layer because of the forced pressing of the peaks by the burnishing ball. This technique uses a stainless-steel tool that can be adapted to a numeric control machine, making it an easily automatable process, but presents some challenges when the totality of the part needs to be treated, or when the part has difficult-to-access zones. With abrasive shot peening or abrasive shot blasting using glass beads, the surface roughness has improved by 20% to 25% (*Ra* = 10.83 μm, *Rz* = 3.57 μm). The use of white corundum as blasting media has resulted in a more moderate improvement of around 10% (*Ra* = 13.40 μm, *Rz* = 62.54 μm). A noticeable fact is that the standard deviation of the roughness measurements after these finishing processes were applied to Ultem parts is considerably higher than with the other treatments, which indicates more limited repeatability as they depend on the experience of the operator performing the treatment, the contact time, and the wear of the beads, amongst others. In comparison with the 50% improvement in surface roughness reported by Valerga Puerta et al. [45] when they corundum blasted FFF PLA parts using analogous treatment time conditions as the ones reported in the present study, Ultem appears to have an increased initial toughness that could explain such more moderate improvements.
