Surface roughness reduction in electron beam powder bed fusion additive manufacturing of super duplex stainless steel 2507: investigating optimisation techniques and face orientation‑dependent irregularities

The use of additive manufacturing in metals by powder bed fusion via electron beam (PBF-EB) is increasing for fabricating high-quality parts meeting industrial standards. However, high surface roughness poses a consistent challenge in PBF-EB. This study investigates two novel approaches to optimise surface roughness for a given machine and powder combination. Using machine control software’s recently introduced research mode functionality, we develop customised beam control code to effectively explore a vast parameter space. Additionally, we explored the impact of beam travel direction and spot morphology on surface roughness. Line-melt-based contours were explored by specimen manufacturing with layer-wise parameter change, whilst spot-melting-based samples were built using a full factorial design of experiments with four factors at three levels. Initial sample characterisation was done using a stylus-based contact profilometer, followed by detailed evaluation using focus variation microscopy. Results reveal that increasing beam power and spot energy exacerbate surface roughness. We also find that a well-defined energy distribution at the spot's edge contributes to smoother surfaces. Whilst the influence of beam travel direction on surface roughness remains uncertain, our findings underscore the importance of parameter selection in achieving optimal results. By adjusting contouring parameters, we achieve a vertical roughness of Ra17.7 ± 0.9 (Sa 21.6), significantly lower than in the current literature. These findings advance our understanding of surface roughness optimisation in PBF-EB and offer practical insights for improving part quality in industrial applications. By harnessing tailored beam control strategies, manufacturers can enhance the capabilities of additive manufacturing technologies in producing metal components.


Introduction
Metal additive manufacturing (AM) is increasingly widespread throughout today's manufacturing industry.The ability to fabricate parts of high complexity and customised microstructure that can be used directly is unrivalled by traditional manufacturing methods.Electron beam powder bed fusion (PBF-EB) is a method of metal AM in which an electron beam is used to melt a powder-based feedstock to form parts in a layer-by-layer process.PBF-EB has long been considered a process that results in built surfaces of high roughness [1][2][3] compared to laser-based powder bed fusion (PBF-LB).The melting of parts in PBF-EB can be divided into two separate stages: the first is the hatch melting that melts the bulk material of the part; the second stage is the contour melting that is responsible for the outermost sections of a part, and therefore, the majority of the resulting surface roughness derives from the second melting stage.However, the denomination of the first and second melting stages is unrelated to the order in which they occur, where the contours are often melted first.The melting strategy typically applied to contours is spot melting, where the beam is rapidly repositioned between several positions along the part's contours.The method is by machine manufacturer Colibrium (formerly GE-Additive), denominated as "Multispot".A specific distance is melted at each position, typically in the millimetre range, defined by combining the spot time and beam speed parameters.Due to the beam being non-stationary, the melting strategy is, in reality, short-line melting or quasi-spot melting.
Surface characteristics of parts manufactured using powder bed fusion technologies, such as PBF-EB and PBF-LB, differ from those of traditionally manufactured components.In traditional manufacturing, the surface roughness is mainly driven by factors like the feeding and geometry of the cutting tool, where each pass of the tool leaves a profile on the surface.In PBF-EB, the factors responsible for the resulting surface roughness are more complex.The surface is ordinarily anisotropic with visible features perpendicular to the build direction originating in the layer-upon-layer nature of the manufacturing method.Powder feedstock grain size distribution [4], processing parameters [4], part geometry [5], part position in the building envelope [6], and melting strategy [1] are other factors that have been shown to influence surface roughness.Whilst several studies exist on the subject, it is unclear to what degree each different mechanism impacts the resulting surface roughness.The interdependency of various factors and phenomena makes conclusions even more complex.
A high surface roughness has several drawbacks: the sometimes required post-processing adds extra time and cost, the part's aesthetic appearance may be reduced, and the increased roughness sometimes results in reduced resistance to pitting corrosion [7].Surface roughness is also a strong predictor of a part's expected fatigue properties, where a higher surface roughness tends to result in reduced fatigue life [3,8].
Surface topography characteristics have three different origins [1].The staircase effect is caused by the layer-wise nature of PBF-EB processing whilst building sloped geometry.The resolution of the slope is dependent on the layer thickness, where an increased layer thickness results in lower resolution and a higher surface roughness.Powder grains adhere to the surface by partial melting and process deviations.Due to high heat diffusion, powder grains in contact with the surface can partially melt and adhere to it.Since powder feedstock designed for PBF-EB is usually within the 50-150 µm range, adhered partially melted powder grains will significantly affect the surface roughness.
Poorly optimised process parameters with insufficient energy input can cause porous surfaces, lack of fusion defects or other directly related defects.Conversely, if the energy input is too great, molten material can overflow the edges and cause a wavelike surface on the vertical sides.Improperly high-energy input can also cause melt-pool instability and movement [9], causing porosity and ejecta.
Traditionally, surface roughness has been performed by profilometry perpendicular to the cutting direction of the machining tool to characterise the resulting surface after machining.The method is performed by traversing a stylus across the surface whilst recording the movement of the stylus, producing a 2D surface profile.Surface roughness typically consists of a waviness on which the roughness is overlaid.The cut-off length parameter used in stylus profilometry determines the length of each measurement section and is used to filter out the waviness of the resulting 2D profile.The cut-off length must match the surface characteristics to measure only the roughness effectively.The standard methods for stylus profilometry measurements are described in ISO 21920 [10].With the introduction of AM, the question has been raised about whether the traditional method of surface roughness characterisation adequately describes the surface roughness of AM parts.Townsend et al. [11] describe how stylus-based profilometry is still the most common method for AM surface characterisation whilst the areal characterisation approach of, for instance, focus variation, is gaining acceptance as the current best practice method for characterising a three-dimensional surface.Focus variation is a method that utilises a microscope with a limited field of depth to capture a series of images whilst moving the lens vertically along the optical axis.Each pixel in the image stack is then evaluated to determine the height along the optical axis that has the maximum contrast, at which the surface for the specific pixels x-y position is determined to be located in the z-position.The process is repeated for every pixel, resulting in a 3D topographical map [12].
Amongst the various parameters available to quantify surface roughness, Ra (arithmetic mean roughness) and Sa (arithmetical mean height) are widely recognised and utilised due to their robustness in representing surface characteristics.Ra provides a straightforward measure of the average surface roughness, typically used in quality control processes where consistent monitoring is necessary for adhering to product specifications.Sa, the areal counterpart to Ra, provides a more detailed analysis covering an area instead of a line, providing a more holistic understanding of the surface topography.Whilst several other surface metrics exist, such as peak to valley within the cut of length (Rz) or total height of the roughness profile (Rt), Ra and Sa are the most robust in terms of being affected by outliers and are the most commonly used.Table 1 shows a representative selection of published work containing Ra and Sa surface roughness measurements parallel to the building direction of parts built using PBF-EB.
The current literature shows that contouring parameter development is typically done by trial and error, with samples manufactured for each parameter to test.Although robust, this approach is time and resource-consuming, and the machine control software often limits the number of parameters possible to test in each build.
With the introduction of research mode to the EBM-control software, it is now possible to run custom beam control code.Beam control code is generated using scripts typically written in Python.The resultant code has similarities to the traditional g-code used by milling machines and lathes.However, it provides lines of code in the format (X coordinate, Y coordinate, beam current, focus offset, dwell time) for spots and in the format (X-coordinate start , Y-coordinate start , beam current, focus offset, beam speed) (X-coordinate end , Y-coordinate end ) for lines.This structure provides precise control over melting parameters and is not affected by the limitation of the EBMcontrol software regarding the maximum number of parameter sets possible to use simultaneously.
The study introduces a Python script-based approach to generate beam control code, aiming to explore a vast processing parameter space for both line melting and spot melting of contours.A key novelty lies in the spot-melting strategy, which deviates from the traditional quasi-spot approach.Instead of short-line melting with a moving beam, this study employs a stationary beam whilst melting, presenting a novel approach to spot melting of contours.Furthermore, investigating the beam behaviour upon exiting and entering a melt pool and the resulting effect on surface roughness adds another layer of novelty to the research.Additionally, the study investigates a novel approach for characterising spot morphology, yielding results that contribute to explaining observed surface roughness variations.These novel methodologies collectively aim to effectively identify processing parameters that produce parts with low surface roughness, a highly desirable property in various applications.

Powder feedstock
In this study, gas-atomized Super Duplex Stainless Steel (SDSS) 2507 (EN 1.4410, UNS S32750), supplied by Sandvik (Sandvik, Sandviken, Sweden), was utilised.This powder has a nominal composition of 25Cr-7Ni-4Mo, and the manufacturer's specified grain size range is 45-106 µm.Various tests were conducted to characterise the properties of the powder: a scanning electron microscope (SEM) was used to analyse the powder's morphology.Flowability was examined using a calibrated Hall flow funnel per ASTM B213.The powder's package ratio was determined according to ASTM B212 using a scale with a resolution of 10 -4 g.Notably, the powder used in the study had undergone several recycling cycles via the standard Arcam powder recovery system, followed by sieving through a 150 µm mesh prior to each reuse.

Build setup, software and process parameters
The samples were manufactured using an Arcam A2X system equipped with EBM-Control 6 software and operating in Research Mode.A set of process parameters developed in a previous study [23] was used for hatch melting the bulk of samples.A 50 µm layer height was used with the standard triple raking of the powder feedstock.The build chamber's target pressure level was 2.0 × 10 −6 bar.Previous studies on spot melting of SDSS 2507 for bulk material [24] have given an understanding of suitable spot melting parameters for the current powder-machine combination and were used as a basis for selecting contour parameters used in the design of experiments.Processing parameters for the contours were then widely selected to have a measurable impact on the resulting surface finish.

Line melting of contours
One build consisting of 28 rectangular specimens, with a 15 mm × 15 mm cross section, arranged on a 170 mm × 170 mm substrate (Fig. 1a) was manufactured to investigate the line melting of contours.Each specimen was assigned a specific line energy between 3 and 0.05 J/mm contour line energy.Specimens 1-4 were assigned 3 J/mm down to 1.5 J/mm, in 0.5 J/mm decrements.Specimens 5 to 28 were assigned 1.2 J/mm down to 0.05 J/mm line energy in 0.05 J/mm decrements.For build heights 5 mm to 30 mm, corresponding to level 1, a 12 mA beam current was used (Fig. 1b).The beam current was decreased for each level in decrements of 2 mA per level, reaching 2 mA at level 6.
The focus offset was increased by 0.05 mA for each 50 µm build layer, with each level starting at 5 mA focus offset and ending at 30 mA, resetting to 5 mA at the start of the next level.In this way, each layer represents an individual set of processing parameters, hereafter referred to as a theme, yielding 3000 themes per specimen, for a total of 84,000 themes in the single build.An overview of the processing parameter range used is presented in Table 2.

Spot melting of contours
One build comprising 27 specimens with 15 mm × 15 mm cross sections, each composed of three levels, was used to examine the impact of varying spot-melting parameters on surface roughness (Fig. 2).The specimens were built on support structures for easy removal from the start plate.A full factorial design of experiments was used with four factors at three levels.The parameters (factors) manipulated in this experiment were beam current, spot distance, dwell time, and the number of spots in sequence between the melting of adjacent spots.For this experiment, the focus offset   was fixed at 5 mA.The beam current was set to 10 mA, 7 mA, and 4 mA for specimen levels 1, 2, and 3, respectively (Fig. 2), yielding 81 different themes used in the build.Since each level had a constant processing setting, each level possessed sufficient area for surface analysis, unlike the previous line contour melting experiment.The process parameters are summarised in Table 3. Surface roughness was measured using stylus needle profilometry, eight themes were selected based on the lowest Ra value, and additional samples were manufactured using selected process settings for further analysis.

Melt pool entry/exit vectors
The experiment was conducted using a single specimen of interest with a 15 mm × 15 mm cross section, specimen 5, combined with a group of surrounding specimens, acting as homogenisers of the build environment and intermediate melting locations (Fig. 3).
A melting strategy was defined in which one side of the specimen (side D) was initially melted with the beam entering and exiting the melt pools from within the bulk of the specimen itself, and the other side (side B) had the beam entering and exiting the melt pools in a direction away from the specimen (Fig. 4). Figure 5 depicts detailed spot melting order.In level 1, part 6 is used for intermediate melting and to direct the beam in the desired manner.In level 2, part 4 is used for intermediate melting and directing the beam.
At level 2 of the build, the directions were switched using Part 4 instead of Part 6. Side D in part 5 now had the beam moving in and out from outside the specimen, whilst side B in part 5 had the beam entering and exiting the melt spots from within the specimen (Fig. 6).
By switching sides halfway through building part 5 instead of using two different parts, much of the variability caused by localised beam irregularities can be disregarded since experimental variance is induced without relocation of the melting.Build temperature, processing parameters, and layer geometry were kept constant.To capture any build height-dependent variation in surface roughness, sides A and B in sample 9 were used for control measurements.

Surface roughness measurements
Two surface roughness measurements were performed: stylus profilometry measuring Ra values and focus variation measuring Sa values.A skid-type semi-automatic stylus profilometer (Surtronic 3 + , Taylor Hobson Ltd.Leicester, UK) with a 2 µm radius tip was used to assess the surface roughness rapidly.The cut-off length was 2.5 mm, and the evaluation length was 12.5 mm for each measurement.For the spot melting samples, three measurements centred on the sample surface in the building direction.They were evenly distributed across the surface area and were performed for each of the four sides of each of the three levels for all 28 samples, resulting in 1008 stylus profilometry measurements.Areal surface measurement values and calculations are described in ISO 25178 [25].A mean plane is formed by calculating the average height of all measured points; from this plane, the average distance is calculated to create a Sa value.Sa measurements by focus variation were performed using an InfiniteFocus XL 200 focus variation machine (Bruker Alicona, Graz, Austria).Alicona Measure Suite 5.3.5 software was used for data collection, processing, and generating figures.A 5 × magnification optical lens provided a square analysis area with 2.8 mm sides.Using the image stitching functionality embedded in the software, the measurement area was expanded in the building direction to 10.2 mm for a total of 28.56 mm 2 measuring area located at the centre of each surface.For samples 1-8 in Table 5, each side's surface roughness was measured once over the 10.2 mm × 2.8 mm area.
For sample 5 in the spot melting experiments investigating the effect of beam entering and exiting vectors on surface quality, each section to be analysed was measured three times, centred in height, and evenly distributed across the surface width.

Statistical analysis
The "Statistical Package for Social Sciences" (SPSS, IBM, NY, USA) software was used to analyse surface roughness.The roughness measurements were used to build 5 linear regression models.Models 1-4 were bi-variate regression models used to investigate the effect on surface roughness by the time to adjacent spot, beam speed, line energy and spot energy.Model 5 combined the fundamental parameters of beam current, Spot distance, Dwell time, and number of spots as independent variables into a multivariate regression model.

Electron beam quality
To investigate the spot morphology and provide clues about the beam energy distribution in the A2X system, the machine was carefully calibrated using the EBM-Control calibration routine prior to the experiments.The step size with which the focussing and astigmatism coils are adjusted was reduced to achieve a better, more precise calibration than is typically the case.After calibration, a series of spots were melted on a steel substrate at room temperature.The beam current was set to 10 mA, corresponding to 600 W, and the focus offset was varied from 0 to 30 mA.Two line melts across the steel plate were performed to stabilise the beam current before melting spots.The spots were melted using increasing spot time.The first two spots had a duration of 50 µs and 100 µs.From 100 to 1000 µs, the spot times were incrementally increased by 100 µs.The selection of a suitable spot time setting for beam characterisation by spot imprint analysis was based on the requirement of the beam not forming a complete melt pool but only slightly melting the surface to create an imprint.
A suitable spot duration time was selected from microscopic analysis.Before the spot melting samples were built, a 2 mm × 2 mm square grid of spots was melted over the entire area of the start plate, providing localised information about spot morphology and energy distribution close to the surface of the later built samples.

Line melting
28 specimens of SDSS 2507 were successfully built using line melting as the outer contour melting strategy (Fig. 7).
During processing of the line energy and focus offset variation build, the outer contours started to swell for specimens 1 to 18. Specimens 1 to 18 were processed with line energies between 3.0 and 0.55 J/mm.Outer contour melting was, therefore, switched off in EBM-C, and the swelling subsided.Specimens 1 to 18 were built without outer parameters to act as reference surfaces for specimens built without outer contours.Contrary to swelling, specimens 26-28 showed no difference from the reference specimens in terms Fig. 7 Image of contour build showing visible variations in surface roughness caused by the layer-wise increase in beam defocus of surface appearance, and it was concluded that the corresponding line energies, 0.15-0.05J/mm, were insufficient to melt the powder at the outer contour.Side A of specimens is shown in Fig. 8.
The line melting build included a band of processing parameters, resulting in contours that did not swell and received sufficient energy to melt.A sharper beam generally resulted in a surface that appeared to have lower surface roughness when viewed through a microscope.Still, at the same time, differences between line energies were not as apparent.Since every layer utilises a unique focus offset, examining a single parameter set is difficult.To thoroughly investigate which parameters yield the lowest surface roughness, selecting processing parameters corresponding to areas of low apparent roughness and manufacturing samples using selected parameters would be necessary.Since numerous studies have shown that line melting of contours is inferior to spot melting of contours, this initial experiment is a proof of concept for efficiently exploring a vast parameter space and finding a suitable region for contour melting.

Spot melting
28 specimens of SDSS 2507 were successfully manufactured using spot melting as the outer contour melting strategy (Fig. 9) The results of the profilometry measurements show an orientation-dependant discrepancy in surface roughness between surfaces using the same processing parameters.For a few sides, the maximum difference in surface roughness was more than 10 µm.Side A (facing towards the door) of the specimens possessed the lowest surface roughness in most cases; in cases where side A did not have the lowest surface roughness, side A still possessed nearly the lowest roughness.Therefore, side A was used to select the processing parameters for further evaluation.
Scatterplots of surface roughness and processing parameters (Fig. 10) show trends where an increase in beam current, line energy, and spot energy tends to increase roughness.Based on these trends, an increased beam speed logically leads to decreased surface roughness.Since samples were melted in a non-randomised order starting at number one, level one and specimen one, and ending with number 81, level three of specimen 27, a scatter plot of surface roughness against sample number is presented.The plot shows a downward trend in roughness as the sample number increases, which is reasonable considering that the beam current is lowered for each level.A non-linear relationship exists between surface roughness and spot distance, with an optimum spot distance value between 0.1 and 0.5.
The statistical analysis results (Table 4) show that the best predictor of resulting surface roughness is model 1, containing the spot energy variable with an adjusted R 2 value of 0.178, meaning that the variation in spot energy can explain 17.8% of the variation in surface roughness.The positive variable coefficient indicates that a rise in spot energy leads to an increase in surface roughness, which can also be inferred from the Spot energy vs Surface roughness scatter plot in Fig. 10.Models 2 and 3 have adjusted R 2 values of 0.099 and 0.087, respectively.Model 4 has an adjusted R 2 value of 0.010, though it fails to reach significance.Model 5, containing the base processing parameters altered in the DOE, has an adjusted R 2 value of 0.149; however, most included parameters fail Fig. 8 Specimens of SDSS 2507 were built using line melting as the outer contour melting strategy.Specimens 19 to 25 represent the processing space regarding line energy and focus offset, yielding outer contours that do not swell yet receive sufficient energy to melt.The visible gradient in samples 19 to 25 is caused by the layer-wise increase in beam defocus Fig. 9 28 SDSS 2507 specimens manufactured using spot melting as the outer contour melting strategy Fig. 10 Scatter plots of surface roughness (Y-axis) and processing parameters data (X-axis) of surfaces using spot-melting as contour melt strategy The statistical analysis shows that more parameters have to be included in the models to fully statistically explain the relationship between processing parameters and the resulting part surface roughness.However, implementing more parameters results in an exponential growth in the number of samples needed for a full factorial study.
The factors influencing surface roughness can be divided into direct, indirect, and hardware factors.Direct factors are process parameters, such as layer thickness, spot size (focus offset), and distance between contour lines/hatch.Indirect factors are building environmental factors that can indirectly influence surface roughness.Indirect factors are overall build temperature, powder quality and size distribution, proximity to other parts, part geometry, melting order, and build design.These all affect a part's temperature or local build environment and, thus, the initial temperature when contour melting occurs.The hardware factors are based on machines relying on electron beams requiring careful calibration to function correctly.Regardless of whether this calibration is performed automatically or manually, as done in the A2X machine used for this study, the spot geometry is unlikely to be completely uniform over the entire build envelope and also unlikely to be completely uniform during the whole lifespan of the electron-emitting cathode/filament.The energy distribution profile of an electron beam is also likely not to be completely Gaussian nor consistent for the entire lifetime of the cathode/filament.The level of influence from each factor would require additional studies.One consideration that must be kept in mind is that the result of this study is only viable within the parameter space presented.If trends are extrapolated outside of the parameter space, the dominance of factors would be shifted, and other phenomena dominate.For instance, if the spot energy is increased by a factor of 10, the material would likely evaporate or swell.
The same applies inversely where an insufficient spot energy level leaves the contour porous or incompletely melted.
Eight sets of contour processing parameters were selected based on the results of the stylus profilometry measurements.It is good practice to keep the beam current as steady as possible when melting samples, so the selected processing parameters were sorted based on melting current, and then new samples were manufactured.Table 5 presents the eight parameters chosen for further investigation, the results from the focus variation measurements, and the stylus profilometry roughness results.
It must be borne in mind that the stylus profilometry measurements were performed along the building direction, meaning that no information about roughness is collected in the perpendicular direction.Looking at the topography maps from the focus variation analysis (Fig. 11), results show how spot distances that are too large relative to the spot diameter negatively impact the roughness.Samples 2 and 5, having 0.5 mm spot distance, show a roughness more than twice that of the other samples, 57.6 and 61.9, respectively, originating in the vertical, ridge-like structure seen in Fig. 11c.Each ridge is the outcome of a single spot position on the contour where spots are stacked in each layer.A layer-wise shift in spot position on the contour using a beam power and spot time sufficient to melt through several layers could theoretically resolve the issue.However, it is possible to envisage deteriorating quality concerning leaning surface quality and overhangs resulting from such a melting strategy.Currently, the strategy is to use proper spot distances for the chosen melting strategy, and experiments on alternating spot positions will be saved for future research.
Comparing the results presented in Table 5 to the ones presented in Table 1, all samples from this study have a lower surface roughness.The results are not directly comparable since this is the first surface roughness study of SDSS 2507, and the powder properties, such as grain size and distribution, influence the surface roughness.However, it is remarkable that a single DOE build produces a variety of processing parameters, all producing surface roughness values below that of the current literature, whilst there is also room for further optimisation.

Beam direction vector
A beam direction vector experiment using spot melting was successfully executed using research mode in EBM-Control.SEM images of surfaces reveal that having the beam enter and leave the contour melting inwards (Fig. 12b) lessens the amount of partially melted particles adhered to the surface compared to entering and exiting outwards (Fig. 12a).By comparing side B at levels 1 and 2 and side D at levels 1 and 2, it was observed that the beam direction vector (inwards or outwards) affected side B. Figure 13 shows the topographical maps of level 1(b) and level 2(a) of side B in sample 5.No significant difference in roughness was detected for side D.
Table 6 shows measured Sa roughness values of sample 5 sides B and D at levels 1 and 2. Sides A and B of sample 9 were measured as a reference for detecting build heightinduced differences.
The observed difference in adhered powder grains and surface roughness can only be explained by the single factored alteration between surfaces: the direction of beam travel.It could be argued that the spot morphology changes over time as the filament in the electron gun degrades.However, in the short timeframe of this experiment, that seems highly unlikely.The beam current is kept constant between spots, and melting is avoided by the beam's high traversing speed.When performing more extended distance relocations, a ramping effect occurs where the beam traversing velocity incrementally slows before arriving at the spot melt location.This incremental decrease in velocity means that the beam will deposit more energy closer to the edge thus heating the powder grains adjacent to the melting spot as it arrives, possibly making the powder grains more prone to adhere to the surface.After the melting of a spot, the outward exiting beam deposits more energy in direct proximity to the still liquid melt pool, possibly extending the boundary of the melt pool outwards.Alternatively, partially melting adjacent grains that sequentially solidify and adhere to the edge.An observation that supports the extension of the melt pool is the step that forms at the transition between levels 1 and 2, displayed in Fig. 14.The corresponding shift is also Side B exhibits an effect whilst side D exhibits only a negligible effect, which could be explained by differences in spot morphology, considering the surface's roughness difference.A sharper gradient in spot energy at side D could make the melt pool less sensitive to beam direction travelling vectors.
Another factor to take into consideration is the beam alignment.Beam alignment in the A2X systems is performed manually, and great care is taken to ensure the beam is aligned correctly.However, were the electron beam slightly misaligned in the column, the electron optical effect of focal shift occurring pre-travel would cause the spot to  relocate, most probably affecting the surface finish.It is possible that such an effect could counteract the detrimental effects of the beam entering the spots from outside the sample and exiting the spots outwards, thus masking the impact on side D.

Spot morphology
Melting of the surface layer to form a beam imprint without forming a deeper melt pool was successfully achieved.The initial spot imprints design of experiments showed that 5 mA focus offset and 10 mA beam current necessitates a dwell time of 50-100 µs, a longer duration than 100 µs initiated melt pool formation, and < 50 µs failed to form an imprint properly.An intermediate 75 µs was used to melt the 80 mm × 80 mm grid of spots with 3 mm spacing across a polished 10 mm substrate.Representative spot imprints are presented in Fig. 15.Sequentially, imprints were melted onto a start plate.The start plate was removed, and imprints surrounding the location where a specimen was later to be built were examined (Fig. 16).The start plate was later repositioned inside the A2X build chamber, and a build was completed.The following investigation showed that surface roughness values were higher on sides of a sample where greater length from energy centre to spot edge was present.
Different regions formed in imprints indicate higher energy levels in the central regions of the beam.Worth noting is that the high-energy areas are decentralised in the imprint, yielding transition zones of different lengths in different directions.It could be observed that a sharper transition of beam power yielded a lower roughness.
Reith et al. [26] describe how the beam diameter grows with increased current, indicating that the spot properties are unequal throughout the beam power spectrum.A DOE where the spot size is required to be equal should, therefore, be designed so that a shift in focus offset is implemented based on the beam current.The beam would, however, need to be carefully characterised to determine the compensation required to maintain the spot size across all beam powers.
Finally, the work presented in this paper aims to reduce the use of financial and material resources and facilitate a faster industrial implementation by reducing the work needed to develop process parameters.The same applies to research institutions where the often-limited supply of powder feedstock can be used to reach a higher readiness level of the material for PBF-EB processing compared to the traditional approach of testing one parameter per sample.

Limitations and future research
The main limitation of the present study is that it focuses solely on vertical surfaces.The methodology is, however, equally applicable to developing process parameters for overhanging surfaces.Future work includes investigating the relationship between vertical surface roughness and overhanging surface roughness via the presented methods for the SDSS 2507 material.
The limitation of the spot characterisation methodology lies in the selection of dwell time used for the spots to be analysed.The method already provides information about the shape and energy distribution of the beam.However, since the beam energy distribution is typically Gaussianshaped, an increase in dwell time will increase the apparent spot size since more energy, in absolute terms, will be added to the spot, thus increasing the area receiving sufficient energy to form an imprint.A methodology to accurately find the commonly used full-width half max, FWHM, size of the spot needs to be developed and verified.
More parameters can be optimised.The position of the contours relative to the 3D-model geometry (offset to model), the inner contour parameters, melting order of hatch and contour, and layer thickness are some.The magnitude of these parameters' impact on the surface roughness remains to be investigated.
For future research in AM-process development for new materials, the approach presented will significantly improve the results from shorter exploratory studies by providing fast, good results with small quantities of feedstock material.The method presented can also be used for extensive

Conclusions
In this study, a novel method for finding a contour processing parameter space resulting in low surface roughness has been developed.The resulting surface roughness presented was lower than that obtained using processes and methods described in existing literature.Some conclusions can be drawn from the data: • Statistical analysis shows that an increase in beam current leads to an increase in surface roughness by 0.29 µm Ra per mA increase in beam current.Spot energy showed a similar behaviour but failed to reach statistical significance.• Sa measurements by focus variation microscopy better represent the surface roughness than Ra measurements performed in the building direction by contact stylus profilometry.• Spot morphology and beam energy intensity distribution impact the resulting surface roughness.• For spot melting, the direction of travel to and from the spot impacts the final surface roughness, whereas inward travel away from the contour is beneficial for reduced roughness.
• The method presented effectively generated the best surface roughness of Ra 17.4 ± 1.1 µm and Sa 21.6 µm from a single-build experiment, lower than values found in the published literature for vertical surfaces.Using the parameter gradient approach presented in this paper, a vast processing parameter space can effectively be investigated within a single build by analysing the gradient results.Suitable parameters can be extracted, and samples can be manufactured for thorough analysis.Utilising this approach for process parameter development presents an opportunity to save financial and material resources by reducing the number of builds required to find satisfactory parameters.This aids in the faster industrial implementation of new materials whilst reducing the environmental impact of material usage.
This study has demonstrated the respective impacts of spot morphology and beam travel direction on surface roughness.However, the relationships between spot morphology, travel direction, and surface roughness require more research to be fully understood.

Fig. 1
Fig. 1 Build layout of line-melting of contours experiment.a Placements and numbering of specimens in build volume.b Single specimen description of levels, each layer in each level in each specimen is assigned a unique set of contour processing parameters for a total of 84,000 parameter sets in a single build

Fig. 2
Fig. 2 Build layout of spot-melting of contours experiment.a Placements and numbering of specimens in build volume.b Single specimen description of levels, each level in each specimen is assigned a unique set of contour processing parameters for a total of 81 parameter sets in a single build

Fig. 3 Fig. 4 Fig. 5 Fig. 6
Fig. 3 Build layout of melt pool entry/exit vectors experiment.a Placements and numbering of specimens in build volume.b Single specimen description of levels in specimens 4-6.Specimen 5 is the specimen of interest in the experiment, whilst specimens 4 and 6 facilitate beam control.Specimens 1-3 and 7-9 are single-level specimens used to achieve a homogenous build environment

Fig. 11
Fig. 11 Focus variation topography maps with corresponding SEM images.a Sample 6 (0.1 mm spot distance), b Sample 7 (0.25 mm spot distance), c sample 5 (0.5 mm spot distance) showing the ridge-like structure caused by excessive spot distance

Fig. 12 Fig. 13
Fig. 12 SEM images of surfaces of side B in beam direction vector experiment.a Depicts the lower half of the sample where the beam entered and exited the contour spot melt from outside the melting

Fig. 14
Fig. 14 SEM image of the step between levels of sample 5, side B. The corresponding shift is also present on side D but not on side A or C

Fig. 15
Fig. 15 SEM images of representative beam imprints on polished stainless steel substrate.Processing parameters: Process parameters: 5 mA focus offset, 10 mA beam current, and 75 µs dwell time.

Fig. 16
Fig. 16 Optical microscopic images of beam imprints on a stainless steel substrate.Process parameters: 5 mA focus offset, 10 mA beam current, and 75 µs dwell time.A, B, C, D, indicating position relative to the sample later built

Table 2
Line

Table 4
Linear regression models 1 through 5

Table 5
Spot melting process parameters selected for further investigation based on the initial experimentsThe results from stylus profilometry and focus variation measurements are also presented

Table 6
Results from experiments on the impact of electron beam direction vector on surface roughness