Are your borders secure?

The outer surfaces of AM components are more than just a cosmetic wrapper; they are critical to part quality. They encapsulate and protect the solid 'bulk' material from the external environment and provide a functional interface to neighbouring components. In structural applications, borders also play an important role in determining the component's strength and durability. Any defects on or just below the surface can reduce tensile strength and lead to premature fatigue failure. It is vital, therefore, that our borders are secure.

This article looks at the special conditions that prevail around the periphery of a component made with a laser powder-bed fusion (LPBF) process. We will see how the hatching process used to produce the bulk of the part can leave minor inconsistencies at the periphery. We will then look at techniques to tidy these up to eliminate surface and sub-surface porosity to produce high-integrity components.

Building the 'bulk'

AM parts are built up layer-by-layer. Each slice through the component is produced by moving a focused laser spot across the powder bed surface. The intense spot melts the powder and forms a small, fast-moving melt pool, which solidifies to produce a weld track, the width of which is typically around twice the laser spot diameter.

The majority of the slice is produced by moving the spot back and forth, an activity commonly known as hatching. This hatched region is then finished off with one or more peripheral or border scans.

The hatch pattern is produced in parallel stripes, a bit like mowing a grass lawn. The straight weld tracks partially overlap one another to progressively produce a solid region of material. The weld tracks extend down into the previous layer, re-melting it so that the newly solidified material is securely attached to the layer below. Weld track depths are typically around twice the layer thickness.

It is good practice to rotate the hatch direction from one layer to the next. This technique ensures that the scanning pattern from one layer is not repeated on the next layer, preventing small defects in one layer from propagating through subsequent layers.

The hatching process is generally performed using one of three scan strategies:

1. Meander - in which the spot is moved back and forth from one side of the part to another, which results in scan tracks of varying lengths
2. Stripes - in which the length of the hatch lines are more consistent. The stripes must overlap one another by a small amount to ensure that they are securely attached to one another, especially when different lasers are used to melt neighbouring stripes.
3. Chessboard - in which squares of hatch lines in alternating orientations are produced. In a similar manner to stripes, overlap regions are required around the periphery of each square.

Anatomy of a melt track

If we look at a section through a moving melt pool, we see the incident laser energy creates a shallow cavity. The laser heats the front face of this cavity as it moves, generating a metal vapour plume that is ejected normal to the surface – i.e. upwards and backwards. Heat energy is conducted into the melt pool, which experiences a degree of turbulent flow due to the high temperature gradients within it and the consequent variations in surface tension. This flow will result in some matter being ejected in the form of weld spatter. 

The moving vapour plume creates an environment around the melt pool that is analogous to a weather system. It can entrain powder from next to the weld track, drawing it towards the laser beam through the Bernoulli effect and then ejecting it outwards. Some of this material will be heated as it passes through the laser, whilst other material is blown around by the induced gas flow in the form of 'winds' adjacent to the laser beam. For more details, refer to X marks the spot.

As the laser proceeds across the powder bed, it 'drags' the melt pool along behind it. As soon as the laser spot moves on, heat is quickly lost through radiation and convection, but mostly through conduction downwards into the substrate.

Cooling and solidification

In this way a solidification front follows the laser spot at the same speed as the laser is moving. The initial rate of cooling once the laser spot passes by is very rapid - as much as one million degrees C per second. As the molten metal cools, the melt pool turbulence abates and it begins to solidify.

The solidification process starts with dendritic growth at the outer regions of the melt pool where the molten material is next to solid, cooler substrate. As heat flows from the weld track into the substrate, the dendrites extend inwards and towards the top surface, progressively thickening until solidification is complete, as seen in this section across a weld track.

Temperature profile of a melt pool

Material in any one location therefore undergoes a temperature profile, experiencing rapid heating to its melting point as the laser passes, followed by rapid cooling during and after solidification. The size of the melt pool affects how long this cooling process takes - a larger, deeper melt pool has more thermal energy to dissipate and so takes longer to cool and solidify. Process parameter choices are important here. Experiments by a team at NIST illustrate these effects, showing how an nickel super-alloy 625 melt pool can remain molten at the surface for between 0.2 and 0.6 mm behind the laser spot, and cools at different rates thereafter dependent on the chosen process parameters.

Image above - melt pool dimensions and temperature profiles in nickel super-alloy 625 at a range of process parameter combinations. The size and duration of the melt pool, as well as its cooling rate after solidification, varies with the speed of the laser and the power that is applied.

Thermal history and hatching

So far, we have considered the formation of a weld track as a steady state phenomenon, assuming a fixed substrate temperature before the melting occurs. The reality of hatching, where we stitch together multiple weld tracks to fill a region, is more complex than this. Hatching scans move the laser back and forth, with each new weld track overlapping the previous one.

The thermal history of the previous weld track will therefore influence the conditions in which the new melt pool forms. If the previous weld track is still hot, then the new melt pool will have a higher thermal energy and will thus be larger.

The temperature of the previous weld track will depend on how long it has been since the laser last passed by that location. When the hatch lines are reasonably long, there is plenty of time for the previous weld track to cool towards the ambient substrate temperature. Shorter weld tracks, which we find in thin-walled regions of the part, will have less time to cool, and so the melt pool in these regions will have more thermal energy. One of the goals of using stripes or chessboard scanning strategies is to keep melting conditions more consistent by reducing this variation in the length of hatch lines.

Start of hatch lines

But even with these scan strategies, we still see variation in melting conditions at hatch reversal points. This is where the laser finishes one weld track, steps across and reverses direction in readiness to melt the next hatch line. The laser is turned off briefly (typically for 0.2 to 1.0 milliseconds) during this reversal process to enable the optical system to prepare for the new scan.

At the start of the new hatch line, the laser is melting next to where it was previously working. The previous weld track will still be hot and may even still be molten. But as the laser moves along the new hatch line, it is melting next to material that was last heated longer and longer ago. We therefore have different conditions at the start of each melt track than we do once the scan reaches equilibrium.

These local differences in melting conditions can lead to differences in melt pool size, resulting in the formation of a 'bulb' - a wider, deeper track in the early stages of a hatch vector. The larger melt pool will also generate different amounts of spatter, will cool at a different rate, and produce subtly different microstructure. The metallurgy of these regions at the start of hatch lines can therefore be somewhat different to the centre of the bulk of the part.

One way around this is to scan in one direction only, returning to the same end of the hatch region after each scan rather than reversing direction. This is a little slower than hatching back and forth as the jumps between scans are longer. Whilst this leads to more consistent thermal conditions, it does mean that the ends of the scans are all on one side of the part, which can be a concern as we will see.

Ends of hatch lines

We have seen how the melting conditions at the start of hatch lines can be different to the middle. The ends of hatch lines also experience different conditions, since they are subject to the laser energy being suddenly removed at the finish point. As soon as the laser is switched off, then cooling and solidification begins, but now this is happening not just at the rear of the melt pool, but also at the front. The shrinking melt pool is pulled around by changing surface tension as it cools and can move from its original location. It is common to observe some retraction of the end of the melt track, leaving a narrow 'tail'.

This retraction effect is exacerbated at higher laser powers and speeds, which produce a longer melt pool with more molten material that can move around under these forces.

An observant reader will have noticed that this effect offsets the thickening of the start of the melt track that we discussed earlier. The start 'bulbs' and finish 'tails' of alternating weld tracks can indeed compensate for one another produce a fully-dense hatch region.

Image above - etched nickel superalloy 718 sample showing the ends of hatch lines in alternating directions. The darker starts to hatch lines are thicker than the lighter ends in between.

Uni-directional hatching will leave all the 'tails' on one side of the part, which is likely to result in sub-surface porosity along that edge.

Hatch and border interaction

Another source of sub-surface porosity can be found in the gaps between successive hatch lines when they meet the border of the part at a shallow angle. In these instances, each new hatch line protrudes beyond the end of its neighbour and there can be a considerable distance between them (see the upper border in the image above).

The periphery of the hatch region where these shallow angles are present will be wavy in form, whereas a crisper edge will be produced where the hatch lines are more perpendicular. Examples of wavy edges of hatch regions with shallow hatch angles are shown right. Sizeable ripples are created in the gaps between the ends of neighbouring hatch lines.

Border scans may struggle to close up such gaps between oblique hatch lines. The hatching process will likely denude the region around its periphery of powder, so there may be insufficient material available to plug the gap during the border scan, leaving a defect. An example of this is shown in this image where the border and hatch are almost parallel.

If our part has square edges then it might be possible to limit our hatch angles to prevent these shallow angles, but most AM parts are complex and thus in practice these are very difficult to avoid altogether. So, we need a method to heal such defects.

Border integrity

Border scans encapsulate the part with a smooth skin that seals the bulk whilst preserving the detail of the surface features.

Conventional borders comprise one or two profiles that enclose the hatch region. Typically, one scan will run around the periphery of the hatch - commonly called the 'additional border' - whilst the outermost surface is produced by a second 'border' scan, often with different melting parameters to optimise surface roughness. The positioning of these scans relative to the ends of the hatch lines and the spacing between them are controlled by parameters.

The integrity of the border region is affected by the roughness of its outer surface, any porosity within the border scans themselves, and also by sub-surface porosity at its inner edge where it meets the hatch region. These stress raisers will reduce the tensile strength and fatigue performance of the component.

Impact on tensile properties

Often when we assess the tensile performance of AM materials, we test relatively thick samples with machined surfaces. This focuses on the integrity of the bulk of the component, but in doing so we eliminate the effect of surface roughness and sub-surface defects. Whilst it is helpful to understand these bulk properties to get an assessment of the potential strength of the material that we are producing, this does not always represent the way in which real AM parts are used.

AM parts often have thin walls, and in many cases, we do not want to remove or modify their surfaces. And, as we have already seen, most of the defects in the hatch region are to be found around the periphery.

To assess the impact of these border regions on part strength, we can produce tensile 'dog bones' that are tested without surface modification. In this instance, the specimens are made from nickel superalloy 718, have a width of 9 mm and thicknesses of either 5 mm or 1 mm. The parts are built on a RenAM 500Q multi-laser AM machine, one laser per part, 67 degree hatch rotation and are tested in the 'as-built' state.

Image above - comparison of median strength and ductility of various tensile specimens of nickel super-alloy 718. The 6 mm diameter machined samples, in which the surface regions are removed before testing, show the highest strength and ductility. Un-modified surfaces result in significant drop in modulus, tensile strength and ductility, with this effect being most marked in the thinner (1mm) section samples, where the border region forms a more significant proportion of the cross section.

Border parameters are often tailored to suit the application and can have a significant impact on tensile performance. One way to reduce surface roughness is to use a high-energy border scan. This forms a relatively large melt pool upon which surface tension acts to smooth the surface as it cools. There is a downside to this approach, namely a loss of detail at corners and sharp edges, but this may not be a concern in many cases. An example of the impact of border parameters on tensile performance is shown below.

Image above - comparison of median strength and ductility of 1 mm thick tensile specimens of nickel super-alloy 718, produced without surface modification but with different border parameters. High-energy borders produce a smooth (5 um Ra) surface, whilst lower energy borders preserve more details but produce a rougher (10 - 15 um Ra) surface. The smoother surface provides a notable improvement in ductility, but with an appreciable penalty in terms of strength.

Fill contours

Another useful technique is to augment the conventional borders by one or more fill contours, essentially extra peripheral scans positioned inside of the additional border. The fill contours re-melt the periphery of the hatch region and can be used to heal defects before encapsulating the part with the borders.

Image above - thin-walled samples built with a single fill contour, followed by an additional border and border scan. The fill contour provides a strong, consistent edge to the hatch region, allowing the borders to encapsulate the part without sub-surface defects, despite the shallow angle of the hatch to the border in these samples.

Parameters for fill contours must be carefully chosen. Unlike hatching or normal border scans, these contours are re-melting solid material rather than powder. Solid metal reflects a larger proportion of the incident laser energy, and so fill contours scans need more energy (higher power and/or lower scan speed) than hatch scans to produce a melt track of the appropriate depth.

Image above - tensile comparison of thin-wall (9 mm x 1 mm) samples with un-modified surfaces, with and without fill contours. The fill contour samples exhibit improved ductility, with only a modest loss of strength.

Obviously, fill contours come at a cost - additional time is needed to re-melt metal that has already been processed. The typical impact on build time is a few percent.

So, fill contours can be useful to improve the mechanical performance of AM parts with un-modified surfaces, helping to heal some of the residual defects left by the hatching process. Is it sensible to take this a step further and use fill contours throughout the part, all the way to the centre, thus avoiding hatch defects altogether?

'Total fill' may not be a total solution

'Total fill' scan strategies use a series of concentric scan vectors to fill the bulk of the part, starting at the periphery and moving towards the centre. Whilst this eliminates sub-surface porosity due to hatch defects, other problems arise. Where total fill contours began to meet in the middle of the parts they may be unevenly spaced, leaving under-melting in the core of the component.

Another concern is that the scan pattern is likely to change very little from one layer to the next - unlike hatching where the scan vector rotates between each layer - so defects may be repeated from one layer to the next. For these reasons, this technique is not advised for use on structural parts.

Summary

Border 'security' in AM parts can determine the performance of the component. These boundaries are regions of discontinuity in the hatching process that is used to form the majority of each layer. Defects arise from localised variation in the weld track and gaps where the hatch meets the border at a shallow angle.

Border parameter choice is crucial to overcome these limitations. High-energy border scans and fill contours can both enhance the ductility of thin-walled components with un-modified surfaces.

Next steps

Visit www.renishaw.com/amguide for more education resources and to access downloadable versions of LinkedIn articles by Renishaw authors.