Gone with the wind - how gas flow governs LPBF performance

Additive manufacturing (AM) machines featuring multiple lasers are becoming increasingly common as manufacturers seek to use laser powder-bed fusion (LPBF) for series production applications. Multi-laser machines offer the promise of reduced build times and lower part costs, strengthening the AM business case for more applications. To fully realise this potential, we need all of our lasers to be firing all of the time, without compromising part quality.

Effective inert gas flow is imperative to deliver both productivity and component quality. In this article we will explore how gas flow contributes to the consistency and quality of AM parts. We will also consider how the design of our gas flow system affects the way in which we can deploy multiple lasers and how this impacts on build rates. In the process, we will debunk a few myths about machine design, opening up new possibilities for rapid part production.

Readers may also want to refer to my articles X marks the spot - find ideal process parameters for your metal AM parts, AM's dirty little secret and the Oxygen algebra series.

Gas flow requirements

In LPBF machines, the chamber gas flow has three main functions:

1. Firstly, we use an inert gas medium to prevent oxidation of our alloy, since we want to preserve its chemistry and hence its mechanical properties. Generating and maintaining an oxygen- and moisture-free atmosphere is critical, especially when we are processing reactive materials.
2. The second function is to carry process emissions away from the melt pool, preserving an unobstructed path for our laser beam down to the powder bed. Of course, more lasers generate more process emissions, and so gas flow becomes even more important in multi-laser machines.
3. Lastly, we want to transport process emissions away from the powder bed and keep the inside of our build chamber clean, minimising accumulation of particles on the chamber walls, ceiling and optical windows. This reduces cleaning during build changeover and prevents long-term damage to optical surfaces.

1. Preventing oxidation

Many materials used in additive manufacturing readily absorb oxygen, especially when they are hot. This change in chemistry can have an unwanted impact on mechanical properties, embrittling the alloy, reducing its ductility and fatigue performance.

Oxygen pick-up

Any oxygen present in the gas flow across the powder bed can be absorbed by the hot alloy, leading to undesirable changes in the chemistry of the solidified component and the process emissions. This highlights the importance of generating and maintaining an inert atmosphere in the build chamber. Moisture is another potential source of oxygen.

Refer to Oxygen algebra - does yours add up? and Oxygen algebra - when it stops adding up for more details of how oxygen is absorbed and redistributed during the LPBF process.

In reactive powders such as Ti6Al4V and AlSi10Mg, argon is generally used as the inert medium as it does not react with the alloy, even at very high temperatures. Nitrogen, which is used as a shielding gas in some AM systems when processing less reactive materials, is not suitable for reactive metals as it will be absorbed. 

Oxygen and nitrogen atoms are absorbed into the metal alloy as interstitial elements - i.e. they occupy the small spaces in between the larger metal atoms. These solute atoms cause lattice distortions that impede dislocation motion (see image right from Wikipedia), increasing the yield stress of the material through a mechanism known as solid solution strengthening. Interstitial elements restrict the slip planes that allow the metal lattice to stretch under load, and so higher concentrations of these elements reduce the ductility of the alloy.

Generating an inert atmosphere

Our build chamber and gas flow system should be designed to generate and hold an inert atmosphere, ideally with minimal waste of the inert gas. An efficient way to do this is to create a vacuum to remove most of the air from the chamber, and then to back-fill with dry argon.

This 2-step vacuum-purge process reduces the initial concentration of oxygen in the chamber and so less venting is needed to generate a low oxygen atmosphere. This reduces the consumption of argon gas compared to systems that simply purge with argon to displace all the original atmosphere. An inert atmosphere can typically be achieved in 10 minutes using the vacuum-purge technique.

Another key machine design consideration is the powder storage and dosing system. Ideally, we want the vacuum-purge process to apply to the whole powder circuit, forcibly moving dry argon through the powder feedstock to displace any air trapped in the powder. This means that an inert atmosphere is achieved throughout the powder circuit more quickly. The sealing required to withstand a vacuum also means that the machine does not leak, and so it consumes very little argon once it is running.

Image above - powder circuit on a RenAM 500M industrial AM machine. Argon flows round the entire circuit, forcing trapped gas out of the powder feed-stock and quickly generating an inert atmosphere.

Maintaining an inert atmosphere

Once we have achieved a sufficiently low oxygen concentration - typically below 1,000 ppm - then we are ready to build. The machine design has important consequences for what happens next. A vacuum-sealed machine will not draw in air from outside, so the oxygen level inside the machine cannot rise. 

Repeated vacuum-purge cycles, each targeting a progressively lower oxygen concentration, can be used to reduce the oxygen concentration further. After several cycles, followed this with a 'gettering layer' - where we raster the laser back and forth across the build plate to deliberately consume any remaining atmospheric oxygen - we can reach 0 ppm oxygen in our gas flow before we dose any powder. For more details, see Oxygen algebra - minimising oxygen pick-up.

So effective gas flow design for processing reactive materials minimises oxidation, by creating a sealed, oxygen-free atmosphere inside the machine.

2. Clearing process emissions

The second function of our gas flow is to clear process emissions away from the melt pool. The LPBF process is not entirely 'clean' - the rapid heating of the powder bed by the moving laser spot results in matter being ejected from the melt pool region. For more details, refer to AM's dirty little secret.

The laser recoil pressure creates a cavity in the melt pool, heating the front face to create a vapour plume. This vapour condenses into fine (typically sub-micron) particles. The vigorous vapour jet also entrains some nearby powder grains through the Bernoulli effect, casting them up into the gas flow. The final type of emission is weld spatter - amorphous material ejected from the roiling melt pool. For more details see X marks the spot.

Ideally we want our gas flow to carry all of this ejected material away from the melt pool and out of the build chamber. There are two main aspects to this:

1. We want the airborne emissions to move away from the melt pool region so that the laser beam has an unobstructed path to the powder bed
2. We want to minimise the amount of heavier emissions such as spatter particles and agglomerates that land in regions of the powder bed that we have yet to melt.

Gas flow design considerations

To remove emissions from the melt pool region, we need a vigorous flow of shield gas. Most LPBF machines use a linear flow of gas across the build plate at 90 degrees to the motion of the dosing wiper. This flow blows the emissions away from the melt pool and draws them towards an outlet port.

The first requirement of our gas flow system is consistency. We are looking for constant velocity flow across the whole build plate to generate consistent melting conditions in all regions. We don't want to see localised regions with low flow rates or re-circulation.

Ideally, we also want our gas flow to be moving faster than our laser scanning speed, so that emissions are cleared from the melt pool region even when we are scanning 'downwind' - i.e. when the laser spot is traversing the powder bed in the same direction as the gas flow. In practice, this means a flow rate of 2 m/sec or more. However, we also need to avoid excessive flow that disturbs the top surface of the powder bed, a particular problem with lighter powders such as aluminium. Hence the rapid flow is directed a few millimetres above the powder bed over a thin layer of slower-moving gas.

Image right - CFD model of gas flow across a build plate, in this case moving from an inlet on the right to an outlet on the left. Flow rates are carefully calibrated to provide consistent conditions across the bed, maximising clearance of process emissions without removing material from the top surface of the powder bed.

Temperature control is also important. Cooler shield gas encourages faster solidification of the vapour jet and the molten emissions. However, faster flow requires a bigger pump, which imparts more heat to the gas as it compresses it. Therefore, an inter-cooler is beneficial on higher productivity machines.

Processing techniques - single laser

We have already considered 'downwind' scanning and the need for sufficient gas flow to remove emissions from the melt pool when we are 'hatching' the bulk of our component, or creating borders around the edge of the part. As well as avoiding the airborne emissions, we ideally also want to minimise the generation of heavier particles.

It should be obvious from the discussion above that even the best gas flow cannot prevent all process emissions from landing in the powder bed. Whilst many will rise into the rapid transverse gas stream, some particles will be emitted at trajectories that will not result in them being carried away. Some emissions, particularly heavier spatter particles, may be ejected at shallow angles such that they are scattered on the top surface of the powder bed around the melt pool. Most will fall downwind, but some will have sufficient momentum to travel upwind. Agglomerates comprising multiple attached emission particles may also settle on the powder bed surface.

What are the implications of this? Ideally, we want our laser melting process to face consistent conditions, with a defined layer thickness of powder to melt and solidify. If spatter particles and agglomerates are landing on top of the powder bed in areas where the laser has yet to pass, then we will have to melt additional material in those locations. The laser will need to deliver sufficient energy to fully melt both the spatter particle and the powder layer, fusing it to the layer below.

Image above - whilst small spatter particles are fully melted, large spatter particles on top of the powder bed can 'shield' powder below, leading to lack of fusion porosity.

The image (right) shows a side-on view of where a large particle (c. 200 microns in diameter) has shielded the powder below, preventing the laser energy from penetrating to the layer below, leading to a void containing several un-melted grains. This type of porosity is likely to reduce fatigue performance.

We can minimise the frequency at which we encounter spatter particles by sequencing our scanning paths such that we start on the downwind side of our build plate and work upwind. We start melting each layer near the outlet port and move progressively towards the inlet port, so that most spatter falls where we have already finished processing. Note that this does not eliminate the risk of encountering spatter on top of the powder bed altogether, as some particles will travel upwind.

The other consideration here is to avoid the generation of excessively large spatter particles. If spatter is much larger than the powder grains, then it may not be fully melted by the laser, or it could be melted but still act to shield powder underneath from the laser energy. This could lead to 'lack of fusion' porosity in our part. A further problem with larger spatter particles is that they can hamper the dosing process for the next layer, leaving a shadow behind them where insufficient powder is present. Whilst this situation is likely to correct itself due to re-dosing and re-melting on subsequent layers, it is clearly not desirable.

The way to control this is through our process parameters. In X marks the spot - find ideal process parameters for your metal AM parts I explained how a combination of power and scanning speed govern the melting process. To optimise both part quality and productivity, we must find a point along an ideal energy density contour that does not stray too far towards unstable melting conditions.

Spatter generation may limit how far along this contour we can go. Spatter is generated by liquid movement within the melt pool in response to thermal gradients. Higher laser power increases these gradients, and so the volume and size of spatter emissions increases with higher laser power. In some materials, it may therefore be necessary to back away from 'maximum attack' processing to avoid excessive spatter generation.

Processing techniques - multiple lasers

When more than one laser is operating simultaneously, we are generating more process emissions and we face the additional complication of interactions between the lasers. We can manage the spatter issue in the same way that we can for a single laser machine - i.e. using sensible parameters and processing each layer starting at the downwind side and working upwind.

But what about how one laser affects another? Conventional wisdom has it that lasers should not be used downwind of one another, where one laser beam has to pass through the emissions of another. These airborne emissions can act to de-focus the laser beam and may absorb some of its energy, such that melting conditions at the powder bed can be affected.

One way to deal with this concern is to zone the lasers. In some multi-laser machines, each laser is assigned a segment of the bed, with small overlapping regions with neighbouring lasers:

Image above - examples of multi-laser zoning and gas flow configurations

The zones are arranged so that the lasers are never downwind of one another, except in the small overlap region. In dual laser machines, the bed is divided into two stripes that align with the gas flow (left-hand image above). In some 4-laser machines, a quadrant arrangement is deployed with a divergent gas flow using a central nozzle and outlet ports on either side of the bed (right-hand image above). It should be noted that divergent designs produce different gas flow conditions at the centre of the build plate compared to those at the periphery.

Zonal multi-laser systems do not always yield the maximum productivity. If a build is not symmetrically arranged, then one or more lasers can find themselves with nothing to do whilst others complete their work in more heavily-populated zones.

It should be noted that this zoning approach is often driven by the optical system design rather than processing considerations. Where large f-theta optics are used to focus the laser beams, these force the lasers to be separated spatially such that they cannot each address the whole bed, so zoning becomes a necessity.

Image above - dynamic focussing using active optics positioned before the two galvo mirrors (only one of which is shown for simplicity) enables the laser beams to be closely spaced on multi-laser machines. By contrast, the use of large f-theta optics after the galvo mirrors - one per laser - forces the laser beams to enter the build chamber through separate widely-spaced windows, resulting in a zonal configuration. For more details of different optical system designs, see Focussing on laser melting performance.

Zonal multi-laser systems must also tackle the issue of mismatch in the overlap zones. On larger components, border scans will be performed by multiple lasers. If the lasers are not perfectly cross-registered, then discontinuities will be seen in the overlap zones where scans from the neighbouring lasers meet. Cross referencing is established during servicing and so any instability of the relative position of the laser sources due to thermal drift, for instance, can result in such misalignment.

By contrast, dynamic focussing enables the galvo mirrors to be placed closer to one another so that greater laser overlap is possible. On machines with a medium bed size, it is possible to have all lasers addressing the whole build plate - i.e. full overlap. An example of laser fields of view for the Renishaw RenAM 500Q quad laser machine is shown in this image.

Full overlap enables efficient use of all four lasers on each layer, minimising the build time. It also enables border scans on large parts to be performed by a single laser, eliminating surface discontinuities. A single, temperature-controlled galvo mounting (see image) reduces drift in the relative positions of the lasers.

However, overlapping lasers also raise the possibility of downstream processing unless the scan paths are constrained. But is it necessary to constrain them? Does processing downwind of another laser really have a detrimental effect?

Testing at Renishaw has shown no measurable effect in the 'as built' tensile properties of Inconel 625 test specimens with 60 um layer thickness, produced directly downwind of identical specimens with fully synchronised laser processing (i.e. one laser is always directly downwind of its counterpart). Further testing to quantify such effects under more extreme conditions is ongoing, but these first results would seem to imply that a very flexible and productive approach can be taken to scanning strategies in many instances.

Image above - tensile data from six ASTM E8 'dogbone' artefacts, tested 'as built' in the horizontal orientation, located at various positions in the bed on a RenAM 500Q multi-laser machine. Several of these samples are built directly downwind of their neighbours. No significant position-dependent variation can be seen between the measurements.

Image above - tensile data from eight ASTM E8 cylindrical artefacts, tested 'as built' in the vertical orientation on a RenAM 500Q multi-laser machine. Once again, several of these samples are built directly downwind of their neighbours. No significant position-dependent variation can be seen between the measurements.

A word of caution, however. Good tensile properties can be achieved even when significant porosity is present. For applications where fatigue performance is critical, it may still be wise to coordinate multi-laser scanning strategies to avoid downwind processing. Some evidence can be seen of isolated spatter shielding in the downwind samples, leading to a small amount of lack-of-fusion porosity in these artefacts that is not seen in their upwind counterparts. This localised porosity does not affect density and strength, but may reduce toughness and fatigue life. With full flexibility in laser assignment across the whole build plate, however, such situations can be avoided through careful scanning strategy design, at the penalty of a small increase in build time.

3. Maintaining a clean build chamber

The final function of our gas flow is to carry process emissions out of the build chamber and prevent them from building up on surfaces inside the machine. Once the emissions move away from the melt pool, we ideally want them to leave the chamber and pass into the gas flow filtration system for removal, before the clean inert gas is re-cycled back into the chamber. Multi-laser machines, with their additional process emissions, require special consideration.

To keep the chamber clean, it is vital that the gas flows directly across the bed and out of the chamber. We need to avoid eddies inside the chamber where the gas re-circulates and slows down. Slower flow enables heavier particles to drop out of the gas stream. Furthermore, slow-moving process emissions, particularly the fine condensate, will attach themselves to any surfaces that they encounter through Van der Waals attraction. If we have slow-moving and re-circulating flow inside the build chamber, we will see a build-up of coarse material on horizontal surfaces and fine particles on walls and optical windows.

These effects can be countered by careful inlet and outlet port design, coupled with significant top-down gas flow. A large volume of inert gas descending from the roof of the build chamber suppresses re-circulation and accelerates flow near the outlet port, preventing any build-up of heavier particles in this region. Also, finer particles no longer have the opportunity to reach and settle on walls and windows, reducing cleaning time between builds.

Image right - CFD model of flow through multiple inlets in the roof of a build chamber. This generates a large volume of downward-moving gas that prevents re-circulation of process emissions inside the chamber.

A corollary of these improvements to gas flow design for multi-laser machines is that the laser beams now have unobstructed paths to the powder bed. The optical window through which the laser beams enter the chamber is kept clean, and the atmosphere in the upper part of the chamber is clear of airborne particles that could otherwise de-focus the laser beam on its way to the powder bed.

Summary

Inert gas flow is a critical feature of LPBF machines, influencing both part quality and process productivity. A well-designed gas flow system maintains inert conditions to prevent oxidation of our alloy, efficiently clears process emissions away from the melt pool region, and minimises the build-up of emissions inside the build chamber.

Whilst even the best gas flow cannot prevent some heavier process emissions landing in the powder bed, well-chosen parameters and intelligent scanning strategies maximise part quality.

In multi-laser machines, extra process emissions mean that gas flow is even more important. Fully-overlapping lasers enable flexible assignment of scan paths to each laser, enabling process optimisation for build rate and part quality as required by the application.

Next steps

For more details of multi-laser machine design, see this video:

RenAM 500Q: Renishaw's quad laser additive manufacturing system for high productivity

Readers may also find the following articles helpful:

• X marks the spot - find ideal process parameters for your metal AM parts
• AM's dirty little secret
• Oxygen algebra - does yours add up?
• Oxygen algebra - when it stops adding up
• Oxygen algebra - minimising oxygen pick-up