Mobile Navigation

Curing a Critical Problem Plaguing 3D Printing

Understanding the physics of additive manufacturing of metal parts has helped unearth the causes of a minute problem that can lead to huge problems.

The laser head passes over the powder-bed, pass after pass, building a part from the micron-sized particles below it, a Phoenix of a part rising from the powder, creating the metallic parts that will one day go into engines, hard tooling, structural components … utilitarian, structurally strong metal widgets … but hidden in that part, there is a problem lurking, a problem so small that it can’t be seen by the naked eye.

The problem is porosity of the metal. Small pores and gaps that can lead to defects and failure of parts that were made by the laser powder-bed fusion process, one of the most common additive manufacturing processes used to create metal products.

Researchers at the Lawrence Livermore National Laboratory (LLNL) may have discovered the causal effect of laser melting metal powders that leads to porosity. LLNL researcher Manyalibo Matthews and his team discovered that gas flow, due to evaporation when the laser irradiates the metal powder, is the driving force that clears away powder near the laser’s path during a build. This “denudation” phenomenon reduces the amount of powder available when the laser makes its next pass, potentially causing tiny defects in the finished part.

This is a problem peculiar to the laser melting process. Unlike laser sintering in which heat levels only need to fuse powder particles, the driving force behind this denudation effect, this powder displacement, is evaporation associated with the melting process. “We tend to get the melt so hot that we’re vaporizing the metal, leading to a vapor stream directed away from the surface at such a speed that it creates a low-pressure zone at the melt pool. That low-pressure zone then pulls in surrounding argon gas and that argon entrains the cold powder around the melt pool and casts the powder along the direction of the vapor stream due to the Bernoulli effect (in which pressure falls as fluid velocity increases).

As the powders are “cast about” the area from which the powder has been displaced will have less powder to melt, which can lead to these porosity problems.

Schematic depicting the action of evaporated metal flux on the flow pattern of the surrounding Ar gas and displacement of particles in the powder bed. As shown in the left diagram,particles are either drawn into the melt pool, adding to melt pool material consolidation, or are ejected upward (and rearward).

There are a couple different scales of pores: there are pores on a “micron-ish” scale, due to vapor being trapped. They’re less of a concern than the larger pores, over 50 micron-size pores, that are created either through what’s called keyholing, where you have a hot spot in the build, and the vapor pressure from the evaporating pushes a little pocket down, and that pocket can get trapped, or from incomplete melting, he said.

According to Matthews, incomplete melting can be caused by an irregularly shaped or large particle cast off by the welding process, or surface non-uniformities from missing powder. That’s what this research was about, studying how powder can go missing, through this process called denudation effect. Those are in the tens of microns size range, and can cause part failure.

Lawrence Livermore’s facility for metal-based additive manufacturing houses five powder-bed, laser-based machines. Fine powders (5-50µm) are used to build parts layer by layer. A powder spreader spreads a thin layer of powder on the build platform. The laser melts the powder in locations where the part is to be. When the layer is complete, the build platform is moved downward by the thickness of one layer, and a new layer is spread on the previous layer. The melting and spreading process is repeated as the part is built up.

Graphic (a): Wide field image of denuded zones around melt tracks created by LPBF as a function of laser power and at a scan rate of 2 m/s. The melted track appears as a shiny semi-continuous line. The denuded zone surrounds each track and appears dark in contrast above the track and light in contrast below the track. Graphic (b): Measured denudation zone (DZ) and resolidified track widths as a function of laser power, scan rate and ambient Ar pressure.

Matthews said that as the laser melting process begins, the temperatures approach near to, or at, the boiling point of the metal, so there is a strong vapor flux emitted from the melt pool. Utilizing a microscope setup custom-built for the laboratory, a vacuum chamber, and an ultra high-speed camera from LLNL’s High Explosives Applications Facility, researchers were able to observe the ejection of metal powder away from the laser. Employing computer simulation and the principles of fluid dynamics, they were able to build models explaining exactly how the particles were moving and affecting the printing process. 

The scan strategy used, which is a combination of laser power, beam size, scan speed, and hatch spacing, effects porosity and void generation. While beam size and other factors were important, hatch spacing is the more critical aspect of the process.

Montage of 1.2 0.25 mm optical micrographs (top) and height maps (bottom) of the solidified melt track within a powder layer following scanning laser exposure at 225W and 1.4 m/s as a function of ambient Ar pressure. Three distinct regions can be identified near the laser path center, namely track accumulation zone, the denuded zone (DZ), and the background powder zone.

“What is referred to as scan strategy involves exactly how you’re taking the laser and you’re scanning it over the powder,” said Matthews. “You can scan in a serpentine pattern, where you go back and forth, or you may scan just in one direction. However you do it, you’re necessarily scanning next to a track that you just created. Then, you use that scanning [pattern] over and over until you fill up a region.

“Hatch spacing is the spacing between the tracks. ‘Are you going to space them 10um apart? 100µm apart? What’s the right spacing?’ If you don’t choose that wisely, you end up with these little channels, where there’s no powder in between two tracks. Along with the spacing, you also want to rotate your pattern so that you’re not compounding surface morphologies associate with each layer. If you do that, you end up with very tall single tracks next to little valleys. You end up with troughs in between your scan tracks. By rotating it, you reduce the effect of powder being displaced.”

Answering these questions will help the LLNL researchers build statistical models from which to compare and better understand future part manufacture.

To develop a broader range of statistical models, the researchers used a variety of powders with different characteristics, such as boiling points. “We were using Titanium 6-4, and stainless steel through 316L, and also aluminum 4032. Aluminum has a low melt point, but Titanium has a high melt point.”

According to Matthews, all three types of material had their differences. One unique finding, not modeled in previous research, dealt with powders being pulled into a melt pool. “The powder for steel gets drawn in and melts,” he said, “but in the case of Titanium, the powders are drawn in and can stick to the top of the track and do not melt under conditions which tend to lead to good weld tracks.”

An in situ, high speed video recording of the powder bed fusion process for a single layer of metal powder. The videos were recorded at 500k frames per second and show the displacement of powder particles due to metal vapor flux and induced Ar gas flow.

Another unique finding dealt with aluminum. Because of its high reflectivity and high thermal conductivity, the researchers had to increase power to create continuous weld tracks. “The power was such that we would melt the track, but as the powder is displaced, the powder would actually melt as it traveled towards the melt pool, in mid-flight for aluminum. We didn’t see that so much in titanium and steel.”

While there were differences, and each material behaved a little differently, the main effect that was observed was a vapor plume, and displaced powder.

And, that data is being built into these statistical models that may make for better parts, and eventually be used to develop closed-loop systems that can adjust the scanning strategies on fly. “What we were going after was the detailed physics to understand the process and validate models,” Matthews said. “By understanding the physics, we can improve the models and not just improve the process directly, through empirical and phenomenological methods, but we want to be able to take our high-performance computing codes and predict behavior, predict processes, and optimize our builds.”

This might include a closed-loop system, said Matthews, but that would be a long-term goal. “We need to know what we’re correcting, what to correct, how it scales,” he said. “The best way to do that … because you don’t want to run a thousand experiments for every material and every geometry you’re going to come across, is a model. Once you train it, you know that you validated it with good data, to be able to get you there. To predict—based on the material, and the geometry, and the laser parameters—what the optimal build process is. After you have a process going, and it goes off-normal, it goes off the ideal, you want to know how to correct it.”

Yes, modeling is very important for closed-loop systems, but researchers are not quite there yet, he says. At some point, utilizing thermal sensors, a control system might automatically detect the brightness of the thermal emission coming off the melt table, and determine if a pore was created, or might be created, and then stop the build to fix it. Until then, researchers continue to look at how these small voids are created to help determine best practices for fixing them.

Lawrence Livermore National Laboratory