By Michele Pavan, Research Engineering Manager, Materialise
A KU Leuven research team has demonstrated something the metal AM industry has been pursuing for years: the ability to locally influence microstructure within a single printed part during the build itself, without relying on post-build heat treatment. The results suggest a broader role for multi-laser powder bed fusion than increasing throughput.1
Geometry is solved. Microstructure isn’t
Laser powder bed fusion has largely delivered on its geometric promise of near-net shapes, internal channels, and topology-optimized structures. The case for geometric freedom is well made and widely accepted. The harder question, we all know, is what happens to the material inside those shapes.
Microstructure is what determines whether a metal part performs. Think about strength, ductility, fatigue life, and corrosion resistance. These are not geometric properties but material properties, and in Laser Powder Bed Fusion (LPBF), they are largely governed by the thermal history the material experiences during printing, a history that, until recently, has been something manufacturers endured rather than designed.
The consequence is particularly stark in duplex and super duplex stainless steels, materials prized in oil and gas, chemical processing, and marine applications for their combination of strength and corrosion resistance. That performance depends on a carefully balanced microstructure: roughly equal parts ferrite and austenite. But the high cooling rates inherent to LPBF suppress austenite formation almost entirely. The result, in the as-built state, is a predominantly ferritic part, strong but brittle, with compromised corrosion performance. A furnace heat treatment after printing has been the only practical route to the right microstructure, a practice which adds cost, time, and distortion risk and, crucially, delivers a uniform microstructure across the whole part, when what engineers actually need is the ability to vary properties locally.
“A furnace delivers a uniform microstructure across the whole part. That has never been what engineers actually needed.”
First attempts to solve this problem
Researchers have not been idle on this problem. Previous work has shown that some degree of microstructural control during LPBF printing is possible, such as adjusting process parameters to alter cooling rates, exploiting thermal gradients to steer grain morphology, or nudging phase fractions in simpler alloys under tightly controlled conditions. But these approaches have shared the same ceiling: limited spatial resolution, poor transferability across materials, and little evidence that they scale to industrially relevant part sizes or build rates.
Super duplex stainless steel, precisely because its microstructure is so sensitive to thermal history and so critical to performance, has proven particularly resistant. The material that most needs in-situ microstructure control has been the hardest to achieve. This has remained a challenge until a recently published study from KU Leuven made a substantive step beyond earlier demonstrations.
When the second laser becomes a metallurgist
While looking for the solutions, the team at the KULeuven found that the answer was already on the machine.
Multi-laser LPBF are now common in production, but their additional beams have so far been used to improve printing speed. Multiple lasers working in parallel reduce build time by processing different regions of the platform simultaneously. Meaning, the second laser has been used primarily as a productivity tool rather than an engineering tool for the material.
The research team at KU Leuven saw that differently. Their approach, published in Additive Manufacturing in October 2025, uses a dual-laser configuration in which the second laser serves as a trailing heat source. It follows the primary melting laser at a controlled offset, reheating the just-solidified material to a precise temperature window before it cools further. For super duplex stainless steel, the critical window is 800–1200 °C, the range in which austenite can nucleate and grow from the primary ferritic structure. Extend the material’s residence time in that window long enough, and the in-situ phase transformation happens. No furnace required.
The concept is not entirely new: the use of a trailing laser for in-situ heat treatment has appeared in the literature before, but three factors make it impractical. First, the thermal dynamics are extremely complex, with the relationship between trailing laser power, offset distance, beam size, scan speed, and the resulting thermal trajectory through the material is highly non-linear, and the parameter space is effectively infinite. Second, without a way to pre-screen that parameter space computationally, the experimental cost of finding viable settings is prohibitive. And third, even on machines physically equipped with two lasers, the software infrastructure to define truly custom synchronised scan strategies, specifying exactly how two beams relate to each other, vector by vector, simply was not accessible.
The KU Leuven study tackles each of these directly.
Modelling first, printing second
The first challenge, navigating the parameter space, was solved analytically. The team employed a fast thermal model based on the Goldak double-ellipsoidal heat source to compute how candidate combinations of trailing laser power and offset distance affect the temperature field at different depths in the material. Crucially, the model was calibrated against experimental melt pool measurements from single-track scans, then used to generate process maps showing which parameter combinations maximize residence time in the 800–1200 °C window, not just at the surface, which gets remelted in the next layer anyway, but at the depths that will persist through the build.
This approach compressed what would have been an exhaustive experimental campaign into a targeted search. Thousands of parameter combinations were screened computationally; a handful of promising candidates were taken forward for experimental validation, first on single tracks and then on bulk 8 mm cube samples. But translating those parameters into an actual synchronized dual-laser build strategy required something the machine’s standard interface could not provide.
Controlling the laser means owning the toolpath
Solving the third challenge, Software access, required a different kind of infrastructure. The machine used for the research was a dual-laser Prima Additive PrintGenius 150. But executing the kind of synchronized, parameter-specific strategies this research demanded, the software needed to operate at the vector level, specifying, for every individual scan path, which laser fires, at what power, with what timing, and how it relates to the beam running alongside it. That level of access is not available through a standard machine interface, nor is it achievable through customer APIs bolted onto a closed system.
It requires a build preparation environment that is genuinely open and programmable from the ground up. In practice, this meant two things. First, real-time laser coordination at the microsecond level, the fast control layer that determines how the machine actually executes a scan strategy. Second, the ability to script fully custom toolpaths: defining trailing-laser offsets, synchronised vector timing, and the dependencies between primary and secondary beam paths that allow one laser to follow the other with precision.
For the KU Leuven team, both were provided by the Materialise Control Platform and Build Processor SDK, which the researchers acknowledge directly in the published paper as central to making the experimental work possible.
Fast control operations ran on the Materialise Control Platform (MCP), the hardware and software layers that manage real-time laser coordination at the microsecond level. On top of that, the Materialise Build Processor SDK (BP-SDK) enabled the research team to define fully custom dual-laser scan strategies, including trailing-laser offsets, synchronized vector timing, and the dependencies between the primary and secondary beam paths required by the experimental matrix.
Without this open, programmable software layer, the research would have stalled at the hardware. The MCP and BP-SDK were not peripheral to the study but bridged the gap between a materials science hypothesis and an executable build strategy on a real machine.
What the experiments showed
By tuning the trailing laser’s power and offset, the team could dial austenite content across the full range of practically relevant microstructures, from 0% under high-speed single-laser conditions to 48% using optimised dual-laser settings, a level of in-process control that has not been demonstrated before on super duplex stainless steels.
The results also revealed an unexpected finding regarding hardness. Even in samples where the overall austenite content changed only slightly, dual-laser processing produced a substantial drop in hardness compared to equivalent single-laser builds. The in-situ reheating from the second laser appears to do more than promote austenite: it modifies the ferrite itself, dissolves brittle nitride precipitates at grain boundaries, reduces residual stress, and homogenises the distribution of alloying elements. The edge cracking visible in hardness indentations of single-laser samples was largely eliminated in dual-laser counterparts.
A logo printed in microstructure
The most striking demonstration of the approach was a visual rather than a statistical one.
To prove that the phase control achieved in bulk samples could be applied with spatial precision at the sub-millimeter level, the team printed two showcase samples. In the first, a smiley face pattern was embedded along the build direction. In the second, the KU Leuven AM Institute logo was reproduced along the scanning direction. In both cases, the image is not visible in the part’s geometry. It appears that only when the sample is sectioned and the microstructure examined, the logo regions print with a near-duplex microstructure, the background with a fully ferritic one.
The interfaces between single- and dual-laser zones are sharp and well-defined, no defects, no microstructural gradients at the transition. The part’s density is unaffected. Combining two different microstructural strategies within the same build does not compromise the integrity of either.
The image is not composed in terms of geometry but in terms of phase. That distinction is the point.
What it means for the industry
The industrial implications run in several directions.
The most immediate is heat treatment. Solution annealing is a significant cost and risk driver for duplex steel component, and it delivers a uniform microstructure across the whole part, whether that uniformity is needed or not. If the right microstructure can be achieved during the build, post-processing requirements can be reduced or removed entirely, with direct impact on lead time, distortion, and part-level cost.
The long-term implication is more transformative because it fundamentally changes the design space when microstructure is a controllable output of the build strategy rather than a side effect of the cooling rate. Think about corrosion-resistant outer surfaces alongside higher-strength internal cores. Magnetically tuned zones. Wear-resistant localised regions. Properties that today require either multiple materials, multiple manufacturing steps, or both can be achievable, in principle, within a single LPBF build, and material properties become as designable as geometry.
There is also a system-level implication on how the industry thinks about multi-laser hardware. As metal AM platforms evolve toward four-, six-, and twelve-laser architectures, the conversation about what those additional lasers are for will need to broaden. The KU Leuven research makes a credible case, backed by experimental evidence, that a second laser used as a precision thermal instrument opens a fundamentally different space.
An honest trade-off
The research is direct about what it has not yet solved. Achieving the near-balanced duplex microstructure currently requires a low scan speed, which substantially reduces the build rate. The optimal dual-laser parameters are efficient metallurgically, but they are not yet efficient industrially. Future work will focus on strategies that apply the slow, microstructure-targeted parameters only where they are needed, using faster single-laser parameters in regions where a predominantly ferritic structure is acceptable, and scaling the approach to larger, more complex geometries where heat accumulation introduces new variables.
The next laser frontier
Metal additive manufacturing has spent two decades establishing that it can produce complex shapes. The KU Leuven research points toward the next question: whether it can produce complex materials, with designed properties, at designed locations, engineered into the build strategy rather than applied afterward.
The answer, based on this work, is that it can. The combination of fast analytical modelling, rigorous experimental validation, and open programmable process software proved sufficient to navigate a problem that had resisted solution through hardware alone. The tools exist. The framework is demonstrated. The path ahead is scaling it.
When thermal history becomes a design variable rather than a process side effect, the boundaries of what metal AM can produce, and what it can promise to the industries that depend on high-performance metal parts, expand significantly. This research is a meaningful step in that direction.
Materialise continues to invest in the open and flexible software infrastructure with its Control Platform, Build Processor SDK, and CO-AM ecosystem to make research like this translatable into production environments. This research from KU Leuven demonstrates what is possible when deep materials science expertise meets programmable, open-process-control software.
Dr. Michele Pavan manages the R&D team at Materialise, focusing on optimizing Laser Sintering processes. His field of expertise ranges across Materials Science, X-ray Computed Tomography and Powder Bed Fusion AM processes.
Journal Reference:
Additive Manufacturing, Vol. 112 (2025), Article 104994. Received 7 May 2025 · Revised 15 Sep 2025 · Accepted 13 Oct 2025 · Available online 13 Oct 2025. DOI: https://doi.org/10.1016/j.addma.2025.104994
Research Credits
Authors (KU Leuven):
Michele Vanini, KU Leuven, Dept. of Mechanical Engineering & Dept. of Materials Engineering; Flanders Make@KU Leuven. Lead author: conceptualization, methodology, investigation, data curation, writing. · [email protected]
Samuel Searle, KU Leuven, Dept. of Mechanical Engineering; Flanders Make@KU Leuven. Methodology, investigation, writing.
Lars Vanmunster, KU Leuven, Dept. of Mechanical Engineering; Flanders Make@KU Leuven. Investigation, writing.
Kim Vanmeensel, KU Leuven, Dept. of Materials Engineering. Supervision, resources, methodology, funding. [Shared last authorship]
Bey Vrancken, KU Leuven, Dept. of Mechanical Engineering; Flanders Make@KU Leuven. Supervision, resources, methodology, funding. [Shared last authorship]
3D Printing Industry is inviting speakers for its 2026 Additive Manufacturing Applications (AMA) series, covering Energy, Healthcare, Automotive and Mobility, Aerospace, Space and Defense, and Software. Each online event focuses on real production deployments, qualification, and supply chain integration. Practitioners interested in contributing can complete the call for speakers form here.
To stay up to date with the latest 3D printing news, don’t forget to subscribe to the 3D Printing Industry newsletter or follow us on LinkedIn.
Explore the full Future of 3D Printing and Executive Survey series from 3D Printing Industry, featuring perspectives from CEOs, engineers, and industry leaders on the industrialization of additive manufacturing, 3D printing industry trends 2026, qualification, supply chains, and additive manufacturing industry analysis.
Featured image shows the KU Leuven AM Institute logo reproduced entirely through in-process 3D local microstructural control, in the as-built state, shown under two optical contrast settings. Photo via KU Leuven.
