Revealing the Five Melting Regimes of Aluminium LPBF: A Synchrotron Journey to Defect-Lean 3D Printing

How in-situ X-ray imaging at 40,000 frames per second uncovered the optimal processing window for single-track AlSi10Mg — and the mysterious 'flickering' keyhole in between

Published: July 2026

Source Research: Additive Manufacturing, Volume 127 (2026), Article 105273

Keywords: Laser powder bed fusion, synchrotron radiography, process map, AlSi10Mg, keyhole dynamics, melting efficiency

Aluminium alloys are the darlings of lightweight engineering — essential for aerospace structures, automotive components, and the intricate thin-wall heat exchangers that keep modern electronics from overheating. Laser powder bed fusion (LPBF) promises to manufacture these components with geometries that no machining operation could ever produce. Yet, if you ask any additive manufacturing engineer about printing aluminium, you are likely to get a sigh in response.

The reasons read like a rap sheet of thermophysical crimes: aluminium reflects up to 91% of the near-infrared laser radiation commonly used in LPBF systems. It conducts heat away from the melt zone two to three times faster than titanium. Its molten viscosity is among the lowest of any engineering alloy — roughly 2.2 mPa·s at the liquidus for AlSi10Mg — making the melt pool prone to violent instabilities. Its surface tension is exceptionally high (~1050 mN/m for pure Al), and its affinity for oxygen creates a tenacious surface oxide that complicates wetting and fusion. Layer on a dramatic drop in hydrogen solubility between liquid and solid aluminium (a factor of roughly 16–20×), and you have a near-perfect recipe for porosity.

Despite these challenges, AlSi10Mg remains one of the most widely used aluminium LPBF alloys, prized for its excellent castability, good thermal properties, and the ability to produce thin-walled structures at high resolution. Yet until now, no one had systematically mapped the full landscape of melting behaviours — from stable conduction to violent keyhole collapse — under the industrially relevant conditions that commercial machines actually use.

A multinational collaboration between University College London, the European Synchrotron Radiation Facility (ESRF), the European Space Agency, Renishaw, and HiETA Technologies set out to change this. Their weapon of choice: the Quad-ISOPR (Quad-laser In Situ and Operando Process Replicator) — a physical twin of Renishaw's RenAM 500Q industrial LPBF system, transplanted into the ID19 beamline at ESRF in Grenoble, France.

The Quad-ISOPR does something remarkable. It takes the exact same 500 W ytterbium fibre lasers, galvanometer mirrors, and optical scan head from a commercial LPBF machine and places them inside a synchrotron beamline — enabling researchers to watch the laser-matter interaction in real time at up to 504,000 frames per second, with an effective pixel size of just 4.3 μm. This is not a simplified laboratory analogue. It is the real industrial process, rendered transparent by polychromatic hard X-rays.

Over a series of beamtimes, the team scanned single tracks of AlSi10Mg powder across additively manufactured substrates across a wide matrix of laser power (200–500 W) and scan speed (400–1600 mm/s) — producing the first-ever synchrotron-derived process map for this alloy under continuous-wave laser conditions.

The radiographs revealed something no prior AlSi10Mg study had captured: five fundamentally different melting regimes, each with its own characteristic keyhole morphology, defect profile, and practical implications for print quality. These are summarized below:

★ Recommended optimal regime for single-track thin-wall builds

The quasi-stable keyhole regime — where a persistent but non-collapsing vapour cavity forms — emerged as the clear winner. Here, melt depths reached 380 μm (6–12× typical layer thickness), melting efficiency hit a maximum of 25%, and energy consumption per unit mass melted dropped by approximately 30% relative to the commonly used conduction and vapour depression regimes. Equally important: no keyhole porosity was observed during steady-state melting.

Perhaps the most scientifically intriguing discovery was the identification of a transitional melting regime — keyhole flickering — that had not been systematically documented in aluminium LPBF before. Located in a narrow band of linear energy density (219–417 J/m), this regime manifests as a rapid, semi-periodic switching between keyhole and conduction melting modes, with cycle times on the order of 1 millisecond.

To resolve the flickering cycle at sufficient temporal resolution, the team pushed their imaging to 504,000 frames per second — a rate at which individual frames capture sub-microsecond dynamics. The mechanism they uncovered is elegant: when a keyhole forms, the associated vapour pressure and surface tension gradients generate a capillary wave that propagates backward along the melt pool surface toward the rear solid-liquid boundary. A partial reflection from this irregular boundary creates a resonant oscillation. When constructive interference overcomes the stabilising recoil pressure, the keyhole collapses into conduction mode. As laser energy continues to be deposited, absorptivity shifts, temperature rises, and the keyhole re-opens — restarting the cycle.

The researchers validated this capillary wave hypothesis through both experiment and high-fidelity multiphase simulation. Crucially, they found that flickering only occurs within a specific melt pool length window of approximately 300–600 μm, corresponding to Strouhal numbers in the range of 0.6–1.2 — a balanced regime where capillary wave and bulk flow transport timescales are well matched. In shorter pools (St > 1.5), capillary waves oscillate too rapidly with insufficient amplitude to destabilise the keyhole. In longer pools (St < 0.6), bulk flow sweeps away surface disturbances before they can develop.

One of the most operationally valuable contributions of this work is a quantitative defect taxonomy linked directly to process parameters. Three defect modalities were systematically characterized:

• Keyhole porosity. In the unstable keyhole regime, the deep, high-aspect-ratio vapour cavity periodically collapses at its tip, trapping a void that is captured by the advancing solidification front. The number of such pores (NKHP) increases linearly with linear energy density: NKHP = 0.096 × LED − 48.97 for LED ≥ 500 J/m (R² = 0.97). The stability transition boundary is best predicted by the front keyhole wall angle: angles exceeding ~83° signal the onset of unstable behaviour.

• Hydrogen porosity. Small (<25 μm), spherical pores form near the solidification front due to the 16–20× drop in hydrogen solubility between liquid and solid aluminium. These were observed in all keyhole-regime tracks but were notably absent in conduction-mode scans — likely because the much smaller melt pools (7% of unstable keyhole volume) and shorter solidification times (0.3 ms vs. 1.6 ms) trap hydrogen in supersaturated solid solution before bubbles can nucleate.

• End-of-track depression (ETD). When the laser switches off at the end of a scan vector, the sudden removal of recoil pressure causes rapid solidification before the melt pool surface fully recovers, leaving a hollow in the track surface. ETDs are most severe in the quasi-stable vapour depression regime, reaching depths of up to 140 μm — a thickness equivalent to several powder layers — and can seed lack-of-fusion defects in subsequent layers.

By overlaying the defect severity distributions onto the P-v process map, the team identified a clear defect-lean window — the quasi-stable keyhole regime at 800–1000 mm/s and 300–400 W — where both keyhole pores and ETDs are simultaneously minimized while maintaining deep, efficient melting.

Beyond defect mitigation, the study makes a compelling economic and energy-efficiency argument for operating in the quasi-stable keyhole regime. The melting efficiency, η — defined as the ratio of energy absorbed for melting to total input laser energy — averaged just 0.14 across conduction and shallow vapour depression regimes. In the quasi-stable keyhole regime, it rose to an average of 0.22, with a maximum recorded value of 0.25.

This ~57% relative improvement in melting efficiency is attributed to the non-linear absorption enhancement that occurs inside a keyhole: the laser beam undergoes multiple reflections within the vapour cavity, and at the front keyhole wall, the angle of incidence approaches the Brewster angle of approximately 84° — maximizing energy coupling into the liquid metal. For a flat aluminium surface, absorptivity is a meagre ~0.09; inside a keyhole, the effective absorptivity can more than double.

The researchers also established that melt pool volume follows a predictable exponential decay relationship with scan speed at constant laser power (R² = 0.95–0.98), providing a robust predictive tool for process designers.

What sets this study apart from prior synchrotron LPBF research is its deliberate grounding in industrial reality. The Quad-ISOPR uses components taken directly from a commercial RenAM 500Q system — including the full optical train and control software. The substrates were printed from the same AlSi10Mg powder on the same machine model, ensuring that the in-situ observations reflect the thermal and microstructural conditions of genuine multi-layer builds, not idealised laboratory coupons.

The process map is therefore not an academic curiosity — it is a ready-to-use engineering reference. For manufacturers printing thin-wall AlSi10Mg heat exchangers, lattice structures, or any geometry where single-track quality determines bulk part integrity, the recommended parameter window translates directly to machine settings on existing commercial equipment.

The dimensionless stability criteria derived in this work — particularly the normalized enthalpy product threshold of H/hmLth = 18 and the front keyhole wall angle boundary of ~83° — provide transferable design rules that can guide parameter selection for other aluminium alloys and potentially other reflective, high-thermal-conductivity materials.

As with any frontier study, this work opens as many questions as it answers. The keyhole flickering phenomenon, while convincingly explained through capillary wave resonance, was observed over relatively few complete cycles per scan. Longer-duration observations and direct measurement of melt pool surface tension under flickering conditions would strengthen the mechanistic picture. The role of scan pattern and inter-layer remelting — both ubiquitous in real LPBF builds — remains to be systematically investigated. And while the defect-lean window was validated for single tracks, its performance in multi-layer, multi-hatch geometries is the next critical test.

Nevertheless, the framework is now in place. The five-regime process map provides a common language for researchers and engineers to discuss aluminium LPBF melting behaviour. The quantitative defect correlations give process designers defensible parameter selection criteria. And the melting efficiency analysis offers a clear economic rationale for operating in the quasi-stable keyhole regime — a sweet spot that, until this study, had never been so precisely defined for the most widely used aluminium LPBF alloy.

The study by Lambert-Garcia et al. represents the most comprehensive synchrotron-based process mapping of AlSi10Mg LPBF to date — and the first to capture it under genuinely industrial conditions. Its five-regime classification, discovery of the keyhole flickering instability, quantitative defect correlations, and identification of a 30%-more-efficient optimal processing window collectively constitute a step change in our understanding of how to print aluminium reliably, efficiently, and without the porosity that has long plagued this material.

For the additive manufacturing community, the message is clear: the quasi-stable keyhole is not something to fear — it is something to harness.

Reference: Lambert-Garcia, R., Fitzpatrick, M., Ruckh, E., et al. (2026). Revealing optimal melting regimes for defect lean single-track aluminium laser powder bed fusion builds. Additive Manufacturing, 127, 105273.