🔬 materials science

Coupled Infrared Imaging and Multiphysics Modeling to Predict Three-Dimensional Thermal Characteristics during Selective Laser Melting

This paper presents an integrated experimental and computational framework that couples high-speed infrared imaging with 3D multiphysics modeling to accurately predict and validate the transient thermal conditions, solidification dynamics, and resulting microstructure evolution during selective laser melting of MAR-M247.

Original authors: Vijay Kumar, Kaitlyn M. Mullin, Hyunggon Park, Matthew Gerigk, Andrew Bresk, Tresa M. Pollock, Yangying Zhu

Published 2026-09-04

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Vijay Kumar, Kaitlyn M. Mullin, Hyunggon Park, Matthew Gerigk, Andrew Bresk, Tresa M. Pollock, Yangying Zhu

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine trying to bake a perfect cake, but the oven is so hot that the batter melts into a tiny, fleeting puddle of liquid metal before freezing solid again in a fraction of a second. This is the reality of a manufacturing process called selective laser melting, where a powerful laser beam fuses metal powder layer by layer to build complex parts for airplanes and medical devices. The quality of the final part depends entirely on what happens inside that tiny, molten puddle, known as a melt pool, as it cools down. If the metal cools too fast or too slow, or if the heat spreads unevenly, the resulting material can develop hidden cracks or weak spots that might cause it to fail later. For years, scientists have struggled to see exactly what is happening inside these molten pools because they are too small, too hot, and change too quickly for standard cameras to capture.

To solve this puzzle, a team of researchers at the University of California, Santa Barbara, developed a new way to peer inside the action. They combined high-speed infrared cameras, which can see heat, with a sophisticated computer simulation to create a three-dimensional map of the temperature inside the metal. Instead of guessing what happens beneath the surface, they used the actual heat patterns measured on the surface to drive a computer model that calculates the invisible conditions below. This approach allowed them to see how fast the metal was freezing and how the heat was moving through the liquid, all in real time. By doing this, they could predict the size and shape of the microscopic structures that form as the metal solidifies, which directly determines how strong the final part will be.

The researchers tested this method on a nickel-based superalloy, a material often used in jet engines because it can withstand extreme heat. They fired a laser across a piece of this metal at different speeds and power levels, creating single tracks of melted metal. Using a specialized infrared camera, they captured the heat radiating from the surface of the melt pool thousands of times per second. They then fed this data into a computer program that simulated how heat travels through the metal. Because the camera provided the exact temperature on the surface, the computer didn't need to make many guesses about how the laser interacted with the material. It simply calculated how that heat would flow downward and sideways, revealing the shape of the liquid pool and the boundary where the liquid turned back into solid.

One of the most surprising discoveries was that the heat transfer inside the pool is not uniform. While the bottom of the pool cools steadily, the top surface behaves differently. The researchers found that the liquid metal at the very top moves around due to surface tension changes, creating a flow that mixes the heat. However, they also found that for the conditions they tested, the movement of the liquid was not strong enough to completely dominate the cooling process; the heat still moved primarily by conduction, or direct transfer, through the metal. This meant that their method, which relied on surface temperature measurements, could accurately predict the conditions deep inside the pool without needing to model every tiny swirl of the liquid, saving a massive amount of computer time.

By knowing exactly how fast the metal was freezing and how steep the temperature changes were, the team could predict the size of the tiny crystal cells that form as the metal solidifies. They compared these predictions to actual slices of the metal they cut and examined under a microscope. The match was remarkably close. When they increased the speed of the laser, the metal cooled faster, and the crystals became smaller and more tightly packed. When they slowed the laser down, the crystals grew larger. The computer model predicted these changes with high accuracy, showing that the method could reliably forecast the internal structure of the metal based on the laser settings used.

The study also revealed that the conditions at the very end of the melt pool, where the liquid finally solidifies, are much more chaotic than in the middle. Here, the speed at which the metal freezes changes rapidly, and the direction in which the crystals grow becomes less consistent. This area is where defects are most likely to form, such as cracks or uneven grain structures. The researchers found that the variations in temperature and freezing speed were significantly higher at the tail of the pool than at the front. This suggests that the end of the laser track is the most critical region to monitor if a manufacturer wants to avoid weak spots in the final product.

This work demonstrates that by combining real-world measurements with computer modeling, scientists can see inside the invisible world of metal manufacturing. The method does not require expensive or complex equipment like X-ray machines, which can only see a single slice of the process at a time. Instead, it uses the heat signature on the surface to reconstruct the entire three-dimensional story of how the metal cools. This gives engineers a powerful tool to understand how changing the laser settings will affect the strength and durability of the parts they are building. While the study focused on a specific type of metal, the approach could be applied to other materials and manufacturing processes, offering a clearer path to designing stronger, more reliable components for the future.

Technical Summary: Coupled Infrared Imaging and Multiphysics Modeling for 3D Thermal Characterization in Selective Laser Melting

Problem Statement
Laser heating during additive manufacturing (AM), specifically Selective Laser Melting (SLM), induces extreme, transient thermal conditions that dictate microstructure evolution and mechanical properties. A central challenge in the field is accurately resolving these conditions with sufficient spatiotemporal accuracy. Existing experimental techniques, such as pyrometry and two-wavelength thermometry, often lack the spatial resolution to capture accurate thermal gradients or are limited to surface measurements. While high-speed infrared (IR) imaging offers better resolution, converting radiation signals to true temperatures for high-melting-point alloys remains difficult due to emissivity uncertainties and the limitations of the Stefan-Boltzmann law when applied to narrow spectral bands. Conversely, high-fidelity numerical simulations can provide three-dimensional (3D) data but are computationally expensive, time-consuming, and often require artificial parameter adjustments to match experimental conditions due to unknown thermophysical properties. Furthermore, techniques like X-ray radiography are limited to single cross-sectional planes. Consequently, there is a gap in obtaining experimentally validated, 3D, transient solidification conditions (thermal gradients and solidification velocities) at the solid-liquid (S-L) interface.

Methodology
The authors developed an integrated experimental-computational framework to reconstruct dynamic sub-surface temperature distributions. The methodology involves:

High-Speed IR Imaging: Experiments were conducted on a wrought nickel-based superalloy (MAR-M247) using a high-speed IR camera (>10,000 fps) to capture transient melt pool surface radiation.
Custom Calibration: A custom calibration was created by integrating Planck's law (integrated over the 3–5.5 µm range) with specific experimental parameters, including optics transmission, sensor exposure time, and in-band irradiance, to convert radiation energy to surface temperature maps.
Coupled Multiphysics Modeling: The experimentally derived, time-resolved 2D surface temperature maps were imported as transient boundary conditions into a 3D heat conduction and solid-liquid phase change model (COMSOL). This approach bypasses the need to explicitly model complex phenomena like laser absorption, Marangoni flow, and surface radiation, as these are inherently captured by the measured surface temperature.
Validation and Analysis: The model's predictions of melt pool geometry and solidification conditions were validated against ex situ scanning electron microscopy (SEM) and electron backscatter diffraction (EBSD) maps. The solidification velocity ( $R$ ) and thermal gradients ( $\nabla T$ ) at the S-L interface were calculated and correlated with microstructural features using the Trivedi model to predict cell spacing.

Key Results

3D Melt Pool Reconstruction: The coupled approach successfully reconstructed the 3D melt pool profile and sub-surface temperature distribution. The model-predicted melt pool depths and cross-sectional profiles showed strong agreement with ex situ SEM measurements across various laser processing parameters (e.g., 200 W at 500 mm/s).
Robustness to Property Variations: Sensitivity analysis demonstrated that the predicted melt pool depth varied by less than 1.5 µm despite a 10% variation in input thermophysical properties, highlighting the method's utility for materials with poorly characterized properties.
Fluid Flow and Marangoni Effects: While thermal Marangoni flow was found to slightly homogenize the internal temperature distribution, the study determined that conduction remains the dominant heat transfer mechanism within the bulk of the melt pool (average Peclet number < 1). Consequently, the simpler conduction-only model, driven by experimental surface data, was sufficient to capture the essential solidification conditions without the computational cost of full fluid dynamics modeling.
Solidification Kinetics: The study revealed significant spatial and temporal variations in solidification velocity ( $R$ ). $R$ varied by three orders of magnitude, from $10^{-3}$ m/s at the pool edges to $\sim 100$ m/s at the tail. Notably, maximum solidification velocities near the surface exceeded the laser scan speed ( $R_{max} \approx 1.5 \times V_{laser}$ ), a phenomenon attributed to strong convective effects and local thermal gradients, contradicting simple geometric approximations that cap $R$ at the scan speed.
Microstructural Prediction: The calculated thermal gradients and solidification velocities were used to predict cell spacing via the Trivedi model. The predicted cell spacing distributions (median $\sim 0.7$ µm for 500 mm/s and $\sim 0.52$ µm for 1000 mm/s) closely matched experimental EBSD and SEM measurements. The model successfully captured the trend of decreasing cell spacing with increasing scan speed.
Temporal Variability: Analysis of standard deviations in thermal gradients and cell orientation revealed that thermal conditions are most stable at the beginning of solidification but exhibit significant variability toward the tail of the melt pool. This region of high variability coincides with areas prone to microstructural transitions (e.g., columnar-to-equiaxed transition) and defect formation.

Significance and Claims
The paper claims that this integrated experimental-computational approach is crucial for realizing in situ prediction and optimization of microstructures in commercial AM. By leveraging high-speed IR imaging as a boundary condition, the method achieves high-fidelity 3D thermal characterization with significantly lower computational costs and fewer assumptions than pure numerical simulations. The authors assert that this framework provides a data-rich platform capable of:

Validating and refining microstructural models (e.g., determining material-dependent constants for cell spacing).
Predicting microstructure size and orientation driven by processing parameters.
Identifying regions susceptible to defects and microstructural heterogeneities based on temporal variations in thermal conditions.
Extending to new materials and manufacturing scenarios (such as powder feedstock) where thermophysical properties may be uncertain, as the method relies on experimental boundary conditions rather than assumed material parameters.

The study concludes that while current state-of-the-art techniques provide 2D information or require expensive 3D simulations, this coupled approach bridges the gap by delivering experimentally validated, spatiotemporally resolved 3D solidification conditions essential for controlling AM processes.

More like this