Abstract
For over a century, the Abraham-Minkowski controversy has questioned whether photon momentum increases or decreases upon entering a material medium. While most radiation pressure experiments support the Minkowski model (momentum increase), this study demonstrates that the presence of absorption—either within a half-immersed object or the surrounding medium—fundamentally alters this behavior. Through full-wave simulations and theoretical analysis of Mie objects at air-liquid interfaces, we show that while lossless systems exhibit Minkowski-like pulling forces, lossy systems (plasmonic or absorbing dielectrics) transition to Abraham-like pushing forces. Furthermore, we present a machine learning framework capable of predicting these momentum transfers with over 97% accuracy, offering a rapid alternative to complex electromagnetic simulations.
Introduction
In 1908 and 1909, Hermann Minkowski and Max Abraham proposed two competing theories regarding the momentum of light within matter. Minkowski argued that momentum increases proportionally to the refractive index ($p = n\hbar k_0$), while Abraham contended that it decreases ($p = \hbar k_0 / n$). This debate has persisted for more than a hundred years because experimental evidence often supports both sides depending on the specific physical context.
Recent interest in "tractor beams"—light beams capable of exerting a pulling force—has provided a new lens through which to view this controversy. This paper investigates the momentum transfer to a Mie-sized object (comparable to the wavelength of light) positioned at the interface between air and a liquid. By analyzing the transition between pulling and pushing forces, we identify the physical parameters that dictate whether light follows the Abraham or Minkowski model.
Methodology
Full-Wave Simulations and Stress Tensor Analysis
The research team utilized COMSOL Multiphysics to conduct high-fidelity electromagnetic simulations of a Mie object half-immersed at an interface. To determine the mechanical effect of the light, they calculated the time-averaged Minkowski stress tensor. A critical component of their analysis involved deriving the relationship between the external force and the internal conducting force, specifically focusing on how material "loss" (absorption) influences the net momentum transfer.
Machine Learning Integration
To streamline the identification of force behaviors, the researchers developed a predictive model using the WEKA workbench. They generated a dataset comprising eight physical features:
- Particle radius relative to wavelength
- Particle absorption (loss)
- Refractive indices of the particle and the background media
- Background loss
- Thickness of the background layer
This dataset was used to train several classifiers to distinguish between optical pulling (Minkowski-like) and optical pushing (Abraham-like) forces.
Key Insights: The "Loss" Trigger
The study’s most significant contribution is the identification of absorption as the primary switch between the two momentum models.
- The Lossless Regime (Minkowski): In systems where both the particle and the background medium are non-absorbing, the photon momentum increases at the interface. This results in an optical pulling force, supporting the Minkowski interpretation.
- The Lossy Regime (Abraham): The introduction of even a small amount of absorption—whether in a plasmonic particle (like gold or silver) or a lossy dielectric medium—causes the momentum to decrease. This triggers a transition from a pulling force to a conventional pushing force, aligning with the Abraham proposal.
- The Interfacial Effect: The researchers discovered that the momentum change is remarkably sensitive to the interface. A water layer as thin as 10nm is sufficient to trigger a Minkowski-like pulling force in a lossless system, suggesting that the momentum shift occurs almost instantaneously at the boundary.
Results and Discussion
Force Reversal in Plasmonic Objects
The simulations confirmed that while standard dielectric particles in water experience a pulling force, replacing them with plasmonic materials like gold or silver reverses the force to a push. This provides a clear, experimental pathway to verify the Abraham momentum: by observing the behavior of lossy particles in a tractor beam setup, researchers can directly witness the momentum decrease that has historically lacked experimental validation.
Predictive Accuracy
The machine learning component demonstrated that these complex optomechanical interactions are highly predictable. The Random Forest classifier achieved a peak accuracy of 97.27% in identifying the force type. Other models, including J48 and Random Tree, also performed with high reliability (96.49% and 95.71%, respectively). This suggests that ML can effectively bypass the need for computationally expensive full-wave simulations in experimental design.
Conclusion and Future Work
This research provides a nuanced resolution to the Abraham-Minkowski controversy by demonstrating that neither model is universally applicable; rather, the momentum of light is contingent upon the presence of absorption within the system. By identifying the specific conditions under which photon momentum decreases, this study opens the door for new experimental verifications of the Abraham momentum using plasmonic Mie objects.
Future research may expand these findings to smaller Rayleigh particles or macroscopic scales. Additionally, the integration of machine learning into photonics research represents a promising "sidewalk" that could be applied to other complex problems in light-matter interaction, providing experimentalists with rapid diagnostic tools for predicting radiation pressure.
Summary updated