Presented at the 73rd Spring Meeting of the Japan Society of Applied Physics

Matlantis Corporation will be making five presentations at the 73rd Spring Meeting of the Japan Society of Applied Physics, which will be held from March 15th (Sun) to 18th (Wed), 2026.

This conference is one of the largest academic conferences in Japan, bringing together researchers from a wide range of fields, including physics, materials, devices, processes, and informatics.

Our presentations this time spanned different sessions, covering phase diagrams, battery materials, semiconductor processes, and enzyme reactions. We hope that they will serve as examples that demonstrate that the general-purpose machine learning potential (Matlantis PFP) can be applied to biomolecular systems in addition to the fields of physics and materials.

We will introduce an excerpt from Matlantis' presentation from the program of the Japan Society of Applied Physics.

[List of our announcements]

■ Ternary temperature phase diagram simulation using general-purpose machine learning potential

Date and time: March 16, 2026 (Monday) 11:00-11:15
Venue: West Building 8, W8E_308
Presenter: Hiroki Kotaka (Matlantis Corporation)

Using the general-purpose machine learning potential PFP, we constructed ternary temperature phase diagrams for In-Ga-As and In-Ga-Sb, taking into account thermal and configurational entropy, and evaluated the thermal stability of the solid solution phase. We were able to confirm that the phase separation was stable at low temperatures, but that the In(1-x)Ga(x)As solid solution phase appeared on the convex hull as the temperature increased. We expect that by combining this method with experimental information such as melting point, we will be able to obtain design guidelines for high-quality alloy semiconductors with desired composition ratios.

■Study of a vapor pressure prediction model based on adsorption free energy calculations of ALD precursors using a general-purpose machine learning force field

Date and time: March 15, 2026 (Sunday) 17:15-17:30
Venue: 70A_101 (70th Anniversary Auditorium)
Presenter: Yusuke Asano (Matlantis Corporation)

To predict the vapor pressure of ALD precursors, we performed adsorption free energy calculations using the generalized machine learning force field (PFP). We applied a planar approximation to adsorption data on various surfaces and found a strong correlation between the resulting thermodynamic coefficients and the Antoine parameters (particularly the latent heat of vaporization term, B). A regression model using this method reproduced experimental values for Co precursors, etc. This is useful as a new screening method for predicting bulk properties from adsorption behavior.

■Enzyme reaction mechanism and free energy analysis using a general-purpose machine learning force field

Date and time: March 15, 2026 (Sunday) 14:40-14:55
Venue: West Lecture Building 1 WL1_301
Presenter: Yoshitaka Yamauchi (Matlantis Corporation)

Using a general-purpose machine learning force field, we calculated the free energy profiles of acylation and deacylation for the all-atom system of the PET-degrading enzyme PETase, which includes solvents, which are difficult to handle using conventional first-principles calculations. The analysis revealed that both processes are two-step reactions via a tetrahedral intermediate, and mutant analysis also yielded results consistent with experimental findings. This study demonstrates that a general-purpose machine learning force field can be a powerful tool for analyzing complex enzymatic reactions.

■Analysis of structural stability and irreversible changes in positive electrode materials for Li-ion batteries using a general-purpose machine learning force field

Date and time: March 18, 2026 (Wednesday) 15:30-15:45
Venue: South Building 2, S2_204
Presenters: Kota Matsumoto, Takahiro Hirai (Matlantis Corporation)

Using the general-purpose machine learning force field PFP and GRRM20 with Matlantis, we comprehensively explored structural changes in Li1-xCoO₂ (x = 0, 0.5). As a result, we found that in the charged state (x = 0.5), a phase transition from hexagonal to monoclinic occurs with an extremely low activation barrier, a result consistent with experimental findings. We also succeeded in identifying the reaction pathway of the irreversible structural change in which Co ions migrate to the Li layer. We demonstrate the effectiveness of this analysis method and report results on its application to other materials.

In addition, Daisuke Okanohara, CEO of Matlantis, will also be giving a lecture.

■ [Open to the public] A must-see for job seekers! Scientific knowledge and the challenges of engineers - The new world of AI and semiconductors

Intelligence created by semiconductor evolution

Date and time: March 16, 2026 (Monday) 13:35-14:35
Venue: West Lecture Building 1 WL1_401
Speaker: Daisuke Okanohara (Preferred Networks, Inc. / Matlantis Corporation)

I look forward to discussing this with you all at the venue on the day.