Large language models can answer scientific questions, yet correct outputs do not guarantee that the model represents or utilizes governing physics.
We demonstrate that materials science mechanism information in the open-weight google/gemma-4-E4B-it model exists in three experimentally separable forms: concepts are readable in individual hidden states, constitutive orientation is conveyed through controlled transformations between states, and selected internal representations causally control engineering answers.
We combine matched direct and Jacobian vocabulary readouts, option-free state geometry, a 60-law counterfactual benchmark, and causal interventions.
In 50 held-out materials descriptions, three independently fitted Jacobian lenses reproduced concept ranks, and target-free word sets from both readouts enabled blinded identification of 9 out of 10 mechanism families.
A separate 72-prompt benchmark produced mechanism-specific hidden-state neighborhoods, but an exact graph audit revealed that this apparent physical organization was equally explained by numerical comparison.
We then compared otherwise identical prompts where only the direction of physical input was reversed, asking whether the resulting hidden-state movement adhered to the supplied constitutive law.
These state transformations ordered direct, physically neutral, and inverse laws across 60 frozen relations and correctly oriented 39 out of 40 directional laws, while lexical controls were near chance.
Bidirectional interventions shifted answer probabilities toward or away from the physically appropriate outcome across all 12 matched cases, while counterfactual state patches transferred opposing decision signals across mechanisms and answer formats.
Thus, physical relationships were more visible in controlled state changes than in absolute states alone.
Blogger's Review: This article illustrates the potential applications of large language models in material science through multidimensional analysis, emphasizing the recognizability and controllability of physical mechanisms, providing new perspectives and methodologies for future scientific research.