Pairwise collisions between terrestrial embryos are the dominant means of accretion during the last stage of planet formation. Hence, their realistic treatment in N-body studies is critical to accurately model the formation of terrestrial planets and to develop interpretations of telescopic and spacecraft observations. In this work, we compare the effects of two collision prescriptions on the core−mantle differentiation of terrestrial planets: a model in which collisions are always completely accretionary (“perfect merging”), and a more realistic model based on neural networks that has been trained on hydrodynamical simulations of giant impacts. The latter model is able to predict the loss of mass due to imperfect accretion and the evolution of nonaccreted projectiles in hit-and-run collisions. We find that the results of the neural network model feature a wider range of final core mass fractions and metal−silicate equilibration pressures, temperatures, and oxygen fugacities than the assumption of perfect merging. When used to model collisions in N-body studies of terrestrial planet formation, the two models provide similar answers for planets more massive than ≈0.1 M _⊕ (Earth masses). For less massive final bodies, however, the inefficient-accretion model predicts a higher degree of compositional diversity. This phenomenon is not reflected in planet formation models of the solar system that use perfect merging to determine collisional outcomes. Our findings confirm the role of giant impacts as important drivers of planetary diversity and encourage a realistic implementation of inefficient accretion in future accretion studies.

在行星形成的最后阶段，地球胚胎之间的成对碰撞是吸积的主要方式。因此，他们在N 中的现实处理- 天体研究对于准确模拟类地行星的形成以及开发望远镜和航天器观测的解释至关重要。在这项工作中，我们比较了两种碰撞处方对类地行星核心-地幔分化的影响：碰撞总是完全增生（“完美合并”）的模型，以及基于神经网络的更现实的模型。接受过巨大撞击的流体动力学模拟训练。后一个模型能够预测由于不完美的吸积和非吸积弹在撞上逃跑碰撞中的演变而导致的质量损失。我们发现神经网络模型的结果具有更广泛的最终岩心质量分数和金属硅酸盐平衡压力、温度、和氧逸度比完美合并的假设。当用于模拟碰撞时类地行星形成的N体研究，这两种模型为质量大于 ≈0.1 M _⊕（地球质量）的行星提供了类似的答案。然而，对于质量较小的最终天体，低效吸积模型预测了更高程度的成分多样性。这种现象没有反映在太阳系的行星形成模型中，该模型使用完美合并来确定碰撞结果。我们的研究结果证实了巨大撞击作为行星多样性的重要驱动因素的作用，并鼓励在未来的吸积研究中现实地实施低效吸积。