Scientists’ leaving scientific work at an age when they should have continued generating more scientific output inevitably leads to limited success in their careers as scientists. However, this phenomenon has received relatively little attention in science of science and scientometrics, relative to how frequently it occurs in academia. Motivated by this gap, this study seeks to identify and characterize scientists who leave the profession before the expected time in science, i.e., scientists whose productivity digresses from the normal pattern at a stage when they should be capable of publishing more papers. We first argue that simply using the termination of publishing as the sign of leaving the profession is inappropriate. Using termination of publishing as the only standard also leads to an underestimation of the overall productivity of certain types of scientists. Thus, we designed a novel measure to evaluate whether a scientist has left science. Utilizing a large-scale bibliographic dataset from Microsoft Academic Graph (MAG) in the field of mathematics, we identified over 10,000 scientists who left before their 20^th career year, and paired each of them with a scientist who survived for longer than 25 years. We found that the characteristics in the first five years of the careers of the scientists who left early can be summarized as incompetency in their research abilities and a lack of collaboration with senior or highly cited authors. The implications of the findings are discussed.

科学家在他们应该继续产生更多科学成果的年龄离开科学工作不可避免地导致他们作为科学家的职业生涯有限的成功。然而，相对于它在学术界发生的频率，这种现象在科学科学和科学计量学中受到的关注相对较少。受这一差距的推动，本研究试图识别和描述在科学预期时间之前离开该行业的科学家，即在他们应该能够发表更多论文的阶段，其生产力偏离正常模式的科学家。我们首先认为，简单地将终止出版作为离开该行业的标志是不合适的。将终止发表作为唯一标准也会导致对某些类型科学家的整体生产力的低估。因此，我们设计了一种新的衡量标准来评估科学家是否离开了科学界。利用来自 Microsoft Academic Graph (MAG) 在数学领域的大规模书目数据集，我们确定了 10,000 多名在 20 岁之前离开的科学家职业生涯的^{第 1}年，并将他们每个人与一位存活超过 25 年的科学家配对。我们发现，早期离职的科学家在职业生涯前五年的特征可以概括为研究能力不足，缺乏与资深或高被引作者的合作。讨论了研究结果的影响。