In this study, we perform a diachronic analysis of the temporal trends of scholarly articles over a fairly long time scale. Through the retweeting dynamics of 298 highly retweeted articles, we find that nearly 57% of the articles have multiple peaks on diffusion curves after their initial exposure on social media. We characterize this phenomenon by measuring the time interval between peaks, the height and width of peaks, diffusion performance and the similarities between peaks. We discover that bursts of multi-peak articles usually have short durations and small coverage and that the length of the peak interval determines correlations between peaks. We also find that overlapping users between peaks can act like weak ties and spread information from one group to another, catalyzing information to experience a renewed burst of retweets. Based on the presence or absence of overlapping users, we summarize the occurrence mechanism of the multi-peak phenomenon into two types: (1) Cross-community diffusion through overlapping users. (2) Bursty diffusion of key opinion leaders at different times.

在这项研究中，我们对相当长的时间范围内的学术文章的时间趋势进行历时性分析。通过298篇高转发文章的转发动态，我们发现近57％的文章在初次暴露于社交媒体后在扩散曲线上具有多个峰值。我们通过测量峰之间的时间间隔，峰的高度和宽度，扩散性能以及峰之间的相似性来表征这种现象。我们发现多峰文章的爆发通常持续时间短且覆盖范围小，并且峰区间的长度决定了峰之间的相关性。我们还发现，高峰之间的重叠用户可以像薄弱的纽带一样，将信息从一组传播到另一组，从而催化信息经历新的转发。基于重叠用户的存在与否，我们将多峰现象的发生机理归纳为两种类型：（1）重叠用户之间的跨社区扩散。（2）关键意见领袖在不同时期的迅速传播。