Community Identity in Astrophysics

Communities of the Dark Matter citation network [Duffee, 2018, p97] visualised with Gephi's Force Atlas algorithm, showing the overlap of the citation communities. This network consisted of 47 187 nodes and 161 870 links created from citations to and by 2659 papers on the subject of Dark Matter or MOND (MOdified Newtonian Dynamics, an alternative theory to Dark Matter), lodged in arXiv between 1992 and 2009.

The nodes are coloured according to the largest seven communities found by the Louvain algorithm with, in decreasing order of size:
- purple traditional astrophysics (11.57%),
- green high-energy physics (11.11%),
- blue cosmology and gravitational lensing (10.92%),
- red Dark Energy and alternate theories (10.45%),
- black detection of particles and high-energy physics (8.15%),
- orange Dark Matter models and galaxy clusters (8.15%),
- turquoise MOND, an alternate theory to Dark Matter (7.53%)

Network statistics:
modularity 0.617
average degree 3.43
average clustering co-efficient 0.067

The top seven communities were labelled by examining the titles of the papers in the community and the label was chosen to best represent the character of the majority of the titles. The layout is determined by the citations between papers and it is intuitive to find related areas adjacent to each other, particle physics next to high-energy physics and cosmology close to astrophysics. This layout places high-energy physics between particle physics and astrophysics,
a good location for theoretical work connecting astronomical observation with laboratory experiment. It also places MOND, one of several alternative theories, as overlapping
traditional astrophysics and Dark Energy + alternative theories. The separation of galaxy clusters and Dark Matter models from astrophysics could be due to a strongly self-contained research community but would need more investigation to verify that claim.

reference:

Duffee, Boyd (2018) 'Quantifying textual similarities across scientific research communities' Doctoral thesis, Keele University.

基于暗物质（Dark Matter）引文网络的社群结构[Duffee, 2018, 第97页]，通过Gephi的Force Atlas算法可视化，展示了引文社群的重叠性。该网络包含47187个节点与161870条连接，其数据来源于1992年至2009年间提交至arXiv的2659篇关于暗物质或MOND（修正牛顿动力学，MOdified Newtonian Dynamics，暗物质的替代理论）的论文的相互引文。

节点颜色依据Louvain算法识别出的前七大社群划分，按社群规模从大到小依次为：
- 紫色：传统天体物理学（占比11.57%）
- 绿色：高能物理学（占比11.11%）
- 蓝色：宇宙学与引力透镜（占比10.92%）
- 红色：暗能量与替代理论（占比10.45%）
- 黑色：粒子探测与高能物理学（占比8.15%）
- 橙色：暗物质模型与星系团（占比8.15%）
- 青绿色：MOND——暗物质的替代理论之一（占比7.53%）

网络统计指标如下：
模块化度：0.617
平均度：3.43
平均聚类系数：0.067

上述前七大社群的标签通过检视社群内论文的标题确定，选取最能代表多数标题核心特征的表述作为标签。可视化布局基于论文间的引文关系构建，相邻节点对应相关研究领域，符合直观逻辑：粒子物理学与高能物理学相邻，宇宙学紧邻天体物理学。该布局将高能物理学置于粒子物理学与天体物理学之间，成为连接天文观测与实验室实验的理论研究的理想区位。同时，MOND作为暗物质的替代理论之一，与传统天体物理学及暗能量+替代理论社群存在重叠。星系团与暗物质模型社群与天体物理学社群的分离，可能源于该领域拥有高度自洽的研究群体，但该推论仍需进一步研究验证。

参考文献：
Duffee, Boyd (2018) "Quantifying textual similarities across scientific research communities" 博士学位论文，基尔大学（Keele University）。