By extracting all authors of articles published in the 3 most popular international journals of physics education for a period of ten years, some collaboration (co-authorship) networks have been constructed based on the criterion that if any two authors have published a paper together, they are labeled as connected or linked by research collaboration in physics education. We have then obtained numerical values for the average number of collaborators, size of clusters formed between physics teachers as well as values of the clustering coefficient. On comparison with the co-authorship networks of scientists in physics, high-energy physics, biomedical science and computer science, our results show that the physics teaching communities are highly fragmented on a global scale by forming many small clusters (or local groups) around the world. However, the internal collaboration within each cluster is astonishingly strong. In this paper, we shall furthermore discuss the implications of those differences on the world-wide dissemination of new or innovative teaching ideas and practices from one physics teacher or educator to the rest in this teaching profession.