Visualizing the chemical structure and dynamics of particles has been challenging for many students; therefore, various visualizations and tools have been used in chemistry education. For science educators, it has been important to understand how students visualize and represent particular phenomena--i.e., their mental models-- to design more effective learning environments. This study aimed to investigate and compare students' static and dynamic representations of mental models for a fundamental concept of chemistry, atomic structure. Static representations of mental models were expressed as drawings and explanations given on paper, with "dynamic" ones being generated by using animation-developing software. This mixed-method study was implemented in three parts. A total of 523 10th (N = 277) and 11th (246) grade high school students participated in a workshop where they first learned how to use one of three animation-developing software programs (K-Sketch, Chemsense or Pencil; N = 162, 204, 157, respectively), and then prepared an animation of an oxygen atom using that program. Before and after creating the animation, students were asked to draw the structure of the atom and to storyboard the oxygen atom for three seconds. After students generated their animations they were asked to explain their animations in 2-3 minute interviews (N = 324). The static and dynamic representations of mental models were compared statistically by the Wilcoxon Signed Rank Test within each group, and they were compared by the Kruskall Wallis Test between the groups. The results of the analysis showed that in all the groups, a significant difference (p = 0.000) between the initial and final static representations of mental models suggested that students modified their mental models towards a more refined and accurate representation of the atomic structure. Regardless of the software program used, students included significantly more dynamic features (p = 0.000) in their static representations of mental models after generating animations than they did initially. No significant difference (p > 0.05) between any of the features was conveyed in static representations of mental models of students who worked with different software programs. In addition, student-generated animations revealed some misconceptions, such as the movement of the parts of the atom or the atom itself besides electrons, which were not detected on paper.