Recently, diverse phase transition (PT) types have been obtained in multiplex networks, such as discontinuous, continuous, and mixed-order PTs. However, they emerge from individual systems, and there is no theoretical understanding of such PTs in a single framework. Here, we study a spin model called the Ashkin-Teller (AT) model in a mono-layer scale-free network; this can be regarded as a model of two species of Ising spin placed on each layer of a double-layer network. The four-spin interaction in the AT model represents the inter-layer interaction in the multiplex network. Diverse PTs emerge depending on the inter-layer coupling strength and network structure. Especially, we find that mixed-order PTs occur at the critical end points. The origin of such behavior is explained in the framework of Landau-Ginzburg theory. Recently, multiplex networks have become a platform for research in network science because in real-world systems, networks are intertangled and function together. Examples include infrastructure in everyday life such as power stations, transportation systems, information networks, and water supply systems. Multiplex networks can be fragile because failure in one network can cause failure in another, leading to cascading back-and-forth failures. Then, the entire system can exhibit a discontinuous percolation transition [1].Mixed-order phase transitions (PTs) (or hybrid PTs) are also observed in multiplex networks [2]. The size of a giant viable cluster in a multiplex network shows such a PT as a function of the fraction of undamaged nodes. Here, a mixed-order PT means that while the order parameter exhibits a discontinuous PT, the mean cluster size (susceptibility) diverges. Consequently, features of both continuous and discontinuous PTs appear at the same transition point. Mixed-order PTs have been found in several physical models, such as the bootstrap model [3], jamming percolation [4,5], the Ising model with long-range interactions [6], and the synchronization model [7]. However, the mechanism underpinning such mixed-order PTs is not fully understood.Effort has been made to understand the diverse patterns that emerge from cooperative phenomena in complex networks using spin models in thermal equilibrium. For example, opinion formation and the spread of epidemics have been studied using the Ising [8] and Potts [9] models, respectively. Such studies of spin models give some insight into what might happen in complex systems. Motivated by this idea, the Ising model was studied on a double-layer network [10]. Interestingly, it exhibited a discontinuous PT, whereas it shows a continuous PT in a mono-layer network. These results led us to speculate that the type of PT can be changed systematically by controlling the inter-layer coupling strength; in the above examples, zero coupling strength (no connection) results in a continuous PT, whereas finite coupling strength leads to a change in PT type.To investigate the origin of such diverse types of PT in a single theoretical framework systematically, w...