Cluster synchronization is a phenomenon in which a network self-organizes into a pattern of synchronized sets. It has been shown that diverse patterns of stable cluster synchronization can be captured by symmetries of the network. Here we establish a theoretical basis to divide an arbitrary pattern of symmetry clusters into independently synchronizable cluster sets, in which the synchronization stability of the individual clusters in each set is decoupled from that in all the other sets. Using this framework, we suggest a new approach to find permanently stable chimera states by capturing two or more symmetry clusters-at least one stable and one unstable-that compose the entire fully symmetric network.Synchronization is a collective network behavior in which the states of the interacting units evolve in step with each other [1], as observed in animal flocking [2], the coordinated firing of neurons [3][4][5], and the synchronous operation of power generators [6]. Beyond the complete synchronization of all units, significant progress has been made on understanding more complex forms of synchronization, including cluster synchronization [7][8][9][10][11][12][13] and chimera states [14][15][16][17][18][19][20][21][22][23][24][25][26][27]. In particular, cluster synchronization (CS), in which clusters of nodes exhibit synchronized dynamics, has seen recent breakthroughs: rigorous relations based on group theory have been established between patterns of synchronous clusters and the symmetries of the network structure [11,12]. Network symmetry can be used to explain various forms of collective behavior, such as remote synchronization, in which two nodes are synchronized despite being connected only through asynchronous ones [28], and isolated desynchronization, in which the synchronization of some clusters is broken without disturbing other clusters [11,12,29,30].Chimera states, which are characterized by the coexistence of both coherent and incoherent dynamics within the same state, are also intimately related to symmetry. Since the initial discovery [14] and subsequent analysis [15] of such states, numerous studies have found-numerically, analytically, and experimentally-that chimera states can be observed in a wide range of systems (see the review in Ref. [22] and the references therein). However, it was recently found that chimeras in finite-size networks can be long-lived but transient states [17] (i.e., the system will eventually settle onto a simpler state, such as complete synchronization). This raised a fundamental question: are permanently stable chimeras possible with a finite number of oscillators [31]? Evidence for the affirmative answer has so far been limited to numerical simulations [21,23,26] (notable exceptions are two case studies with stability analysis: one for "weak" chimeras in bistable populations of phase oscillators [27] and the other for a four-node network of delay-coupled opto-electronic oscillators [25]). Our approach for addressing this problem is to identify symmetry-based "templates" fo...