Consider growing a network, in which every new connection is made between two disconnected nodes. At least one node is chosen randomly from a subset consisting of g fraction of the entire population in the smallest clusters. Here we show that this simple strategy for improving connection exhibits a phase transition barely studied before, namely a hybrid percolation transition exhibiting the properties of both first-order and second-order phase transitions. The cluster size distribution of finite clusters at a transition point exhibits power-law behavior with a continuously varying exponent τ in the range 2 < τ (g) ≤ 2.5. This pattern reveals a necessary condition for a hybrid transition in cluster aggregation processes, which is comparable to the power-law behavior of the avalanche size distribution arising in models with link-deleting processes in interdependent networks.PACS numbers: 64.60. De,64.60.ah,89.75.Da Transport or communication systems grow by adding new connections. Often certain constraints are imposed by society and if these constraint involve global knowledge about connectivity, the transition to a percolating system can become first order, as happens for instance when suppressing the spanning cluster, when imposing a cluster size [1] or when favoring the most disconnected sites [2]. Typically this effect is accompanied by the loss of critical scaling making these abrupt transitions less predictable and thus more dangerous. We will show here, that for a specific case, namely a variant of the model introduced in Ref.[3], critical fluctuations and power-law distributions can prevail and for the first time identify a hybrid transition in explosive percolation.Hybrid phase transitions have been observed recently in many complex network systems [4,5]; in these transitions, the order parameter m(t) exhibits behaviors of both first-order and second-order transitions simultaneously aswhere m 0 and r are constants and β is the critical exponent of the order parameter, and t is a control parameter. Examples of such behavior include k-core percolation [6,7], the cascading failure model on interdependent complex networks [8,9], and the Kuramoto synchronization model with a correlation between the natural frequencies and degrees of each node on complex networks [10,11], etc. For the models in [6-9], a critical behavior appears as nodes or links are deleted from a percolating cluster above the percolation threshold until reaching a transition point t c . As t is decreased infinitesimally further as 1/N beyond t c in finite systems, the order parameter decreases suddenly to zero and a first-order phase transition occurs. Thus, a hybrid phase transition occurs at t = t c in the thermodynamic limit. Next we recall discontinuous percolation transitions occurring in generalized contagion models [12][13][14]. Recent studies [15] of a generalized epidemic model [13] revealed that the discontinuous percolation transition turns out to be a hybrid percolation transition (HPT) represented by (1). For this case, a HPT ...