We investigate the impact of droplets of dense suspensions onto a solid substrate. We show that a global hydrodynamic balance is unable to predict the splash onset and propose to replace it by an energy balance at the level of the particles in the suspension. We experimentally verify that the resulting, particle-based Weber number gives a reliable, particle size and density dependent splash onset criterion. We further show that the same argument also explains why in bimodal systems smaller particles are more likely to escape than larger ones.Splashing of liquid droplets upon impact on a solid surface has been investigated for over a century [1][2][3][4][5][6][7][8][9][10][11]. More recently, there has also been a growing interest in what happens to the spreading and splashing if particles are added to the liquid [12][13][14]. On micron scales, ZrO 2 suspensions have been used in studies aiming to optimize ink-jet printing applications [15][16][17][18][19], and on truly macroscopic scales there has been the development of 3D printers that dispense cement slurry [20,21]. In all of these situations, an important concern is to prevent splashing, and particles from escaping, when droplets hit a surface. However, the question of when and why particles are ejected has remained unsettled, and existing experimental studies mostly focus on dilute suspensions.Current models for suspension drop impact associate the onset of splashing with the condition that K = Weexceeds a critical value K 0 , which has been the traditional criterion for pure liquid splashing on dry surfaces at atmospheric pressure in a regime independent of surface roughness [4,22,23]. Here the Weber and Reynolds numbers are defined as We d = ρ l r d U 2 /σ and Re d = ρ l r d U/µ, with r d the droplet radius, U the droplet impact velocity, and ρ l , σ and µ the liquid density, surface tension and dynamic viscosity, respectively.In these models, the addition of particles has been captured by replacing µ with an effective viscosity µ e that increases with packing fraction [12,[24][25][26][27][28]. This predicts that a droplet of a pure liquid that would splash under certain conditions should not splash after adding enough particles. To our knowledge, there exists no systematic study that confirms this prediction. In fact, Nicolas observed [12] that adding particles, instead, lowered the splashing threshold K 0 .To investigate the influence of added particles, we depart from the dilute limit described above, and instead focus here on the limit of dense granular suspensions with volume packing fractions φ = 0.62 ± 0.03, where the discrepancy with the above droplet-scale splash onset criterion is most pronounced.In pure liquid droplets, the size of the ejecta depends on either the destabilization of a thin liquid sheet [5,[29][30][31][32][33][34] or, in the case of prompt splashing, on an instability at the moving contact line [5,10,35]. At splash onset in a suspension, on the other hand, the ejecta are the solid particles (see Fig. 1), which implies a built-in...