We report clear experimental signatures of the theoretically unexpected gas-liquid transition in the first three monolayers of 3 He adsorbed on graphite. The transition is inferred from the linear density dependence of the γ-coefficient of the heat capacity measured in the degenerate region (2 ≤ T ≤80 mK) below a critical liquid density (ρc0). Surprisingly, the measured ρc0 values (0.6∼0.9 nm −2 ) are nearly the same for all these monolayers in spite of their quite different environments. We conclude that the ground-state of 3 He in strict two dimensions is not a dilute quantum gas but a self-bound quantum liquid with the lowest density ever found. PACS numbers: 67.30.hr, 67.30.ej, 67.10.Db, 67.30.ef Matter can in principle be in either gas or a liquid phase at absolute zero if the quantum parameter, the zero-point kinetic energy divided by the potential energy, is large enough. The two-dimensional (2D) helium-3 ( 3 He) system has long been thought as the only material which stays gaseous at the ground state [1]. This system is experimentally realized in 3 He monolayers adsorbed on an atomically flat and strongly attractive graphite surface. Most previous theories based on the variational calculations [2-4], the diffusion Monte Carlo calculation [5] and the Fermi hypernetted chain method [6] support the absence of self-binding of 3 He in 2D. Indeed, no signature of the gas-liquid (G-L) transition was experimentally observed in the first and second layer 3 He on graphite down to T ≈ 3 mK and to areal density ρ = 1 nm −2 [7]. This is in sharp contrast to monolayer 4 He with smaller quantum parameter on graphite. It is well established experimentally [8] and theoretically [9] that in this system the G-L transition takes place at temperatures below 1 K and the self-bound liquid density at T = 0 (ρ c0 ) is 4 nm −2 .The first experimental address to this problem was made by Bhattacharyya and Gasparini [10], who found a kink or small discontinuity near 100 mK in the heat capacity (C) of submonolayer 3 He floated on a thin superfluid 4 He film adsorbed on a Nuclepore substrate. They attributed this to a puddle formation of 3 He in 2D. It is to be noted, however, that in this system the indirect 3 He-3 He interaction mediated by ripplons in the underlying 4 He film, which is not considered in most theoretical works, might be important. In addition, Nuclepore is believed to be a much less uniform substrate than graphite.Recently, Sato et al. [11] found the G-L transition with ρ c0 ≈ 1 nm −2 in the heat-capacity measurements on the third layer of 3 He on graphite down to T = 1 mK. This was inferred from a linear ρ−dependence of γ, the coefficient of the leading T -linear term of C in the degenerate region, as well as a kink at γ ≈ γ ideal . Here2 ) is the γ value of an ideal Fermi gas spreading over the whole surface area (A) of the substrate, and m is the bare mass of 3 He. Note that γ depends only on A and m not on the number of particles in the 2D case. One possible explanation for their result, which contradicts exi...