We introduce a simple model of SO(N ) spins with two-site interactions which is amenable to quantum Monte-Carlo studies without a sign problem on non-bipartite lattices. We present numerical results for this model on the two-dimensional triangular lattice where we find evidence for a spin nematic at small N , a valence-bond solid (VBS) at large N and a quantum spin liquid at intermediate N . By the introduction of a sign-free four-site interaction we uncover a rich phase diagram with evidence for both first-order and exotic continuous phase transitions.The destruction of magnetic order by quantum fluctuations in spin systems is frequently invoked as a route to exotic condensed matter physics such as spin liquid phases and novel quantum critical points [1][2][3]. The most commonly studied spin Hamiltonians have symmetries of the groups SO(3) and SU(2) which describe the rotational symmetry of 3-dimensional space. Motivated both by theoretical and experimental [4] interest, spin models with larger-N symmetries have been introduced, e.g. extensions of SU (2) to SU(N ) [5][6][7][8] or Sp(N ) [9].The extension of SO (3) to SO(N ) is an independant large-N enlargement of symmetry, with its own physical motivations [10]. While there have been many studies of SO(N ) spin models in one dimension [11][12][13], our understanding of their ground states and quantum phase transitions in higher dimension is in its infancy. To this end, we introduce here a simple SO(N ) spin model that surprisingly is sign free on any non-bipartite lattice. This model provides us with a new setting in which the destruction of magnetic order can be studied in higher dimensions using unbiased methods. As an example of interest, we present the results of a detailed study of the phase diagram of the our SO(N ) anti-ferromagnet on the two-dimensional triangular lattice.Models. -Consider a triangular lattice, each site of which has a Hilbert state of N states, we will denote the state of site j as |α j (1 ≤ α ≤ N ). Define the N (N −1)/2 generators of SO(N ) on site i asL αβ i with α < β; they will be chosen in the fundamental representation on all sites:L αβ j |γ j = iδ βγ |α j − iδ αγ |β j . Now consider the following SO(N ) [14] symmetric lattice model for N ≥ 3,where the "·" implies a summation over the N (N − 1)/2 generators and ij is the set of nearest neighbors. To see thatĤ J does not suffer from the sign problem, define a "singlet" state on a bond, |S ij ≡ 1 √ N α |αα ij and the singlet projectorP ij = |S ij S ij |. Using these operators and ignoring a constant shift we find the simple form [15],We make four observations: First, it is possible to create an SO(N ) spin singlet with only two spins for all N (in contrast to SU(N ) where N fundamental spins are required to create a singlet); Second Eq. (1) being a sum of projectors on this two-site singlet is the simplest SO(N ) coupling, despite it being a biquadratic interaction in the generatorsL αβ ; Third, since the singlet has a positive expansion,Ĥ J is Marshall positive on any lattice; ...