We study the free energy of the worm-like-chain model, in the constant-extension ensemble, as a function of the stiffness λ for finite chains of length L. We find that the polymer properties obtained in this ensemble are qualitatively different from those obtained using constant-force ensembles. In particular we find that as we change the stiffness parameter, t = L/λ, the polymer makes a transition from the flexible to the rigid phase and there is an intermediate regime of parameter values where the free energy has three minima and both phases are stable. This leads to interesting features in the force-extension curves.PACS numbers: 87.15.-v, 05.20.-y, 36.20.-r, 05.40.-a The simplest model for describing semiflexible polymers without self-avoidance is the so called Worm-LikeChain (WLC) model [1][2][3]. In this model the polymer is modeled as a continuous curve that can be specified by a d−dimensional (d > 1) vectorx(s), s being the distance, measured along the length of the curve, from one fixed end. The energy of the WLC model is just the energy due to curvature and is given bywhereû(s) = ∂x/∂s is the tangent vector and satisfieŝ u 2 = 1. The parameter κ specifies the stiffness of the chain and is related to the persistence length λ defined through û(s).û(s ′ ) = e −|s−s ′ |/λ . It can be shown thatThe thermodynamic properties of such a chain can be obtained from the free energy which can be either the Helmholtz's (F ) free energy or the Gibb's (G) energy. In the former case one considers a polymer whose ends are kept at a fixed distance r [one end fixed at the origin and the other end atr = (0, ...0, r)] by an average force f = ∂F (r, L)/∂r, while in the latter case one fixes the force and the average extension is given by r = −∂G(f, L)/∂f . It can be shown that in the thermodynamic limit L → ∞ the two ensembles are equivalent and related by the usual Legendre transform G = F −f r. For a system with finite L/λ, the equivalence of the two ensembles is not guaranteed, especially when fluctuations become large. We note that real polymers come with a wide range of values of the parameter t = L/λ [e.g. λ ≈ 0.1µm for DNA while λ ≈ 1µm for Actin and their lengths can be varied] and fluctuations in r (or f ) can be very large. Then the choice of the ensemble depends on the experimental conditions. Experiments on stretching polymers are usually performed by fixing one end of the polymer and attaching the other end to a bead which is then pulled by various means (magnetic, optical, mechanical, etc.). In such experiments one can either fix the force on the bead and measure the average polymer extension, or, one could constrain the bead's position and look at the average force on the polymer. In the former case, the Gibb's free energy is relevant while it is the Helmholtz in the second case. This point has been carefully analyzed by Kreuzer and Payne in the context of atomic force microscope experiments [4]. Theoretically, the constant-force ensemble is easier to treat, and infact an exact numerical solution has been o...