The mechanical response of solids depends on temperature because the way atoms and molecules respond collectively to deformation is affected at various levels by thermal motion. This is a fundamental problem of solid state science and plays a crucial role in metallurgy, aerospace engineering, energy. In glasses the vanishing of rigidity upon increasing temperature is the reverse process of the glass transition. It remains poorly understood due to the disorder leading to nontrivial (nonaffine) components in the atomic displacements. Our theory explains the basic mechanism of the melting transition of amorphous (disordered) solids in terms of the lattice energy lost to this nonaffine motion, compared to which thermal vibrations turn out to play only a negligible role. It predicts the square-root vanishing of the shear modulus G ∼ √ Tc − T at criticality observed in the most recent numerical simulation study. The theory is also in good agreement with classic data on melting of amorphous polymers (for which no alternative theory can be found in the literature) and offers new opportunities in materials science.PACS numbers: 81.05. Lg, 64.70.pj, 61.43.Fs The phenomenon of the transition of a supercooled liquid into an amorphous solid has been studied extensively and many theories have been proposed in the past, starting with the Gibbs-DiMarzio theory [1]. All these theories focus on the fluid to solid aspect, coming to this glass transition from the liquid side (supercooling). However, it is only one facet of the problem. For the reverse process, i.e. the melting of the amorphous solid into a liquid, no established theories are available.The problem of describing the melting transition into a fluid state [2-9] is complicated in amorphous solids by the difficulties inherent in describing the elasticity down to the atomistic level (where the thermal fluctuations take place). It is well known that the standard (Born-Huang) lattice-dynamic theory of elastic constants, and also its later developments [10], breaks down on the microscopic scale. The reason is that its basic assumption, that the macroscopic deformation is affine and thus can be downscaled to the atomistic level, does not hold [11]. Atomic displacements in amorphous solids are in fact strongly nonaffine [11][12][13][14], a phenomenon illustrated in Fig. 1. Nonaffinity is caused by the lattice disorder: the forces transmitted to every atom by its bonded neighbors upon deformation do not balance, and the resulting non-zero force can only be equilibrated by an additional nonaffine displacement, which adds to the affine motion dictated by the macroscopic strain.Recently, it has been shown [15] that nonaffinity could play a role in the melting of model amorphous solids, although the basic interplay between nonaffinity, thermal expansion, and thermal vibrations remains unclear. With a number of other models of the glass transition, such as mode-coupling theories, there is also an issue when they rely on liquid-state theory for relating the stress tensor to local ...